No I was never really a fan of the footage I'd seen from the FS100 due to its highlight handling. Even if the FS700 is not quite up to the FS100 in low light it looks dam close and more than enough for real life shooting.
Ro.
Awesome FS700 test of the FS700
- Thread starter FelixGER
- Start date
No I was never really a fan of the footage I'd seen from the FS100 due to its highlight handling. Even if the FS700 is not quite up to the FS100 in low light it looks dam close and more than enough for real life shooting.
Ro.
Nirv
Carbonite Member
The highlight performance of the FS100 is fine and a lot of the blown highlights you have seen are down to people clipping super brights in post or during decompression by video sharing sites.
The LCD of the FS100 is not up to snuff and if you expose by eye then you're likely to clip. The LCD is useless as there's much more clean information being recorded on the card than the LCD lets you see. The histogram isn't much cop either but at least shows when you're clipping white in the far right column.
I use the standard gamma curve and record values between 0-109 ire and balance in post. I never suffer from clipped highlight issues or with those horrid coloured posterisation effects. The FS100 is so noise free that you can easily save highlights and bring up the blacks (with no noise or macro blocking) in post. My mantra is look after the highlights and the blacks will look after themselves. Which they do, because of the incredibly high SNR of the FS100 which should not be forgotten.
I should be shooting a yacht video soon that will hopefully prove there's plenty of life left in the FS100.
The problem with the some of the FS700 gamma curves is they reduce the contrast in the brighter registers so much that skin tones look muddy and unnatural. There is too much of a knee slope (for my liking). They're just less contrasty versions of the CineTone 2 curve of the FS100. Mid tones look better with the FS700 curves but the highlights are crushed to b*ggery. In 8 bits theres no way to put any contrast back in there.
thekeygun
Well-known member
LOW LIGHT handling of the fs700 vs. the fs100.
http://www.youtube.com/watch?v=bCgCcNTBDec&list=UUr1jF09uL7J_evqa_BgoSiQ&index=8&feature=plcp
http://www.youtube.com/watch?v=bCgCcNTBDec&list=UUr1jF09uL7J_evqa_BgoSiQ&index=8&feature=plcp
paulcurtis
Veteran
Aside from being impossible to tell what the noise patterns are like on youtube that is one of the videos that pushed me over onto the 700 simply because of the way the candles at the end show the highlight differences. One situation that the 100 struggles is under tungsten lighting (or warmer in this case) and highlights, under daylight the highlight handling is much better.
Of course it's possible that Andy set up the 100 with default PP making it worse but even so the handling is much better in that scenario.
cheers
paul
Barry_Green
Moderator
Charts don't lie. People who read them may choose to misinterpret them or may not know what they're looking at, but the charts tell the real story.
Your eyes, on the other hand, are the things that can and frequently are fooled. The whole reason 4:2:0 works at all is because the eye can't see (or prioritize) a difference that very much is there.
For example - look at the squares "A" and "B". Your eyes tell you they're a different color. Go into photoshop and sample them -- they are exactly the same color.
Taken from the article at http://scienceblogs.com/startswithabang/2010/02/21/weekend-diversion-a-question-o/
So I will always trust an impartial, scientific, properly-designed, properly-used chart to determine what something is capable of. Especially in comparisons, one against another.
Barry_Green
Moderator
Definitely charts can be misinterpreted. Clearly. There's people saying the 5D Mark III resolves 890 lines(!) when the charts show actual resolving power of around 600 lines, before aliasing sets in. So someone reading a chart has to know what they're doing, and what the shooting scenario was, before being able to render a proper reading.
But I still say eyes are far more trickable than charts. Look at the Zacuto shootout, where people are using their eyes to judge a GH2 on par with an Alexa...
But I still say eyes are far more trickable than charts. Look at the Zacuto shootout, where people are using their eyes to judge a GH2 on par with an Alexa...
Barry_Green
Moderator
Well, that's the problem -- they're not using the same chart. The FS100 has been tested repeatedly to deliver 10.5 stops (even if Abel insists on oddly numbering it "11.5", if you count the actual stops shown, when using the DSC 18-stop dynamic range chart. Which seems reasonable.
But the FS700 has been tested to 14 stops, using a brand new self-backlit DSC chart.
Is it that the charts are revealing different numbers? I used a Stouffer chart to test the GH2/7D early on, and after posting my results, I decided to upgrade to a much better chart, the DSC backlit 18-stop. The Stouffer isn't designed for measuring camera dynamic range, it's a step chart for printer evaluation, so it was basically the wrong tool for the job. I use the DSC 18-stop for all my dynamic range evaluation now (which is, I think, the same one Abel used in their FS100 and FS700 testing, and they show the same dynamic range for both, 10.5 stops).
In short, I simply don't think that there's any dynamic range difference between the FS100 and FS700 based on the most similar chart testing I've seen (Abel's). But Adam Wilt is an impressively smart guy and knows very well what he's doing, and if he got 14 stops out of an FS700 using the new chart, then there must be some way to get 14 stops out of it. But my question would be, what would an FS100 show on that same new chart? What would an AF100 show? What would an Alexa show? We don't have a base of information to compare, we have only that one test on that one chart, and to me, that's way too early to go calling it definitive.
During the SCCE they tested all the cameras they had (from Alexa on down to 7D) using one consistent testing methodology and using one specific chart, and that is the right way to do it. The only way to answer how the FS700 compares to the FS100, in reality, is to test both of them using the same methodology on the same chart. As far as I know, that has been done by Abel, and they declared "The FS700 has much the same dynamic range as the FS100", and their chart backs it up (even though, for no discernable reason I can fathom, they call it "11.5" stops when there's clearly 10.5 on the waveform display...
Nirv
Carbonite Member
The FS100 presets are unusually warm and need to be tamed.
I also built a FCPX Effect in Motion that colour corrects the highlights and applies a knee to the highlights. I rigged a levels filter with the Open EXR Tonemap filter because it has a nice knee control. This is exactly what the FS700 is doing in camera and what other manufacturers have been doing for a while. Even my XHA1 colour corrects the highlights.
I'm still working on the effect and want to see if I can add more controls and restrict the corrections to a key i.e. just skin and then I'll make it available for download for those with FCPX.
One of my friends made a documentary for BBC4 and his film was littered with posterised highlights and he didn't use a FS100 either. It was a XDCAM of some description set to REC709. When I asked why the highlights were blown he was rather defensive. Turns out there's no money to colour grade and what comes out the camera hits the screen not even legalising super brights just let them clip ugly!
Glad I'm not working on BBC Factual crap.Hey Barry, maybe this is wishful thinking, but didn't Abel use an early pre-production model while Adam used the actual production model? Maybe some tweaks happened during that time? I saw Mathew Allard's review and he thought the FS700 had at least 1 stop more than the FS100.
Last edited:
Barry_Green
Moderator
Not sure on the pre/pro status, but I've used lots of preproduction units and haven't found that substantial a difference between the preproduction and the final.
The problem I have with someone "thinking" that some unit has "at least" so-many-stops more, is that it's totally subjective and, unless compared in a controlled circumstance, useless. Again, I point to the Zacuto shootout, where an 8-stop DSLR (GH2) is being rivaled with a 14-stop Alexa. That kind of thing simply isn't helpful in determining what *is* and what *isn't*.
Only way to know for sure is to put them side by side, in the same conditions, by the same operator, judging by the same criteria, and then you should know. I don't really see a reasonable argument why an FS700 should have any more dynamic range than an FS100, if judging by pixel size, and certainly can't see why it should have three and a half stops more. I mean, reality trumps theory and maybe it does, but just from armchair speculation I certainly wouldn't have thought that that would be the logical outcome, and again, Abel's testing shows it to not be so.
nyvz
Veteran
I would then suggest that I don't see any reasonable argument why an FS700 SHOULDN'T have any more dynamic range than an FS100... What are you basing your statement on? These are different cameras with different sensors. One of which (FS100) uses a sensor from the F3 which is known to have additional highlight latitude available that was deliberately removed in the software implementations available in the non-SLOG F3 and FS100. I agree we should all be skeptical of most claims, especially ones throwing around numbers of stops without much quantification of noise levels. On the other hand, it seems fairly clear that cinegamma4 will almost definitely give us something that was not possible on an FS100, and that is an additional stop of overexposure latitude even if it does not indicate anything in particular about the full dynamic range.
Whether or not there is more dynamic range available to the FS700, the practical reality is that it may nor may not be useful if the gamma curves do not allow us to take advantage of it in the way we want. Fortunately the addition of extra cine gammas will probably make the range more useful as we will have more options to utilize it in the way we want to capture a particular scene. Of course we will still be limited to some extent if there is a lot of shadow range that cannot easily be brought up in camera.
Last edited:
nyvz
Veteran
Abel tends to judge dynamic range from a relatively small waveform that at best has a resolution of about 100lines, whereas Adam Wilt's claim of 14 stops came from images from the camera where blocks on the 21stop chart were actually visible especially after boosting the signal. Since most digital imaging devices tends to have a large part of their dynamic range compressed into the very low shadows due to their linear light response that is only corrected by a gamma curve, as opposed to a log curve which would actually fully decompress the dynamic range. Much of that compressed shadow detail is too low in contrast to show up as discrete steps on a low resolution waveform without a signal boost. It is quite likely both tests saw the same thing, they just read it diffferently. Whether that low contrast range can be pulled out in camera or in post as useable dynamic range is a matter of workflow and of subjective tolerance to noise.
morgan_moore
Major Contributor
None of the DR tests seem to show any colour shifts (being black and white!)
I think they dont really show how these cameras work in the field
Dont think im against accurate testing - Im a big fan of keeping it scientific - but Im just not convinced on the methodology of any of these tests
I think DR is over-rates (esp for 8 bit) what is important to me is the transitions to clipped
I think much better testing comes from evaluation of sidelit smooth cylinders, it would just be nice to formalise it - here is a very informal example.. https://vimeo.com/27878451
S
I think they dont really show how these cameras work in the field
Dont think im against accurate testing - Im a big fan of keeping it scientific - but Im just not convinced on the methodology of any of these tests
I think DR is over-rates (esp for 8 bit) what is important to me is the transitions to clipped
I think much better testing comes from evaluation of sidelit smooth cylinders, it would just be nice to formalise it - here is a very informal example.. https://vimeo.com/27878451
S
Last edited:
paulcurtis
Veteran
That's true.
That sounds very similar to what i want to do, i'd love to see how it works when you're ready. You know i was banging on about reversing the camera curves. However the bulk of my work is going to be in CS6 (it's also very cost effective to get others working in CS6 too with a few months of cloud subscription) so i'm looking to build solutions within that workflow.
Although in theory the 700 could do a better job simply because it's working with pre compressed image data and has more to work with.
One of my friends made a documentary for BBC4 and his film was littered with posterised highlights and he didn't use a FS100 either. It was a XDCAM of some description set to REC709. When I asked why the highlights were blown he was rather defensive. Turns out there's no money to colour grade and what comes out the camera hits the screen not even legalising super brights just let them clip ugly!
Ah, the BBC....
cheers
paul
Nirv
Carbonite Member
That's exactly how I have set up my camera. There is a row of houses with chimney pots that catch the sun (when there is some) in the morning and evening right outside my office window. They have been invaluable in learning how to tweak the knee and to cool off the colour.
The Auto knee also works impressively well and is well worth using to learn how to setup the knee in different lighting. I'm not normally an advocate of auto anything but when I was shooting some tests on a cloudy day with the subject going in and out of the shade the auto knee results were very impressive.
Having compared FS100 and FS700 footage it seems to me the FS700 sees into the dark area more than the FS100. The FS100 seems to just stop seeding detail at a particular low level where as the FS700 sees lower but they are much noisier. Now these noisy stops appear on test charts leading to numbers from 12-14 stops but are these extras 1-2 stops of any value other than to the marketing department?
The biggest improvement I can see are the Gamma curves which appear much more usable than the ridiculous CineTone2 curve. Whoever thought that curve was a great idea needs a kick in the goolies. The best FS700 curves look like they are a combination of the standard gamma low and mid tone performance and CineTone highlight roll off. Something that Sony could've given FS100 users in the firmware update but chose not to.
I would really like to see a consistent test comparison between the FS100 and 700 by the same person using the same test methodology just to see exactly how far apart these cameras are. The pedant geek in me demands this happens.
Nirv
Carbonite Member
Ah, the BBC....
It appears I have offended a friend by daring to suggest, in the nicest possible way, that his documentary looked like sh1te.
I took him to task because to an outside production house you're hit with a whole raft of stringent quality demands from cameras to post post production and yet internally they produce quality no better than 24hr rolling news. What a joke.
chromeboy007
Well-known member
Very nice footage. Imagine if you had a tilt shift lens at 2:25!
cuervo
Veteran
I hardly hope to come from the level of experience expressed here. I have an FS700 on order and an EX1 and 7D sitting on my equipment shelf. I fully intend to set up all three, simultaneously, with a step wedge, when my FS700 arrives. Should relativistic DR be revealed in such a test? And if so, why hasn't anyone already done that?
Last edited: