PDA

View Full Version : FS100 and AF100 compared



Pages : [1] 2

Barry_Green
08-30-2011, 10:31 PM
Sony and Panasonic have both launched large-sensor video cameras, targeted (presumably) at the market of indie filmmakers and video professionals who want to create cinematic-looking footage, as an alternative to the DSLR with its various shortcomings. I was an early adopter of the AF100, and I have been evaluating the FS100, to see how they stack up against each other. Here are some of my findings.

1. General Summation: both are large-sensor, interchangeable-lens, shallow-DOF AVCHD 1080p camcorders with XLR audio inputs and headphone jacks, recording to SDHC memory cards, in the about-$5,000 price bracket. In that respect, both manufacturers are to be applauded for responding to the changing market that has been demanding such a product.

2. Nutshell summation: the Sony FS100 uses a somewhat larger sensor and provides higher potential gain settings for the ability to see in darker conditions. The Panasonic AF100 offers more professional features, frame rates, recording modes, and connectivity, while still providing very shallow DOF and excellent low light performance, on par with the FS100 up to the AF100's maximum 3200 ISO.

3. Design differences: Even a cursory examination of the two products will reveal that they were designed by very different groups who had highly different priorities in mind. The FS100 has a very DSLR-oriented design; it really does seem like a DSLR with XLR inputs and a headphone jack – which is, frankly, what many DSLR shooters had been asking for. It lacks many of the common and expected features and functionality of a conventional video camera (such as SDI output and ND filters, things that DSLRs also lack). The AF100, on the other hand, is from the design school of “give us a pro video camera but stuff in a larger sensor.” It has the features and functionality of their AG-HMC150, HPX170, and new AG-AC160, with the primary difference from them being that it has a large sensor and interchangeable lenses. (For purposes of disclosure, I was invited to attend one of their early focus group meetings, as did several other DVXuser members, cinematographers, ASC members and IATSE Local 600 members. Panasonic asked us what we wanted, and we basically told them to give us an HPX170 with a large sensor.)

These design differences crystallize what is, in my opinion, the primary way to determine which of these is the right choice for any given buyer. If one absolutely loves the way the DSLR works, but just wishes that it had a little better aliasing performance and some XLR ports, then the Sony is aimed towards that buyer. If, instead, a buyer is wanting all the features, connectivity, and capabilities of a professional video camera, but with shallow DOF and interchangeable lenses, the AF100 was designed for that buyer.

4. Performance: I'm sure this will surprise some folks, but I've been exhaustively testing these units, and the result is: these cameras are so close in performance that it's really not even worth doing a whole raft of charts and graphs. I mean, I've done those tests, and really it's nitpicking to choose between them; especially after grading the footage, you can make them look virtually identical. And besides, people usually complain about charts and want to see “real world” footage.

Accordingly, here's an example: this is a scene that was set up for, designed for, and lit for, an FS100. The FS100 had optimal image settings dialed in as per the DIT and director of photography. This scene was actually borrowed from a short film that was being shot on the FS100; I piggybacked onto the film set to get an equivalent shot from the FS100 and the AF100 in a "cinematic" setting. I had previously set the AF100 to image settings that largely matched the FS100; I then graded the shot slightly in post to get them even closer. Here are the results:

http://vimeo.com/28382392
Yes, those really are from the two different cameras. As I said, the results are nearly identical in general shooting. There is no gulf of performance difference between these two cameras, just as a reasonable person would expect – considering that both companies hire the best camera engineers in the world, and given the constraints of the price tag, it's not surprising that the net result is performance that is generally comparable. Configured for maximum range, the AF100 shows 10.5 stops of dynamic range; I got the FS100 to do a tiny bit better, it shows 11 stops on the same chart. Rolling shutter performance is practically identical. Chroma clipping and highlight rolloff are extremely comparable. In terms of sharpness, they're once again very much on par with each other; the AF100 delivers observed alias-free resolution of 700h x 680v; the FS100 delivers alias-free resolution of 670h x 720v, with the net effect being equivalent overall sharpness.

5. Sensitivity and Gain: Overall the cameras perform very similarly, but there are two notable image-performance differences: gain performance, and aliasing/moire. Let's first examine the gain/noise/sensitivity issue. For reference, the Panasonic is calibrated in either decibels of gain or in film-standard ISO values, and its default 0dB rating is 400 ISO. I tested the FS100 and found it responds at exactly 320 ISO. Note – other testers have rated the Sony FS100 at 500 ISO, but I can't replicate that. In my testing I put the AF100 and the FS100 both on Rec709 gamma and Rec709 color matrix, put equivalent Zeiss ZF's on them, set them at the same shutter speed of 1/60th, and pointed them at an identically-lit chart, and set both lenses to exactly f/11. In order to deliver identical brightness, I had to set the AF100 on 320 ISO, meaning that if we're comparing these two, 0dB on the FS100 equates to 320 on the AF100.)

In terms of noise performance, the AF100 is basically noise-free and grainless to ISO 1600, and very low noise up to 3200. You can see my prior test of AF100 ISO performance in this video.

http://vimeo.com/19082240
For comparison, here's the link to the Sony FS100 gain/noise video. I set up the same scenario, same chart, same lens, and configured for lower-noise settings, to create a comparable video.

http://vimeo.com/28379557

By my viewing, the FS100 exhibits low/no grain up to about 15 to 18dB. Anything beyond that has noticeable noise. Coincidentally, an FS100 at 18dB is an ISO equivalent of close to 3200 – which puts these two, once again, on par – at least as far as the AF100's ISO goes. The AF100 tops out at 3200, but the FS100 can go further. It can be turned up to 30dB, which is two stops more sensitivity than 18dB. Yes it results in more noise and grain, but it's still surprisingly manageable even at that level. Compared to cameras of just a few years ago, the AF100 is comparatively nearly grainless throughout its ISO range, and the FS100 is nearly grainless up to about 15dB and picks up from there; even so, at its maximum of 30dB, it's about on par with a traditional camcorder at maybe 9dB.


For a more direct comparison, I re-shot the AF100 at 3200 ISO vs the FS100 at +18dB of gain. Here's a direct pixel-to-pixel side-by-side as well as a 2x blowup version.

http://vimeo.com/28417902


Maximum gain is one thing; a usable range is something else. The Panasonic is designed to operate in an ISO range from 200 to 3200 (or 3 to 3200 when combining its ND filters); the Sony ranges from 320 to 10,240. This results in an interesting difference – the Sony's range is biased heavily towards low/no-light situations, but is basically useless in broad daylight without some manner of heavy external ND filtration. The Panasonic's ISO range is targeted towards all-purpose shooting scenarios from noonday sun down to one candle. (For a lark, I had previously done a shot on the AF100 to show just how much brightness it can get out of a single candle; this shot took place in a pitch-dark room and the only source of illumination was a single candle):

http://vimeo.com/24598583

I've never felt limited by the 3200 ISO of the AF100; after all, I'm used to shooting on film, which maxxes out at 500 ISO, and hey, an entire interview setup on a single candle... well, I'm satisfied that the AF100's ISO range is quite adequate. But there are scenarios where more would always be appreciated; and if you really need more than a clean 3200 ISO, the Sony can deliver more. Its gain switch is targeted towards low light sensitivity and in that regard it performs quite well. If you were in a scenario such as a dimly lit reception hall and had only a slow kit zoom lens, the Sony's additional gain might prove to be quite handy. If, on the other hand, you're in a lit environment such as a film set, the additional gain is complete overkill, at the expense of the ability to lower the sensitivity. And sometimes you really need lower sensitivity! Film goes down to 25 ISO for use outdoors, for good reasons, but the Sony's lower limit is 320. To use a car example, the Sony is like a sports car that can go faster than 200 mph – but can't go slower than 60. Whereas the Panasonic can go from 0 to 120mph. It has a range that is applicable to nearly all shooting scenarios, whereas the Sony has bragging rights to the most gain, even if it makes it an impractical camera for brightly lit scenarios. Someplace where I can see where that additional gain would be handy would be trying to shoot a wedding or similar event in a dark interior, with only a slow kit lens that doesn't open up to more than maybe f/4. In that case, the additional gain could be quite useful.

6. Aliasing: The other image performance difference I found worthy of note is the aliasing performance between the two. Aliasing is the bane of the DSLR; aliasing (and its brother moire) are drawbacks that are part and parcel of the way a DSLR samples its sensor. These cameras have greatly minimized the effect as compared to a DSLR. The Panasonic has been tuned to eliminate all chroma aliasing and nearly all luma aliasing; the Sony appears to be tuned more for apparent sharpness at the expense of exhibiting more aliasing in color and, in certain cases, it can exhibit the chroma moire patterns that are typical of the Canon DSLRs, especially on fine weave fabrics or on bricks. In terms of actual resolved detail they are extremely closely matched, but the Sony can sometimes have the “appearance” of some extra sharpness due to the false detail of aliasing, whereas Panasonic has tuned their sensor to minimize or eliminate as much aliasing as possible. The result is that the Sony can show colored aliasing and chroma moire patterns on some finely detailed patterns, whereas the Panasonic just never will.

Here is an example of the FS100 showing the chroma moire pattern on some bricks (at normal size and then at 3x magnification). The Vimeo compression kind of squishes out all the aliasing on the wide shot, which is why I included the 3x shot. If you want to see it in the wide shot, download the original file instead of the vimeo-recoded streaming version.

http://vimeo.com/28383119
It's not that big of a deal, but it's there, even though it certainly is reduced as compared to the Canon DSLRs. In other words, there are many scenarios where the Canons would show the rainbow patterns, where the Sony will avoid or resist showing it; however, on the finest detail and repeating patterns the Sony can and does exhibit the same type of artifact, so a prudent Sony shooter will still need to keep an eye on fine patterned fabrics or bricks or other fine repeating patterns to make sure that they're not exhibiting moire. In terms of aliasing, this is one place where I found a difference worth showing: they both show some luma aliasing beyond their measured limits, but the big difference is that the Panasonic is immune to the colored purple/orange chroma moire, whereas the Sony shows quite a bit more aliasing, especially in color, as shown in the previous bricks pattern but also can be seen in this comparison of the DSC Labs “Wringer” chart:

http://vimeo.com/28383331

The Wringer is a color resolution chart designed to reveal the limits of a camera's imaging capabilities. For this test I recorded directly to the NanoFlash, to get the full-resolution 4:2:2 signals. The differences are pretty huge -- the AF100 handled the chart much better than the FS100, with much less aliasing. Of the four quadrants in the video, the FS100 handled the yellow & black (lower right) quadrant best, it shows some aliasing in the cyan/purple quadrant (upper right), and shows substantial aliasing and lower resolution in the two quadrants on the left. The AF100 performs much better, showing higher resolution and way less aliasing.

(Be sure to watch all these videos in full-screen mode on a 1920x1080 monitor; if you're watching it downrezzed you won't be seeing the actual pixels or the actual effect.)


7. Sensor size and Wide-Angle shots: One question I've heard asked several times is – doesn't the bigger sensor give you a wider angle field of view on the Sony? How do you get wide enough on the Panasonic? The Sony does have a substantially larger sensor, and in terms of field of view, for any given lens, the Sony will show a noticeably wider field of view, and the Panasonic will show a tighter field of view. The relative crop factor between them is approximately 1.33x. For example, when the Sony is using a 35mm lens, the Panasonic can match its field of view by using a 28mm lens. This doesn't result in any change in perspective at all, of course; you will achieve identical perspective between the two shots when the camera is at the same distance to the subject . However, it does mean that there will be a slight depth of field difference between the two shots, because of the focal length difference between the two. You can compensate for that by opening up the AF100's lens by approximately 1/2 to 2/3 of one f-stop. If you can use a 1.33x wider lens and open up the AF100's lens by 1/2 to 2/3 of an f-stop, you can match the depth of field, field of view, and perspective of the Sony exactly. Now, that won't always be possible, of course – if you're using an 8mm Ultra Prime on the Sony at f/2.8, it's not exactly likely that you're going to find a 6mm f/2.2 lens to put on the Panasonic! But that's an extreme example. Again, I refer back to the restaurant scene (http://vimeo.com/28382392)at the start of this article – that was shot using a 35mm lens at f/2.8 on the Sony, and a 28mm lens at f/2.2 on the Panasonic. Same field of view, same depth of field, just as wide, and the same perspective – basically identical footage. In short, they can both do the job. Both these cameras have large sensors that are near to 35mm cinema film in size. The Panasonic's sensor is about 18% smaller than 35mm cinema film, and the Sony's is about 15% larger than cinema film.

In general, it's true that any given lens will show a wider field of view on the bigger sensor, but the question of whether you can get “wide enough” on the Panasonic is easily answered by the excellent and relatively inexpensive Lumix 7-14mm. As an AF100 owner I've found that lens answers every wide-angle question I've encountered, as it delivers a field of view that's wider than what a Tokina 11-16 would show on the Sony. It's as wide as a 14mm on a full-frame SLR. It's not overly fast, at f4, but the gain on both these cameras is plenty clean enough to allow for boosting it by 6dB to get the relative brightness up to the equivalent of f/2.8. The point being: wide-angle solutions exist for either camera. For any given lens the Sony will have the wider field of view and the Panasonic will have a more telephoto field of view, but you can get wider lenses for the Panasonic, and more telephoto lenses for the Sony, so the net result is that either can deliver the field of view you're looking for.

8. Grievous shortcomings: each camera has an inexcusable omission or two, in my opinion. In the case of the Panasonic, it's the lack of magnified focus assist. While its focus-in-red is workable in most scenarios (especially when combined with turning the viewfinder to B&W mode), it's just silly that the magnified focus assist is missing. It's been a staple of video cameras since at least 2005! It should be there, and it isn't. (By comparison, the Sony does offer magnified focus assist.)

The Sony has three glaring omissions which I find egregious: the lack of neutral density filters, the lack of HD-SDI output, and the total lack of any way to view what the camera's doing when it's placed at eye level or higher. First, the ND filter is such a basic staple of the professional camera, that its lack is just glaring and punitive. The FS100 has no negative gain, and no way to lower brightness below its default 320 ISO. How are you supposed to shoot in daylight? You can't. The only thing you can do is crank up the shutter to some brutally short speed and/or stop the lens iris down to f/22 or more, which will result in stuttery/jagged motion and the loss of all shallow DOF and may also cause soft blurry footage because of diffraction from using such a tiny aperture. Not acceptable. (By comparison, with its ND filters, the Panasonic can go to easily 25 ISO equivalent, and down as low as 3 ISO; by combining 200 ISO with the strongest six-stop ND filter, it delivers an equivalent ISO of 3). That means you can easily shoot at a wide-open f/2 at 1/50th of a second, even in the brightest noonday sun on the Panasonic.

Second, the Sony's use of HDMI instead of HD-SDI is a consumer-oriented cop-out and unacceptable in a world of SDI recorders and SDI monitors. HDMI doesn't even carry timecode, although Sony has created their own nonstandard variant of HDMI to wedge timecode into it; the result is that your HDMI device may or may not work with the Sony's output. Just silly. I mean, their own NX5U costs $1,000 less and has SDI output on it -- there's no justification for it being missing on the FS100. And the lack of SDI means there's a lack of standardized timecode, which means there's no synchronizing of cameras for a multi-camera shoot, no 24PsF output, and reliance on an HDMI cable and connector that were never designed or built for the rigors of video production. (By comparison, the Panasonic offers both HDMI and HD-SDI, including 24PsF output, timecode, and record start/stop flags for external recorder use.)

And, finally – who approved the placement of the LCD? It's ridiculous. By putting the LCD on top and squarely in the middle, it makes it obscured or even invisible if you've got the camera at eye level or higher. The little GH2 was roundly praised for having an articulating LCD that could be seen when the camera was at any angle, as opposed to the earlier Canon DSLRs with their fixed flat screen on the back; but at least with the Canon screen you could at least see something when the camera was higher than eye level; the Sony's screen is completely blocked by the camera body. Won't be a problem if you always shoot with the camera waist-level like it's a Hasselblad or something, but anyone who's ever covered live events, sports, news, press conferences, conventions, docs, or in any other type of less-controlled environment is going to be cursing this design. (By comparison, the Panasonic offers both an eye-level viewfinder and an articulating LCD panel that rotates to accommodate viewing with the camera at any height).

Some of these shortcomings can be overcome with third-party accessories; the Panasonic's lack of focus assist and the Sony's lack of ability to be seen from various angles are both easily solved by adding an external monitor. That's an extra-cost option, but equally necessary on both if you want to overcome those particular shortcomings. The Sony's lack of ND filters can be worked around by using external ND filters, but that brings its own share of complications – external glass is another source of potential reflections, flares, dust, and fingerprints; screw-in filters need to be moved to every lens that you use, or you need to bulk up the camera rig with a mattebox and rails in order to use ND filters. Then you'll also need to carry multiple filters, or spend for a proper variable ND filter such as the Heliopan, which is a $450 item, and will still need to be moved from lens to lens every time you change a lens. As for the lack of SDI, that just can't be made up. Yes, you can get a $500-ish box that converts from HDMI to HD-SDI, but that won't compensate for the questionable HDMI connector (http://vimeo.com/24972137), or the lack of standard timecode, or the lack of 24PsF, or the lack of simultaneous record flags (present in SDI, but missing in HDMI). In the end, it's up to you to determine as to what you can live with, what you can't live without, what you can work around, and what workarounds you don't want to put up with.

9. So how do you decide? Deciding between these two cameras seems to be overly complicated by emotional reactions put forth on the internet and through forums, as has always been the case and will likely always be the case. And many times people would rather believe what they've been told or what's been frequently repeated, rather than just evaluate the products on their individual merits side-by-side. In my evaluation I've satisfied myself that the footage between them (in any reasonable shooting scenario) can be graded to be nigh unto identical, so it's not like one's going to have some big massive advantage over the other – they are both excellent cameras that both deliver excellent footage. Perhaps the biggest question is: do you really need faster than 3200 ISO, and are you willing to give up the rest of the features the Panasonic offers, in order to get that faster ISO? As I come away from evaluating these two, to me, that is the primary question. The major advantage to the Sony is its faster top speeds; in nearly every other way the Panasonic has the advantage in terms of features and usability and connectivity. So determine what's important to you, what you need, what you can work around, and what you can't live without.

As for what is different between them, I went through the owner's manuals of both and catalogued the differences. As such, here are two lists. The first is a list of things that the Panasonic
does, that the Sony doesn't. The second is a list of things that the Sony does, that the Panasonic doesn't.

Things the AF100 does that the FS100 doesn't:

The AF100 has three physical neutral density filters, the FS100 has none.
The AF100 has SD-SDI and HD-SDI output, the FS100 has no SDI output at all.
The AF100 is a “world” camera with 50/60Hz PAL/NTSC switchability, the FS100 is a 60Hz-only camera.
The AF100 offers a waveform monitor; the FS100 has only a histogram.
The AF100 offers a vectorscope, the FS100 has nothing comparable.
The AF100's LCD is rotatable and articulated; the FS100's is only top-mounted. That makes it impossible to use the FS100 overhead for high-angle shots.
The AF100 is immune from rainbow moire; the FS100 can exhibit rainbow moire similar to (but much less frequently than) the Canon DSLRs
The AF100 can record 1080/60p overcranking on a Class 6 memory card; the FS100 requires a Class 10 card to do the same.
The AF100 has a wide variety of frame rates for overcranking and undercranking, including over 2 dozen frame rates. The FS100 is extremely limited; it has only 8 frame rates, and only two slow-motion frame rates. The AF100 offers slow motion at 25, 26, 27, 28, 30, 32, 34, 36, 37, 40, 42, 44, 45, 48, 50, 54, and 60 fps. The FS100 offers only 30 and 60fps.
The AF100 can change its sensitivity via changing either the ISO or the gain; the FS100 doesn't have any reference to ISO; it's gain only.
AF100 offers interval recording for time-lapse.
AF100 offers a pre-record function, which is buffering up to three seconds of footage prior to pressing the “record” button. FS100 doesn't have any pre-record capability.
The AF100 has 1080/24PsF output, the FS100 doesn't.
The AF100 offers a 3-year warranty, the FS100 is only a 1 year warranty
AF100 has two levels of zebras, the FS100 has only one.
The AF100's zebra range is wider, it goes down to as low as 50 IRE; the FS100's minimum zebras are 70.
The AF100 has a Y Get Marker (“spotmeter”) function, the FS100 doesn't.
The AF100 has a nearly-infinitely-variable Syncro Scan shutter system, the FS100 has nothing like that. No ClearScan, no Syncro Scan.
The AF100 has a timecode output/input port for synchronizing timecode with an LTC-compatible device such as a broadcast camera or a timecode slate; the FS100 has no way to synchronize timecode.
AF100 offers a Skin Tone DTL function for smoothing out skin for more flattering portraits, FS100 doesn't offer anything like that.
The AF100 can save its scene files (“picture profiles”) to a memory card, which can then be recalled later or used in another camera to synchronize look settings. The FS100 can't do anything like that.
The AF100 can save its operational settings to a User File on a memory card, so an operator can keep his/her favorite settings and configure another camera to operate like s/he prefers, instantly; the FS100 has nothing like that.
The AF100 has a separate viewfinder and an LCD; the FS100 only has an LCD (which can have an added viewfinder tube to make it a viewfinder; but adding the viewfinder tube means nobody else can monitor from the camera, only the operator can see.)
The AF100 can set its shutter angle in terms of a film camera's shutter, meaning you can select the angle of degrees of the shutter and the shutter will compensate exposure based on the frame rate selected. The FS100 doesn't have any type of comparable capability.
The AF100 provides a distance readout for the focus position of the lens, in feet and inches (or meters); the FS100 only provides “infinity” and “minimum” focus markers.
The AF100 can report a zoom lens's focal length in exact millimeters. The FS100 only has a bar or 0-99 number.
The AF100 has a slower minimum shutter speed, 1/2 second vs. the FS100's 1/3 second.
The AF100 has seven gamma curves; the FS100 has four gammas.
The AF100 can perform relay recording from one card to the next, the FS100 can't.
The AF100 can fade to black or white, the FS100 can't.
The AF100 can capture a single-frame still either from a live feed or during recording; the FS100 can only extract a still during playback.
The AF100 allows you to group clips together and sort them on playback (i.e., grouping by marked shots, or same format); FS100 doesn't.
AF100 provides low sensitivity down to 200 ISO; FS100's minimum sensitivity is 320 ISO.
AF100 can adjust sensitivity in 1/3-stop increments, FS100 adjusts sensitivity in 1/2-stops.
AF100 has a 2.35:1 widescreen frame guide, FS100 doesn't.
AF100 has automatic face detection and face focus tracking (when used with an appropriate lens); FS100 doesn't.
The AF100 has an “area” function for prioritizing the area of the screen to focus on or expose to
The AF100 has a Dynamic Range Stretching function; FS100 has no equivalent.
AF100 can shoot in B&W mode, FS100 can't.
The AF100 offers a Last-Clip Delete function, FS100 doesn't.
The AF100 offers backlight or spotlight compensation functions, FS100 doesn't.
The AF100 has a Focus Bar focus assist graph, the FS100 doesn't have anything like that.
The AF100 has an Iris Meter over/under exposure aid, the FS100 has nothing comparable.
The AF100 has two card slots, the FS100 has only one.
The AF100 has an internal iris for automatic black balancing; the FS100 requires you to physically cap the lens to do a black balance.
The AF100 has a built-in stereo microphone; the FS100 has no built-in mic but does include an external mic.
The AF100 can input sound from either an external mic, the internal mic, or a mix of both; the FS100 can only input sound through an external source.
The AF100 has a built-in speaker, the FS100 has none.
The AF100 displays its remaining battery time down to the minute; the FS100 instead only has four graphic icons that show when the battery is below 20% capacity, below 40%, below 60%, or greater than 60%.
The AF100 doesn't overheat. In the FS100 manual there are warnings on page 2 about not using it in direct sunlight, because the battery pack or camcorder might overheat. I have not encountered any overheating yet, but the manual does warn about it.

Things the FS100 does that the AF100 doesn't
It's slightly smaller and lighter
Its sensor is bigger, thus yielding a wider field of view
It has magnified focus assist; the AF100 offers focus-in-red and the focus bar but no magnified focus assist.
Higher possible gain (up to +30dB for effectively 10,240 ISO, vs. the AF100's max of 3200 ISO.)
It has 4:4:4 RGB output on its HDMI port
It has a touchscreen for playback
It has analog component output; the AF100 has SDI/HDMI/Composite but no analog component
Can use Memory Stick Pro Duo cards in addition to SDHC/SDXC
Can record Standard Def; AF100 is only HD.
It can record 1080/60P as 60P (meaning, the “live” mode, with sound), at 28mbps. The AF100's 1080/60p is only for slow-mo and doesn't record sound, although it records at 35mbps.
Max shutter speed = 1/10,000; AF100's max = 1/2000
It has the capability for auto audio level control, as well as manual; AF100 is manual-only
Mic input level is adjustable from -60 to -30 in 6dB steps, as opposed to AF100's -50 or -60
More knee control with separate point and slope adjustments, the AF100 has a fixed slope and three adjustable points
Individual color channel controls for R/G/B/C/M/Y
White Balance shift controls include individual red/blue gain
More extensive control over detail including h/v balance
Has a GPS for tagging location in footage metadata
Has 120fps mode, even though it's very low quality and very short duration
Rec Check plays the whole last clip, not just the last 2 seconds
Six user buttons instead of the AF100's three. There aren't really any more functions, but there are more buttons to customize.
Can make subclips of clips that have been recorded, AF100 can't.
Can automatically adjust gain, and you can set the limit of how much gain is allowed. The AF100 can also automatically adjust gain to a limit you set, but that capability is only available when the camera is in fully-automatic mode.
Can do date/time stamp burned into the footage, AF100 can't.
Has wind reduction feature for the microphone
Can offload footage to a USB drive, and play back footage from a USB drive

dustylense
08-31-2011, 01:55 AM
Great write up Barry. Thanks for this. In the end both camera are great. I will agree that the top side LCD on the Fs100 is really a design flub.

simonpwood
08-31-2011, 02:26 AM
Super article. People will find this invaluable to help them decide which camera best suits their specific needs. Looks like the AF is really the one for me - it seems Panasonic were paying attention to the wish-list of pro DSLR users...

Sticky?

alaskacameradude
08-31-2011, 02:56 AM
Nice write up for sure, very thorough. I personally would add that I think the 'integration' of the Sony with the kit lens has some benefits to
'video' shooters who are not film or DSLR people familiar with DSLR lenses. Smooth iris adjustment not 'steppy' and really good image stabilization
and autofocus.....they seem to be better than what the AF 100 has. The biggest issues with the FS 100 in my opinion is the lack of ND (I had to spend
$400 on the Heliopan, which is really nice, but an added expense) and the LCD placement, both which you referenced. In the interest of trying to help
other FS 100 owners, the work around I have come up with for shooting high angles, is to put the LCD screen flat against the body of the camera
(facing up), and hold the camera upside down (so that you can see the LCD screen right above your head) and then you have to flip the image in post.
Pain in the butt, but unless you want to buy an external EVF (which is a good idea if you can come up with the money....again you are spending extra
money to get features the AF 100 has built in...), it is the only workaround I have been able to come up with. One other thing that you referenced in
your list, the 'world camera' has been confirmed by Sony, that the FS 100 will receive a free firmware update to make it into a 'world camera' as
well....it was on Philip Bloom's blog a couple weeks ago.

Postmaster
08-31-2011, 04:01 AM
Great write up Barry - as always.

Personally I don´t care about build in ND filters and the monitor placement, - don´t used to have that on film cameras too and nobody saw that as a problem,
but yeah, no SDI is a real bummer.



By the way, IMHO the second shot on the restaurant scene was the FS100.
Until that was not introduced in post, the redish shadows tell the tale.

Really looking forward to the "world camera" update though.

Frank

diegocervo
08-31-2011, 04:27 AM
I choose the AF101 over the FS100 and bought it yesterday, after long time thinking which was better to my style of working.
Add to this that I found a special offer kit on the AF101 that made me save lots of bucks.

maarek
08-31-2011, 05:13 AM
I'm gonna go out on a limb and bet that the first shot is FS100. Why? It is naturally sharp, not oversharpened like the last. Both have been cc'd so I'm not basing it on color.

Barry_Green
08-31-2011, 09:28 AM
I'm gonna go out on a limb and bet that the first shot is FS100. Why? It is naturally sharp, not oversharpened like the last. Both have been cc'd so I'm not basing it on color.
The FS100 shot was not CC'd. It is presented exactly as it was shot, using the settings that the DP and DIT had determined would give them the look that they wanted. The AF100 shot was CC'd to try to match the natural/un-CC'd shot from the FS100.

nyvz
08-31-2011, 09:50 AM
I'm gonna go out on a limb and bet that the first shot is FS100. Why? It is naturally sharp, not oversharpened like the last. Both have been cc'd so I'm not basing it on color.

Agreed, clearly FS100 is first, AF100 is second. Having shot on both a lot, it seems obvious from the increased and coarser noise and clippy yellow-shifted highlights that the second video was the AF100. I have a lot of trouble shooting on the AF100 after using the FS100 since I get so used to the lower noise and great looking highlights (cinematone1) on the FS100. The AF100's gamma curve, noise level, and sensor size always seem like a step backwards from 5D/7D, though the resolution is nice.

I'm not really sure how the AF100 could be called noise-free at any gain setting, let alone iso1600. I've had a director complain about noise on the AF100 at iso640. This isn't surprising since the AF100 has photosites about the same size as my EX1, and accordingly I've seen pretty similar noise performance between the two, whereas the FS100 uses photosites about 7x the size (potentially a 3stop improvement in capturing light). Actually from the candle test, it looks like AF100 at iso3200 has a surprising amount of blue banding noise, but that may just be exacerbated by the low color temp white balance required by the candle.

Barry_Green
08-31-2011, 10:26 AM
I'm not really sure how the AF100 could be called noise-free at any gain setting, let alone iso1600.
Did you watch the noise comparison video I posted? Using the low-noise settings, there's just ... no noise. At all.


This isn't surprising since the AF100 has photosites about the same size as my EX1, and accordingly I've seen pretty similar noise performance between the two, whereas the FS100 uses photosites about 7x the size (potentially a 3stop improvement in capturing light).
Yet there is no 3-stop improvement in sensitivity, or in noise. They are, instead, nearly exactly the same. In terms of base sensitivity and noise performance, the FS100 is practically identical to the AF100 at the same ISO/gain setting. That raises questions about what Sony means when it says that the FS100 has only 3.43 million "effective" pixels.

Whether the claim is that the pixels are actually that much larger, or they're confusing the issue and calling a red/green/blue/green bayer block an "effective" pixel, or whatever, I cannot currently answer. But in any case, I am quite satisfied that whatever the claimed number of pixels, the actual performance is absolutely directly on par and equivalent.

Cory Braun
08-31-2011, 10:27 AM
I'm pretty sure the AF100 is the first take. You can tell by the way the skin looks.

Also, I agree that the AF100 noise problems have been over exaggerated. As long as you know how to control the picture profile settings (or whatever the equivalent is called on the AF100) you can get pretty clean images at 3200. I just shot a music video which had a pool scene that we shot at midnight and it was pretty damn clean.

nyvz
08-31-2011, 11:08 AM
Did you watch the noise comparison video I posted? Using the low-noise settings, there's just ... no noise. At all.


Yes, it is a bit difficult to tell anything given that the AF100 chart was significantly overexposed compared to the FS100, and it was scaled to 720p (FS100 was 1080p) and appeared significantly softened by comparison. Low contrast charts that fill the screen with no information falling in shadow will tell almost nothing about the noise performance of a particular gain setting.

The candle-lit scene seems far more indicative of the AF100 noise performance because it shows the full luminance range available to the sensor. iso3200 on the AF100 is obviously not useable for most HD work, and I generally do not consider iso1600 to be either. Perhaps they are, however, useful for overexposing a low contrast scene, but that is rarely what high gain settings are used for, in fact it is usually just the opposite.

Edit: I just noticed the one-candle video also went through post noise reduction, so I suppose one should expect even more banding noise in camera

paragasd
08-31-2011, 11:35 AM
Completely agree with nyvz. When it first came out, I was totally won over by the af100 based on it's feature set which I still think are great for the price. But after extensive use all over the world for commercial jobs, the image just did not hold up and I sold the camera after only a few months. In fact, i tried to sell to nyc rental house and they said they did not want it because they never rent it but would buy an fs100 in a minute, if that says something. Admittedly the omission of pro features on the fs100 is annoying but image wise it is a much much better camera. Hands down. I have had several union DPs reject the af100 for commercial jobs I was directing based on the blown out highlights and noise even at low isos and we ended up renting another camera. This is real world experience not charts. I myself have decided to go for the f3 because it has all the functionality and image quality for commercial production. But I am aware it is at a different price point. Can someone make a camera with the features of the a f100 and the sensor of the fs100.

Barry_Green
08-31-2011, 12:01 PM
Yes, it is a bit difficult to tell anything given that the AF100 chart was significantly overexposed compared to the FS100, and it was scaled to 720p (FS100 was 1080p) and appeared significantly softened by comparison.
Those charts were shot at different times, the AF100 test is an older test. But there was no resizing, they were posted as pixel-for-pixel extracts.

So, just to satisfy the curious, I re-shot just a couple of minutes ago. Same chart, same lighting, same tripod position, equivalent lenses. FS100 at 18dB, AF100 at 3200 ISO. Pixel-for-pixel and 2x blowups. The noise level, as you'll no doubt observe, is extremely comparable. The noise texture is different, but the amount of noise is about the same.

Here's a mediafire download of the 1080p render straight from the timeline, no CC, no noise reduction, just raw pixel-for-pixel video.
http://www.mediafire.com/?o579xuvkiy3z309

And for those who want to cut to the chase, here's the vimeo:

http://vimeo.com/28417902

Any way you slice it, there's no magical 7x bigger pixel effect happening. These are comparable generation technology products, and they are performing at a comparable level.I don't know why you would say 1600 or 3200 ISO on the AF100 is "not usable for most HD work", I certainly think that 1600 is very clean and totally usable, and 3200 is usable if necessary -- and a light pass of Neat Video noise reduction would render it extremely usable. But that's beyond the point -- the point is, gain for gain, these two are extremely evenly matched.

Jean-Philippe
08-31-2011, 12:18 PM
In this test I can definitly see more grain on the AF100 in the darker areas, or at least a less "pleasant" noise. The noise on the AF100 is magenta and you can see it "move" around, which is somewhat distractive. However, the noise on the FS100 doesn't exhibit any "chroma" funkiness, and it is somewhat uniform. Overall, I can't tell for sure where there is more noise, I am not an expert on that. But I can definitly said that, as a viewer, I am less distracted by the noise exhibited by the FS100.

nyvz
08-31-2011, 12:19 PM
Whether the claim is that the pixels are actually that much larger, or they're confusing the issue and calling a red/green/blue/green bayer block an "effective" pixel, or whatever, I cannot currently answer. But in any case, I am quite satisfied that whatever the claimed number of pixels, the actual performance is absolutely directly on par and equivalent.

They are not confusing, all manufacturers report their photosites in megapixels or some synonym regardless of color filtering, Sony is no different. All we care about are effective pixels. Sony has made it clear in the presentations I've seen and reps I've talked to that their imaging elements cover the full sensor. Panasonic on the other hand, does not report their effective pixels anywhere and it is not clear how they are utilizing the 16MP sensor in their AF100, and no one has ever gotten them to say as far as I have heard. The FS100 may seem to be an evolution from DSLRs in terms of its design, but it takes its sensor from the $13k Sony F3 whereas the AF100 sensor seems to share almost all of its specs with the sensor from the $1k GH2, which has always bothered me a bit.

Barry_Green
08-31-2011, 12:26 PM
They are not confusing, all manufacturers report their photosites in megapixels or some synonym regardless of color filtering, Sony is no different. All we care about are effective pixels. Sony has made it clear in the presentations I've seen and reps I've talked to that their imaging elements cover the full sensor. Panasonic on the other hand, does not report their effective pixels anywhere and it is not clear how they are utilizing the 16MP sensor in their AF100. The FS100 may seem to be an evolution from DSLRs in terms of its design, but it takes its sensor from the $13k Sony F3 whereas the AF100 sensor seems to share almost all of its specs with the sensor from the $1k GH2, which has always bothered me a bit.
Well, there's marketing, and then there's actual performance. I am more interested in what the products actually DO, than what the manufacturers say about them.

In my evaluation, I have found them to be extremely comparable in performance. Whether one manufacturer claims "it's the sensor from the F3" or another says that it's a new sensor or whatever, doesn't matter. What DOES matter is: how do they perform? And to that question, after extensive evaluation, I have found the answer -- they perform directly comparably. They are equally sensitive (as far as their ISO ranges go), they deliver comparable sharpness, they deliver comparable dynamic range (FS100 ever so slightly better, but it's a difference of 5%).

Alan Roberts made some fascinating discoveries in his evaluation of the F3, and concluded that it was probably a 12.9 megapixel sensor. Not the claimed 3.3 megapixels. So yeah, it may sound like good marketing for Sony to say "same sensor as the F3!" but in reality, it performs just like the sensor in the AF100. Either the AF100's sensor is dramatically better than people have been claiming (when they disparage it by saying it's "from the $1k GH2") or the FS100's sensor is nowhere near as good as the vaunted F3, I cannot say. But what I can say, definitively, is that the AF100 and the FS100 perform comparably.

Mestizo Devon
08-31-2011, 12:47 PM
Is there an agenda here? Having owned both, I think it is easy to see that the FS100 provides a cleaner picture. Not that the Af100 is garbage, just easy to see in real world usage. It seems like much energy is being placed into placing the af100 above the fs100.

Just my thoughts!

Barry_Green
08-31-2011, 12:55 PM
All I did was put 'em side by side. I've been using the FS100 for a couple of weeks, I'd heard all the praise and all the excitement, and I kept looking for this magical super-sensitive amazing 3.4-megapixel image boost, and wasn't finding it. So I put the two cameras under the microscope, to get to the bottom of what is actually happening. And posted my results. That was my agenda.

The FS100 sensor is bigger, which gives it a wider field of view. And it has higher potential gain settings, so for those who need to go into darker conditions, it has the ability to gain up more. Other than that, they are quite comparable in imaging. The main differences between them are in features and design, not in actual video performance.

My methodology and results are posted. If anyone disagrees, I'd be glad to see their test results, including documenting the settings they used etc., so that I can figure out why they got different results than I did. I hear lots of anecdotal things such as "wow, the gain is so clean" and such, but rarely does anyone put them side by side and quantify the exact differences. That's what I attempted to do here.

nyvz
08-31-2011, 01:07 PM
Is there an agenda here? Having owned both, I think it is easy to see that the FS100 provides a cleaner picture. Not that the Af100 is garbage, just easy to see in real world usage. It seems like much energy is being placed into placing the af100 above the fs100.

Just my thoughts!

Agreed. I have found the same in my experience after using the AF100 for months and then the FS100, any findings saying somehow they perform the "same" seem a bit suspect. That is not to say that you cant get very nice looking images from the AF100 in certain situations. I've just been incredibly impressed with the way the FS100 allowed me to not worry about most of the imaging issues I've had with the AF100,7D,etc. Clean shadows, lots of useable gain options, very smooth highlights, no color shifting, no oversensitivity to green/magenta, best in class latitude, what more could I ask for? And the ridiculously good battery life doesnt hurt.

As for Alan Roberts' paper, who knows what happened with his test, but it has enough factual errors in it that it would make me seriously question its validity or at least its usefulness.

I think it may be difficult to really quantify what some people will feel is too much noise. I looked at that side-by-side video, especially in the black areas on the far sides of the chart and the resolution indicated by the trumpets, and it looks like about what I would expect, the AF100 seems to have quite a bit more noise that is also much more coarse, and you can also see that it lost quite a bit of sharpness because the codec is tied up compressing noise and of course is to some extent smoothing the noise and detail to improve its compression. This softening effect can be seen in the significant discrepancy between the resolution as indicated by the trumpets on the chart. The FS100 on the other hand has very fine noise that seems much more pleasing. It is likely this difference would be even more visible in the uncompressed output of both cameras as the noise and detail might not be softened so much by the encoder. The FS100 seems to also have more latitude in that situation, partly because of the lower noise floor and also partly in that the whites seem further from clipping. Latitude comparisons in these high gain modes is difficult to tell without more contrast, though.

I agree that it is not as simple as saying the FS100 has the same sensor as the F3, even if it does, because we know both the FS100 and F3 were crippled out of the factory in the sense that unlocking the top 2-3stops of latitude for the sensor can only be done by buying the F3 with the slog upgrade. In this way, some of the performance of the sensor is certainly hindered in both Sony models. What's impressive is that despite crippling their cameras, the FS100 still performs noticeably better in noise performance and in latitude than the AF100 at pretty much all gain modes. For some these are not the most important attribute of a camera. For me, it happens to be very important along with highlight handling.

TheReverend
08-31-2011, 01:08 PM
Is there an agenda here? Having owned both, I think it is easy to see that the FS100 provides a cleaner picture. Not that the Af100 is garbage, just easy to see in real world usage. It seems like much energy is being placed into placing the af100 above the fs100.

Just my thoughts!

This seems like your statement is a generalization, even though your experience isn't wrong. I think the AF100 is more difficult to get low noise out of (specific profiles, etc.), but I agree with Barry and the obvious results he posted here. They ARE comparable. There is 5% or less difference between relative image quality in 100% of circumstances. This definitely seems a case of to each his own. Philip Bloom's comparison hits the same issues on the head as well.

Personally, I'm glad the AF100 has ND filtering that allows f1.4-2.8 in sunlit conditions without any issues. That is super useful. Than again, I'd love to have an FS100 for those night time shots that look cool to the natural eye, but never seem to translate to video.

Barry_Green
08-31-2011, 01:11 PM
Agreed. I have found the same in my experience after using the AF100 for months and then the FS100, any findings saying somehow they perform the "same" seem a bit suspect.
Casting aspersions is of course not appreciated, but I again refer to the footage. I have put them on the charts, and in the real-life scenarios. With a little grading you can get them nigh unto identical.T


As for Alan Roberts' paper, who knows what happened with his test, but it has enough factual errors in it that it would make me seriously question its validity or at least its usefulness.
What factual errors are you referring to?

Barry_Green
08-31-2011, 01:15 PM
I think the AF100 is more difficult to get low noise out of (specific profiles, etc.), but I agree with Barry and the obvious results he posted here. They ARE comparable. There is 5% or less difference between relative image quality in 100% of circumstances. This definitely seems a case of to each his own. Philip Bloom's comparison hits the same issues on the head as well.

For point of clarification -- if someone just grabbed both cameras and pointed them randomly at some scene, they may not get the results I got. I know the AF100 intimately and know how to configure it to extract the performance that it's capable of. I have spent quite a while with the FS100 trying to get competent with it before delving into showing side-by-sides.

I can make either of these cameras look pretty bad. But that is not my intention. My intention is to show what they are capable of, when massaged to perform their best. That's why in the "cinematic" scene, I had an FS100 owner and cinematographer and DIT determine what the image settings should be, to optimize the FS100's look. I then attempted to match it with the AF100. Whether a 100% perfect match or not, I think they're pretty darn close.

Chris Messineo
08-31-2011, 01:40 PM
Great article.

Two wonderful cameras. Filmmakers are lucky to have so many great choices these days.

I'm very happy with my AF100, but I think I would probably have loved the FS100 too.

Isaac_Brody
08-31-2011, 02:00 PM
Nice write up Barry. Should be pretty useful for those looking to compare. I don't think of the omissions with each camera as bad things, it is obvious that both companies geared their cameras at different user brackets.

nyvz
08-31-2011, 02:15 PM
What factual errors are you referring to?

Various issues have been debated pretty extensively in the past, but the clearest seems to be that he took only the base sensitivity specification of iso800 of the F3 and used that to conclude that the photosites were 5um based on similar 0db sensitivity of 2/3" HD cameras and decided that therefore the sensor was 12MP.

Barry_Green
08-31-2011, 02:30 PM
Posts 45,636
http://www.dvxuser.com/V6/images/icons/icon1.png
Right, but my question is -- why is that a factual error? It may very well be true. There are several things about the F3 and FS100 that point to the notion that it simply can't be a 3.4 megapixel sensor. Alan observed the performance, observed the ratio of green to blue to red, and declared that it had to have a lot more pixels to produce the kind of results he was seeing. Which was further extrapolated upon by his noise comparison.

Which brings us back to "effective" pixels. Note they don't say "actual" or "active", they say "effective". I would not be surprised in the least to find out that the F3 and FS100 do indeed use a 13-megapixel sensor, and that the marketing department is calling each group of r/g/b/g pixels an "effective" pixel in order to make it sound more impressive than it actually is.

I mean, in simple math, it just doesn't add up. If it's a Bayer pattern, and there are truly only 3.4 huge megapixels on the chip, then that leaves a maximum red/blue resolution of 25% of 3.4, or 0.85 megapixels. According to Sony's published specs, the chip is supposed to be 2464 x 1394, which means the red and blue resolution would be (in a Bayer pattern) 1232 x 697. That would be adequate for 4:2:0, but would make it impossible to get 4:2:2. Yet the camera clearly resolves more color than 4:2:0. And if the red and green are 1232x697, then it makes the idea of 4:4:4 RGB output just silly.

If it's an RGB Stripe sensor, it's worse... that would mean that each channel (r, g, b) would be 821 x 1394. Barely enough horizontal res for standard-def. Clearly such an arrangement would make no sense.

Whereas, if we took Alan Roberts figures, and infer that Sony's counting "r/g/b/g blocks" as "effective pixels", then that means that in terms of individual photosites, there's actually 3.4 x 4, or ... wait for it ... 13.6 megapixels on the chip. And Alan had come to an esimation that the chip had 12.9 megapixels. And with 13.6 megapixels on the chip, that provides enough individual r/g/b resolution for full 1080p resolution at 4:4:4, something that the F3 and FS100 both claim to offer (or, well, the F3 can be upgraded to it, of course).

It all adds up.

Furthermore, it explains why the FS100 and AF100 have such similar imaging capability -- the AF100 is using a 12.4-megapixel chip (or, if we wanted to refer to it in Sony-speak, we could call it 3.1m "effective" pixels). If the AF100 is using 12.4 and the FS100 is using 12.9, then the observed results about sharpness, sensitivity, and noise performance all make perfect sense.

What I would love to do is find a way to verify that, once and for all.

Barry_Green
08-31-2011, 02:47 PM
Actually, this is interesting, Juan Martinez of Sony seems to be supplying confirming information. In a video presentation for the FS100, he says:

We're grossly oversampling for HD, and we're not allowed to say the specification of the sensor, but I can say that we are grossly over-sampling high definition

He says he's "not allowed" to quote what the real sampling is, but "grossly oversampling" doesn't sound like a 3-megapixel chip for a 2.2-megapixel frame. Rather it sounds like a 12.9-megapixel frame. Which, again, backs up everything else that's been said.

Rick Burnett
08-31-2011, 02:47 PM
I have spent a great deal of time with both cameras as well and have done plenty of my own tests. There are aspects of both cameras which I like VERY much, and things that I detest.

Pertaining to features, there are some omissions and mistakes above:

1) The FS100 has an internal speaker. You know this when you turn it on or hit record and hear the chimes. Also, if you play back material with audio, it will play out. Most of the time I don't have a mic hooked up to the FS100 so it was a surprise to find that.

2) The battery life per battery weight on the FS100 is FAR superior to the AF100. This is a huge plus. It's an extremely efficient camera.

3) The flash module that attaches to the FS100 is not just a usb device. It's specific to Sony. What is nice about it is that you can simultaneously record to both it AND and SDHC card. You cannot do simultaneous recording with the AF100 in the same way.

4) The screen on the FS100 can articulate in 2 directions; just 2 different directions then the AF100. While I agree that vertically the AF100 gives you more freedom of position, horizontally you can view the LCD from 2 different sides of the FS100. This has been really helpful when running around on a Blackbird stabilizer. Both cameras lack a moveable EVF solution (as in position).

5) The focus assist color can be changed on the FS100.

Pertaining to errors in Robert's reports, I can say that his testing of the AF100 rolling shutter was incorrect. The rolling shutter was much faster then he reported and I even reverse engineered his provided pictures and came up with more appropriate numbers. I tried to contact him to discuss those results, but I never got a response. I've spent a lot of time testing rolling shutters on my cameras and comparing the results with other professionals who've reported their results and feel very confident with the methodologies I use. For the rest of his report on the AF100 and/or if he did one on the FS100 (I've not seen it) I did not evaluate.

Pertaining to the noise on either camera, my findings were not the same as Barry's. I spent a great deal of time shooting with the AF100 on a few professional shoots and I spent a lot of time tweaking the AF100 to get the most out of it. I have also done the same thing with the FS100.

My views come from not just what I got in camera, but pushing around the footage in post. The biggest difference to me on the AF100 is I had to make a choice, sharpness or noise. I don't like the sharpening algorithm in the AF100 and had reduced it to -5 detail/vdetail. At or around 0 gave a noise signature in the greys that was blocky and typically had a magenta feeling to it like mentioned earlier. Also, when pushing footage further by raising up the bottom end, that noise would also come out. By reducing the detail to -5, and the effects of the 8-bit compression, that noise would go away and I was VERY happy with the footage.

With the FS100, I don't have that problem. I think the noise signature is cleaner (and denser) in the image I see and I don't have to reduce the detail at all. This is why I find the FS100 sharper because I can have both worlds (to an extent) and not have to pick one or the other. I spent a lot of time pointing both cameras at a low light scene and tweaked the settings back and forth to see this.

Why this matters to me is I am not perfect. I have accidentally underexposed scenes on the AF100 because I was in a hurry and out of time shooting it and I wasn't as happy with the results. Trying to push the scenes just a little bit and I saw a lot of noise come in. I've recreated those conditions in general with the FS100 and have more freedom to push the footage. In addition, given how fine the noise detail I see is on the gain settings I use, I find that cleaning up that noise with a plugin has VERY good results, and not something I saw with the more blotchy noise I was getting with the AF100.

All that said, I have terrific footage off of the AF100 when I nailed the exposure, I was extremely happy with the footage. I say this because all this talk about the noise is in specific cases to me. Only a few shots did I have this issue and moving forward, I would have known to be more careful in those conditions. I also cannot say one camera is overall better than the other because, as many of the points Barry has pointed out, there are good and bad on both. This is a pretty good summary of the cameras (other than of course my different feelings! Haha).

What I hate the most is there is no aperture control for EOS lenses yet from ANYONE (and I mean the aperture in the lens, not one in the adapter).

bruceallen
08-31-2011, 02:58 PM
Thanks for doing this, Barry!

Both look great. My 2c - in the cinematic example, whichever camera came first seems significantly higher resolution and also aliases a lot less vertically.

The second camera has three problems:
- at the end of the move, there seems to be some aliasing in her hair details at the top right (almost as if it is being deinterlaced or something?
- some video oversharpening thing happening with the metal highlights on the girl's bracelets, rings, etc.
- much less detail in the face

The first camera does have something slightly funky going on as well though with subtle detail. But good enough!

BTW - I don't have a favorite. I am not buying either camera out of protest for both of them totally lacking genlock / 3D sync! How can two manufacturers make 2 cameras that would be perfect size/weight/performance-wise for 3D mirror rigs... then not bother to give any sync options? Ironically, F3 has the genlock & 3D sync options... but then you're saddled with a bulky camera when you could have been using the same sensor in the FS100! It's a joke. Don't they want to sell 2 cameras instead of 1? Don't Sony and Panasonic both want the 3D ecosystem to succeed with good content?

Bruce Allen
www.boacinema.com (http://www.boacinema.com)

Barry_Green
08-31-2011, 03:36 PM
But 3D doesn't benefit from shallow DOF. Both manufacturers are making small-chip cameras that are more suitable for the 3-D effect.

Rick Burnett
08-31-2011, 03:47 PM
But 3D doesn't benefit from shallow DOF. Both manufacturers are making small-chip cameras that are more suitable for the 3-D effect.

I wondered about that. I was watching a 3D scene in a movie with really shallow DoF and it just felt really wrong to my eyes, they even started watering.

Stephen Mick
08-31-2011, 03:49 PM
I wondered about that. I was watching a 3D scene in a movie with really shallow DoF and it just felt really wrong to my eyes, they even started watering.

Yeah, Rick. That's why you cried at the end of Harry Potter. :D

Rick Burnett
08-31-2011, 03:54 PM
Yeah, Rick. That's why you cried at the end of Harry Potter. :D

ROFL. In truth, I have not seen the last 3 HP movies. Not that I don't want to see them, just haven't had time. And I doubt I will see them in 3D as I *ONLY* go to IMAX now a days to see films.

Which brings up an interesting question more on topic. Has anyone compared deep depth of field with a large sensor as opposed to a smaller sensor with regards to noise performance and light sensitivity? For situations which require deep DoF, it would be interesting to see where that inflection point is between small and large sensor of CURRENT cameras.

bruceallen
08-31-2011, 04:02 PM
But 3D doesn't benefit from shallow DOF. Both manufacturers are making small-chip cameras that are more suitable for the 3-D effect.

For 3D it's much better to have shallow DOF and a noise-free image than deep DOF and a noisy or poorly-exposed image.

Shooting 3D on a mirror rig you are fighting lack of light at all times. I don't want a camera with a chip below 2/3" for a mirror rig. And for proper 3D... you need a mirror rig (or adjustable interocular).

The only "cheap" stereo mirror rig ready compact small-chip cameras I can think of are:
- SI2K... but it's a whole price bracket up.
- Canon XF100... but it's not as compact as a FS100, doesn't have comparable low-light performance and doesn't do 60p at 1080.

Did I miss something?

BTW, if shallow depth of field made bad 3D why would Peter Jackson and James Cameron have bought collectively 50 Epics for their respective 3D productions?



Which brings up an interesting question more on topic. Has anyone compared deep depth of field with a large sensor as opposed to a smaller sensor with regards to noise performance and light sensitivity? For situations which require deep DoF, it would be interesting to see where that inflection point is between small and large sensor of CURRENT cameras.
Yes. Big sensor kicks the ass of small sensor when it comes to noise performance and light sensitivity. Hence my wanting to use a big sensor for 3D.

Sorry to drag this OT.


I wondered about that. I was watching a 3D scene in a movie with really shallow DoF and it just felt really wrong to my eyes, they even started watering.

No, that's because the convergence was done wrong. That movie was converted from 2D anyway.

For good shallow DOF 3D, look at Avatar - a bunch of very good examples in the CG parts for example.

Bruce Allen
www.boacinema.com (http://www.boacinema.com)

Rick Burnett
08-31-2011, 04:30 PM
By 'in a movie' I wasn't referring to Harry Potter. :) I was referring to mostly all the 3D films I have seen. Avatar looked great, but anytime there was REALLY shallow DoF, I didn't like it. It just felt wrong to my eyes. But I will DEFINITELY agree that I DO NOT like converted from 2D. The few I have seen are horrible.

Steve Kahn
08-31-2011, 05:28 PM
Thanks for the great test Barry. Interesting to see the aliasing on the FS100. Perhaps the camera does not have the same chip as the F3???

After using the AF100 for 8 months I can forgive the lack of focus magnification. I've learned to work with that. I can live with focus in red.

What's been more difficult for me has always been the 8bit limitation.

Ian-T
08-31-2011, 05:35 PM
By 'in a movie' I wasn't referring to Harry Potter. :) I was referring to mostly all the 3D films I have seen. Avatar looked great, but anytime there was REALLY shallow DoF, I didn't like it. It just felt wrong to my eyes. But I will DEFINITELY agree that I DO NOT like converted from 2D. The few I have seen are horrible.

I know what you mean. If the movie were strictly to be viewed in 3D (only) then the right thing to do would be to leave the Shallow DOF out and let our eyes adjust like in real life. But....because most folks end up watching it in 2D (DVDs, BlueRay etc) I guess they only had one way to go.

However, rotoscoping and fake background blur could have been one solution.

Barry_Green
08-31-2011, 05:45 PM
Thanks for the great test Barry. Interesting to see the aliasing on the FS100. Perhaps the camera does not have the same chip as the F3???

Well, Alister Chapman showed the same kind of aliasing on his F3 when he first got it.

But saying the FS100 has "the same chip" as the F3 doesn't really mean a whole lot. What counts a lot more is what's behind the chip -- the processing horsepower that wrangles the signal that's coming off the chip. We know, for example, that even though Sony says they're "the same chip" -- the F3 is rated at 800 ISO, and the FS100 is 320 ISO. What's behind that?

Well, it's what's behind the chip, is what's behind that. Think about the price difference -- there's a $9,000 price difference between an FS100 and an F3. Now, let's just arbitrarily throw some of that out for SDI ports and SxS slots and whatnot, say, randomly, $1500. That leaves $7,500 difference. Where does it all go? It goes in the processing. It goes in the computers that take up the F3's much bigger body. And how much computer processing power can you buy for $7,500? A hell of a lot. I mean, a fully-equipped 12-core top-of-the-line Mac Pro costs $4,999. So imagine if you had an entire 12-core Mac Pro processing each and every frame that comes off that sensor. Could it do some noise reduction? I believe it could. I think you could do some amazing real-time noise reduction and sharpening with that much processing hardware behind it, dedicated to doing nothing but processing that image signal.

Or, forget the Mac Pro, just look at this thing...
http://www.newegg.com/Product/Product.aspx?Item=N82E16814195102

That's an ATI graphics card with 1600 processing cores(!) It's the most expensive graphics card on Newegg's site. And it's $2799 -- meaning, for the price difference between an F3 and an FS100, you could cram THREE of those in the body of the F3.

That could work miracles on your footage.

This isn't a trivial difference, it's a huge difference. That much power could pull 14 stops out of a sensor that the little processing in the FS100 only handles 11 from. It could pull sharpness out of a better demosaic algorithm. And it could easily do enough noise reduction to put the F3's rated sensitivity at 800 vs. the FS100's 320. In Alan Roberts' report, he said that his testing revealed significant amounts of in-camera noise reduction going on -- which makes perfect sense in context of all that we've seen.

Now, I'm spitballin' here, but I think it makes sense. I'd like to get an F3 and an FS100 side by side and see just how much a difference that additional processing power the F3 has, really makes to the image.

_JD_
08-31-2011, 07:36 PM
Well written, thought-provoking post Mr. Green.

nyvz
08-31-2011, 08:00 PM
Actually, this is interesting, Juan Martinez of Sony seems to be supplying confirming information. In a video presentation for the FS100, he says:

He says he's "not allowed" to quote what the real sampling is, but "grossly oversampling" doesn't sound like a 3-megapixel chip for a 2.2-megapixel frame. Rather it sounds like a 12.9-megapixel frame. Which, again, backs up everything else that's been said.

In the pre-NAB announcement of the FS100, he was not yet allowed to discuss the specifications of the camera including its pixel count. He said the same thing at a Sony presentation I saw in March. In many later presentations and spec sheets, however Sony confirms that it is 3.34MP, which is in fact significant oversampling for 2.07MP output. Even in the pre-NAB video you quoted, he also specifically says that the pixel size is 4x that of the typical DSLR, mainly referring to the 5D/7D. The average modern DSLR has a pixel size of 4-6um, so that would mean the FS100 has a ~10um pixel size, which is exactly what was revealed when the FS100 specs were released (23.6mm horizontal / 2464 horizontal pixels = 9.5um). So the pixels in the FS100 are 4.88x the size of those in the AF100.

nyvz
08-31-2011, 08:23 PM
Well, Alister Chapman showed the same kind of aliasing on his F3 when he first got it.

But saying the FS100 has "the same chip" as the F3 doesn't really mean a whole lot. What counts a lot more is what's behind the chip -- the processing horsepower that wrangles the signal that's coming off the chip. We know, for example, that even though Sony says they're "the same chip" -- the F3 is rated at 800 ISO, and the FS100 is 320 ISO. What's behind that?

Well, it's what's behind the chip, is what's behind that. Think about the price difference -- there's a $9,000 price difference between an FS100 and an F3. Now, let's just arbitrarily throw some of that out for SDI ports and SxS slots and whatnot, say, randomly, $1500. That leaves $7,500 difference. Where does it all go? It goes in the processing. It goes in the computers that take up the F3's much bigger body. And how much computer processing power can you buy for $7,500? A hell of a lot. I mean, a fully-equipped 12-core top-of-the-line Mac Pro costs $4,999. So imagine if you had an entire 12-core Mac Pro processing each and every frame that comes off that sensor. Could it do some noise reduction? I believe it could. I think you could do some amazing real-time noise reduction and sharpening with that much processing hardware behind it, dedicated to doing nothing but processing that image signal.

Or, forget the Mac Pro, just look at this thing...
http://www.newegg.com/Product/Product.aspx?Item=N82E16814195102

That's an ATI graphics card with 1600 processing cores(!) It's the most expensive graphics card on Newegg's site. And it's $2799 -- meaning, for the price difference between an F3 and an FS100, you could cram THREE of those in the body of the F3.

That could work miracles on your footage.

This isn't a trivial difference, it's a huge difference. That much power could pull 14 stops out of a sensor that the little processing in the FS100 only handles 11 from. It could pull sharpness out of a better demosaic algorithm. And it could easily do enough noise reduction to put the F3's rated sensitivity at 800 vs. the FS100's 320. In Alan Roberts' report, he said that his testing revealed significant amounts of in-camera noise reduction going on -- which makes perfect sense in context of all that we've seen.

Now, I'm spitballin' here, but I think it makes sense. I'd like to get an F3 and an FS100 side by side and see just how much a difference that additional processing power the F3 has, really makes to the image.

That does not make sense. First of all, rated sensitivity is just the middle grey level for a particular gamma curve, and most have placed the FS100 around iso500-800 depending on your picture profile settings. Second, I understand that you are just spitballin' but making any assumptions about the F3's internal processing based on its price relative to the FS100 is pretty out there considering all the differences they have in terms of cinealta factory vs nex, cinealta internal software vs prosumer, XDCAM EX recording system vs AVCHD, and most importantly market positioning. And the prices of general purpose commodity computer hardware probably has little to do with the chips they put in these cameras.

Yes the F3 probably has more components and more expensive components than the FS100, but whether that added expense comes from fewer customers for cinealta cameras vs NEX or expensive XDCAM processing systems or expensive dual hd-sdi, tc, 3d link, sxs, etc processing and ports or from higher spec'd components is anyone's guess.

TimurCivan
08-31-2011, 08:37 PM
Barry,

On the FS, is it similar to an F3 in that its ISO range lies to you? The F3 says 0dB is ISO400, but with a light meter its ISO800. Infact, -3dB on the F3 is most likely equivalent to "0" on the FS if its reading it as 320iso. If this is the case, FS100 users should shoot at +3 dB to get the maximum dynamic range.

Any FS users out there know how to perform this test? You need a chart, a bright light, and an afternoon to kill....

TimurCivan
08-31-2011, 09:44 PM
Posts 45,636
http://www.dvxuser.com/V6/images/icons/icon1.png
Right, but my question is -- why is that a factual error? It may very well be true. There are several things about the F3 and FS100 that point to the notion that it simply can't be a 3.4 megapixel sensor. Alan observed the performance, observed the ratio of green to blue to red, and declared that it had to have a lot more pixels to produce the kind of results he was seeing. Which was further extrapolated upon by his noise comparison.

Which brings us back to "effective" pixels. Note they don't say "actual" or "active", they say "effective". I would not be surprised in the least to find out that the F3 and FS100 do indeed use a 13-megapixel sensor, and that the marketing department is calling each group of r/g/b/g pixels an "effective" pixel in order to make it sound more impressive than it actually is.

I mean, in simple math, it just doesn't add up. If it's a Bayer pattern, and there are truly only 3.4 huge megapixels on the chip, then that leaves a maximum red/blue resolution of 25% of 3.4, or 0.85 megapixels. According to Sony's published specs, the chip is supposed to be 2464 x 1394, which means the red and blue resolution would be (in a Bayer pattern) 1232 x 697. That would be adequate for 4:2:0, but would make it impossible to get 4:2:2. Yet the camera clearly resolves more color than 4:2:0. And if the red and green are 1232x697, then it makes the idea of 4:4:4 RGB output just silly.

If it's an RGB Stripe sensor, it's worse... that would mean that each channel (r, g, b) would be 821 x 1394. Barely enough horizontal res for standard-def. Clearly such an arrangement would make no sense.

Whereas, if we took Alan Roberts figures, and infer that Sony's counting "r/g/b/g blocks" as "effective pixels", then that means that in terms of individual photosites, there's actually 3.4 x 4, or ... wait for it ... 13.6 megapixels on the chip. And Alan had come to an esimation that the chip had 12.9 megapixels. And with 13.6 megapixels on the chip, that provides enough individual r/g/b resolution for full 1080p resolution at 4:4:4, something that the F3 and FS100 both claim to offer (or, well, the F3 can be upgraded to it, of course).

It all adds up.

Furthermore, it explains why the FS100 and AF100 have such similar imaging capability -- the AF100 is using a 12.4-megapixel chip (or, if we wanted to refer to it in Sony-speak, we could call it 3.1m "effective" pixels). If the AF100 is using 12.4 and the FS100 is using 12.9, then the observed results about sharpness, sensitivity, and noise performance all make perfect sense.

What I would love to do is find a way to verify that, once and for all.

Alexa gets 1080p 444 from a similar 3.6ish MP count.

Barry_Green
08-31-2011, 10:11 PM
however Sony confirms that it is 3.34MP, which is in fact significant oversampling for 2.07MP output.
But -- it in fact isn't. Because in the common vernacular, 3.34 megapixels would mean monochrome pixels. Sensors are inherently monochrome. If you were talking about a color sensor I'd agree -- but there isn't any such thing (discounting Foveon). So 3.34 megapixels is more like the bare minimum that you would need, in order to deliver a 2.07MP color output. And even then, it would be just a 4:2:0 color signal, nowhere near adequate for the claimed 4:4:4 output!

It simply does not add up.

If you wanted to have a sensor that could deliver 4:4:4 output, and "grossly oversample" it, you would need a minimum of, at least, 9 megapixels, and more like 12.

Unless Sony's changed the definition of what a "pixel" is, and they are instead referring to an R/G/B/G block as an "effective" pixel. If the sensor has 3.34 million r/g/b/g blocks, then that would indeed represent gross oversampling, and would provide for 4:4:4 color on a 1080p output. But it'd also mean that the sensor's total number of photosites would be more than 12 million -- which, again, makes sense and is reflected in the performance of the unit.



Even in the pre-NAB video you quoted, he also specifically says that the pixel size is 4x that of the typical DSLR, mainly referring to the 5D/7D. The average modern DSLR has a pixel size of 4-6um, so that would mean the FS100 has a ~10um pixel size, which is exactly what was revealed when the FS100 specs were released (23.6mm horizontal / 2464 horizontal pixels = 9.5um). So the pixels in the FS100 are 4.88x the size of those in the AF100.
They can't be. Either that is a misrepresentation, or Sony is the most incompetent sensor engineer in the world, because their sensor performs exactly on par with the 12.4-megapixel AF100 sensor, so if the best they can do is get exactly the same performance from 5x larger pixels, they'd have to be incompetent.

I don't believe they are. I believe that the evidence is pretty clear -- the pixels are comparable size, and the performance is comparable.

I would love to be able to confirm this somehow. Maybe get the sensor under a microscope and count the pixels. But I'm willing to bet a thousand bucks, cash money, right now, that there are far more than 3.34 million photosites on the FS100 chip. And I am quite convinced that it's more on the order of 12-13 million. Just like the F35 before it, just like the AF100, and just like it performs.

TimurCivan
08-31-2011, 10:13 PM
I agree that it is not as simple as saying the FS100 has the same sensor as the F3, even if it does, because we know both the FS100 and F3 were crippled out of the factory in the sense that unlocking the top 2-3stops of latitude for the sensor can only be done by buying the F3 with the slog upgrade. In this way, some of the performance of the sensor is certainly hindered in both Sony models. What's impressive is that despite crippling their cameras, the FS100 still performs noticeably better in noise performance and in latitude than the AF100 at pretty much all gain modes. For some these are not the most important attribute of a camera. For me, it happens to be very important along with highlight handling.

ITs not crippling the camera, its making it functional in the 8bit XDcam world that alot of its users would stick to. compressing 14 stops into an 8 bit codec means color correction goes out the window in post.

They hadto make it function with the on board codec. XDcam is designed for REC709. Thats a 8 stop signal. Squeezing the F3's 12.5 into that is dangerous enough. Squeezing in and additional 2 via log, would render the 8 bit aspect of the camera useless.

Also, this concept of "the firmware should be free cause RED does it...." needs to go away. RED updates their firmware for "free" cause their camera is buggy, and they are fixing things.

Barry_Green
08-31-2011, 10:14 PM
That does not make sense. First of all, rated sensitivity is just the middle grey level for a particular gamma curve, and most have placed the FS100 around iso500-800 depending on your picture profile settings.
Picture profile settings can affect where middle gray is set. I didn't test that way. I put both of them in Rec 709 gamma, where they rendered images of extremely similar tonality. I then set them to record the whites at exactly 90 %. I then changed the ISO on the AF100 to where at the identical f-stop, they were delivering identical brightness. And the proper setting for that is 320 ISO.


Yes the F3 probably has more components and more expensive components than the FS100, but whether that added expense comes from fewer customers for cinealta cameras vs NEX or expensive XDCAM processing systems or expensive dual hd-sdi, tc, 3d link, sxs, etc processing and ports or from higher spec'd components is anyone's guess.
Obviously I'm not suggesting there's an actual Mac Pro inside the box. What I am saying is that they have plenty of room in that price tag to include more powerful hardware to extract better performance from the chip. Are you saying that you disagree with that assumption?

Barry_Green
08-31-2011, 10:20 PM
Alexa gets 1080p 444 from a similar 3.6ish MP count.
The specs I've read put the Alexa at more like 4.7 megapixels. 2880 x 1620.

Barry_Green
08-31-2011, 10:22 PM
Barry,

On the FS, is it similar to an F3 in that its ISO range lies to you?
I assessed the AF100 via light meter and found its ISO ratings to be spot-on.
The FS100 doesn't list ISO. It only shows decibels of gain. So there's no claimed ISO at all. But based on the AF100, which I verified that when set to 400, it is exactly 400, I then compared against the FS100 by (as said before) shooting an identical chart at identical exposure levels and identical iris and identical shutter speed. And in order to get the exposure identical, I had to set the AF100 to 320 ISO, meaning that the FS100 in Rec 709 gamma is 320 ISO.

Barry_Green
08-31-2011, 10:25 PM
What's impressive is that despite crippling their cameras, the FS100 still performs noticeably better in noise performance and in latitude than the AF100 at pretty much all gain modes.
However, this is categorically untrue. It absolutely does not. In side by side testing the FS100 performs nigh unto identical in noise performance (in terms of quantity of noise) and it has a very little bit more latitude, about 5%. Assuming that the FS100 has the same basic pixel count as the AF100, in terms of actual counted photosites regardless of their definition of the term "pixel", that dynamic range difference could probably be accounted for due to the larger sensor size.

Barry_Green
08-31-2011, 10:54 PM
Y'know, I think I should clarify something here. I imagine there will be folks who are shaking their heads in disbelief that I'm challenging the notion of what Sony claims their pixel count is. There are reasons for that. First and foremost, the math just doesn't add up. But also, they have done this kind of thing before, and it stuck in my mind as a fishy kind of thing to do. When the EX1 came out, their brochure listed the LCD as having 1920x480 resolution. Here's a picture from the brochure saying so (I enhanced the relevant line):

39345

Now, that sounds like a really, really high-res LCD, right? Especially because everyone else was offering LCDs more in the range of 640x480, so to have 1920x480 gave them a decided marketing edge. Unless, of course, you read the fine print that they did at least print in the back of the brochure, which confirms that the LCD is REALLY 640x480:

39346

So here we have a case of where they changed the definition of "pixels" to mean something different from what the industry typically meant. They referred to individual red, green, and blue components as different "pixels", when the traditional definition for LCD monitors was to count a red/green/blue unit as one comprehensive "pixel". And there's no question, the LCD is not a 1920x480 LCD! It shows a 640x480 image. The claim of 1920x480 was simply marketing spin.

Because of this, when Sony claims 3.4 million "effective" pixels, I am not prepared to assume that they are meaning the same thing as what we would assume they mean, which is (in the case of sensors) the pixel count refers to the individual photosites. I don't believe that's what they've done here. I am quite convinced that they are doing the same thing, in reverse -- they believe that it tells a better marketing story to claim fewer/larger pixels, and so the marketing is claiming "effective pixels".

I could be wrong, but I don't think I am. I think the performance of the chip, and the sharpness and noise and dynamic range, are consistent with a comparable photosite size to the existing 12.4-megapixel AF100 chip, and furthermore Alan Roberts' testing led him to conclude that he thought it was a 12.9-megaphotosite chip (I'm discarding the word "pixels" because of the apparent confusion).

nyvz
08-31-2011, 11:08 PM
However, this is categorically untrue. It absolutely does not. In side by side testing the FS100 performs nigh unto identical in noise performance (in terms of quantity of noise) and it has a very little bit more latitude, about 5%. Assuming that the FS100 has the same basic pixel count as the AF100, in terms of actual counted photosites regardless of their definition of the term "pixel", that dynamic range difference could probably be accounted for due to the larger sensor size.

I guess we'll just have to agree to disagree on this one :)

Your side by side test to me showed significantly greater and coarser noise and noticeable loss of sharpness in the AF100 to the extent that I would call the FS100 useable and the AF100 not at all useable.

Why would you make that assumption when there is no evidence to support it and all the specification sheets say otherwise?

nyvz
08-31-2011, 11:18 PM
When the EX1 came out, their brochure listed the LCD as having 1920x480 resolution. Here's a picture from the brochure saying so (I enhanced the relevant line):


That is a very common way for LCD manufacturers to report their resolution that has nothing to do with image sensors. Panasonic also reports its LCD and EVF resolutions in "dots", claiming 920,000 dots for the AF100 LCD in its manual which is actually only 307,200 pixels (640x480). This is not something new and while some manufacturers may use this to mislead the customer, some also do it just to be able to compete with the other manufacturers' high numbers and most just do it because it is standard practice. I wouldn't read into it or think it has anything to do with counting photosites.

ian_h
08-31-2011, 11:20 PM
Barry, thank you for this exhaustive testing process and in-depth analysis. And to all the others chiming in. DVXuser is an extraordinary resource, both in terms of sheer education, and in terms of assisting informed purchase decisions when you live in a place where you can't rush out to the nearest rental store and try out all possible options before committing to a purchase.

There is a tremendous amount of anecdotal 'evidence' floating around the web, and everyone's an expert. I have friends who are passionate about their brand of lager, but brand affinity and what they think it does or doesn't say about them is often revealed as a powerful force when you stick a blindfold on them and ask them to really tell you which is which. I really can't say much - I've shot on an AF, but not an FS.

I'd be fascinated to see a series of shots of the same scenes, by a talented DOP, in a range of different lighting conditions, between the AF and the FS using, say, a 35 and 28mm CP.2 to remove the perspective difference as much as possible... and see whether in blind presentation people really could be so sure about the differences that they imagine to be there.

Thanks for the attempts at scientific rigor, and let the debates, the arguments and the flame wars rage on!!! :)

Rick Burnett
08-31-2011, 11:24 PM
I guess we'll just have to agree to disagree on this one :)

Your side by side test to me showed significantly greater and coarser noise and noticeable loss of sharpness in the AF100 to the extent that I would call the FS100 useable and the AF100 not at all useable.

Why would you make that assumption when there is no evidence to support it and all the specification sheets say otherwise?

The video shows my exact experience with the AF100 and FS100. I decreased my sharpness on the AF100 down to between -5 and -7 mostly when I shot and at that setting, the noise was much more acceptable at higher gains. That said, if you watch the AF100 noise, you see these patches of denser noise that come in and out, whereas, I don't get that sort of issue with the FS100. On the 2X blowup it is VERY apparent. This is what I was talking about with regards to having a harder time with post processing and noise reduction because the distribution isn't that smooth in the noise.

The other issue, as you've also pointed out, is the loss of sharpness. To really test the noise between the two, I'd match the FS100 to what the AF100 sharpness shows, and then do the reverse. When I increased detail on the AF100 to zero, not only does the noise at higher gain values get more pronounced on my tests, the aliasing also pops out more. (And the luma aliasing between both of them was about the same).

bruceallen
08-31-2011, 11:26 PM
Am I the only idiot who doesn't know which camera is which in the first scene?

I like the 1st camera a lot more. It seems to have more detail. The 2nd camera seems oversharpened and has weird "deinterlacing-like" vertical aliasing in the bright strands of the girl's hair at the end. The first camera seems to have some aliasing throughout as well but it's not as objectionable to my eyes.

Do other people feel the same way or is it just me?

The only thing people commented on was my side-tangent about 3D...

BTW did the whole look of the DVXuser board change or am I hallucinating?

ALSO: Thank you so much for doing this test, Barry!

Bruce Allen
www.boacinema.com (http://www.boacinema.com)

nyvz
08-31-2011, 11:29 PM
Picture profile settings can affect where middle gray is set. I didn't test that way. I put both of them in Rec 709 gamma, where they rendered images of extremely similar tonality. I then set them to record the whites at exactly 90 %. I then changed the ISO on the AF100 to where at the identical f-stop, they were delivering identical brightness. And the proper setting for that is 320 ISO.

That sounds a bit odd. Alexa fits 15 stops of latitude into its REC709 gamma curve, the AF100 and FS100 fit 10-11ish, older cameras have fit fewer. And that says nothing in terms of discrepancies in terms of how over/under latitude is handled in these cameras that all call it REC709. It does not seem that the luminance value that equates to 90% video level would be particularly indicative of a camera's sensitivity since such iso values are primarily used for light meter readings which from what I understand are calibrated around a standard grey level. I could certainly see there being significant debate over what is a proper middle grey video level for certain types of shooting and certain gamma curves, but to throw that out entirely and use a 90% luma value for sensitivity measurement seems like it would not be especially useful. In my experience the FS100 compresses more highlight latitude into its highlight range than the AF100, so I imagine that would certainly throw off your ISO reading.

Mike Harvey
08-31-2011, 11:31 PM
Am I the only idiot who doesn't know which camera is which in the first scene?


No, I couldn't tell either... but I haven't used either, either (was that even grammatically correct?). And honestly, I thought they both looked good. I could tell they were different cameras, but I didn't think one looked superior to the other. Just different.

nyvz
08-31-2011, 11:39 PM
Am I the only idiot who doesn't know which camera is which in the first scene?

I like the 1st camera a lot more. It seems to have more detail. The 2nd camera seems oversharpened and has weird "deinterlacing-like" vertical aliasing in the bright strands of the girl's hair at the end. Though it has a little bit of aliasing, it's not as objectionable to my eyes.

Do other people feel the same way or is it just me?

The only thing people commented on was my side-tangent about 3D...

BTW did the whole look of the DVXuser board change or am I hallucinating?

ALSO: Thank you so much for doing this test, Barry!

Bruce Allen
www.boacinema.com (http://www.boacinema.com)

Maybe :) It is clear to me that the first is the FS100, I know the AF100 highlights and noise all too well, though they are a bit hidden in that scene if you aren't looking for it or dont know what to look for :)

Agree about the detail, I've generally found this with the FS100, it seems to have more, finer detail than the AF100, though the AF100 is certainly no slouch

I agree with you that shallower DOF can be very nice in 3D if done properly, personally that was one of the many revelations that I had while watching avatar, shallow depth of field is so rarely done in 3D that when it is, if done right is a wonderfully interesting way of seeing

Yeah I noticed some changes in DVXuser's color scheme and design today, too. It's a surprisingly subtle change though!

HansSteinert
08-31-2011, 11:41 PM
I wondered about that. I was watching a 3D scene in a movie with really shallow DoF and it just felt really wrong to my eyes, they even started watering.

I've read that many times, such as in Transformers 3, that close up shallow DOF shots are shot mono and then converted to 3D in post because rotoscoping a shallow DOF clip is much easier or something.

So maybe that's why they look weird? It's artificial?

nyvz
09-01-2011, 12:02 AM
ITs not crippling the camera, its making it functional in the 8bit XDcam world that alot of its users would stick to. compressing 14 stops into an 8 bit codec means color correction goes out the window in post.

They hadto make it function with the on board codec. XDcam is designed for REC709. Thats a 8 stop signal. Squeezing the F3's 12.5 into that is dangerous enough.

I've heard this before but I am very skeptical of it. I understand that more latitude into the same bits might be more likely to cause posterization artifacts, but I don't think I've ever heard anyone say, "my camera has too much latitude, its posterizing too much, please reduce the available latitude"... The tradeoff is very minor since lets say the camera has 11stops (~FS100) its fitting into 16-235 to be conservative, in gamma-encoded or log space that should give about 20 levels per channel per stop, pushing it to 14 stops only reduces it to 16 levels per channel per stop. Remember that 20 levels per channel is still 8000 levels per pixel and 16 is still 4096 levels per pixel. The difference between 12.5 and 14 stops would be even less. I've seen real HDR converted to 8bit and never noticed any posterization. While 8bit is not 10 bit or 12bit, I think many people still underestimate it. Remember it's still basically 16Million colors and 4x more colors than most LCDs can display (dithering helps). I'd take more latitude any day.

The F3 fitting 12.5 stops into 8bits may seem dangerous, but I'm sure I've never heard anyone complain about the extra latitude.

You could always reduce your latitude in the picture profile or choose a contrastier gamma anyway.

TimurCivan
09-01-2011, 01:06 AM
I've heard this before but I am very skeptical of it. I understand that more latitude into the same bits might be more likely to cause posterization artifacts, but I don't think I've ever heard anyone say, "my camera has too much latitude, its posterizing too much, please reduce the available latitude"... The tradeoff is very minor since lets say the camera has 11stops (~FS100) its fitting into 16-235 to be conservative, in gamma-encoded or log space that should give about 20 levels per channel per stop, pushing it to 14 stops only reduces it to 16 levels per channel per stop. Remember that 20 levels per channel is still 8000 levels per pixel and 16 is still 4096 levels per pixel. The difference between 12.5 and 14 stops would be even less. I've seen real HDR converted to 8bit and never noticed any posterization. While 8bit is not 10 bit or 12bit, I think many people still underestimate it. Remember it's still basically 16Million colors and 4x more colors than most LCDs can display (dithering helps). I'd take more latitude any day.

The F3 fitting 12.5 stops into 8bits may seem dangerous, but I'm sure I've never heard anyone complain about the extra latitude.

You could always reduce your latitude in the picture profile or choose a contrastier gamma anyway.

They never said "my camera has too much latitude, im getting posterizing" because it was never available to them before. I just bought sLog today.... like 14 hours ago.

Loaded it up, and recorded to XDcam its full sLog glory. Trying to CC back into something, and its impossible. The macroblocking, banding and XDcam artifacting is dragging all this garbage out of the depths of the signal. When shooting log, youre "middle grey" is now 35IRE. imagine dragging 35 up to the mid 50's on XDcam... The point being, when shooting sLog you're supposed to be recording externally to soemthing at LEAST 10bit. The Alexa can smush 15 stops into a "REC709" space (this what LOG is for) because its 12bit PRORES4444. the gradations are so fine you can squish all 14 stops into a 30 IRE band of the signal and still get something out of it.

bruceallen
09-01-2011, 01:17 AM
OK, I just checked out the other tests (which are awesome - thank you so much Barry for doing this - and ESPECIALLY for finding a Wringer chart - why is it so hard to find a chart that shows color not luma resolution?).

To my (currently tired and grumpy) eyes the weird vertical de-interlacing style aliasing is all over the Wringer test on the AF100 side. Sure there's some color aliasing and minor luma aliasing on the FS100 side and it doesn't look like it's resolving a full 1080p - but the resolution for the AF100 looks like 1920x540 or something! What the...? But on the trumpets noise chart the aliasing doesn't seem quite as bad - still a lot worse than the FS100 though.

Sorry if I haven't been keeping up - I thought the AF100 was based on the GH2 sensor - and I'm sure I didn't see those problems on the GH2?

Bruce Allen
www.boacinema.com

nyvz
09-01-2011, 01:18 AM
They never said "my camera has too much latitude, im getting posterizing" because it was never available to them before. I just bought sLog today.... like 14 hours ago.

Loaded it up, and recorded to XDcam its full sLog glory. Trying to CC back into something, and its impossible. The macroblocking, banding and XDcam artifacting is dragging all this garbage out of the depths of the signal. When shooting log, youre "middle grey" is now 35IRE. imagine dragging 35 up to the mid 50's on XDcam... The point being, when shooting sLog you're supposed to be recording externally to soemthing at LEAST 10bit. The Alexa can smush 15 stops into a "REC709" space (this what LOG is for) because its 12bit PRORES4444. the gradations are so fine you can squish all 14 stops into a 30 IRE band of the signal and still get something out of it.

Hah! I stand corrected.

Are you certain you need to push 35ire to 50ire though? I mean with the extra highlight latitude slog unlocks, seems like you shouldn't have to be pulling the image up more in post, you could be pretty safe to expose a bit over and if anything bring it down and add a bit of contrast in post and it might avoid artifacting a bit more? Just a guess though. I'd be very interested to see the results you are getting if you could post uncompressed frames.

ian_h
09-01-2011, 01:31 AM
Am I the only idiot who doesn't know which camera is which in the first scene?

I like the 1st camera a lot more. It seems to have more detail. The 2nd camera seems oversharpened and has weird "deinterlacing-like" vertical aliasing in the bright strands of the girl's hair at the end. The first camera seems to have some aliasing throughout as well but it's not as objectionable to my eyes.



I also prefer the first one. But in isolation I'm pretty sure I'd think the second one looked fantastic.

Sumfun
09-01-2011, 09:42 AM
Barry, thank you for putting together another great test for us. Your work and dedication is one of the main reasons that DVXUser sis uch a great source of information.

I never noticed the quality of noise before, until nyvz and Rick pointed it out here. Now I have to agree with them that the noise on the AF100 at ISO 3200 looks splotchy and has a tendency towards red. Although the FS100 has about the same amount of noise, the noise has finer grain, and may be easier to eliminate with a noise reduction program.

In the cinematic comparison, both pictures look good, but the first one looks a little sharper, so I prefer that one. It's been several days now. Will you tell us which one was shot with which camera?

In the wringer color chart and the list of features, I think the AF100 wins hands down.

So in the end, I agree that the cameras are pretty evenly matched, each having its own advantages and disadvantages. When it comes to choosing between these 2, I think you just have to decide what qualities or features are more important to you and go with that. If you want the best of both cameras, there's always the F3 at about 3 times the price. I really see an opening for Panasonic here. If they can fix the noise/softness issue, add a 10-bit AVC intra codec, and maybe a little more DR, they'd have a winner. And if they can price it in the $10k range, they would sweep the market.

nyvz
09-01-2011, 09:52 AM
How are you determining the "amount of noise" for comparison? It is tough to talk quantitatively about something that is being judged qualitatively. Unless you are using some particular measurement tool, I would tend to say, the qualitative assessments of splochier, coarser, etc in terms of noise would have to factor into any talk about "amount of noise". Everyone seems so careful to say the noise is the same amount but in a different quality, but I would say that difference in quality is exactly what determines the amount in this kind of analysis. If the noise is splochier, coarser, and causes more havoc to the encoder, then that is "more" noise even if they both have some visible noise.

Osslund
09-01-2011, 09:56 AM
Thanks Barry for taking time to do this comparison. AF vs FS is just one of the hardest $5K decisions right now.

Barry_Green
09-01-2011, 10:04 AM
but to throw that out entirely and use a 90% luma value for sensitivity measurement seems like it would not be especially useful.
Well, look -- you weren't there. The image looked identical in all ways. It wasn't about finding 55 IRE on a gray card. The brightness and range on this chart, when viewed on the waveform monitor, were absolutely identical.

I mean, it's not like I don't know what I'm doing. This was a full range chart, not a thin line of gray.

I am extremely confident in my evaluation that they were appropriately rated for ISO in the way I tested it.

nyvz
09-01-2011, 10:26 AM
To my (currently tired and grumpy) eyes the weird vertical de-interlacing style aliasing is all over the Wringer test on the AF100 side. Sure there's some color aliasing and minor luma aliasing on the FS100 side and it doesn't look like it's resolving a full 1080p - but the resolution for the AF100 looks like 1920x540 or something! What the...? But on the trumpets noise chart the aliasing doesn't seem quite as bad - still a lot worse than the FS100 though.


I hadn't looked at it too carefully before, but even in the 720p vimeo version it is very interesting how different the nature of the aliasing in these cameras is. The FS100 appears to have colored moire in the red-blue areas, whereas the AF100 seems to have that weird steppiness in the yellow-blue area that looks quite undersampled and appears more like traditional pixel aliasing toward the higher frequencies. Too bad vimeo isnt letting me download the 1080p original.

I'm amazed that highlight handling isnt a big part of the FS100 vs AF100 discussion. This was the single greatest reason I prefer the FS100 over the AF100 and even prefer the 5D/7D/GH2 over the AF100 for a lot of work. The AF100 has nasty yellow-shifted clippy highlights in every gamma mode (Cine V the least, but its still pretty bad), whereas the FS100 has super smooth highlights (especially using cinematone1 gamma and cinematone2 matrix) . For the first time I found myself intentionally overexposing on the FS100 because the highlight rolloff is so great looking.

nyvz
09-01-2011, 10:31 AM
Well, look -- you weren't there. The image looked identical in all ways. It wasn't about finding 55 IRE on a gray card. The brightness and range on this chart, when viewed on the waveform monitor, were absolutely identical.

I mean, it's not like I don't know what I'm doing. This was a full range chart, not a thin line of gray.

I am extremely confident in my evaluation that they were appropriately rated for ISO in the way I tested it.

Ah ok, that makes more sense, but whats wrong with the old grey card method?

I imagine the grey card method would be best for comparing two cameras with different gamma curves, though. How the cameras compare in REC709 modes may not mean much to the majority of people who are using Cine V/D on the AF100 or Cinematone1 on the FS100.

Barry_Green
09-01-2011, 10:57 AM
Ah ok, that makes more sense, but whats wrong with the old grey card method?
There's nothing "wrong" with it, of course, but it's highly dependent on the gamma curve itself. If you're using cinematone 1, versus "high" on an AF100, it can make you think that there's almost a full stop of sensitivity difference between them. But it's just due to the way the gamma curve pushes or pulls up the gray; it's not changing the sensitivity of how the camera sees the incoming light.

By using a comparable gamma curve on both, you eliminate that variability. And what I was trying to do was compare relative sensitivity, to see how they compare to each other. It seemed like the Rec709 method was the most identical gamma, and therefore it would eliminate that variable from consideration.

TimurCivan
09-01-2011, 12:44 PM
Hah! I stand corrected.

Are you certain you need to push 35ire to 50ire though? I mean with the extra highlight latitude slog unlocks, seems like you shouldn't have to be pulling the image up more in post, you could be pretty safe to expose a bit over and if anything bring it down and add a bit of contrast in post and it might avoid artifacting a bit more? Just a guess though. I'd be very interested to see the results you are getting if you could post uncompressed frames.

Will start a new thread, in the F3 foum tonight. Doing log testing today.

dcloud
09-01-2011, 01:11 PM
Barry what are your settings on the af100?

Barry_Green
09-01-2011, 01:33 PM
For what -- in general? Or for the comparison shot, or ?

nyvz
09-01-2011, 01:43 PM
There's nothing "wrong" with it, of course, but it's highly dependent on the gamma curve itself. If you're using cinematone 1, versus "high" on an AF100, it can make you think that there's almost a full stop of sensitivity difference between them. But it's just due to the way the gamma curve pushes or pulls up the gray; it's not changing the sensitivity of how the camera sees the incoming light.

By using a comparable gamma curve on both, you eliminate that variability. And what I was trying to do was compare relative sensitivity, to see how they compare to each other. It seemed like the Rec709 method was the most identical gamma, and therefore it would eliminate that variable from consideration.

I understand what you mean, but that gamma variability likely exists no matter whether you choose REC709 or any other gamma. In either case, you cannot expose the way in which the recorded, gamma-corrected images were created from the ~14bit(?) linear RAW sensor data.

What I mean is that a comparison of sensitivity between the two cameras in REC709 modes is a good indicator for someone who uses REC709 and is choosing between using REC709 on AF100 or REC709 on FS100, but not a good indicator of anything specific about the sensitivity of the sensor or of the camera when used with a different gamma mode. It will not help you compare "how the camera sees the incoming light" in general, only how that similarly-intended gamma performs compared to the similar gamma on the other camera.

The only attribute I'd think you could really use to compare the cameras' sensitivity in this way that might transcend various gamma implementations would be by comparing the hard white clip point for 5000k white (might avoid differences in uneven color channel clipping). Even that is difficult in these cameras since I doubt even white clip point remains constant between gamma/knee modes, but it could be meaningful if you find the mode with the highest white clip and use that. But that might be a somewhat meaningless comparison for most people except for someone using a spot meter on speculars or bright areas to expose to the right without blowing out, which is a more gamma-agnostic way of exposing digital cameras.

Steve Kahn
09-01-2011, 02:25 PM
Both camera end result images look similar to my eyes. When I feel I may be in trouble with highlights on the AF100 I dial down the chroma, pop it into Cine-D and try my best not to overexpose. To me this is not an issue. Nor is the slight aliasing on the FS. It seems minimal.

The real issue is that both cameras are 8bit and to me is the real limitation which would steer me away from using them on a feature film or extended shoot.

Most people will not notice yellow blown highlights or mild aliasing (hell in the dSLR craze some people even claimed to LIKE aliasing). I can guarantee, though, no one will not see color banding - or like it.

Barry_Green
09-01-2011, 02:44 PM
Both camera end result images look similar to my eyes. When I feel I may be in trouble with highlights on the AF100 I dial down the chroma, pop it into Cine-D and try my best not to overexpose. To me this is not an issue. Nor is the slight aliasing on the FS. It seems minimal.

The real issue is that both cameras are 8bit and to me is the real limitation which would steer me away from using them on a feature film or extended shoot.
Agreed with all of the above, except -- 8-bit has been used in feature films before. Star Wars Episode II was all shot on HDCAM, which was 8-bit (and only 1440x1080, at that). In fact, if you search the IMDB for HDCAM you'll find lots of quite well known films that were shot in 8-bit for theatrical release, including Crank, Sin City, Spy Kids 2, Once Upon A Time In Mexico...

Steve Kahn
09-01-2011, 02:53 PM
Agreed with all of the above, except -- 8-bit has been used in feature films before. Star Wars Episode II was all shot on HDCAM, which was 8-bit (and only 1440x1080, at that). In fact, if you search the IMDB for HDCAM you'll find lots of quite well known films that were shot in 8-bit for theatrical release, including Crank, Sin City, Spy Kids 2, Once Upon A Time In Mexico...

And I agree. It can be done. The problem I have is the lack of flexibility one has color correcting in post. When you have a big budget and time this is not an issue. One knows the look they are going to get will be set when shooting. On a low budget set things are vastly different though. Often times they are rushed. Many times you can't light they way you would love to. For mild color correction 8bit is fine. But for fixing big mistakes it just doesn't work. And, on a low budget set there are bound to be big mistakes.


(other than this I love the AF. It's a great, great little camera)

Sumfun
09-01-2011, 03:47 PM
I'm amazed that highlight handling isnt a big part of the FS100 vs AF100 discussion. This was the single greatest reason I prefer the FS100 over the AF100 and even prefer the 5D/7D/GH2 over the AF100 for a lot of work. The AF100 has nasty yellow-shifted clippy highlights in every gamma mode (Cine V the least, but its still pretty bad), whereas the FS100 has super smooth highlights (especially using cinematone1 gamma and cinematone2 matrix) . For the first time I found myself intentionally overexposing on the FS100 because the highlight rolloff is so great looking.

I would be interested in seeing a comparison of the highlight handling as well. Barry, if you still have both cameras, maybe you can shoot a chart, but expose the whitest strip at 110 IRE?

mcgeedigital
09-01-2011, 03:52 PM
And I agree. It can be done. The problem I have is the lack of flexibility one has color correcting in post. When you have a big budget and time this is not an issue. One knows the look they are going to get will be set when shooting. On a low budget set things are vastly different though. Often times they are rushed. Many times you can't light they way you would love to. For mild color correction 8bit is fine. But for fixing big mistakes it just doesn't work. And, on a low budget set there are bound to be big mistakes.


WHat kind of production are you shooting where this would possibly be the case?

Ryan Patrick O'Hara
09-01-2011, 03:55 PM
And I agree. It can be done. The problem I have is the lack of flexibility one has color correcting in post. When you have a big budget and time this is not an issue. One knows the look they are going to get will be set when shooting. On a low budget set things are vastly different though. Often times they are rushed. Many times you can't light they way you would love to. For mild color correction 8bit is fine. But for fixing big mistakes it just doesn't work. And, on a low budget set there are bound to be big mistakes.


(other than this I love the AF. It's a great, great little camera)

You never have time. Ever. Not when you're an amateur and certainly not when you're a pro. :P

It's not an issue of time, but big budgets hire pro's and pro's don't make big mistakes. That's the difference.

Steve Kahn
09-01-2011, 04:02 PM
You never have time. Ever. Not when you're an amateur and certainly not when you're a pro. :P

It's not an issue of time, but big budgets hire pro's and pro's don't make big mistakes. That's the difference.

1st part right.
2nd part VERY WRONG.

I was an actor on a crappy Keanu Reeves film where because of a BIG mistake producers had to come back A YEAR LATER with 50 extras and re-shoot on a huge set. A YEAR LATER!!!


And, yes, there are always mistakes - big and small. And pros make big mistakes all the time. Hell, most of the movies made in HWood today are big mistakes from beginning to end!


But, certainly an AF shooter is probably more apt to make a big mistake than an Arri shooter - which goes to my point exactly.

alaskacameradude
09-01-2011, 04:12 PM
I hadn't looked at it too carefully before, but even in the 720p vimeo version it is very interesting how different the nature of the aliasing in these cameras is. The FS100 appears to have colored moire in the red-blue areas, whereas the AF100 seems to have that weird steppiness in the yellow-blue area that looks quite undersampled and appears more like traditional pixel aliasing toward the higher frequencies. Too bad vimeo isnt letting me download the 1080p original.

I'm amazed that highlight handling isnt a big part of the FS100 vs AF100 discussion. This was the single greatest reason I prefer the FS100 over the AF100 and even prefer the 5D/7D/GH2 over the AF100 for a lot of work. The AF100 has nasty yellow-shifted clippy highlights in every gamma mode (Cine V the least, but its still pretty bad), whereas the FS100 has super smooth highlights (especially using cinematone1 gamma and cinematone2 matrix) . For the first time I found myself intentionally overexposing on the FS100 because the highlight rolloff is so great looking.

Highlight handling is one of the reasons I chose the FS100 over the AF100.....I see the same thing you do as far as differences in the way these two cameras handle highlights.

dcloud
09-01-2011, 05:04 PM
For what -- in general? Or for the comparison shot, or ?for the noise free

nyvz
09-01-2011, 05:40 PM
Highlight handling is one of the reasons I chose the FS100 over the AF100.....I see the same thing you do as far as differences in the way these two cameras handle highlights.

Actually almost all of the content I've shot on AF100 has eventually been edited in FCP (by some editor or another, never me) with no color done at all and to varying degrees it always seems like the highlights get clipped an shadows crushed somewhere along the line so the color shifts are not so visible in the final products, but the clipping sure is.

The worst of them being this one: http://video.aol.com/video/landline-romantic-comedies-in-60-seconds/72057659435892184
They also forgot to put the 2.35:1 letterbox on... :P
I assure you I did not hard clip those skin tones in camera, even if they were a bit hot

That's other highlight thing I like about the FS100, I can set up my picture profile up to be stay broadcast safe so this does not happen. AF100 always uses superwhite no matter your gamma or knee setting, so it seems quick edit no-CC projects could always be a bit o a risk.

nyvz
09-01-2011, 05:47 PM
I was an actor on a crappy Keanu Reeves film where because of a BIG mistake producers had to come bad A YEAR LATER with 50 extras and re-shoot on a huge set. A YEAR LATER!!!


Can I ask what the big mistake was?

jdv
09-01-2011, 06:06 PM
WHat kind of production are you shooting where this would possibly be the case?

Every indie film I've ever worked on for one.

;)

Steve Kahn
09-01-2011, 06:18 PM
Can I ask what the big mistake was?

You think anyone in HWood admits to anything? But all I know is that they called back all the actors a year later and everyone got paid twice for doing the same (lavish) scene.

Steve Kahn
09-01-2011, 06:19 PM
Every indie film I've ever worked on for one.

;)

Yep. Every film. Every project. Kubrick made crews wait for days until the light was good enough for him. You think any other director gets that luxury??? Nope.

nyvz
09-01-2011, 06:33 PM
You think anyone in HWood admits to anything? But all I know is that they called back all the actors a year later and everyone got paid twice for doing the same (lavish) scene.

Perhaps they realized in the edit that they needed to rewrite a scene or wanted different/additional coverage? I'd think lots of productions would love to be able to do that if they had a budget, but I'm not sure its necessarily the same thing as a big mistake like screwing up something a bit more technical like lighting, exposure, whitebalance, or losing media/rolls.

Steve Kahn
09-01-2011, 06:52 PM
Perhaps they realized in the edit that they needed to rewrite a scene or wanted different/additional coverage? I'd think lots of productions would love to be able to do that if they had a budget, but I'm not sure its necessarily the same thing as a big mistake like screwing up something a bit more technical like lighting, exposure, whitebalance, or losing media/rolls.

You're probably right. These kind of things happen all the time. They say you tell a story three times: when you write it, when you shoot it, and when you edit it.

Again, to my point a big studio has the budget to reshoot a year later. A crew shooting an AF100 would never have that luxury and most probably it would be up to a talented editor to make things work in post. Without good color grading flexibility this job is made all that much harder.

The Keanu film was shot on film 5 years ago. Now anything shot digital is being edited on set. I just was at a wrap party for "The To Do List" and was speaking to the director the day after final wrap. The editors already had gotten a rough edit to her! (It was shot on an Alexa)

Luis Caffesse
09-01-2011, 07:16 PM
Kubrick made crews wait for days until the light was good enough for him. You think any other director gets that luxury??? Nope.

Terrence Malick.


But that's neither here nor there.
Great write up, Barry. As always.
To me these test prove one thing for certain - both cameras are capable of nearly identical performance.
If anyone is seeing a major difference between them in regards to image quality, then I would guess the differences are coming from improper setup or some other operator introduced variable.

lexicon
09-01-2011, 08:08 PM
I would love to be able to confirm this somehow. Maybe get the sensor under a microscope and count the pixels. But I'm willing to bet a thousand bucks, cash money, right now, that there are far more than 3.34 million photosites on the FS100 chip. And I am quite convinced that it's more on the order of 12-13 million. Just like the F35 before it, just like the AF100, and just like it performs.

Barry, thanks for the FS100 and AF100 'comparison' piece, another opportunity for clarification and learning as we are used to with all your articles.

Speaking about the F3/FS100 pixel-count conundrum, I recall that, some time ago, an F3 user commented on the website dvinfo that he was given access to the service manual of the camera to remove a stuck pixel on their F3. Interestingly, it was found in the service manual that you can address pixel manually to mask them. And he found that the 'pixel' positions go from 1 to 2468 Horizontally and 1 to 1398 vertically.

In his own words:

"I called Sony and they were nice enough to send me the info to fix it via the service menus (the black balance procedure in the user menu didn't take care of it).

What I found interesting was the info about the pixel addressing for the RPN procedure. There is a section where you can go in and manually select the location of the bad pixel if the "auto" RPN procedure in the service menu doesn't work. The limits for H and V pixel positions are as follows H = 1 to 2468 and V = 1 to 1398. This is directly out of the service manual.

Based upon a simple multiplication of the numbers that comes out to 3.45 megapixels. The service manual specifically uses the word "pixels" in the procedures."

Alister Chapman commented on that thread that the "RPN will almost certainly be working at the photo site level, as it will have an impact on the De-Bayer process. The camera most likely simply addresses an adjacent photo site as opposed to the defective one. So pixels almost certainly equals photosites on the F3."

And David Heath added that "Early on, Sony were saying the sensor was about 3.36MP effective, which works out to be about 2456x1372. I'd expect "effective" to be a little less than the actual count, so 2468 and 1398 are pretty well exactly what may be expected of the actual count."

http://www.dvinfo.net/forum/sony-xdcam-f3-cinealta/496135-imager-info-service-manual.html#post1651346

All of this ties in nicely with the published specifications of the F3 sensor at 3.45 Million Pixels. So, assuming that the FS100 has the same sensor than the F3 (probably true), are you still "willing to bet a thousand bucks, cash money, right now, that there are far more than 3.34 million photosites on the FS100 chip?" Because I want to tell you that I am ready for that bet :happy:

However, I agree with you that if this photosite count of 3.45 million turns out to be true, this sensor simply cannot deliver 'true 444', meaning by that, one pixel sample for each colour at every point within the image. So for a true 444, 1920 image, that would require 1920 Red, 1920 Green and 1920 Blue photosites. For a Bayer sensor that would imply a sensor with at least twice as many horizontal and vertical pixels as the desired resolution, something like 3840x2160 photosites for a 1920x1080 image. If the F3/FS100 sensor is only 2468 x 1398, this is clearly not enough for true 444. It's probably enough for true 422 but not enough for true 444.

Still we have to concede that the 444 configuration of the camera is reading more information than the 422 configuration, and more even than the 420 configuration (xdcam internal recording). So the images from the so-called 444 configuration are still going to be better than the ones recorded at 422, and those, better than the ones at 420. So, maybe, the meaning of 444 in this context is not 'true 444', but just a way to say that it is better than 422, a way to say that is the most information that you can get from that sensor. On this terms, Alexa is probably also not true 444 camera, nor the new Epic X with a sensor of only 5K photosites for a 4K output image.

So SONY is apparently not 'speaking marketing' to us on the pixel count arena (the camera might truly have only 3.45 million photosites), but on the color subsampling specifications (specifically, when they talk about 444 capabilities).

It seems that SONY engineers were trying to find a sweet spot in terms of photosites count on the F3 sensor. Large enough numbers to get proper HD resolution, but not too many so they can keep the noise levels low. Large enough numbers to get better color subsampling than 'true 422', but willing to sacrifice 'true 444' to keep the photosite count (hence noise) low. And we all agree that the F3 levels of noise are very low, something consistent with a lower photosite count on that sensor.

So why did you find comparable noise peformance between the FS100 and the AF100? (and based on some comments on this thread by owners of both cameras, this is still a controversial issue). Well, maybe it is not because they have similar pixel counts, but because other factors. Presumably, both the F3 and the FS100 have the same sensor --i.e. same number of photosites,-- however the F3 is still less noisy than the FS100. As you mentioned before there are reasons like differences on the internal processing of the signal between the two cameras that can explain differences on the output image (for example, the AD conversion is done with less bit-depth on the FS100 than on the F3). Given the large number of pixels that the AF100 sensor contains (12 million?), and the very acceptable levels of noise you get in many shooting situations, well, this only speaks of how good was the job Panasonic engineers did on that camera.

mcgeedigital
09-01-2011, 08:15 PM
Yep. Every film. Every project. Kubrick made crews wait for days until the light was good enough for him. You think any other director gets that luxury??? Nope.

That's a CHOICE you decide to make.

It has nothing to do with the camera you use.

nyvz
09-01-2011, 08:44 PM
So why did you find comparable noise peformance between the FS100 and the AF100? (and based on some comments on this thread by owners of both cameras, this is still a controversial issue). Well, maybe it is not because they have similar pixel counts, but because other factors. Presumably, both the F3 and the FS100 have the same sensor --i.e. same number of photosites,-- however the F3 is still less noisy than the FS100. As you mentioned before there are reasons like differences on the internal processing of the signal between the two cameras that can explain differences on the output image (for example, the AD conversion is done with less bit-depth on the FS100 than on the F3). Given the large number of pixels that the AF100 sensor contains (12 million?), and the very acceptable levels of noise you get in many shooting situations, well, this only speaks of how good was the job Panasonic engineers did on that camera.

So now we are saying that the AF100 has the same noise as the FS100 AND that the FS100 has more noise than the F3? Based on what? Neither sound right. From everything I've seen, it appears that the AF100 has more noise than the FS100 and the F3 has similar noise levels to the FS100. Could you provide some evidence otherwise?

bruceallen
09-01-2011, 11:12 PM
OK... after thinking about this for a while and doing a bit more reading, here's my 2c:

Not sure if this is old news to you folks... but I'm beginning to think that the AF100 has a GH1 sensor in it (with a tweaked OLPF), not a GH2 sensor (which in all the footage I've seen is crisper and doesn't have that weird vertical resolution / deinterlacing issue the AF100 does)?

The vertical aliasing stuff I saw in the girl's hair in the real-life shoot can also be seen on the AF100 chart here:
http://provideocoalition.com/index.php/awilt/story/ag-af100_and_pmw-f3_on_the_charts/P1/

...which looks damn identical to the GH1 chart here:
http://provideocoalition.com/index.php/awilt/story/overview_several_first-generation_versacams/P1/

Not that this is a bad thing. It DOES make me look forward to an AF200 based on a GH2 sensor though! And they can up the bitrate and add 3D stereo sync while they're at it :P

I think it is wrong to suggest that the image qualities are the same. To me, just resolution-wise:
- FS100 has significantly higher resolution, at the cost of some overall luma aliasing in all directions and some pretty bad chroma aliasing if you push it the wrong way (but not 5DMk2-level bad).
- AF100 has real vertical resolution issues (managing to look both blurry and aliased - like half of what a 1080p signal should have) but has zero chroma issues

This is a significant difference, which IMHO shouldn't be glossed over.

Why not just say that the AF100 has consistent (but very tolerable) vertical aliasing / deinterlacing / resolution issues... but no nasty chroma moire surprises...

More importantly, just like the HVX200 - which was a great camera even though it was "only" 720p - it is "sharp enough" to make great stuff with.

But to say that the two images look the same seems a bit odd to me?

Personally, the vertical resolution issues would drive me nuts and I'd prefer to shoot FS100... but then if you care that much about resolution over other factors, why not shoot on a RED?

AF100 looks like a great compromise. But yeah, after checking this out, I do want to see an AF200 with full resolution though :) Heck, just go to 4K! I saw that JVC 4K prototype camcorder - looked good! Let's do it!

BTW - Barry, I'm really enjoying reading all of your articles. I need to spend more time on DVXuser - tons of useful stuff. Thank you for all of the countless hours of work and research you've done and written about!

Bruce Allen
www.boacinema.com (http://www.boacinema.com)

reem12
09-01-2011, 11:37 PM
Barry just let me first say you guys are animals with your production skills. I think the first one has a slight more cinematic look then the second but believe also that the first is the af100.

dcloud
09-02-2011, 04:01 AM
OK... after thinking about this for a while and doing a bit more reading, here's my 2c:

Not sure if this is old news to you folks... but I'm beginning to think that the AF100 has a GH1 sensor in it (with a tweaked OLPF), not a GH2 sensor (which in all the footage I've seen is crisper and doesn't have that weird vertical resolution / deinterlacing issue the AF100 does)?

similar generation? maybe. same exact sensor? no.
gh1 (17.3x13) has a bigger sensor than the af100 (17.8x10)

from my experience, AF100 looks a lot detailed in real life than on charts

dcloud
09-02-2011, 04:19 AM
They never said "my camera has too much latitude, im getting posterizing" because it was never available to them before. I just bought sLog today.... like 14 hours ago.

Loaded it up, and recorded to XDcam its full sLog glory. Trying to CC back into something, and its impossible. The macroblocking, banding and XDcam artifacting is dragging all this garbage out of the depths of the signal. When shooting log, youre "middle grey" is now 35IRE. imagine dragging 35 up to the mid 50's on XDcam... The point being, when shooting sLog you're supposed to be recording externally to soemthing at LEAST 10bit. The Alexa can smush 15 stops into a "REC709" space (this what LOG is for) because its 12bit PRORES4444. the gradations are so fine you can squish all 14 stops into a 30 IRE band of the signal and still get something out of it. with that logic, looking at 10bit avcintra 4:2:2 you could get a lot more stops there if panny ever introduces a P2-AF100? it wont be 12bit 4:4:4 but definitely it could add 2-3 stops if they made a panny-log or something ?:)

nyvz
09-02-2011, 12:04 PM
similar generation? maybe. same exact sensor? no.
gh1 (17.3x13) has a bigger sensor than the af100 (17.8x10)

from my experience, AF100 looks a lot detailed in real life than on charts

You are misinformed about gh1 sensor size. Gh1, gh2, and af100 all have same pixel size and physical sensor dimensions, just Af100 uses a cropped portion for its effective imaging area.

Jan_Crittenden
09-02-2011, 03:06 PM
You are misinformed about gh1 sensor size. Gh1, gh2, and af100 all have same pixel size and physical sensor dimensions, just Af100 uses a cropped portion for its effective imaging area.

Not exactly.The pixel size is correct, but the sensor dimension is slightly different 17.8 vs. 17.3. the imager in the GH1 is different altogether in its makeup than the GH2, different chip altogether. The AF100 and the GH2 have a similar starting place at the silicon but the rest of the camera is very different. The AF100 imager is slightly wider and of course is 16:9 so the crop part is true vertically.

This has been an interesting thread to read and I know that Barry will eventually reveal to you all which is which on the cinematic shot. I know and I guessed incorrectly the first time through. Point I see in this whole exercise is that both cameras can make cinematic images. If you don't believe it, look through the collection of videos that I have in my Vimeo group. many look very film-like. These have been done by some really talented people from all over the world. It really goes down to the reality that talent behind the camera, talent in front of the camera and a good script go a long way. I invite anyone to take a look: http://vimeo.com/groups/72874/videos/sort:newest there are about 158, clips, short films and tests over there.

The deal is both cameras offer the ability to give you a shallow depth of field, and if that was the only thing to drive the decision, well it would be difficult because both cameras can do it. Take a look at the rest of the feature sets of both cameras. which one has the tools YOU need. That's the decision maker.

Best regards,

Jan

Michael Olsen
09-02-2011, 04:41 PM
1. Thanks for all the time and effort put into this test, Barry. It's always interesting to read and watch your pieces. Secondly, thanks for all the time and effort you put into answering questions about the test.

2. I think the second portion of the "cinematic scene" is the FS-100.

3. Not to be at all rude but, nyvz, if you so soundly disagree with just about everything regarding this specific test, perhaps you should shoot a test yourself? I'm sure we would all appreciate another set of results, even more so if they contradict these results - nothing is more exciting than anomalies after all. Just make sure to do as Barry did and let us know all the particulars.

4. Thanks also to Jan for dropping in an putting her voice to the choice, as it were. It's always a pleasure to have an official representative around.

dcloud
09-02-2011, 05:53 PM
You are misinformed about gh1 sensor size. Gh1, gh2, and af100 all have same pixel size and physical sensor dimensions, just Af100 uses a cropped portion for its effective imaging area.heheh. i searched for the specs before making that post and its fairly accurate. they both have the same pixel count but different size of sensor.. how could they both be the exact same sensor? Jan answered it for me and thats coming from a panasonic representative

Jan_Crittenden
09-02-2011, 06:46 PM
a panasonic representative

Correction, Panasonic Product Manager, part of the Design Team. Keep in mind that a good number of features that the AF100 has came from this group.

Best,

Jan

TimurCivan
09-02-2011, 07:25 PM
with that logic, looking at 10bit avcintra 4:2:2 you could get a lot more stops there if panny ever introduces a P2-AF100? it wont be 12bit 4:4:4 but definitely it could add 2-3 stops if they made a panny-log or something ?:)

Well, if panaLog was a part of the next AF100, and the horsepower to process it was in the next AF, it would need AVCintra, and P2 media again. IT would be a spectacualar SMASH because you could shoot LOG, IN CAMERA!!!!!!!

I'm thinking the camera would be perhaps double the price, but OH SO WORTH IT! imagine getting LOG performance, with Panasonic Colors, fully functional and record able without all the Gemini, Kipro, Samurai stuff.... it would be an AMAZING AMAZING AMAZING camera.... ( cough* Jan cough*)

Mike Harvey
09-02-2011, 07:42 PM
I'm thinking the camera would be perhaps double the price, but OH SO WORTH IT! imagine getting LOG performance, with Panasonic Colors, fully functional and record able without all the Gemini, Kipro, Samurai stuff.... it would be an AMAZING AMAZING AMAZING camera.... ( cough* Jan cough*)

Double the price is still just south of $10K. For those that need it, that sounds like an absolute steal.

reem12
09-02-2011, 08:41 PM
I think I will hold out until all 3 companies are done milking this 8 bit. and cams become available at around the $10,000 price point for 10 bit or more. I'm seeing that all these 8 bit contraptions are delivering a great image in good hands so it all comes down to what features one needs to accomplish their end result. For me it boils down to image, and I know the workarounds and limitations of my tool so it doesn't warrant me to spend an additional $3000 to get audio and less banding when I can wait and spend the extra money for what's really important to me and thats 10 bit.

dcloud
09-02-2011, 09:14 PM
Correction, Panasonic Product Manager, part of the Design Team. Keep in mind that a good number of features that the AF100 has came from this group.

Best,

Janthat makes it even more accurate & reliable :)

mcgeedigital
09-02-2011, 09:52 PM
Well, if panaLog was a part of the next AF100, and the horsepower to process it was in the next AF, it would need AVCintra, and P2 media again. IT would be a spectacualar SMASH because you could shoot LOG, IN CAMERA!!!!!!!

I'm thinking the camera would be perhaps double the price, but OH SO WORTH IT! imagine getting LOG performance, with Panasonic Colors, fully functional and record able without all the Gemini, Kipro, Samurai stuff.... it would be an AMAZING AMAZING AMAZING camera.... ( cough* Jan cough*)

I would buy this camera.

ian_h
09-02-2011, 11:11 PM
You're definitely not the only one :)So will I.

BrianMurphy
09-03-2011, 06:13 AM
I think I will hold out until all 3 companies are done milking this 8 bit. and cams become available at around the $10,000 price point for 10 bit or more. I'm seeing that all these 8 bit contraptions are delivering a great image in good hands so it all comes down to what features one needs to accomplish their end result. For me it boils down to image, and I know the workarounds and limitations of my tool so it doesn't warrant me to spend an additional $3000 to get audio and less banding when I can wait and spend the extra money for what's really important to me and thats 10 bit.

I totally agree! A 10 bit camera is where it is at for me too. 8 bit is not there for most broadcasters and not everybody wants to hang a recorder off the side of a camera. And I think it would be fair to say that Red will soon have an entry in the over $10K market that will certainly change the game for everyone.

Jan_Crittenden
09-03-2011, 06:18 AM
I totally agree! A 10 bit camera is where it is at for me too. 8 bit is not there for most broadcasters and not everybody wants to hang a recorder off the side of a camera.

Well I am not going to say that 10 bit isn't great, but you have to realize that much of what you have seen in the last 10 years including Planet Earth and any film shot on an F900 or a Varicam was all recorded in 8 bit. 8 bit has been, in reality, the only level of HD recording in camera up until the intro of AVC-Intra. And this is still true. Great results are still very possible with 8 Bit.

Best,

Jan

dcloud
09-03-2011, 06:53 AM
Well I am not going to say that 10 bit isn't great, but you have to realize that much of what you have seen in the last 10 years including Planet Earth and any film shot on an F900 or a Varicam was all recorded in 8 bit. 8 bit has been, in reality, the only level of HD recording in camera up until the intro of AVC-Intra. And this is still true. Great results are still very possible with 8 Bit.

Best,

Janbut still, panasonic should be the first to release a camera with a large sensor & 10bit 4:2:2 AVCINTRA under $8k ;-)

BrianMurphy
09-03-2011, 09:05 AM
I couldn't agree more with your accurate statement about the past. I shot a one hour special that will air around the world this winter on an 8 bit HMC 150. However, I have been told in no uncertain terms by QC people at networks in the USA and Canada that they will not turn a blind eye to future acquisitions that are 8 bit and 1/3 inch is an issue as well. Their position as explained to me was that the material they are buying now for broadcast and that will become part of their inventory will need to stand up to what lies ahead and currently far too much substandard production work knocking on the door. We all know that broadcasters and narrowcasters now recycle their inventory of programming constantly this is understandable as they try to maintain their share of the market as it diversifies to the web and elsewhere.
The issue for me is I want a camera that allows me to work in that market and as younger producers and directors come on the scene they want the look they got when the shot their first indy films, even if we are shooting some guy cleaning a sewer up to his ears in ,,,,dirty water. My 1/3" 8 bit won't cut it, micro 4/3 8 bit has the "look" but...
So the options right now in a less than $20K camera are ? unless you have a surprise up your sleeve for us Jan?
I am a Panasonic fan, I admit it and I have had great stuff come out of my three Panasonic cameras which were quite a switch for me coming from a full size Ikegami beta cam. This is Sony (CBC-Radio Canada) country and I spend and have spent lots of time selling my cameras over EX1 and EX3s. Now I will have to compete with an F3 or who knows.


Well I am not going to say that 10 bit isn't great, but you have to realize that much of what you have seen in the last 10 years including Planet Earth and any film shot on an F900 or a Varicam was all recorded in 8 bit. 8 bit has been, in reality, the only level of HD recording in camera up until the intro of AVC-Intra. And this is still true. Great results are still very possible with 8 Bit.

Best,

Jan

Steve Kahn
09-03-2011, 10:04 AM
Great results are still very possible with 8 Bit.


Without a doubt. I can get images from my AF100 that are STUNNING!

Would they look worse than an exact shot in 10bit or 12bit or 16bit? Well probably very minimally and in most instances you wouldn't even see the differences unless you did a side by side and even then when does that happen in real life?

Same for sharpness. Is the AF as good as an Alexa? No. But, I bet no one would notice or think about that unless they saw the exact images shot on both to compare. But again, who would really care? At some point, and we're getting closer and closer, it really doesn't matter anymore. (Like how CPU power just isn't a factor like it was 5 years ago)

Same for dynamic range. The zone system has 11 discreet zones. At 10.5 stops the AF comes pretty close. Do I really want to have HDR or sLOG to see all the details in the shadows? To be honest. Not especially. Much of the time I want shadows to be dark and help focus the viewer on the subject. Sure in some cases it would be nice but not in most controlled shots.


But as we've been getting at the problem with these two cameras at the moment is dealing with 8bit in post - not straight out of the camera. I can barely push images out of my AF100 in post. I can do mild color to correct imperfect white balances and that's about the extent of it. Anything more and they fall apart and band like crazy. This is the real issue here so let's not confuse matters with minor aliasing or minor resolution or minor dynamic range differences.

Barry_Green
09-03-2011, 03:44 PM
To my (currently tired and grumpy) eyes the weird vertical de-interlacing style aliasing is all over the Wringer test on the AF100 side. Sure there's some color aliasing and minor luma aliasing on the FS100 side and it doesn't look like it's resolving a full 1080p - but the resolution for the AF100 looks like 1920x540 or something! What the...?
Haven't been on in a couple of days and will barely be on for a while more, but I figured I'd answer this one directly. Look at the Wringer -- the AF100 is clearly showing much higher resolution in that chart. Look at the rings -- you can see more resolved rings on the AF100 version. I'm saying that in color resolution, it's higher.

What you're noticing is the crunchy edges caused by having the detail turned up too high for that particular shot. If I can find time I'll reshoot it with the detail lower. Detail doesn't affect the actual resolution of the camera, it only affects the acutance - well, the edges of transitions, making the edge transition sharper. So what you're seeing isn't "1920x540" resolution or "interlacing" or anything like that. It's just the crunchy edges caused by the detail setting.

snowleopard
09-03-2011, 05:00 PM
...Kubrick made crews wait for days until the light was good enough for him. You think any other director gets that luxury??? Nope.

Terrence Malick.

Michael Cimino.

Uh, wait a minute...

reem12
09-03-2011, 11:18 PM
I myself don't necessarily have a big gripe with 8 bit. I'm just stating that I'm seeing nothing that majorly different in image quality that would have people constantly dumping a freshly bought camera when another one comes out just because of great marketing hype. I for one was constantly dumping cameras in search for that elusive film look since the inception of the dvx and 35mm adapters. but now that I have gotten as close to my image goal with the cannon dslrs then I thought I would, I won't be budging until there is something substantial to replace it in it's price range. All these test and shoot outs have proved that all these cams can be made to look amazing. 10 bit at sub $10,000 is what I'm waiting on. Not a repackaged version of what I already have.

jonatha
09-04-2011, 04:05 AM
Well, Alister Chapman showed the same kind of aliasing on his F3 when he first got it.

But saying the FS100 has "the same chip" as the F3 doesn't really mean a whole lot. What counts a lot more is what's behind the chip -- the processing horsepower that wrangles the signal that's coming off the chip. We know, for example, that even though Sony says they're "the same chip" -- the F3 is rated at 800 ISO, and the FS100 is 320 ISO. What's behind that?

Well, it's what's behind the chip, is what's behind that. Think about the price difference -- there's a $9,000 price difference between an FS100 and an F3. Now, let's just arbitrarily throw some of that out for SDI ports and SxS slots and whatnot, say, randomly, $1500. That leaves $7,500 difference. Where does it all go? It goes in the processing. It goes in the computers that take up the F3's much bigger body. And how much computer processing power can you buy for $7,500? A hell of a lot. I mean, a fully-equipped 12-core top-of-the-line Mac Pro costs $4,999. So imagine if you had an entire 12-core Mac Pro processing each and every frame that comes off that sensor. Could it do some noise reduction? I believe it could. I think you could do some amazing real-time noise reduction and sharpening with that much processing hardware behind it, dedicated to doing nothing but processing that image signal.

Or, forget the Mac Pro, just look at this thing...
http://www.newegg.com/Product/Product.aspx?Item=N82E16814195102

That's an ATI graphics card with 1600 processing cores(!) It's the most expensive graphics card on Newegg's site. And it's $2799 -- meaning, for the price difference between an F3 and an FS100, you could cram THREE of those in the body of the F3.

That could work miracles on your footage.

This isn't a trivial difference, it's a huge difference. That much power could pull 14 stops out of a sensor that the little processing in the FS100 only handles 11 from. It could pull sharpness out of a better demosaic algorithm. And it could easily do enough noise reduction to put the F3's rated sensitivity at 800 vs. the FS100's 320. In Alan Roberts' report, he said that his testing revealed significant amounts of in-camera noise reduction going on -- which makes perfect sense in context of all that we've seen.

Now, I'm spitballin' here, but I think it makes sense. I'd like to get an F3 and an FS100 side by side and see just how much a difference that additional processing power the F3 has, really makes to the image.

Well said,thanks for your great comparison

Jan_Crittenden
09-04-2011, 08:28 AM
I couldn't agree more with your accurate statement about the past. My 1/3" 8 bit won't cut it, micro 4/3 8 bit has the "look" but...

But what? The AF100 is $5,000. The F3 is $16,000. And the F3 is still 8Bit, unless you go to an external device, which you could also do with the AF100.


So the options right now in a less than $20K camera are ?

There are no cameras with a large imager that is 10 bit recording under $20,000.


Now I will have to compete with an F3 or who knows.

The F3 is 8Bit MPEG2. That isn't competition for MPEG4. Check out Barry's other article on XDCAM EX vs AVCHD. You have to get to 50MBs to be close to AVCHD in MPEG2. So nice to compete with the F3, but you will be spending 3X the cash.

Best,

Jan

Jan_Crittenden
09-04-2011, 08:33 AM
I can barely push images out of my AF100 in post. I can do mild color to correct imperfect white balances and that's about the extent of it. Anything more and they fall apart and band like crazy. This is the real issue here so let's not confuse matters with minor aliasing or minor resolution or minor dynamic range differences.

Steve what system are you editing on? And what is your intermediate codec? How did you get it there?

Thanks,

Jan

bgundu
09-04-2011, 08:51 AM
I would buy this camera.

If there are any left over, I'll grab one too!

BrianMurphy
09-04-2011, 09:01 AM
Thanks for taking the time to respond. I will wait and see what comes down the pike and count on buying an SD PIX240 in any case. Just seem counter-productive to make a camera that is not broadcast standard and has other issues. I am comfortable spending more for a camera and will wait for the right one and continue to ask questions though sometimes its not popular to do so.
Best
Brian

Grug
09-04-2011, 09:26 AM
But as we've been getting at the problem with these two cameras at the moment is dealing with 8bit in post - not straight out of the camera. I can barely push images out of my AF100 in post. I can do mild color to correct imperfect white balances and that's about the extent of it. Anything more and they fall apart and band like crazy. This is the real issue here so let's not confuse matters with minor aliasing or minor resolution or minor dynamic range differences.


Are the issues you're having with 8-bit 4:2:2 to an external recorder, or with 8-bit 4:2:0 to the internal AVCHD codec?

dop16mm
09-04-2011, 09:40 AM
When will avchd be 10bit 422? version3,4. There is no reason why it isn't possible. AVC-intra is 10 bit at 100Mb/s, and that is for discreet frames. Surely a long gop codec could rival that quality at less than half the data rate.

I think an ideal compromise for in camera acquisition would be to create a long gop version of the 10 bit codec that could be losslessly decoded to the intraframe version without padding bits or color space.

alaskacameradude
09-04-2011, 10:53 AM
But what? The AF100 is $5,000. The F3 is $16,000. And the F3 is still 8Bit, unless you go to an external device, which you could also do with the AF100.



There are no cameras with a large imager that is 10 bit recording under $20,000.




Best,

Jan

Well, I'd have to disagree somewhat. This link:
http://www.bhphotovideo.com/c/product/743863-REG/Sony_PMW_F3L_PMW_F3L_Super_35mm_Full_HD.html

lists the F3 at $13,600. Now I'm guessing that the difference between that and your stated $16000 price, is that this does NOT come
with the 3 sony primes. For some reason, I could not find the package with the lenses listed at B&H....but my guess is it would cost
$16,000 which is your quoted price for the F3. However, lets say you bought the F3 for $13,600. Then say you buy a external recorder
for it. You are going to be under 20 grand and be capable of 10 bit recording. But, you say, you have no lens! This is true....and
if you want to stay under 20 grand, you may have to adapt some SLR lenses for the F3, but it would certainly be doable for under 20 grand.
And while you could buy a external recorder for the AF100, it would only be 8 bit, because that is all the AF100 is capable of outputting.
Now of course this would cost more than the AF100, but to the poster that was asking for a large sensor camera
outputting 10 bit under 20K.....I'd say your only choice would probably be the Sony F3.

BrianMurphy
09-04-2011, 11:12 AM
http://www.bhphotovideo.com/c/product/743866-REG/Sony_PMW_F3K_PMW_F3K_Super_35mm_Full_HD.html

$19,890.00 with three primes

F3 or Epic S if it is out in time unless the replacement for the HPX 500 is something different.

alaskacameradude
09-04-2011, 11:43 AM
http://www.bhphotovideo.com/c/product/743866-REG/Sony_PMW_F3K_PMW_F3K_Super_35mm_Full_HD.html

$19,890.00 with three primes

F3 or Epic S if it is out in time unless the replacement for the HPX 500 is something different.

Well, my mistake, it is NOT $16,00 with lenses, it is almost 20. Still, it is $13,600 without lenses, so you could buy that, an external recorder
and some SLR lenses and be under 20 grand. I think that is your only option at this point....

TheReverend
09-04-2011, 12:59 PM
I still don't think that AF100 bands that badly. I have more doubt in FCP7 and the transcode to ProRes than doubt in the AF100, AVCHD and 8bit. FCP7 gamma shifts, and any "correction" that you do in FCP7 tends towards poor results. It's passable very often, but there is definitely issues. If you are having trouble grading AF100 footage, I recommend trying external recording and checking your postproduction workflow.

nyvz
09-04-2011, 01:32 PM
Not exactly.The pixel size is correct, but the sensor dimension is slightly different 17.8 vs. 17.3. the imager in the GH1 is different altogether in its makeup than the GH2, different chip altogether. The AF100 and the GH2 have a similar starting place at the silicon but the rest of the camera is very different. The AF100 imager is slightly wider and of course is 16:9 so the crop part is true vertically.

This has been an interesting thread to read and I know that Barry will eventually reveal to you all which is which on the cinematic shot. I know and I guessed incorrectly the first time through. Point I see in this whole exercise is that both cameras can make cinematic images. If you don't believe it, look through the collection of videos that I have in my Vimeo group. many look very film-like. These have been done by some really talented people from all over the world. It really goes down to the reality that talent behind the camera, talent in front of the camera and a good script go a long way. I invite anyone to take a look: http://vimeo.com/groups/72874/videos/sort:newest there are about 158, clips, short films and tests over there.

The deal is both cameras offer the ability to give you a shallow depth of field, and if that was the only thing to drive the decision, well it would be difficult because both cameras can do it. Take a look at the rest of the feature sets of both cameras. which one has the tools YOU need. That's the decision maker.

Best regards,

Jan

Sorry I misspoke, yeah the GH1 has a different pixel size with it's different pixel count, but none of these cameras has a 17.3mm imager width since they use the "multi-aspect" sensor rather than the standard 4:3 aspect 4/3" sensor, so I'm not sure what you are referring to when you mention 17.3. The AF100, however, has a smaller imaging area than the GH1 and GH2.

nyvz
09-04-2011, 02:37 PM
3. Not to be at all rude but, nyvz, if you so soundly disagree with just about everything regarding this specific test, perhaps you should shoot a test yourself? I'm sure we would all appreciate another set of results, even more so if they contradict these results - nothing is more exciting than anomalies after all. Just make sure to do as Barry did and let us know all the particulars.


Not at all, I understand what you mean. My disagreement was not with the test itself, I certainly appreciate that Barry made these test videos available to us and is willing to put up with our thoughts and even criticisms. Part of the reason I'm not doing my own test right now is that I trust the materials Barry has provided us with, but it seems I tend to disagree on the with the analysis of those test materials. For some reason we are all looking at the same comparison videos and I say yes clearly that shows the AF100 has more noise, and he says yes clearly it shows that the AF has equal noise but different type of noise. It may just be a matter of our previous experiences driving what we think we see, or just difference in perspective on how to evaluate "amounts" of something that are not being measured quantitatively. It doesn't mean one analysis is right and one is wrong or that another test would make one righter or wronger.

Steve Kahn
09-04-2011, 03:29 PM
Are the issues you're having with 8-bit 4:2:2 to an external recorder, or with 8-bit 4:2:0 to the internal AVCHD codec?

8-bit 4:2:0 to the internal AVCHD codec.

But, perhaps my issue is more with 8bit than anything else.

Also, perhaps when you try to jam more and more steps of latitude into an 8bit codec banding and posterization becomes more an issue, as Timur seems to suggest.

For the record I think AVCHD is a fantastic, fantastic codec!

Steve Kahn
09-04-2011, 03:32 PM
Steve what system are you editing on? And what is your intermediate codec? How did you get it there?

Thanks,

Jan

Hi Jan

Premiere CS5 editing AVCHD straight out of the camera, onto a hard drive then imported into Premiere.

Love the camera, love AVCHD just don't really love how far I can push footage. Perhaps I'm doing something wrong here???

Thanks, btw, for always being a force on this board. Much appreciated.

mcgeedigital
09-04-2011, 05:45 PM
You might want to try taking a short clip and transcoding it to a 10bit intermediate codec like ProRes or DNxHD and then doing a a/b comparison while grading.

Steve Kahn
09-04-2011, 06:13 PM
You might want to try taking a short clip and transcoding it to a 10bit intermediate codec like ProRes or DNxHD and then doing a a/b comparison while grading.

I'll give it a shot. Maybe it will help?

Matt, do you edit using AVCHD or transcode?

mcgeedigital
09-04-2011, 06:52 PM
I transcode everything to Avid DNxHD.

Eta, what are you using to grade?

Mike Harvey
09-04-2011, 07:37 PM
You can also transcode to AVC-Intra since you're using CS5

Jan_Crittenden
09-05-2011, 04:48 AM
Well, I'd have to disagree somewhat. This link:
http://www.bhphotovideo.com/c/product/743863-REG/Sony_PMW_F3L_PMW_F3L_Super_35mm_Full_HD.html
lists the F3 at $13,600.



Actually the F3 has a list price of $16,000. The $13,600 is B & H's price. And you assume that everyone sells the camera for this. And there are not as many adapters for DSLR glass for the F3 as there are the




And while you could buy a external recorder for the AF100, it would only be 8 bit,


Actually it would be a 10 bit recording. And you would be able to push the signal around much like you would the 10 bit output.


[QUOTE=alaskacameradude;2425046]Now of course this would cost more than the AF100, but to the poster that was asking for a large sensor camera
outputting 10 bit under 20K.....I'd say your only choice would probably be the Sony F3.

And I pretty much stand by what I said before because unless you go with the sLog option on the F3 there isn't enough of a difference between it and the AF100 to warrant the difference in price. And don't forget the recorder to go with that.

Best regards,

Jan

Jan_Crittenden
09-05-2011, 04:49 AM
[QUOTE=Steve Kahn;2425208]
Premiere CS5 editing AVCHD straight out of the camera, onto a hard drive then imported into Premiere. Perhaps I'm doing something wrong here???
/QUOTE]
Transcode to a 10 bit intermediate codec, your issues will go away.

Best,

Jan

Jan_Crittenden
09-05-2011, 04:54 AM
Sorry I misspoke, yeah the GH1 has a different pixel size with it's different pixel count, but none of these cameras has a 17.3mm imager width since they use the "multi-aspect" sensor rather than the standard 4:3 aspect 4/3" sensor, so I'm not sure what you are referring to when you mention 17.3. The AF100, however, has a smaller imaging area than the GH1 and GH2.

17.3 is the width of the GH2/GH1 sensor. The number correct on the AF100 is not 17.3 it is 17.8. So while you are correct we are not as tall as the GH2 chip set, we are wider, and I do not need the height of the 4:3 imager as I am only recording HD.

Best,

Jan

maarek
09-05-2011, 06:26 AM
It's quite clear (to me anyway) that the FS100 does have about 3.34 megapixels. So why can the AF100 get close to the sensitivity with 12mp? The AF100 also has a lot of pixels to downsample so it won't be THAT much behind. People forget that the pixels are not 100% and they get a lot of sensitivity if the downscaling is done well. The 7d does scaling really badly skipping a lot of pixels, but if it would do it properly, the 18mp would not be that much of a hindrance.

dcloud
09-05-2011, 06:36 AM
I think the main bottomline with these cameras (AF100 & FS100) is to properly expose your shots. no compromise. post correction may help but you shoot to grade. you grade to enhance not to correct a wrong.

Osslund
09-05-2011, 06:46 AM
I took a photo of the sensor on a F3 and the AF100 and they seamed to have the same proportions so this shows the AF sensor is not like the G/GH cameras sensors which are multiformat.

nyvz
09-05-2011, 07:53 AM
17.3 is the width of the GH2/GH1 sensor. The number correct on the AF100 is not 17.3 it is 17.8. So while you are correct we are not as tall as the GH2 chip set, we are wider, and I do not need the height of the 4:3 imager as I am only recording HD.

Best,

Jan

No it is not, that's only for 4:3 mode. 16:9 is wider on the gh1/gh2 at 18.8mm.

Jan_Crittenden
09-05-2011, 09:04 AM
No it is not, that's only for 4:3 mode. 16:9 is wider on the gh1/gh2 at 18.8mm.

No it isn't. It is 17.3 on the 16:9, at least that is what my engineers are saying.

Best,

Jan

dcloud
09-05-2011, 09:29 AM
previously i read this according to panasonic GH1 & GH2 website 17.3x13 in 4:3

according to DPReview
GH1 (18.89 x 14.48 mm)
GH2 (18.89 x 14.48 mm)

Panasonic Brochure
AF100 (17.8 x 10)

so there this settles it.
GH1 cant have the same sensor as the AF100.
Jan, the next af100 should be 18.9mm ;-)

alaskacameradude
09-05-2011, 10:58 AM
Actually the F3 has a list price of $16,000. The $13,600 is B & H's price. And you assume that everyone sells the camera for this. And there are not as many adapters for DSLR glass for the F3 as there are the



Actually it would be a 10 bit recording. And you would be able to push the signal around much like you would the 10 bit output.




And I pretty much stand by what I said before because unless you go with the sLog option on the F3 there isn't enough of a difference between it and the AF100 to warrant the difference in price. And don't forget the recorder to go with that.

Best regards,

Jan

Umm....Ok, lets look around a little.

http://omegabroadcast.com/fmp/feature-detail.php?fmp-manuRecID=firstpage&-recid=759

http://www.abelcine.com/store/Sony-PMW-F3L-Super-35mm-XDCAM-EX-Camera-body-only/

I quickly looked up two other reputable sources of professional video equipment. Guess what Jan? They
BOTH have the F3 at $13,960. A little more than B&H, but still a couple grand south of your 16 grand
'list price'. Most of us that run a business care more about what we can actually BUY the camera for
(street price) than the (list price). I used B&H as my first example, because as a business person,
they are who I ALWAYS go to when I am pricing equipment.

Second thing. Are you telling me that the AF100 will output 10 bit to an external recorder? That's
what it kind of sounded like. Or are you using 'marketing speak' to say that it will somehow record
10 bit in the camera but only output 8 bit 'padded'? If the AF100 will actually output 10 bit, please
let us all know. I contend that the AF100 will only output 8 bit to an external recorder, and regardless
of if you can 'push it around in post much like you would 10 bit output', it is NOT true 10 bit output,
which you CAN get from the F3! So let's not mislead people here into thinking the AF100 can
output TRUE 10 bit to an external recorder unless it actually can.

I understand you work for Panasonic and want to defend your very fine camera. That's ok.
And honestly...it doesn't need much defending. A big sensor camera that is under 5 grand
(including a lens at least for a limited time)? I mean come on......that's amazing. But
I still say the 10 bit output of a F3 to an external recorder looks better than the external
8 bit output of an AF100 to an external recorder.....even without S Log. As it should.
The F3 costs $13, 600 (or $13,960) depending on where you buy it, to the AF100's $4,795.

I have often said, the first thing you should do when shopping for a camera, is to figure out
your budget.....that is most important. If your budget is 5 grand....why bother looking at the
F3? Look for the best 5 grand camera you can buy. However, the OP was asking for options
for 10 bit output with a big sensor camera, for under 20 grand. And I still stand by MY statement
that for that budget, and those requirements....the F3 will beat the AF100...no doubt in my mind
about it (not in a mean way mind you...that's just my personal opinion :-)

TimurCivan
09-05-2011, 11:19 AM
OK folks cool it.

Jan comes here to answer question directly, not because she has to, but because by interacting with the consumers, she can help design a better product FOR US. So lets keep the interactions civil and polite. No finger pointing, and for the love of god, don't dissect her statements and put words in her mouth.

It is our PRIVILEGE to interact with someone at Panasonic that can help us by answering questions, ad take our input to make a new camera. Remember that.

alaskacameradude
09-05-2011, 11:29 AM
OK folks cool it.

Jan comes here to answer question directly, not because she has to, but because by interacting with the consumers, she can help design a better product FOR US. So lets keep the interactions civil and polite. No finger pointing, and for the love of god, don't dissect her statements and put words in her mouth.

It is our PRIVILEGE to interact with someone at Panasonic that can help us by answering questions, ad take our input to make a new camera. Remember that.

I am sorry, but when I see statements that appear to be misleading, I'm going to challenge them....regardless of if they come from a Panasonic rep or not.

First there is the price thing. I was merely showing, that Jan may be quoting 'list price' but actual 'street price' is significantly
lower. I think the price you can actually buy the camera for (street price) is what we should be using when we are discussing this. I was attempting
to help someone who wanted a large sensor camera with 10 bit output for under 20 grand. I have found THREE places you can buy the F3 for well under
the 16 grand 'list price' (enough under, that you could add the external recorder and be around 16 grand!) So, let's use the real price here.

Second, is the AF100 output 10 bit or not? It appeared that Jan was saying it was....or not. I'm not 100% sure which.
Her quote was

'It would be a 10 bit recording. And you could push it around in post much like a 10 bit output'.

If that's not confusing, I don't know what is. My opinion is this should be cleared up lest people think the AF100 has 10 bit output to an
external recorder. And if it does PLEASE let us know that as well. Again, my posts are attempting to be helpful to the 'poster' who was asking for
options for a big sensor camera with 10 bit output with a 20 grand budget. I don't think the AF100 should be in the discussion, and
I was pointing out why....the 10 bit output thing. I was also attempting to point out that it IS possible (with DSLR glass and external
recorder) to outfit an F3 for that price, and give the OP what they wanted. To do this, I had to question some assertions.
If you as moderators don't like me questioning the Panasonic rep in this way, I will gladly delete my account and go to the other board.
But I'm not going to hang out here and not question misleading or 'marketing' speaking that I don't think is correct. If I am wrong,
PLEASE tell me, as I would be interested in learning my mistake. My impression is the AF100 output to external recorder is not
true 10 bit but padded 8 bit at best. I've been wrong before, and would be interested in learning if I am mistaken again.

TimurCivan
09-05-2011, 11:34 AM
IT does not. The AF100 outputs 8bit. But if you record it to 10bit codec directly before any compression is applied, it will grade far better.

Fohdeesha
09-05-2011, 11:50 AM
...I thought the af100 had a 10-bit SDI output?

Barry_Green
09-05-2011, 11:50 AM
I think the main bottomline with these cameras (AF100 & FS100) is to properly expose your shots. no compromise. post correction may help but you shoot to grade. you grade to enhance not to correct a wrong.
Bless you.

Barry_Green
09-05-2011, 11:53 AM
I was merely showing, that Jan may be quoting 'list price' but actual 'street price' is significantly lower.
Jan quoted list price for both products. She didn't quote street price for the AF100 and list for the F3, she quoted list for both.

Barry_Green
09-05-2011, 11:55 AM
...I thought the af100 had a 10-bit SDI output?
No, it doesn't -- and neither does practically anything else. EX1/EX3, HPX370, and new HPX250... that's about it. No Canons, no JVCs have 10-bit. Oh, there's an NXCAM that does it... NX5 I think?

10-bit output is talked about like it's commonplace, but in actuality it's extremely rare in the under-$10k price bracket.

alaskacameradude
09-05-2011, 11:56 AM
Jan quoted list price for both products. She didn't quote street price for the AF100 and list for the F3, she quoted list for both.

And I quoted 'street price' for both products. See, there is MUCH more of a difference between 'list price' and 'street price' for the F3 than there is for the AF100.
When someone wants to actually know how much something costs....my opinion is that we should go to B&H or AbelCine or one of the many fine places that actually SELL
the cameras and look at the price. If it is listed at price B....we should say it costs price B, NOT say it costs price C just because that is the 'list price'.
Just my opinion.

alaskacameradude
09-05-2011, 11:58 AM
No, it doesn't -- and neither does practically anything else. EX1/EX3, HPX370, and new HPX250... that's about it. No Canons, no JVCs have 10-bit. Oh, there's an NXCAM that does it... NX5 I think?

10-bit output is talked about like it's commonplace, but in actuality it's extremely rare in the under-$10k price bracket.

It is rare in the under 10k price bracket, I think you've named all cameras who do it under 10k. There was a poster who was asking
about it in the under $20k price bracket (and adding that he wanted a large sensor.) Which was why I included the F3 in the discussion.
Look, I think 8 bit is fine myself. If I've shot something so badly with my FS100 (or my friend's AF100 for that matter) that I can't
grade it....that's my bad for shooting it wrong (in my opinion). But there may be some high end users who want or need 10 bit.
Is it possible to get a large sensor camcorder that outputs 10 bit for under 20 grand? I think the F3 is the only one at this point.
As you said, very rare.

Barry_Green
09-05-2011, 12:32 PM
And I quoted 'street price' for both products. See, there is MUCH more of a difference between 'list price' and 'street price' for the F3 than there is for the AF100.
When someone wants to actually know how much something costs....my opinion is that we should go to B&H or AbelCine or one of the many fine places that actually SELL
the cameras and look at the price. If it is listed at price B....we should say it costs price B, NOT say it costs price C just because that is the 'list price'.
Just my opinion.
That is your opinion and you're entitled to it. But Jan is an official Panasonic employee who is likely required to quote MSRP because that is the official price, and she's a Panasonic official.

I'm not telling you you're wrong for your preferences, I'm trying to explain why (I think) she did what she did.

J Michael
09-05-2011, 01:10 PM
Sony has their dealers under pretty tight control WRT the prices they sell for - look around, the price is pretty much the same anywhere you look if you buy from official source. And I suppose Panasonic does this too, it's pretty difficult for small dealers to make anything on hardware otherwise due to margins being so slim.

Did Panasonic ever come back with an official response to the independent zone plate resolution test that was made earlier this year? I recall Jan saying she was awaiting something official from the factory on that.

Melvin Harris
09-05-2011, 02:17 PM
I looked at this footage... I looked at it on mac, pc, cinema displays, monitors, tvs, I even snuck the footage into staples and best buy and looked at it on their stuff and I honestly have to say that any noise grain and artifacting that I saw out of that footage didn't mean anything. Plus, people who were watching with me couldn't have cared less about it. I think that sometimes we forget that we are not our audiences. These are two incredible cameras and they look good. Thanks Barry for a tremendous test. It more than solidified my decision.

alaskacameradude
09-05-2011, 04:14 PM
That is your opinion and you're entitled to it. But Jan is an official Panasonic employee who is likely required to quote MSRP because that is the official price, and she's a Panasonic official.

I'm not telling you you're wrong for your preferences, I'm trying to explain why (I think) she did what she did.

Barry,

Actually now that you say that, it makes sense that Panasonic may require her to quote MSRP. However, not to come across as some sort of argumentative tool, but if someone
is asking what they can get for $20 grand, it just made sense (in my mind) to correct the perception that it ACTUALLY cost 16 grand for the F3. I understand that Jan may
be constrained in what she can say, but I'm not, and since the actual price a person will pay is different, and I felt that is valuable information for a prospective buyer to
know.....I think it is appropriate for me to jump and and make that clear.

nyvz
09-05-2011, 04:30 PM
I think the main bottomline with these cameras (AF100 & FS100) is to properly expose your shots. no compromise. post correction may help but you shoot to grade. you grade to enhance not to correct a wrong.

It is very preferable to do that, but in practice that may not always be possible. When operating these cameras, we are faced every day with situations that may not have an ideal "proper" exposure. It is certainly diplomatic to say both cameras perform their best when operated correctly, but these tests of what can be pushed and how much have very real implications to those using them on a regular basis. If your particular shooting situation means you are lit to levels below ideal or your contrast is greater than the latitude of the camera, it helps to know which camera will be the right tool for the job and what its limits are.

I have been quite surprised with the level to which I can overexpose the FS100 in bright exterior situations to keep even dark shadow detail more visible without sacrificing the look of my highlights, whereas on the AF100 I find myself constantly underexposing and losing my shadow detail in similar situations because if anything blows out it just starts looking quite problematic. Fortunately, I can usually change the shot a bit to make things work better when shooting AF100 but it can be quite annoying to frequently changing my shots to work with the camera rather than have the camera work for me.

dcloud
09-05-2011, 05:51 PM
It is very preferable to do that, but in practice that may not always be possible. When operating these cameras, we are faced every day with situations that may not have an ideal "proper" exposure. It is certainly diplomatic to say both cameras perform their best when operated correctly, but these tests of what can be pushed and how much have very real implications to those using them on a regular basis. If your particular shooting situation means you are lit to levels below ideal or your contrast is greater than the latitude of the camera, it helps to know which camera will be the right tool for the job and what its limits are.

I have been quite surprised with the level to which I can overexpose the FS100 in bright exterior situations to keep even dark shadow detail more visible without sacrificing the look of my highlights, whereas on the AF100 I find myself constantly underexposing and losing my shadow detail in similar situations because if anything blows out it just starts looking quite problematic. Fortunately, I can usually change the shot a bit to make things work better when shooting AF100 but it can be quite annoying to frequently changing my shots to work with the camera rather than have the camera work for me.definitely. and my statement refers to those who dont know the camera and expect it to perform like something else. its good that you know both cameras weaknesses and strengths.

I also have problems with af100 when it comes to high contrast scenes. but thats a common problem for me for every other camera. most of the time i just expose for the subject. (and thats me im not suggesting anything to you ;P)

ian_h
09-06-2011, 12:46 AM
I'd love to see some side-by-side comparison - actual evidence of the same shot under identical conditions with the same lenses (!!) - of a highlights roll-off comparison between these two cameras. I can't find one on the web. Barry has done a fantastic job of giving us some great, hard tests - there are a lot of opinions and anecdotes floating around, but nothing is a substitute for real tests. Anyone want to shoot them?

I've been invested in the GH1/2 world the last year or so. That means Novoflex adapters (x3 for my different mount lenses) which isn't a trivial cost. I have access to an AF and have shot a fair bit on that, but may well look at buying an AF or FS in the near future - the FS is $1000 more (both with kit lens), and $2000 if I look at adding new adapters (since I'll only deal with Novoflex or MFT or other high quality adapters and they don't come cheap - I've been down that knockoff ebay route), which is not a trivial amount of money relative to the price of the cameras.

Comparisons to the F3 are a little superfluous, let's be honest. At nearly 3x the price (on street prices) it's just simply not in the same class. If I had $20k to spend on a camera would I look at the F3 with S-log and an external recorder? Undoubtedly - it's a significant step up from the AF! But that's really just a silly debate. You could apply the same logic to the F3 and cameras 3x more expensive than that - cameras aren't competing against cameras in different classes unless they're being bought by cost-cutting production companies trying to hide the shortcomings and get away with it. No amount of arguing in favour of the F3 is going to work if you need 4k. No amount of argument in favour of the AF is going to work if you need 10 bit and s-log...

Oh, and PS: Jan, if you're listening, please get Panasonic to make frame rate independent of shutter speed on the AF. I'd love to be able to select 1/100 shutter on 24/60p overcrank without jumping through convoluted calculations on percentage degrees so that I can get maximum slow motion here in PAL land.

Jan_Crittenden
09-06-2011, 04:46 AM
Oh, and PS: Jan, if you're listening, please get Panasonic to make frame rate independent of shutter speed on the AF. I'd love to be able to select 1/100 shutter on 24/60p overcrank without jumping through convoluted calculations on percentage degrees so that I can get maximum slow motion here in PAL land.

Why don't you use the degrees? Rather than shutter speeds, and the other benefit the motion blur will look the same whther you are in 60P or 24P.

Best,

Jan

ian_h
09-06-2011, 04:54 AM
Why don't you use the degrees? Rather than shutter speeds, and the other benefit the motion blur will look the same whther you are in 60P or 24P.

Best,

Jan

Hi Jan - yeah, works great, except if you're shooting 60/24p in PAL land under street lights at 180deg shutter - in which case you get terrible flicker. At 60p/24 you need 1/100 shutter, or at 24p you need 1/50 shutter, and to get there involves a very convoluted calculation of degrees (since otherwise you're only offered 1/60 or 1/120 which causes flicker). Of course I could shoot 50/25p and 25p but that only gives 1:2 slowdown instead of 1:2.5 (every bit counts) and I'm now shooting everything at 24p to provide compatibility with the GH2 as a b-cam since it doesn't shoot 25p (or not properly).

Would be wonderful if we could have all the frame rates and all the shutter speeds selected completely independently to further enhance the AF's fantastic world camera capabilities - currently a big advantage over the FS, although the FS is rumoured to get an update at the end of the year to close this gap.

Jan_Crittenden
09-06-2011, 05:15 AM
So, let's use the real price here.


I am not in the resell business and if you ask me what is the price of the AF100 it is $4995. I do not lead anyone to believe that they can buy something for less than list as that is not my job nor authority. My point was that unless you put the sLog on the F3, there isn't sufficient difference between them to warrant the extra money. So take the camera, add the sLog, the external recorder that will record the extra bandwidth needed to get the benefit of sLog, and lenses and then you have a serious jump in quality. this I believe takes you closer to the $20,000+, although granted not by much. And I think that when I wrote the first reply I may not have included the cost of the sLog, but without it, the F3 seems only marginally better than the AF100 and thus almost 3X time the price.



Second, is the AF100 output 10 bit or not? It appeared that Jan was saying it was....or not. I'm not 100% sure which.
Her quote was:


JCL: previous post: It would be a 10 bit recording. And you could push it around in post much like a 10 bit output'.

Sorry if you found that confusing, but it is true. If you have a recorder that takes an 8 bit input(AF100 or FS100) yet makes a 10 bit recording, it is a 10 bit recording. it will massage in post more like a 10 bit recording than an 8 bit.


I don't think the AF100 should be in the discussion,


This thread is in the AF100-FS100 discussion and frankly by virtue of the name of the thread and discussion it should be here. If you would like to start an F3 discussion and how to price and get the best deal for it, there is always the F3 area on this very same board. You are in a thread that talks about the AF100 vs the FS100, obviously a different price point.



I was pointing out why....the 10 bit output thing. I was also attempting to point out that it IS possible (with DSLR glass and external
recorder) to outfit an F3 for that price, and give the OP what they wanted. To do this, I had to question some assertions.
If you as moderators don't like me questioning the Panasonic rep in this way,


I think the objectionable part is in the tone you use or at least what I find objectionable..



If I am wrong,
PLEASE tell me, as I would be interested in learning my mistake. My impression is the AF100 output to external recorder is not
true 10 bit but padded 8 bit at best.


Well you are wrong, you could improve the output from your FS100 as well by putting a 10 bit recorder on it as well. It will make your post flow better.

Best regards,

Jan

dcloud
09-06-2011, 05:41 AM
Hi Jan - yeah, works great, except if you're shooting 60/24p in PAL land under street lights at 180deg shutter - in which case you get terrible flicker. At 60p/24 you need 1/100 shutter, or at 24p you need 1/50 shutter, and to get there involves a very convoluted calculation of degrees (since otherwise you're only offered 1/60 or 1/120 which causes flicker). Of course I could shoot 50/25p and 25p but that only gives 1:2 slowdown instead of 1:2.5 (every bit counts) and I'm now shooting everything at 24p to provide compatibility with the GH2 as a b-cam since it doesn't shoot 25p (or not properly).

Would be wonderful if we could have all the frame rates and all the shutter speeds selected completely independently to further enhance the AF's fantastic world camera capabilities - currently a big advantage over the FS, although the FS is rumoured to get an update at the end of the year to close this gap.true

ian_h
09-06-2011, 05:47 AM
Haven't got an AF here, but I'm reasonably sure it doesn't work like this. If you've got the cam set to 59.95 hz (24p or 60/24p) it only gives you 1/60, 1/120 etc. shutter, and if you've got it set to 50hz (25p) it only gives you the PAL shutters - i.e. the assumption that if you're in NTSC land you won't want PAL shutter speeds. Works great unless you want to shoot 24p in PAL land, like the above scenario. But I do stand to be corrected...

Jan_Crittenden
09-06-2011, 05:50 AM
Hi Jan - yeah, works great, except if you're shooting 60/24p in PAL land under street lights at 180deg shutter - in which case you get terrible flicker. At 60p/24 you need 1/100 shutter, or at 24p you need 1/50 shutter, and to get there involves a very convoluted calculation of degrees (since otherwise you're only offered 1/60 or 1/120 which causes flicker). Of course I could shoot 50/25p and 25p but that only gives 1:2 slowdown instead of 1:2.5 (every bit counts) and I'm now shooting everything at 24p to provide compatibility with the GH2 as a b-cam since it doesn't shoot 25p (or not properly).

See the shutter speeds relate to time and you are asking me to change time, that doesn't work. But there is a degree in the list that will resolve your issue , look for 172.8. It was put there for this particular situation.


Would be wonderful if we could have all the frame rates and all the shutter speeds selected completely independently to further enhance the AF's fantastic world camera capabilities - currently a big advantage over the FS, although the FS is rumoured to get an update at the end of the year to close this gap.

This degree capability, that AF100 has it, the F3 has it and the FS100 does not have it. This is even more huge. Point is the shutter speed, denotes time, and the degree denotes an opening.

Best,

jan

BrianMurphy
09-06-2011, 07:23 AM
Well, my questions have certainly caused some flurry. My intention was certainly not to inspire conflict but to clarify a couple of issues. The one issue that must remain a fact is that neither camera FS100 nor AF100 will be accepted as a principal camera for television production by most networks. The F3 at 20 K will be. The other offerings since they were mentioned in this thread by Barry Green will also soon not be in the broadcast category by many networks due to the 1/3" chip size. I shoot for broadcast, not indy film or local commercials or other stuff. I have to jump through the network hoops and must comply or I don't work so if you shoot for broadcast, check your regular clients and make sure you know what is coming down the pike. I am also concerned with the number of two and three month old AF100s that are popping up every week on the marketplace here and elsewhere.
In any case Jan Crittenden's point is valid and this is indeed a thread about comparing Panasonic's AF100 to the Sony FS100 and perhaps I should not have raised the issue in this thread.

Jan_Crittenden
09-06-2011, 07:29 AM
Haven't got an AF here, but I'm reasonably sure it doesn't work like this. If you've got the cam set to 59.95 hz (24p or 60/24p) it only gives you 1/60, 1/120 etc. shutter, and if you've got it set to 50hz (25p) it only gives you the PAL shutters - i.e. the assumption that if you're in NTSC land you won't want PAL shutter speeds. Works great unless you want to shoot 24p in PAL land, like the above scenario. But I do stand to be corrected...

Just walked into the lab picked up my AF100, went to 50Hz, 1080/25P, VFR set to 24P, and the SynchoScan set to degrees and dialed in the 172.8. This one is in there just for 24P in PAL land.

Best,

Jan

Jan_Crittenden
09-06-2011, 07:49 AM
Well, my questions have certainly caused some flurry. My intention was certainly not to inspire conflict but to clarify a couple of issues. The one issue that must remain a fact is that neither camera FS100 nor AF100 will be accepted as a principal camera for television production by most networks. The F3 at 20 K will be.

But only if you record to an external device. And then you may make either of the 100s acceptable by doing the same thing. The codec on all three cameras is 8 bit, the F3 is MPEG2, and the others are MPEG4. For some broadcasters even 35Mbs is too low. external devices will resolve all of that for any of them.


The other offerings since they were mentioned in this thread by Barry Green will also soon not be in the broadcast category by many networks due to the 1/3" chip size. I shoot for broadcast, not indy film or local commercials or other stuff. I have to jump through the network hoops and must comply or I don't work so if you shoot for broadcast, check your regular clients and make sure you know what is coming down the pike.

Depends on your networks but here in the US there is a huge, I repeat, HUGE number of AG-HPX170s in broadcast, including NBC, ABC ESPN and others. I anticipate that 1/3" will continue to go up with the newest introductions of the 250. 160 and 130.


I am also concerned with the number of two and three month old AF100s that are popping up every week on the marketplace here and elsewhere.

I would say this is most likely due to the fact that the Original Buyer was not equipped to deal with a cinema style camera. People that come to the camera with the hope and expectation that it will be like an HVX200, may find themselves either delighted as the similarity is there or disillusioned as they really need the ease of an integrated lens. I have seen this with virtually every camera introduction, even the HPX300 and 370.


In any case Jan Crittenden's point is valid and this is indeed a thread about comparing Panasonic's AF100 to the Sony FS100 and perhaps I should not have raised the issue in this thread.

I recognize why you did; what I heard in your question was that you really wanted a camera that cost like the AF100 or FS100 that would suit your purposes, not that you wanted a $20,000 purchase. I fully believe that the $5000 cameras will work for you, provided you take the time to master the control over the camera. Take a look at the clips on the vimeo site, http://vimeo.com/groups/72874/videos/sort:newest, they are varied in application and for the most part show a growing mastery over the camera (there are a few tests there). Remember many of the guidelines that broadcasters put out there is to up the ante in the quality control department. The thinking is that the more you spend on a camera, the longer you have been in the business and you will be handing in better images. If the quality of the images are good, in general there is no complaint.

Best regards,

jan

alaskacameradude
09-06-2011, 10:26 AM
I am not in the resell business and if you ask me what is the price of the AF100 it is $4995. I do not lead anyone to believe that they can buy something for less than list as that is not my job nor authority. My point was that unless you put the sLog on the F3, there isn't sufficient difference between them to warrant the extra money. So take the camera, add the sLog, the external recorder that will record the extra bandwidth needed to get the benefit of sLog, and lenses and then you have a serious jump in quality. this I believe takes you closer to the $20,000+, although granted not by much. And I think that when I wrote the first reply I may not have included the cost of the sLog, but without it, the F3 seems only marginally better than the AF100 and thus almost 3X time the price.


Barry explained that, and it makes sense that you must use MSRP. However, those of us who are NOT reps for some company or another are MUCH more interested in the ACTUAL price of something.
Now for the AF100, there is not much of a difference from your $4995 MSRP to the $4795 'street' price. For the F3 however, there is a substantial difference between your quoted $16,000 MSRP
and the $13,600 'street price'. Seems to me that it would be valuable for a prospective buyer to know this, so I figured I'd point that out, seems like valuable information to me.


Sorry if you found that confusing, but it is true. If you have a recorder that takes an 8 bit input(AF100 or FS100) yet makes a 10 bit recording, it is a 10 bit recording.
it will massage in post more like a 10 bit recording than an 8 bit.


Well you are wrong, you could improve the output from your FS100 as well by putting a 10 bit recorder on it as well. It will make your post flow better.

Yes, I understand that. What my question was, is the external recorder actually receiving a full 10 bit signal or 'padded' 8 bit? I am aware that on the FS100
it is 'padded' 8 bit. I thought the same was true with the AF100. It's kind of like 'transcoding' something from 4:2:0 AVCHD to 4:2:2 ProRes. You can't actually
improve the image that is there. It may make it a little better to play with in post, but it is not as good as an actual 4:2:2 recording. In like manner
a 'padded' 8 bit recording is not as good as a 'true' 10 bit recording. So am I really wrong? I understand it will 'make my post flow better', but my question is if
the AF100 outputs TRUE 10 bit or 'padded' 8 bit. I assert that the AF100 is giving a 'padded' 8 bit signal to any external recorder and I don't think I 'am wrong'.
Earlier in the thread, TimurCivan said it was 8 bit. That's fine, I personally think the 8 bit out of the AF100 is great, but I was just saying, that it wasn't
'true' 10 bit.


This thread is in the AF100-FS100 discussion and frankly by virtue of the name of the thread and discussion it should be here. If you would like to start an F3 discussion and how to price and get the best deal for it, there is always the F3 area on this very same board. You are in a thread that talks about the AF100 vs the FS100, obviously a different price point.

Well, you are certainly correct that I contributed to 'derailing' a thread beyond it's title. But I wasn't the only one.



But what? The AF100 is $5,000. The F3 is $16,000. And the F3 is still 8Bit, unless you go to an external device, which you could also do with the AF100.




The F3 is 8Bit MPEG2. That isn't competition for MPEG4. Check out Barry's other article on XDCAM EX vs AVCHD. You have to get to 50MBs to be close to AVCHD in MPEG2. So nice to compete with the F3, but you will be spending 3X the cash.

These posts were both before I entered the discussion. So is it inappropriate for me to comment
on this comparison since the thread title is 'AF100 vs. FS100'? It didn't seem like it to me as others including you
were commenting on the comparison between the AF100 and the F3. You are right, they are totally different price points
and normally I would not think they would be compared. But it wasn't me who 'started an F3 discussion'....I certainly
plead guilty to contributing to it, but that's all. I think the reason the F3 was brought into it, was a poster was asking
about the availability of a large sensor camera with 10 bit output....that was under 20 grand.



So the options right now in a less than $20K camera are ?

You answered


There are no cameras with a large imager that is 10 bit recording under $20,000.

I still say there is.....the F3. Yeah, you have to buy an external recorder, but even with that cost, it is under 20 grand....and
it is true 10 bit output.


I think the objectionable part is in the tone you use or at least what I find objectionable.

I really don't mean to be. I think the AF100 is a great camera. It has a HUGE feature I wish my FS100 had,
built in ND filters. It only costs $4795 and right now you can get a free lens with it! I mean come on, that's unheard of
just a year ago. The fact that I don't think it compares with the F3 does not mean it is not a GREAT camera in it's own right.
It's just not an F3. And an F3 is not an Alexa.....and so on.

bumkicho
09-06-2011, 11:07 AM
Wow really? Do we really need to go on and on about MSRP vs "actual" price, and 8bit output vs 10bit output? Move on people.

Jan_Crittenden
09-06-2011, 11:27 AM
Yes, I understand that. What my question was, is the external recorder actually receiving a full 10 bit signal or 'padded' 8 bit? I am aware that on the FS100 it is 'padded' 8 bit. I thought the same was true with the AF100.



that is true.



It's kind of like 'transcoding' something from 4:2:0 AVCHD to 4:2:2 ProRes.


Not really. As it never touches the compression engine.



You can't actually improve the image that is there. It may make it a little better to play with in post, but it is not as good as an actual 4:2:2 recording.


It is an actual 4:2:2 recording as it has never seen the compression which is where the 4:2:0 is applied.


In like manner
a 'padded' 8 bit recording is not as good as a 'true' 10 bit recording. So am I really wrong?


Actually you are. There is significant difference in the signal between an 10 bit 4:2:2 recording from the 8 Bit stream and what is recorded on the SD card. There is a difference between that Card recording and the transcode and the direct recording. You are saying that there would be no advantage of recording an uncompressed signal output to 200Mbs versus going to 4:2:0 at 24, or 35 in MPEG2. Frankly anywhere along the chain you can gain an inch you should take it if you are striving for a higher quality signal, and I think that was the OP's desire.


I assert that the AF100 is giving a 'padded' 8 bit signal to any external recorder and I don't think I 'am wrong'.


I have stated probably 600 times since December all over the internet that it was 8 Bit 4:2:2. And if you send an uncompressed 8 Bit 4:2:2 signal to a 10 bit recorder at a higher data rate like 100 Mbs or 200 Mbs. it will be an improvement over the recording in camera. You are only wrong in thinking that it doesn't make a big difference, it does.

nyvz
09-06-2011, 12:10 PM
Sorry if you found that confusing, but it is true. If you have a recorder that takes an 8 bit input(AF100 or FS100) yet makes a 10 bit recording, it is a 10 bit recording. it will massage in post more like a 10 bit recording than an 8 bit.


You're still going to confuse people with that statement. Perhaps you have a different understanding of what people will take "it will massage in post more like 10bit" to mean. Upconverting from 8bit output in the recorder by recording in 10bit or by transcoding to a 10bit format later will likely not give significantly different results. In many cases, especially if the 10 or 8bit is being decoded and manipulated as 16 or 32bit RGB as most pro apps should, there will likely be no benefit from recording or transcoding an 8bit output in a 10bit format.

Jan_Crittenden
09-06-2011, 12:17 PM
there will likely be no benefit from recording or transcoding an 8bit output in a 10bit format.

Sorry but you are incorrect in thinking this. You have gained 4;2:2 and the full 10 bit on the recorder. Take a search around here on DVXuser and on the internet where the tests have been done and it is true.

Best,

Jan

DadinWestchester
09-06-2011, 12:39 PM
"Depends on your networks but here in the US there is a huge, I repeat, HUGE number of AG-HPX170s in broadcast, including NBC, ABC ESPN and others. I anticipate that 1/3" will continue to go up with the newest introductions of the 250. 160 and 130."

Jan, you know these are news or reality cameras and not considered production cams. These were never purchased for studio use or to be integrated in scripted series.

nyvz
09-06-2011, 12:51 PM
Sorry but you are incorrect in thinking this. You have gained 4;2:2 and the full 10 bit on the recorder. Take a search around here on DVXuser and on the internet where the tests have been done and it is true.

Best,

Jan

Ah that is the source of confusion. Some are talking about 10bit vs 8bit specifically, but you are talking about 10bit 4:2:2 uncompressed vs 8bit 4:2:0 compressed AVCHD. Please be more specific when you are talking about a particular workflow, we all seem to be correct but are arguing about different things that do not relate to each other. Of course 4:2:2 uncompressed offers great benefits over 4:2:0 AVCHD, but that does not relate directly to the discussion some of us are having specifically about 8bit vs 10bit recording for 8bit originated material irrespective of compression and color sampling parameters.

nyvz
09-06-2011, 01:16 PM
It is an actual 4:2:2 recording as it has never seen the compression which is where the 4:2:0 is applied.

Actually you are. There is significant difference in the signal between an 10 bit 4:2:2 recording from the 8 Bit stream and what is recorded on the SD card. There is a difference between that Card recording and the transcode and the direct recording. You are saying that there would be no advantage of recording an uncompressed signal output to 200Mbs versus going to 4:2:0 at 24, or 35 in MPEG2. Frankly anywhere along the chain you can gain an inch you should take it if you are striving for a higher quality signal, and I think that was the OP's desire.

I have stated probably 600 times since December all over the internet that it was 8 Bit 4:2:2. And if you send an uncompressed 8 Bit 4:2:2 signal to a 10 bit recorder at a higher data rate like 100 Mbs or 200 Mbs. it will be an improvement over the recording in camera. You are only wrong in thinking that it doesn't make a big difference, it does.

1. He was talking about transcoding from AVCHD to Prores, of course it has seem compression. Are you referring to a different "it"?

2. Again, he wasnt talking about 10bit 4:2:2 vs what is on the SD card, not sure why you keep assuming he is

3. Of course we all know that, but the question posed is essentially whether if you send an 8bit 4:2:2 signal to an 10bit recorder will it be somehow better than sending that same 8bit 4:2:2 to an 8bit recorder with the same basic encoding parameters besides the bit depth per channel? This is the question to answer and the one that I was answering when I said that I thought that no, it would not be better to record 10bit and would not provide the same benefit of recording 10bit from a camera that is designed to output a full 10bit signal vs 8bit. Perhaps it was not clear enough. Thanks

Jan_Crittenden
09-06-2011, 01:24 PM
Please be more specific when you are talking about a particular workflow, we all seem to be correct but are arguing about different things that do not relate to each other. Of course 4:2:2 uncompressed offers great benefits over 4:2:0 AVCHD, but that does not relate directly to the discussion some of us are having specifically about 8bit vs 10bit recording for 8bit originated material irrespective of compression and color sampling parameters.

But if you were to take a 10 bit output and put it into an 10 bit input, and compare that with the 8 bit into 10 bit, here the difference is mor marginal that you would think. Is there a difference? sure, but quantifiable? much more difficult. If you want a higher quality out of any of the $5000 cameras plug in a 10 bit 4:2:2 200Mbs external recorder.

You will see a quantifiable difference. And this is what I have been saying all through this thread. Sorry if you were reading any thing else.

Best,

jan

Fohdeesha
09-06-2011, 02:44 PM
If you have a recorder that takes an 8 bit input(AF100 or FS100) yet makes a 10 bit recording, it is a 10 bit recording. it will massage in post more like a 10 bit recording than an 8 bit.


But if you were to take a 10 bit output and put it into an 10 bit input, and compare that with the 8 bit into 10 bit, here the difference is mor marginal that you would think.


This is what Panasonic reps actually believe.

alaskacameradude
09-06-2011, 02:49 PM
1. He was talking about transcoding from AVCHD to Prores, of course it has seem compression. Are you referring to a different "it"?

2. Again, he wasnt talking about 10bit 4:2:2 vs what is on the SD card, not sure why you keep assuming he is

3. Of course we all know that, but the question posed is essentially whether if you send an 8bit 4:2:2 signal to an 10bit recorder will it be somehow better than sending that same 8bit 4:2:2 to an 8bit recorder with the same basic encoding parameters besides the bit depth per channel? This is the question to answer and the one that I was answering when I said that I thought that no, it would not be better to record 10bit and would not provide the same benefit of recording 10bit from a camera that is designed to output a full 10bit signal vs 8bit. Perhaps it was not clear enough. Thanks

Thank you nyvz for putting it better than I did. I unfortunately am not always great at stating things on forums. You hit the nail on the head, what I am saying is that if you send an
8bit 4:2:2 signal to a 10bit recorder it will not provide the same benefit of recording 10bit from a camera that is designed to output a full 10bit signal. At least one person understood
what I was saying, thanks much for clarifying my rambling thoughts!!

alaskacameradude
09-06-2011, 02:55 PM
But if you were to take a 10 bit output and put it into an 10 bit input, and compare that with the 8 bit into 10 bit, here the difference is mor marginal that you would think. Is there a difference? sure, but quantifiable? much more difficult. If you want a higher quality out of any of the $5000 cameras plug in a 10 bit 4:2:2 200Mbs external recorder.

You will see a quantifiable difference. And this is what I have been saying all through this thread. Sorry if you were reading any thing else.

Best,

jan

Jan,

I not only understand this, I actually agree with you! I don't think you are going to see a huge difference between 'true' 10 bit and 'padded' 8 bit either. I'm not even
sure there will be that big a difference between 4:2:0 AVCHD (which seems to look really good) and any external recorder at all. But that wasn't my point. My point was, if
someone is asking for 10 bit true output, the F3 has it.....the AF100 and FS100 don't. To me, that's no big deal at all! But someone was asking for it, so I thought
I'd try to point them at the camera that had it....they may have some specific reason for needing it. I certainly don't consider it a must have myself, as you have
said, the 8bit output looks awfully good.....so in no way is this intended as a slight at the AF100.....just a fact that I was stating for the help of another poster.

Jan_Crittenden
09-06-2011, 04:14 PM
This is what Panasonic reps actually believe.

And while I know that I believe this, I actually know this which is slightly different. What is it that you meant?

Best,

Jan

Melvin Harris
09-06-2011, 04:17 PM
could we possibly get back to the comparison? I would love to know what workarounds exist for highlight troubles on the fs as well as the af based on this test, and chet like that...

Melvin Harris
09-06-2011, 04:21 PM
... and Jan is the best. I don't think that i could take this much!

Postmaster
09-06-2011, 05:06 PM
I just want to remember you guys, that HDMI 10 bit is not really 10 bit.
And no, it´s not just zeros orl empty information - bit 9 is an "otimizing bit" and bit 10 "DC Balancing".

But even if you have SDI and real 10 bit - do you have a 10 bit monitor and graphic card?

Barry showed us, that both cameras are brilliant and capable of cinematic pictures (in the right hands).
Both the AF100 and the FS100 would have been a dream 5 years ago.
We all better should get out and shoot something with these great new tools, instead of constant pixel peeping.

end of rant

Frank

MadMonkFish
09-06-2011, 05:14 PM
Both the AF100 and the FS100 would have been a dream 5 years ago.
We all better should get out and shoot something with these great new tools, instead of constant pixel peeping.
end of rant
Frank

I said the same thing regarding the 'dream' of an FS100 or AF100 5 years ago, in another thread... must be a European way of thinking... LOL ;)

Pixel Peeping is a sin!

Elton
09-06-2011, 06:06 PM
This thread would make for a good Philip K. Dick-type paranoid/dystopian tale: "Do Cam-droids Dream of a Pixel Peep?

MadMonkFish
09-06-2011, 06:11 PM
This thread would make for a good Philip K. Dick-type paranoid/dystopian tale: "Do Cam-droids Dream of a Pixel Peep?

:cheesy:

Camelspiders
09-06-2011, 08:30 PM
And while I know that I believe this, I actually know this which is slightly different. What is it that you meant? Best, Jan Adding 2 bits of zeros to pad a signal does not, in fact, increase the source quality, and any modern software calculates luma and chroma in floating point regardless of the source, so bringing in a 10-bit file for the sake of "working in 10-bit" is totally null.

Fohdeesha
09-06-2011, 09:04 PM
And while I know that I believe this, I actually know this which is slightly different. What is it that you meant?

Best,

Jan

What I meant is, the fact that you would not only state that "recording 8-bit to a 10-bit file allows more massaging in post", but also state that "the difference between 10-bits recorded to a 10-bit file is marginal" - is a sign of a lack of understanding in the realm of digital cinema, to say the least.

To say these things clearly indicates you don't pipeline any kind of footage other than something straight off camera into a final cut system. One, recording 8-bits into a 10-bit file does not allow more massaging in post. It allows the exact same amount, as the recording contains the exact 8-bit information off camera with yes, zeroes to pad the stream. Loading in a 10-bit file does not make the software automatically jump to working in 10-bit. Any modern software does this already, if not floating point calc for processing work. (Premiere does this, apple color does this, davinci does this, Scratch does this, etc. bringing in 8-bits in a 10-bit wrapper simply to "do calculation in 10-bit space" is totally null and a waste of time. It's when you're writing out your processed work that has been processed in floating point, that it does benefit to go to a higher bit depth.)

Then to go on to say something like the difference between a full 10-bit source into a 10-bit container is of marginal difference than an 8-bit source into a 10-bit container. For a lack of a better word, that is just very uninformed. As I said before, yes, when you record an 8-bit space into a 10-bit space, it is padded, with zeroes in the header. No new extra magical information that was never there to begin with, no "stretching", it's the exact same information, padded to fit in a 0-1024 space. I can provide several examples of moderate grades that hold up just fine and dandy on 10-bit source footage, but quickly fall apart due to banding on an 8-bit source, even though that source has been padded to a 10-bit wrapper. What we all need to remember is that an 8-bit signal in 10-bit space is identical to that 8-bit signal in an 8-bit space. the header information has simply been padded so it is stored correctly. When you load a padded file into something like avisynth that analyzes the sources encode values, it will actually come back and tell you it is 8-bit, as that's the level of steps in signal it finds.



I just want to remember you guys, that HDMI 10 bit is not really 10 bit.
And no, it´s not just zeros orl empty information - bit 9 is an "optimizing bit" and bit 10 "DC Balancing".


...what? you are talking about ancillary data that has nothing to do with video signals, or SDI at all. Also, "optimizing bit" is a made up term that actually does not exist. Also DC balancing, or 8b/10b encoding, which is what I think you meant to say, is a telecommunications method for transferring 8-bit data via a 10-bit medium in the most efficient manner, retaining the ability to modify DC balance so it can be transmitted through a serial data medium with changing voltage, such as something using automatic gain control. Once again, having absolutely nothing to do with the 0-1024 scale the chroma and luma information is stored on. Your talking about ancillary data.

I'm totally open to loading a padded file up in avisynth sometime this week, stripping the header information, and showing you guys how luma and chroma is padded and stored. It's identical to the 8-bit source. I'm also fine with the idea of posting comparison screenshots of how a padded signal holds up to a moderate grade versus a full 10-bit origination. I'm all for staying on topic, but a part of me simply cannot let misinformation like this be passed around to influence these peoples buying decisions. That's a pretty questionable practice at best.

Barry_Green
09-06-2011, 09:32 PM
I'd love to see some side-by-side comparison - actual evidence of the same shot under identical conditions with the same lenses (!!) - of a highlights roll-off comparison between these two cameras. I can't find one on the web. Barry has done a fantastic job of giving us some great, hard tests - there are a lot of opinions and anecdotes floating around, but nothing is a substitute for real tests. Anyone want to shoot them?

I plan on getting to it. The schedule is preposterous right now, but when it opens up I plan on trying to get that shot.

ian_h
09-06-2011, 10:38 PM
I plan on getting to it. The schedule is preposterous right now, but when it opens up I plan on trying to get that shot.

Awesome - look forward to it!

Regarding the bit debates, I come from an audio background, and I remember when 24-bit became the all the rage in reasonably priced recording equipment. Digital recordings prior to that point had been done in 16-bit and it was clearly demonstrated by a number of reviewers that the recordings produced by older, high quality 16-bit recorders far exceeded the quality produced by many of the new-fangled (and cheaper) 24-bit recorders.

The numbers aren't everything. The AF and FS are aimed at the same people (us!) who were shooting on XH-A1s and HVXs and DVXs not very long ago, and I'm sure I'm not alone in being exceeding delighted at the absolutely farking incredible advances that this technology has made in recent years. For independent feature production (which I guess is the dream many of us are chasing) the 8-bit-ness, highlights roll-off, less-than-11-stops and noise levels of either of these cameras are going to be pretty damn close to LAST in the long list of problems and limitations one has to deal with.

Just in case anyone is getting the wrong idea here, these are incredible cameras for the money, limitations notwithstanding, and I for one feel delighted to be part of this revolution!

Barry_Green
09-06-2011, 10:59 PM
The numbers aren't everything.
Way true.


The AF and FS are aimed at the same people (us!) who were shooting on XH-A1s and HVXs and DVXs not very long ago, and I'm sure I'm not alone in being exceeding delighted at the absolutely farking incredible advances that this technology has made in recent years. For independent feature production (which I guess is the dream many of us are chasing) the 8-bit-ness, highlights roll-off, less-than-11-stops and noise levels of either of these cameras are going to be pretty damn close to LAST in the long list of problems and limitations one has to deal with.
Yep. You totally get it. We long ago passed the threshold of "good enough" where someone can make a high quality movie with the products that are available today.

alaskacameradude
09-06-2011, 11:27 PM
Way true.


Yep. You totally get it. We long ago passed the threshold of "good enough" where someone can make a high quality movie with the products that are available today.

Yup, it is unfreakingbelievable that you can get a camera this good, for this little money. I have never owned anything that puts out an image like this.
You can't go wrong with either one in my opinion.....

maarek
09-07-2011, 05:34 AM
Sorry but you are incorrect in thinking this. You have gained 4;2:2 and the full 10 bit on the recorder. Take a search around here on DVXuser and on the internet where the tests have been done and it is true.

I am truly sorry but that is just wrong. As someone who works at a post house and who does stuff with real 10-bit material I can safely say that: An 8-bit signal WILL NOT get any bit better by recording it into a 10-bit codec.

I think it's wrong to confuse people like that.

maarek
09-07-2011, 06:03 AM
But even if you have SDI and real 10 bit - do you have a 10 bit monitor and graphic card?


The point of 10bit is that you have more latitude at post. If you have 8-bit padded to 10-bit, it's no use. There is no additional information.

Jan_Crittenden
09-07-2011, 07:59 AM
What I meant is, the fact that you would not only state that "recording 8-bit to a 10-bit file allows more massaging in post", but also state that "the difference between 10-bits recorded to a 10-bit file is marginal" - is a sign of a lack of understanding in the realm of digital cinema, to say the least.

I am aware of the numbers and the fact that there is more to the signal in 10 bit. 1024 is greater than 256. If I said it too casually, then I am guilty. But I do know that it can make a difference for the 8 bit signal to record to an external recorder, as the color depth is increased and there is going to be less compression which is better than more compression. So you maybe the words became too slim in the full description of what I was advocating. I am not just talking about going from 8 bit to 10 bit but also going from 4:2:0 to to 4:2:2 and from 24 Mbs to 200Mbs. All of this can help in the overall signal survival.


To say these things clearly indicates you don't pipeline any kind of footage other than something straight off camera into a final cut system.

Actually the experience that I reference is not computer based at all, but primarily from machine to machine. 8 bit recordings being dumped to 10 bit recordings, manipulated and then dumped back for final comparison. And in rereading this part of the thread, I can see where if you just took the words, you would find that what I was saying would be misleading.


One, recording 8-bits into a 10-bit file does not allow more massaging in post. It allows the exact same amount, as the recording contains the exact 8-bit information off camera with yes, zeroes to pad the stream. Loading in a 10-bit file does not make the software automatically jump to working in 10-bit. Any modern software does this already, if not floating point calc for processing work. (Premiere does this, apple color does this, davinci does this, Scratch does this, etc. bringing in 8-bits in a 10-bit wrapper simply to "do calculation in 10-bit space" is totally null and a waste of time. It's when you're writing out your processed work that has been processed in floating point, that it does benefit to go to a higher bit depth.)

I do think that we did cover this point in the fact that the statements were made several times that it is an 8 bit signal padded into a 10 bit stream and your last statement actually tells me why we would do it. And as far as each of those pieces of software that you mention, they do it automatically but they also do it differently, much like the differences that one would find in the AVCHD codec implementation done in the AF100 vs the FS100, they are both AVCHD, both are good, but they are implemented differently.

While I can appreciate the fact that the bits do not increase in number they do not become worse through the concatenation that would typically happen when going from 8 bit to 8 bit, codec to different codec. Additionally there does seem to be an increase in the stability of the signal, and this may be attributed primarily to the 4:2:2 and less compression. We often do this when analyzing lesser codecs and comparing, you take them to a higher level codec so the footprint is not there of the higher one and the lower one doesn't degrade. It does have its merit.


I'm totally open to loading a padded file up in avisynth sometime this week, stripping the header information, and showing you guys how luma and chroma is padded and stored. It's identical to the 8-bit source. I'm also fine with the idea of posting comparison screenshots of how a padded signal holds up to a moderate grade versus a full 10-bit origination. I'm all for staying on topic, but a part of me simply cannot let misinformation like this be passed around to influence these peoples buying decisions. That's a pretty questionable practice at best.

When and if you do this, would you please do that in another thread? It might be educational, but might be lost in this thread. And it would be educational to show it from an AVCHD recording into 10 bit sample, uncompressed into 200 MBs and uncompressed 8 bit to uncompressed 10 bit. I think that recording to the 10 bit domain does help an 8 bit signal, but I may have been making the attribution incorrectly and weighting it too heavily on the bits and not enough on the increased color depth and less compression.

Would I put 8 bit up against 10 bit in a deep level of compositing, say 8-10 layers, no, but I have seen it hold for 3-4 fairly easily. Peoples' buying practices will be more easily persuaded by what they can afford and what works for them. If they really need 10 bit they will spend the extra money for the camera that has a 10 bit output and nothing we say here will deter them from doing so. If they cannot afford that, the next step back is one of the $5000 cameras and record to 10 bit external recorder which does have its benefit. And if they don't have the bucks for that they will just buy the $5000 record to 8 bit and edit in 10 bit I-frame only. The puzzle about image quality is not just the bit depth, it is a combination of a number of things.

Helping people understand where the alternatives are in the purchasing tree is not a bad thing. Life is all about little trade-offs. Yes 10 bit to 10 bit is better than 8 bit to 10 bit and the 8 bit recording dumped into 10 bit can also be very good but it too is not as good. It is all about choices and what people can afford.
I don't think that any of the people engaged in this conversation, including myself, were saying that you get the same thing as the 10 bit output camera with the 8 bit output camera. (Although in reading back through the conversation I can see where the heated conversation doesn't fully express everything for a clear understanding, but that is the nature of the internet and forums in general, it isn't a white paper, it is a conversation in print, not all avenues are covered.) The 10 bit output camera costs a whole lot more than the 8 bit output camera by about 3X the price tag.

The primary purpose of any of this discussion is so that people can feel comfortable with what they can afford and get onto creating stories. Some people would allow the fact that they can't afford the $20,000 camera get in the way of their creativity. They shouldn't. They should be happy with what they can afford at that level and move on to telling stories.

Best,

Jan

Jan_Crittenden
09-07-2011, 08:01 AM
could we possibly get back to the comparison? I would love to know what workarounds exist for highlight troubles on the fs as well as the af based on this test, and chet like that...

So the way to deal with the highlight issues on either camera is first to know that it can happen and then control your exposure.

Best,

Jan

alaskacameradude
09-07-2011, 09:19 AM
So the way to deal with the highlight issues on either camera is first to know that it can happen and then control your exposure.

Best,

Jan

This is the approach I take as much as possible....expose for the highlights and let the rest of the image fall where it may. Now obviously there are exceptions,
but I was surprised what I could 'recover' in post, even with the 4:2:0 AVCHD recording.


much like the differences that one would find in the AVCHD codec implementation done in the AF100 vs the FS100, they are both AVCHD, both are good, but they are implemented differently.

Wow, I wasn't aware of this, although I guess it makes sense. I just assumed AVCHD was pretty much the same (as long as the bit rate was the same). Either way, it seems
like a really good codec, better than I would have thought.

Rick Burnett
09-07-2011, 11:08 AM
The AF100 outputs 8-bit, whether that is in a 10-bit container or an 8-bit container doesn't really matter, and touting that the AF100 uses 10-bit is misleading because when people ask about the bit depth, they are referring to the image bit depth, not the container. Frankly, it's a waste of space for recording purposes. Those empty bits do NOTHING for the recorded image. You gain NOTHING for recording those padded bits from a video capture standpoint, period.

In post, it depends on your editing software and what you use as your intermediate storage format. Internally, the image data in many hosts is converted to floating point anyways. If your internal rendering format is a higher bit depth, then having a higher starting bit depth is doing NOTHING for you. I tested this with Premiere Pro CS5. I brought in files that were native H.264 AVCHD and also files converted to ProRes 422. I pushed both pieces of footage by the same plugins, by the same amount, brought them back in and subtracted them. I did not see a difference in the negative aspects of what I saw. Noise, banding, etc all had the same intensity from being pushed around.

Now, if your editing software doesn't work like this, and using the 10-bit output could save the step of transcoding to a deeper format, then yes, you could save time, but you aren't getting any extra quality that you can't get from just moving your footage to a different format before editing. Further, you are eating up more space in your recording for no additional gain. 6 bits per pixel (RGB), by 1920x1080, by 24fps. with no compression, that is 37Mbytes of wasted information per second if it was 4:4:4 I believe. So if we assume 4 bits (which would be 4:2:2 I believe) that is still 25Mbytes per second of zeros you are recording. Which I would personally consider a negative benefit to 8-bit in a 10-bit container.

The other thing I think people argue for and against constantly is that two cameras can capture the same scene in the same way. I think in a huge range of circumstances that both the AF100 and FS100 can in fact capture the same quality image. I've done it. Further, how perfect the conditions are in that realm where they both work well is in deed a true statement.

But there are conditions outside those where both cameras do not perform the same, and as others have pointed out, if you do work in those situations, then you need to be aware of those differences. If you shoot a lot of low light footage, the AF100 does not perform as well in these settings. I've shot the footage myself, both cameras, together, and measured the results. The problem is, noise is not a quantifiable parameter in these measurements because it's how the noise looks, not necessarily how much of it there is. In every case, I found the noise of the FS100 to be denser and more random, less splotchy, and easier to work with in post. The AF100 noise is less forgiving. Pushing footage that was underexposed by my mistake too far led to unsatisfactory results, like I've said before. I recreated the tests, and made sure I understood what was happening. Sure, as people are so happy to point out that exposing correctly is the solution, but mistakes happen, and I'll be the first to admit, having a camera that can give me some more freedom for when I do make mistakes, is a positive for me. Also, artistic differences cause scenes that were exposed for a certain artistic reason to require changes in post, and having more freedom for that is a good thing (which is why I understand the desire for 10-bit). It's about freedom as well. I think part of using the AF100 is knowing this limitation. Do no expect to push into the shadows in post, very far, and you'll get excellent footage.

That said, when used in a more reasonable range, both cameras perform almost the same, it's knowing where they don't perform the same which is how you decide which is right for you. They each have such great features that you can easily point to many things on each that are a must. I can't tell you how much I miss the built in ND of the AF100. Or the waveform which is SO MUCH better than the histogram on the FS100. Yet, the focus zoom on the FS100 is awesome.

Why can't I just have an AFS100 :)

Ed Kishel
09-07-2011, 12:09 PM
reading this thread, I have realized my knowledge of this stuff is a drop in the bucket of knowledge some of you folks have- amazing.

These "this cam vs that cam" arguments are like a train wreck- hurts to watch but you can't look away if you own one of these cameras. Putting aside roll-off, sensitivity, pixels, bits, sensor size, codecs... sometimes just looking at the screen and knowing what looks good is all that matters. If it means I'm not a pro because I don't scrub the screen with my eyeballs looking for a blown yellow highlight- so be it. I guess I have too much audience member left in me and not enough DIT.

alaskacameradude
09-07-2011, 12:47 PM
The AF100 outputs 8-bit, whether that is in a 10-bit container or an 8-bit container doesn't really matter, and touting that the AF100 uses 10-bit is misleading because when people ask about the bit depth, they are referring to the image bit depth, not the container. Frankly, it's a waste of space for recording purposes. Those empty bits do NOTHING for the recorded image. You gain NOTHING for recording those padded bits from a video capture standpoint, period.

In post, it depends on your editing software and what you use as your intermediate storage format. Internally, the image data in many hosts is converted to floating point anyways. If your internal rendering format is a higher bit depth, then having a higher starting bit depth is doing NOTHING for you. I tested this with Premiere Pro CS5. I brought in files that were native H.264 AVCHD and also files converted to ProRes 422. I pushed both pieces of footage by the same plugins, by the same amount, brought them back in and subtracted them. I did not see a difference in the negative aspects of what I saw. Noise, banding, etc all had the same intensity from being pushed around.

Now, if your editing software doesn't work like this, and using the 10-bit output could save the step of transcoding to a deeper format, then yes, you could save time, but you aren't getting any extra quality that you can't get from just moving your footage to a different format before editing. Further, you are eating up more space in your recording for no additional gain. 6 bits per pixel (RGB), by 1920x1080, by 24fps. with no compression, that is 37Mbytes of wasted information per second if it was 4:4:4 I believe. So if we assume 4 bits (which would be 4:2:2 I believe) that is still 25Mbytes per second of zeros you are recording. Which I would personally consider a negative benefit to 8-bit in a 10-bit container.

The other thing I think people argue for and against constantly is that two cameras can capture the same scene in the same way. I think in a huge range of circumstances that both the AF100 and FS100 can in fact capture the same quality image. I've done it. Further, how perfect the conditions are in that realm where they both work well is in deed a true statement.

But there are conditions outside those where both cameras do not perform the same, and as others have pointed out, if you do work in those situations, then you need to be aware of those differences. If you shoot a lot of low light footage, the AF100 does not perform as well in these settings. I've shot the footage myself, both cameras, together, and measured the results. The problem is, noise is not a quantifiable parameter in these measurements because it's how the noise looks, not necessarily how much of it there is. In every case, I found the noise of the FS100 to be denser and more random, less splotchy, and easier to work with in post. The AF100 noise is less forgiving. Pushing footage that was underexposed by my mistake too far led to unsatisfactory results, like I've said before. I recreated the tests, and made sure I understood what was happening. Sure, as people are so happy to point out that exposing correctly is the solution, but mistakes happen, and I'll be the first to admit, having a camera that can give me some more freedom for when I do make mistakes, is a positive for me. Also, artistic differences cause scenes that were exposed for a certain artistic reason to require changes in post, and having more freedom for that is a good thing (which is why I understand the desire for 10-bit). It's about freedom as well. I think part of using the AF100 is knowing this limitation. Do no expect to push into the shadows in post, very far, and you'll get excellent footage.

That said, when used in a more reasonable range, both cameras perform almost the same, it's knowing where they don't perform the same which is how you decide which is right for you. They each have such great features that you can easily point to many things on each that are a must. I can't tell you how much I miss the built in ND of the AF100. Or the waveform which is SO MUCH better than the histogram on the FS100. Yet, the focus zoom on the FS100 is awesome.

Why can't I just have an AFS100 :)

Great post!

Steve Kahn
09-07-2011, 04:30 PM
In post, it depends on your editing software and what you use as your intermediate storage format. Internally, the image data in many hosts is converted to floating point anyways. If your internal rendering format is a higher bit depth, then having a higher starting bit depth is doing NOTHING for you. I tested this with Premiere Pro CS5. I brought in files that were native H.264 AVCHD and also files converted to ProRes 422. I pushed both pieces of footage by the same plugins, by the same amount, brought them back in and subtracted them. I did not see a difference in the negative aspects of what I saw. Noise, banding, etc all had the same intensity from being pushed around.

This is exactly what I though was the case and why people, including me, were cutting in AVCHD without transcoding first. Thanks for doing this test. It confirms my intuition. I also did a similar test in photoshop with 8bit VS 16bit and saw no benefit.

Rick Burnett
09-07-2011, 05:15 PM
It DEFINITELY depends on the host software, so make sure, but on Premiere I tested this. I also edit bringing in the AVCHD directly because of it in both Premiere and After Effects. Now, depending on how much CPU you want to use, that also makes a difference. Different compression (or lack there of) will affect your performance. When working in After Effects, I was NOT impressed by the performance of DNxHD files. I thought it would work better, but it didn't. Some memory/bottleneck performance threshold was an issue. I transcoded to ProRes422 on those files and the performance of After Effects was INSTANTLY better for me. Now, that doesn't mean it is a codec problem, it could very well be an After Effects problem, but it was easily producible for me.

What you need to know is what your software does internally with values. If you pick the intermediate native format that you are editing in, and it is NOT a floating point representation, I'd be DEFINITELY working with a codec that has higher quality than AVCHD. (And more than 8-bit) as you want to reduce your level of quantization errors. So this isn't a golden rule per say, but an application specific question.

Jan_Crittenden
09-07-2011, 05:18 PM
Hi,

I have not have the same experience, but then again, I have seen a difference in the transcoded footage and how it got to be pushed around. It was why I asked you Steve why you were having issues with your footage, I have not seen that on the footage that the consultant's(the guys that shoot for me and own AF100s) footage, but they all do the transcode to ProRes. The point of the earlier discussion was not just 8-10 bit, but also 4:2:2 uncompressed to a higher bit rate recorder. This too will increase the quality of the footage.

Best regards,

Jan

Rick Burnett
09-07-2011, 06:47 PM
I think the biggest problem is transcoded footage is not a quantifiable procedure for the same reason MB/s on video transmission rate doesn't indicate the quality of the material being transmitted, and people want it to be as easy as that, but it isn't. In any field, where quality is concerned, people do try to put a quantifiable measurement on their least accepted quality, but really, when those stakes are placed in the ground, technology continues to get better and many times, those requirements are based on the past and performances that are not as relevant anymore to "modern" technology. However, with all things technology based, there will still be a HUGE infrastructure still using older technology that those requirements will still need to be headed. Therefore, you plan for the worst possible scenario to make sure ALL footage or audio you receive is quality, regardless of where your source comes from. It makes the process easier for acceptance.

With codecs and bit rates settings, and programs performing those tasks, there are a HUGE number of variables to what is happening to your footage. Especially with adaptive codecs, anything that is lossy is going to be losing information, and how that information is lost is something you are going to have to test and see if you can live with what you are losing.

Many codec/applications will change gamma, which will make you think noise has increased, or exposure has changed, but in truth, the software is modifying the data, not just transcoding it. Further, different codecs have different colorspaces and how well those translations happen are important.

Even further, some codec manufacturers I have seen claim they can re-encode color information that is better than the original file. Of course, I don't believe this myself. Not that it might not have a pleasing effect in many instances, but nevertheless, it is guessing on creating data that is not there anymore and many times, those sorts of algorithms have limitations or situation where artifacting occurs. I've seen this with very complex fractal resizing algorithms that sharpen an image based on wavlet technology.

The best thing you can do is test your setup with multiple types of transcoded input files, outputs, etc. I often use the layer composition tools in AE to subtract video from one another then gain up the result to see what is happening between processes. It is very interesting sometimes. Every time you ask, you are going to get a TON of different procedures for how people work. If you haven't tried some of them, try them. You might find something better then what you are doing, you might find something worse. I test these things VERY OFTEN between projects and typically learn A LOT from others using the same camera as me.

I just want to reiterate, a simple change in gamma from 1.8 to 2.2 can make you think a codec is doing something hugely different, and it isn't. Gamma shifts on OSX platform have been a huge source of misconception for a long time, and many people aren't even aware of when FCP or QuickTime do this. It's really annoying. AND, it also depends on a flag in the ProRes files as well as to how QuickTime handles this issue. For all the things I love about my OSX machines, this is something that needs some serious fixing, which Apple never does.

maarek
09-08-2011, 12:52 AM
I was gonna write a long reply but Rick got there first with the exact same words... dammit!

Jan_Crittenden
09-08-2011, 08:52 AM
The AF100 outputs 8-bit, whether that is in a 10-bit container or an 8-bit container doesn't really matter, and touting that the AF100 uses 10-bit is misleading because when people ask about the bit depth, they are referring to the image bit depth, not the container. Frankly, it's a waste of space for recording purposes. Those empty bits do NOTHING for the recorded image. You gain NOTHING for recording those padded bits from a video capture standpoint, period.

First I never said it had 10 bits. But if you read everything I said it isn't just the 8-10bits it is also about less compression. 24mbs vs. 200mbs, i-Frame only and 4:2:2 on the external recorder. There is a difference.



Even further, some codec manufacturers I have seen claim they can re-encode color information that is better than the original file. Of course, I don't believe this myself. Not that it might not have a pleasing effect in many instances, but nevertheless, it is guessing on creating data that is not there anymore and many times, those sorts of algorithms have limitations or situation where artifacting occurs. I've seen this with very complex fractal resizing algorithms that sharpen an image based on wavlet technology.


So you are saying that Cineform, ProRes and the Avid DNX intermediate codecs are all a waste of time and nothing is gained. You might want to take up that argument with them as they build a lot of reputation on just that gain. Intermediate codecs are the perfect platform for editing long GOP recordings as the first thing that happens is that they are changed into I-Frame only. I am not going to address which does what or how well as far as handling the signal but I will say that I have seen favorable results out of each and every one of these, so my experience runs a little contrary to yours.


I just want to reiterate, a simple change in gamma from 1.8 to 2.2 can make you think a codec is doing something hugely different, and it isn't. Gamma shifts on OSX platform have been a huge source of misconception for a long time, and many people aren't even aware of when FCP or QuickTime do this. It's really annoying. AND, it also depends on a flag in the ProRes files as well as to how QuickTime handles this issue. For all the things I love about my OSX machines, this is something that needs some serious fixing, which Apple never does.

Of course a change in gamma is going to make a huge difference in the perception of a clip, but note this is not what I am talking about at all.

Best regards,

Jan

alaskacameradude
09-08-2011, 09:19 AM
First I never said it had 10 bits. But if you read everything I said it isn't just the 8-10bits it is also about less compression. 24mbs vs. 200mbs, i-Frame only and 4:2:2 on the external recorder. There is a difference.



So you are saying that Cineform, ProRes and the Avid DNX intermediate codecs are all a waste of time and nothing is gained. You might want to take up that argument with them as they build a lot of reputation on just that gain. Intermediate codecs are the perfect platform for editing long GOP recordings as the first thing that happens is that they are changed into I-Frame only. I am not going to address which does what or how well as far as handling the signal but I will say that I have seen favorable results out of each and every one of these, so my experience runs a little contrary to yours.



Of course a change in gamma is going to make a huge difference in the perception of a clip, but note this is not what I am talking about at all.

Best regards,

Jan

I don't think Cineform, ProRes or DNX claim that they can 'improve' video that has already been compressed. You can't really 'add' information....well you
can, but it is always 'guessing' how to add info which is not necessarily a good thing. Also, I think every one of those codecs is a 'lossy' codec in
real life, (although they all claim to be 'visually lossless'). What is gained from them, is an edit friendly format that a computer can deal with
easily, without any 'visual loss' in quality, but I have never seen a 'net gain' in quality. With the rise of long GOP codecs, which can be very good
for 'aquisition' as the pack a LOT of really high quality looking imagery, into a small amount of space, there was a market for a 'edit friendly'
codec (or three) to transcode these 'aquisition' codecs into, so that a computer could deal with more layers of video, with effect applied and so on,
without 'choking up'. That is my understanding of the matter.

Not to put words in Rick's mouth, but I think he is saying, in a like manner, once you have a 8 bit video signal, you can't 'improve' it by
padding it out to 10 bit. There might be advantages in other ways, but you aren't going to improve the base quality of the video.
However, as you say, you CAN improve the quality by recording 4:2:2 to an external recorder....you just can't make it 10 bit.
However, honestly the AVCHD codec is really good. I know that Doug Jensen who is kind of Barry's counterpart on the Sony side
(as he does training books and DVD's for Sony cameras) has basically said that you should save your money and not buy an external
recorder for the FS100.
http://www.b-roll.net/forum/showthread.php?t=27503

Rick Burnett
09-08-2011, 09:26 AM
First I never said it had 10 bits. But if you read everything I said it isn't just the 8-10bits it is also about less compression. 24mbs vs. 200mbs, i-Frame only and 4:2:2 on the external recorder. There is a difference.

This is what you said:


Actually it would be a 10 bit recording. And you would be able to push the signal around much like you would the 10 bit output.

And given this thread is about the differences between the FS100 and AF100, there is no gain you are getting from padding those 2 bits out of the AF100 that will be an advantage over the FS100. It's 8-bit. Period. Talking about the bit depth of the pixels is a different issue then talking about the data rate of the compression. You are wasting disk space on the recording medium for no gain in the recorded signal with regards to bit depth. The AF100 is still taking a larger bit depth and mapping it into 8-bit with a curve. You might save time depending on your post flow, with regards to the internal media of the AF100, but this doesn't mean that you cannot recognize the same advantage on the FS100 when recording externally to something like ProRes 422 as well off the HDMI. The bits will just have to be padded on the record side since ProRes is 10-bit.

And yes, I agree that the recorded signal coming out is going to be better in many ways since you are avoiding the compression, but that's a different issue. Both cameras gain from this.

But let's take it a step further. The FS100 can output 4:4:4 8-bit. So you could remove even more subsampling of the color (although given the source is bayer derived, I don't really think 4:4:4 is buying you anything over 4:2:2 other than more lost disk space), but it would be at least something different then just padding.


So you are saying that Cineform, ProRes and the Avid DNX intermediate codecs are all a waste of time and nothing is gained. You might want to take up that argument with them as they build a lot of reputation on just that gain. Intermediate codecs are the perfect platform for editing long GOP recordings as the first thing that happens is that they are changed into I-Frame only. I am not going to address which does what or how well as far as handling the signal but I will say that I have seen favorable results out of each and every one of these, so my experience runs a little contrary to yours.

The key word here is *intermediate* as it depends on what and where the footage is being used. I see people throw this around all over the place with regards to *having* to transcode first before using footage off the camera in something like Premiere as the quality will be different and it is wrong. Like I said before, it depends on the application you are using. I've heard this argument over and over, and every time I test it to see if there is an appreciable difference on heavy grading, I don't see a difference. If one bands, the other bands, just the same.

Where these intermediate codecs are most important is when you switch from ANY program that is altering the data from the camera. So if you go from Premiere to After Effects, you NEED to use an intermediate codec or you are going to degrade the footage. In these cases, after the video has been touched inside Premiere or After Effects, I use ProRes 422 myself as the data gets passed around with the other people I've worked with. Or, if you need to use a codec that is less harsh on CPU/Memory Speed/Drive Speed. There is no quality gain from this, but it can make older hardware still able to edit multi-layer source material.

In fact, I had one client who sent me DNxHD files to send me the raw files off the camera so I could compare as well. I ran BOTH files through some HEAVY effects in After Effects. The reason I did this is the DNxHD files were performing poorly from a computer performance standpoint (I suspect RAM in AE). I had 30+ layers in AE effecting the video. I took video from both and compared the results, cranking up the gamma as well so I could see artifacting easier. The results were the same.

Had they actually applied edits to the DNxHD versions, yes, I believe they'd have been necessary to not cause quantization into a capture format, but they had not, so I continued with the H.264 based files and the results were great.


Of course a change in gamma is going to make a huge difference in the perception of a clip, but note this is not what I am talking about at all.

Best regards,

Jan

This was not with regards to anything you said but in general. Lot's of people compare footage between different codecs and are not aware of some codecs that change the gamma depending on information in the original footage. QuickTime and FCP *try* to be smart about this and do what they think is right, but I don't see this happen. I've seen QT increase the gamma and pop the noise out higher with AF100 footage and someone not knowing this might think that QT was actually adding the noise in. When in fact, it isn't, it's just changing the gamma. Put the gamma back right and the footage looks the same. If you will recall, I spent a lot of time with the AVCHD files out of my AF100 and with ClipWrap as ProRes files that were getting created at the time were getting incorrectly interpreted by QT and FCP (which I believe they have fixed).

Jan_Crittenden
09-08-2011, 09:40 AM
I don't think Cineform, ProRes or DNX claim that they can 'improve' video that has already been compressed. You can't really 'add' information....well you can, but it is always 'guessing' how to add info which is not necessarily a good thing. Also, I think every one of those codecs is a 'lossy' codec in
real life, (although they all claim to be 'visually lossless'). What is gained from them, is an edit friendly format that a computer can deal with
easily, without any 'visual loss' in quality, but I have never seen a 'net gain' in quality. With the rise of long GOP codecs, which can be very good
for 'aquisition' as the pack a LOT of really high quality looking imagery, into a small amount of space, there was a market for a 'edit friendly'
codec (or three) to transcode these 'aquisition' codecs into, so that a computer could deal with more layers of video, with effect applied and so on,
without 'choking up'. That is my understanding of the matter.

Hi, Again I don't wish to start another argument here, but I have seen improvement even on a DVCPRO HD Signal on the Cineform Codec, they might still have this on their website, but to say that they cannot improve the signal is a way too general of a statement and frankly I have seen differently. There are all of people way smarter than me working on these things and they do it because it can make a difference and that difference is marketable. You guys can have your opinion but I too have the right to have an opinion as our opinions are based on our own experiences. My experiences are much different than yours but no less valid.


I know that Doug Jensen who is kind of Barry's counterpart on the Sony side
(as he does training books and DVD's for Sony cameras) has basically said that you should save your money and not buy an external
recorder for the FS100.

And yes there can be improvement by adding an external recorder as long as the recorder can record at a higher bit rate, 4:2:2 and I-Frame. You can look back at my early posts as the AF100 was coming out, and I too had the same opinion as the Sony guy you reference, and was one of the last to be convinced. However I have since seen the difference and there is a difference in a positive way. Taking the 4;2:2 8 Bit HDSDI output and recording to a 10 bit, 4:2:2, 200mbs, I-Frame only codec can and does make a difference.

Best regards,

Jan

Jan_Crittenden
09-08-2011, 09:44 AM
Rick that quote is out of context in that what I was referencing as the recording was on an External recorder. Not on the camera. So it is an 8 bit padded signal on the recorder but it is i-frame, 4:2:2 with less compression and it is better. And in either camera it should make a difference, I think I said that back there as well.

Best,

Jan

alaskacameradude
09-08-2011, 10:12 AM
Hi, Again I don't wish to start another argument here, but I have seen improvement even on a DVCPRO HD Signal on the Cineform Codec, they might still have this on their website, but to say that they cannot improve the signal is a way too general of a statement and frankly I have seen differently. There are all of people way smarter than me working on these things and they do it because it can make a difference and that difference is marketable. You guys can have your opinion but I too have the right to have an opinion as our opinions are based on our own experiences. My experiences are much different than yours but no less valid.


Well, manufacturers can CLAIM anything they want. I call it 'marketing' :-) I have to agree with Rick on this one. I've tested it with a PBS engineer....both
DVCPro HD, XDCam HD, and the AVCHD from my FS 100. We could find no 'increase' in quality between the 'camera original' and the 'transcoded version'.



And yes there can be improvement by adding an external recorder as long as the recorder can record at a higher bit rate, 4:2:2 and I-Frame. You can look back at my early posts as the AF100 was coming out, and I too had the same opinion as the Sony guy you reference, and was one of the last to be convinced. However I have since seen the difference and there is a difference in a positive way. Taking the 4;2:2 8 Bit HDSDI output and recording to a 10 bit, 4:2:2, 200mbs, I-Frame only codec can and does make a difference.

Best regards,

Jan

I am not arguing that there is no quality increase from AVCHD to recording 4:2:2 to an external recorder. I am sure there is. I just think it is a small increase, UNLESS you are
doing greenscreen or something similar. For just general shooting, without heavy grading, it's not going to look hugely different. Many people are put off by the relatively
'low' bit rate of these cameras until they see how good it actually looks.

Rick Burnett
09-08-2011, 10:18 AM
I read the entire thread, and the context wasn't taken away from what you wrote. Your response was confusing people into thinking some quality advantage was being gained from the 10-bit padded versus the 8-bit, and it isn't. The quality comes from other aspects of the higher bit rate recording and the bypassing of the compression. Yes, the external recording is better, but not because of being 10-bit padded.

And pertaining to codecs and quality, I see company marketing claim astounding performance all the time with their codecs. Like I said earlier, yes, there can be things that can happen differently in their codec, but even if they claim to make footage better, they are creating something from nothing as once the data is lost, it is lost, and their method for trying to recreate that material could have adverse reactions to certain situations. It's something that needs to be tested. There are companies that claim that QT converts to ProRes poorly and adds noise to the footage. Does it? I've not seen it. I've looked, but it has never been an issue when I used it.

The problem is that people don't understand what happens to footage once it is inside the host application. How that footage is handling internally makes ALL the difference. If the application you use for instance converts internally all footage to 32-bit float uncompressed, then once the footage is read in, the input format DOES NOT MATTER for anything visual. What does matter is the method for how that video is converted internally from ANY codec into the internal codec, if you will. So now you are taking something that was H.264 -> Intermediate Codec -> Internal Codec. Once inside, it does not matter anymore. Each layers effect will be working on the internal format. What does matter is if the plugins you are using are high bit depth. Internally, a plugin can do whatever it wants to the video data, including, using lower bit depths. Anyone with deep knowledge of AE knows about how all the plugins are not necessarily 32-bit and can really affect your footage if used poorly (at least in the past it has been an issue).

What I can say is that if I use the footage off of the camera, then I know it is not modified. It's as original as I can get. (Or off of an external recorder even.) I've removed one variable into something modifying my footage, and my tests have shown to me that there is no reason not to work like this. I'm satisfied with how Adobe converts my footage to their internal format.

There are people who use DNxHD over ProRes because they think it is superior. I've tested both, I didn't see any difference to me other than the massive amounts of disk space DNxHD used. But, it could be host dependent.

All that said, I am not telling people to not use an intermediate codec, I am saying don't use it blindly and assume you understand what is going on. You need to know what works best for the application you are using. Adobe themselves indicated that transcoding wasn't necessary with Premiere. I tested their advice, I agree with it. If you are having performance issues, try different codecs in your application to see if you can get good performance with a different one. Some differences in how a codec works may be okay for some, and not for others. Some codecs may behave more correctly and have less issues with some hosts on input.

I will say this as well, I do not like the H.264 encoder in Premiere Pro. I took some really nice AF100 footage I shot and it looked great inside Premiere Pro. No matter what I selected for H.264 output, it banded and had lots of compression noise. I output as ProRes422 and compressed out of FinalCut and it looked great.

Ed Kishel
09-08-2011, 11:19 AM
You may see no difference between raw and intermediate during editing, but do you think codecs like cineform hold up better over 2 or 3 generations vs the raw codec from the camera? If I render back out to cineform and 6 months later want to tweak it more, I think a robust codec would hold up better. Or would it not?

Hidef1080
09-08-2011, 11:21 AM
Thanks Rick for your input.

You always give me something to think about!!

Jan_Crittenden
09-08-2011, 11:21 AM
Okay, I think we are saying the same thing. Your ProRes example is what I am talking about. All of my consultants that are on a MAC platform use ProRes and output to AVC-Intra. The stuff looks fabulous each and every time.

If there was confusion introduced based by how my words were read, I apologize, all I can say is that i wasn't confused. ;-). As I said earlier, this is a conversation and not a white paper. Things go shorthand a lot in conversations. Sorry. Can we move on? Do we still have an audience on this thread or have we bored them all to tears?

Thanks,

Jan

Rick Burnett
09-08-2011, 11:36 AM
You may see no difference between raw and intermediate during editing, but do you think codecs like cineform hold up better over 2 or 3 generations vs the raw codec from the camera? If I render back out to cineform and 6 months later want to tweak it more, I think a robust codec would hold up better. Or would it not?

Absolutely, this is only on input to the application from the camera output. Once you do ANYTHING to that footage at all inside the host application, then you are changing values to the data. It is here where having a larger bit depth (and less or no compression) for what you write out is very important. If you don't do this, then compression and quantization will be throwing out information and you do not want this.

When I leave Premiere Pro or AE, I write out to ProRes422 and leave it in this format, even for archive. At this point, I don't want to introduce any destructive processes until I am writing out the final format for a distribution codec. Many times I will write out ProRes422 and then use that to generate the additional formats in case I am having problems with the codec settings and need to compress multiple times till I am happy.

Rick Burnett
09-08-2011, 12:09 PM
Okay, I think we are saying the same thing. Your ProRes example is what I am talking about. All of my consultants that are on a MAC platform use ProRes and output to AVC-Intra. The stuff looks fabulous each and every time.

If there was confusion introduced based by how my words were read, I apologize, all I can say is that i wasn't confused. ;-). As I said earlier, this is a conversation and not a white paper. Things go shorthand a lot in conversations. Sorry. Can we move on? Do we still have an audience on this thread or have we bored them all to tears?

Thanks,

Jan

I don't used AVC-Intra, but I think it is important noting that when the AF100 came out, if one converted to ProRes with ClipWrap they would have thought something was wrong with the footage when brought into FCP or Premiere Pro. This was a HUGE source of confusion on quality which I did a lot of tests on. From what I understand from Divergent was that there was no colorspace flag in the AVCHD files and they just passed that along in the ProRes file. However, it appeared that QT and FCP were not picking a proper colorspace. They told me they had to talk with Apple to find out what the 'default' colorspace was supposed to be. People were jumping the gun claiming H.264 decoder failure and all kinds of things. At the end, it just ended up being gamma shifts and not necessarily a 'colorspace' from what I could see.

And I don't think people are bored, I think there is a lot to learn in this thread. People take a lot of this stuff for granted and don't really understand every bit of it, or question it. I've seen people convert footage into older lossy codecs because they thought they were good, throwing away tons of information.

But then again, I am not just a cinematographer, I am a technical director as well, and for our projects, it's my job to know this stuff inside and out. Especially when I do VFX work and want to make my job easier. Maybe that's why I find it so interesting! :)

Max Smith
09-08-2011, 02:24 PM
I'm following with rapt attention.

Jan, I was also a bit befuddled by you're reference to 10-bit signals and external recorders. Even if you meant differently it didn't read that way. I'm not picking either, it's great when someone in your position takes the time to interact with the community, I only bring it up to point out that nobody else is picking either. There was some genuine confusion and it's cleared up.

And though I've always gone ProRes422 when working in After Effects because I was told it was a good codec to work in, I've ever known that much about how AE handles things internally. Also didn't realize the varience between plug-ins. That's good information!

Rick Burnett
09-08-2011, 03:42 PM
I'm following with rapt attention.

Jan, I was also a bit befuddled by you're reference to 10-bit signals and external recorders. Even if you meant differently it didn't read that way. I'm not picking either, it's great when someone in your position takes the time to interact with the community, I only bring it up to point out that nobody else is picking either. There was some genuine confusion and it's cleared up.

And though I've always gone ProRes422 when working in After Effects because I was told it was a good codec to work in, I've ever known that much about how AE handles things internally. Also didn't realize the varience between plug-ins. That's good information!

When using the plugin browser, the icons will tell you what their bit depth is. For instance 'Bilateral Blur' is 32-bit but 'CC Vector Blur' is 8-bit.

Max Smith
09-08-2011, 03:46 PM
Thanks! One of those things I probably would have noticed if I payed more attention, or read everything I should. Of course then I'd probably never get anything done :happy:

nyvz
09-08-2011, 09:16 PM
You may see no difference between raw and intermediate during editing, but do you think codecs like cineform hold up better over 2 or 3 generations vs the raw codec from the camera? If I render back out to cineform and 6 months later want to tweak it more, I think a robust codec would hold up better. Or would it not?

I doubt anyone ever suggested it would be a good idea to expect highly compressed acquisition formats to be a good format to re-encode to.

Yes uncompressed or high bitrate codecs designed as intermediates hold up quite well with several generations, but camera original format files without any generational losses will always be best. In this day and age, there are many end-to-end workflow options that allow you to complete post without ever having to render or transcode from the camera originals except for final renders. If you are rendering or transcoding, chances are it's because you are using software or hardware or workflow that is at least a full generation out of date.

Rick Burnett
09-08-2011, 09:24 PM
I doubt anyone ever suggested it would be a good idea to expect highly compressed acquisition formats would be a good format to re-encode to.

Yes uncompressed or high bitrate codecs designed as intermediates hold up quite well with several generations, but camera original format files without any generational losses will always be best. In this day and age, there are many end-to-end workflow options that allow you to complete post without ever having to render or transcode from the camera originals except for final renders. If you are rendering or transcoding, chances are it's because you are using software or hardware or workflow that is at least a full generation out of date.

There are options to do that, but it doesn't mean you are a generation behind, it means you are probably using software from multiple companies and have to pick an intermediate file format if you want to go between them.

My editor works in Final Cut. It really doesn't matter what version because I am in After Effects, and if he has edited a a scene together requiring my VFX work, I want his color matched footage from FCP. I then take that material with outputs from Blender, Motion and Maya and bring it all together in AE. That said, I need to get that footage back. I export out in an intermediate codec and give it back.

And this doesn't even include if you are using separate coloring applications as well.

nyvz
09-08-2011, 10:12 PM
There are options to do that, but it doesn't mean you are a generation behind, it means you are probably using software from multiple companies and have to pick an intermediate file format if you want to go between them.

My editor works in Final Cut. It really doesn't matter what version because I am in After Effects, and if he has edited a a scene together requiring my VFX work, I want his color matched footage from FCP. I then take that material with outputs from Blender, Motion and Maya and bring it all together in AE. That said, I need to get that footage back. I export out in an intermediate codec and give it back.

And this doesn't even include if you are using separate coloring applications as well.

Final cut pro 7 is over two years old and is a generation old :)

I certainly understand your situation, but more what I was suggesting is that if it is of importance to you, it is likely that there is a relatively easy to adopt workflow that will avoid generational losses. In your situation, for example, I'm a bit surprised you're doing preliminary color work in FCP when you are compositing and finishing in AE. It imagine would not be difficult to just instead lock your edit in FCP and import XML or however into AE and have your color done there without ever having to render or re-encode anything. Clearly you may choose not to do that for whatever reason, and I'm sure there are plenty of good reasons. I just meant that for most people, if they allow a little flexibility in their workflow and have relatively up-to-date hardware and software it is often quite feasible to never have to deal with intermediate render or transcode steps that degrade their footage, takes up hard drive space, and takes up CPU time.

maarek
09-09-2011, 01:58 AM
Final cut pro 7 is over two years old and is a generation old :)

And FCP 7 was only a very incremental upgrade from 6. So it is indeed, quite old. I've pretty much switched to Premiere cs5.5 though I still do real work with FCP7 as a lot of other post houses use it. If you do a lot of work with FCP & AE, you would be quite surprised how well Premiere and AE work together. I regularly just copy&paste files straight from the timeline in Premiere to AE. No need to transcode.

Ed Kishel
09-09-2011, 11:01 AM
using cs5.5 btw, so software is not out of date- but there are times when I have been asked to take old final renders and make small changes (and I no longer have the raw footage). A title change or re-arranging a sequence perhaps.

Rick Burnett
09-09-2011, 10:17 PM
I still think it is not so simple of a solution if you are working with multiple people. If my editor was using FCPX there would be no difference in what we do. In fact, we'd not have an XML option at all if we wanted to use it. The limitation is the connectivity between the software.

And pertaining to the levels, absolutely I'd rather have that done in FCP before I get the material. This makes the scenes much more consistent for me, and allows me to have less to worry about trying to get the material I am adding colored correctly for the scene. In addition, the color correction at this point is for consistency only, not any sort of look. Again, to help in compositing more than anything since for me, the look is always the same.

Once I am done with the scenes, they go back to the editor and put back into the entire project, THEN the editor applies the look they want to the scene and/or entire film.

After Effects is THE LAST place I would ever edit an entire film. And definitely THE LAST place I would be trying to do color correction for a film. Given many of the scenes I work on have 30-100 layers, there is NO WAY you could edit an entire project. This is why just individual scenes or groups of scenes are brought into AE, worked on independently and then return to either FCP or (whenever I edit) Premiere Pro. I have no choice. With say 20-30 different scenes requiring this much work, AE would fail in a single project. (Some of my scenes now are pushing the limit and requiring a lot of precomposing).

From a generational standpoint I don't see anything wrong with multiple encodes through high bit rate intermediate codecs. I have run material multiple times through ProRes 422 and compared to the original material and I don't see any issues. The quality of the codec is so much higher than the AVCHD material I am putting into it, I am far from seeing propagated losses. And I have looked.

For a lot of people, before the recent release from Panasonic, you had no choice but to transcode before going into FCP from AVCHD as there wasn't a mechanism for reading AVCHD files directly. I work in AE and PP so my experience was different. I have zero issues with intermediate codecs between programs so I see no reason to change. I see no advantage to changing our workflow because the quality of results really speak for themselves.

Rick Burnett
09-10-2011, 08:45 AM
So is the first video on that comparison from the FS100?

I realized I never gave my guess either. :) The first video looks like FS100 and the second AF100 to me. Where I see this is in the display case in the back. If you look at the noise in that area, the first clip is much smoother. The second clip has splotchiness.

TheReverend
09-10-2011, 09:51 AM
With all this posting, did anyone ever say which shot is which? In the OP?

Steve Kahn
09-10-2011, 11:19 AM
I realized I never gave my guess either. :) The first video looks like FS100 and the second AF100 to me. Where I see this is in the display case in the back. If you look at the noise in that area, the first clip is much smoother. The second clip has splotchiness.

That splotchiness looks like Vimeo compression issues to me. I never have that problem in the blacks with the AF100.

Rick Burnett
09-10-2011, 11:33 AM
It's not the blacks, but dark grays. And it's not vimeo that adds this. I've shot the same scenes with my FS100 and my AF100 when I had it and noticed the same things. For most it will be hardly noticeable, but if you push the grade too far, that's when it starts to pop out. When I was shooting on the AF100 I'd shoot the skin tones a little higher and push down in which really looked nice. In honesty, wish I could define my own curves for both cameras. (and I also downloaded the video and didn't watch it on vimeo)

Steve Kahn
09-10-2011, 11:45 AM
It's not the blacks, but dark grays. And it's not vimeo that adds this. I've shot the same scenes with my FS100 and my AF100 when I had it and noticed the same things. For most it will be hardly noticeable, but if you push the grade too far, that's when it starts to pop out. When I was shooting on the AF100 I'd shoot the skin tones a little higher and push down in which really looked nice. In honesty, wish I could define my own curves for both cameras. (and I also downloaded the video and didn't watch it on vimeo)

So you're saying you can push FS100 footage further than AF in the grade - (im guessing you mean internal codecs of both)?

What IRE are you setting your zebras for skin tone highlights? (and what gamma were you using?)

Rick Burnett
09-10-2011, 07:18 PM
With the FS100 I can indeed pull up the lower end more and have more of a range to do this that is usable. This also translates into using higher gain values that are usable as well. I find the noise in the FS100 to be lower, and, I find it to be a finer more random density and it doesn't have dancing splotchiness to it that I've only really seen on the AF100 when pushing a grade too far. I've also found that 3rd party noise reduction plugins, like Neat Video, do a better job as well given how the noise is distributed with FS100 footage.

Also, this is with the internal codecs as well. In fact, having viewed footage from the AF100 through an external recorder, I find that the AVCHD actually compresses out some of the noise that you'll see with a less compressed image. (I am sure the FS100 probably has the same thing as well, just haven't hooked mine up to a recorder). I've seen this with a lot of H.264 compressors. Of course, you trade less noise for the artifacts of macro-blocking, which also prevents how far you can grade (and definitely affects green screen footage). This is why I really only want an external recorder for green screening. I am very happy with the AVCHD else wise.

I spent A LONG time working with my AF100 and I tuned the settings to what I thought looked best. As I said earlier, I think it is the sharpening algorithm in the AF100 that I dislike the most. When I set the detail to between -5 and -6, the noise was never a problem unless I really pushed the footage. (Specifically a scene I underexposed a bit too much). Before this, I had spent a lot of time with the 7D. The problem I have with just exposing skin tones to a particular IRE is that it really depends on the curve you are using in camera. I spent over a year shooting on the 7D and when I exposed skin at the same level on the AF100, I didn't like where the rest of the image was. There are plenty of threads here where me and others discussed this. I found if I exposed the skin a bit higher then pushed it back down in post, I thought the results were great (and much better than the 7D). I hadn't really thought about this before until I noticed others coming from Canon DSLR and getting results they were not happy with at first. Curves make a huge difference.

With the AF100, I don't mind over exposing a bit if I have a difficult scene to film that might exceed the DR of the camera. It happens. I don't always have control of the lighting and sometimes other peoples schedules dictate this. Further, if I am running like B camera, which I like to do, my lighting is not always of primary importance if it's a difficult scene to light. I've learned with my time with the AF100 that I never want to underexpose. Period. I've battled highlight handling on both cameras by reducing saturation as I think both Sony and Panasonic push that stuff up WAY too much (and especially the red on the FS100). Most HL issues on both cameras can be easily fixed in post.

With the FS100 I don't have to reduce sharpness, so overall, I think the image is sharper because I'm not fighting some strange noise enhancing issue. I was okay with the sharpness at -5 on the AF100, but, I do prefer the sharper image. I really just wish I could pick and choose aspects of both cameras and put it in one. I hated having to pick just one to keep.

bgundu
09-10-2011, 07:37 PM
With the FS100 I can indeed pull up the lower end more and have more of a range to do this that is usable. This also translates into using higher gain values that are usable as well. I find the noise in the FS100 to be lower, and, I find it to be a finer more random density and it doesn't have dancing splotchiness to it that I've only really seen on the AF100 when pushing a grade too far. I've also found that 3rd party noise reduction plugins, like Neat Video, do a better job as well given how the noise is distributed with FS100 footage.

Also, this is with the internal codecs as well. In fact, having viewed footage from the AF100 through an external recorder, I find that the AVCHD actually compresses out some of the noise that you'll see with a less compressed image. (I am sure the FS100 probably has the same thing as well, just haven't hooked mine up to a recorder). I've seen this with a lot of H.264 compressors. Of course, you trade less noise for the artifacts of macro-blocking, which also prevents how far you can grade (and definitely affects green screen footage). This is why I really only want an external recorder for green screening. I am very happy with the AVCHD else wise.

I spent A LONG time working with my AF100 and I tuned the settings to what I thought looked best. As I said earlier, I think it is the sharpening algorithm in the AF100 that I dislike the most. When I set the detail to between -5 and -6, the noise was never a problem unless I really pushed the footage. (Specifically a scene I underexposed a bit too much). Before this, I had spent a lot of time with the 7D. The problem I have with just exposing skin tones to a particular IRE is that it really depends on the curve you are using in camera. I spent over a year shooting on the 7D and when I exposed skin at the same level on the AF100, I didn't like where the rest of the image was. There are plenty of threads here where me and others discussed this. I found if I exposed the skin a bit higher then pushed it back down in post, I thought the results were great (and much better than the 7D). I hadn't really thought about this before until I noticed others coming from Canon DSLR and getting results they were not happy with at first. Curves make a huge difference.

With the AF100, I don't mind over exposing a bit if I have a difficult scene to film that might exceed the DR of the camera. It happens. I don't always have control of the lighting and sometimes other peoples schedules dictate this. Further, if I am running like B camera, which I like to do, my lighting is not always of primary importance if it's a difficult scene to light. I've learned with my time with the AF100 that I never want to underexpose. Period. I've battled highlight handling on both cameras by reducing saturation as I think both Sony and Panasonic push that stuff up WAY too much (and especially the red on the FS100). Most HL issues on both cameras can be easily fixed in post.

With the FS100 I don't have to reduce sharpness, so overall, I think the image is sharper because I'm not fighting some strange noise enhancing issue. I was okay with the sharpness at -5 on the AF100, but, I do prefer the sharper image. I really just wish I could pick and choose aspects of both cameras and put it in one. I hated having to pick just one to keep.

Hey Rick, I too wasn't too keen on the sharpening on the Af100 until I read this blog. A +2 Coring over the detail setting makes sense to retaining the sharpness outside of the dark zones.

http://marvelsfilm.wordpress.com/2011/08/08/panny-af100-af101-flat-picture-style-for-cinematic-look/

Rick Burnett
09-10-2011, 08:02 PM
Hey Rick, I too wasn't too keen on the sharpening on the Af100 until I read this blog. A +2 over the detail setting makes sense to retaining the sharpness outside of the dark zones.

http://marvelsfilm.wordpress.com/2011/08/08/panny-af100-af101-flat-picture-style-for-cinematic-look/

You forgot to mention that it was coring that you were referring too. :) They still run with the detail way down at -6 just about where I was. I played with the coring as well, and I believe I was at -4 too.

And a good point that article makes is you still need noise if you want to battle banding. This is why I never reduced detail all the way. It is essential in all 8-bit output compressed cameras. It helps to dither out the transitions in gradients.

bgundu
09-10-2011, 08:10 PM
You forgot to mention that it was coring that you were referring too. :) They still run with the detail way down at -6 just about where I was. I played with the coring as well, and I believe I was at -4 too.

And a good point that article makes is you still need noise if you want to battle banding. This is why I never reduced detail all the way. It is essential in all 8-bit output compressed cameras. It helps to dither out the transitions in gradients.

Oh yes, I've updated my post. As for Noise, depending on the type of shoot, I find the Af100's very acceptable, but I know many don't. I'm a photographer as well, so I am very comfortable lighting my compositions which usually eliminates most noise issues.

Noel Evans
09-10-2011, 08:10 PM
Hey Rick, I too wasn't too keen on the sharpening on the Af100 until I read this blog. A +2 Coring over the detail setting makes sense to retaining the sharpness outside of the dark zones.

http://marvelsfilm.wordpress.com/2011/08/08/panny-af100-af101-flat-picture-style-for-cinematic-look/

Clearly they dont understand how detail coring works. And I disagree with their quoted "panasonic techies".

Great article BTW Barry. A solid and fair comparative. Im thinking I might push onto the F3.... blame Timur.

Steve Kahn
09-12-2011, 12:27 PM
Im thinking I might push onto the F3.... blame Timur.

You may want to wait on that. I was told by a Canon rep that they're coming out with a 4k cam with a "5D like chip" in the F3 price range.

Sumfun
09-12-2011, 12:44 PM
You may want to wait on that. I was told by a Canon rep that they're coming out with a 4k cam with a "5D like chip" in the F3 price range.

Will it be a DSLR or camcorder body? Any time frame?

simonpwood
09-12-2011, 12:57 PM
Will it be a DSLR or camcorder body? Any time frame?

Those rumors have been going around for nearly 2 years now (though they keep getting bigger & better - 4K is a new one!). CanonRumors.com says there will be an announcement in October.

Steve Kahn
09-12-2011, 12:59 PM
He told me "I didn't tell you this" wink wink...

camcorder body
4k
444
I said 5D quickly but I think he said it would be an APS-C sized chip.
early next year.

That's all I could get out of him.

simonpwood
09-12-2011, 01:16 PM
Well I hope he is right, because more competition can only be good for us consumers.
However Canon is notoriously tight about releasing information, and it would be unusual for Canon Reps to know much about new products before Canon releases the official specs (unless they are actually on the design team).

It could be your rep reads canonrumors.com and has been hearing the same rumors we have all been exposed to for the last year and more. Canon certainly will release a big sensor camcorder at some point (they are always last to join the party). The last rumors stated it would feature a 7D style sensor, and use the same 4:2:2 XF codec that Canon has already implemented with the XF300. This as yet vaporware camcorder was meant to have an EF mount, which all makes sense as Canon is already making equipment with all of those features (but just not in the same body)!

bgundu
09-12-2011, 01:31 PM
Remember this?

http://www.engadget.com/2010/09/02/canons-4k-multipurpose-camera-concept-weakens-knees-shoots-vid/

Gary Huff
09-12-2011, 01:36 PM
The Canon blow dryer with 4k watts of power? :cheesy:

TimurCivan
09-12-2011, 02:52 PM
Clearly they dont understand how detail coring works. And I disagree with their quoted "panasonic techies".

Great article BTW Barry. A solid and fair comparative. Im thinking I might push onto the F3.... blame Timur.

I accept blame for nothing. :smile: