Quality of EX1 and HPX500

I've been watching EX footage on a 154" screen, viewed from 9' and driven by a 30-bit 2K projector and frankly I haven't seen anything from film that "annihilates" the better digital productions. And then there's the factor that you might be comparing different cameras under different shooting conditions. I would be very surprised if it were possible to pull more than 10 f-stops out of film without producing a lot of grain and artifacts. Take someone like Philip Bloom, who knows how to shoot with an EX camera and put his work against anything else out there shot on film, or any of the best digital systems, and it would be hard to tell which camera shot which picture. The exposure lattitude I get out of the EX3 seems significantly greater than what I got out of FujiColor or Kodak Ektar films on my 35mm SLR years ago. I have prints and high end scans of negatives I've shot under various lighting conditions, and while the film today does quite well, I think it's met its match in the digital world.
Since I've had my projection system, which produces a far supior contrast ratio to the Barco offering, being 12,000:1 vs Barco's 2000:1, I am seeing details deep into the shadows as well as the highlights, that my LCD panels before it could not differentiate. The EX picture thrives on a big screen with deep color support. I've shot some interior room shots with bright noon-day sun on snow-covered landscape streaming in the window directly in the shot, and no fill lighting and it looked like what my own eyes see. I never could do that with film, without using a flash or photofloods to fill, unless I wanted a black silhouette shot of the subject backlit by the window.
It wasn't until I saw EX footage, practically wrapped around to my peripheral vision, sitting close to a wall-sized screen, that I really started to appreciate the EX picture.
 
With respect there are some strands here that need pulling apart.

Your camera's recording is 8bit. Period. Deep color or not (my projector supports deep color too) - A 30bit projector at the moment is a nonesense.

Contrast ratios are also fraught with ill-conceived battles. A True 2000:1 (on/off) at D65 is good enough, trust me. How is the 12,000:1 you have measured and calibrated? Does it use a dynamic Iris to get there? Is it a genuine D65? You see consumer grade projector bulbs are often deficient in certain colours and to get that 12,000:1 are pumped up so much that their greyscale tracking would give the Hulk a flush.

What I'm trying to say is you're getting lost in marketing specification and the shock of how good your own system is without accepting that there is stuff out there in a different price league that is much better (at a higher price point.)

Now I've seen EX1s on various consumer grade projectors and as good as it looks (it's very sharp) - it does not come close to DCi in terms of dynamic range and colour - the two things that make an incredible difference to picture quality.

Some people would choose 720p and 4:4:4 colour over the higher resolution as we believe this mimics the quality of film better.

And I'm pretty sure PB would admit that as good as he is and the EX1 is - it's not as good as film.

If the EX1 and your projector were as good as you say there would be no need for RED, VARICAM, Sony F35, VIPER and the DCi standards; they could save a load of money by using your workflow and installation.

Now I've got a great 1080p Projector (Calibrated, bat cave etc) and some nice video camera equipment - and I've seen good filmouts and excellent DCi presentations and I can assure you there is a big difference.
 
Yes, it's NOT a dynamic iris, but a true static contrast ratio. You can tell the minute you see blacks on this thing. Where consumer projectors turn it to mud, there is beautiful subtle detail. Skin tones are utterly natural looking. It's got some of the most accurate color I've seen anywhere.
What you get once you get over $225,000 is lumens. 30,000 lumens and up, but at the cost of needing 3-phase 208vac power. Even the Christie at $70K has less contrast ratio and not as nicely deep blacks. But you get the long throw for a 120' projection onto a 60' screen.
What gets written to SxS is only 24-bit color; you have to use HD-SDI to get the 30-bit color. Either way, the gradients are totally smooth on the projector, whereas my 24-bit LCD shows gross banding with a particular sunset scene. I suspect there's a kink in the gamma curve of the LCD that's responsible for this anomaly.
While it's likely that a transfer to a top-notch filmstock may show a difference in quality and resolution with one of the $250K and up cameras, or with direct film optical cinematography, it's all lost by the time it gets to Blu-ray disc. You'd have to be watching a 4K projector, or an optical projector to see a difference. And it would have to be a master from the camera, not a distro print. Distros look about like DVD on a good quality upscaler. I keep wanting to scream at the projectionist to focus the projector, but I can see the grain is clearly in focus, just the picture isn't very strong on detail.

That said, I know that film captures more detail. But what the eye perceives is a particular anomaly of MTF, or Modulation Transfer Function. Film's detail response starts rolling off in amplitude at a very low spacial frequency. Modern digital cameras are flat right up to the Nyquist sampling limit. Our eyes don't perceive the low amplitude fine detail (it is also mixed up with the film's grain structure), but they do perceive the intensity of the slightly less high frequency spacial detail with much greater acuity. The result is that a good 720P picture with flat signal response right up to the antialias filter cutoff will look more detailed than a picture shot at 1080P with a poorly-designed antialias filter, poor optics, etc.

Film needs a sort of 'post emphasis' on the high frequency detail, if it is to show its true potential. But this doesn't work this way in practical use, so it always looks a bit soft. I am not aware of any digital camera system that is capable of the eye's 14 f-stops as of yet, though RED is rumoured to be working on an imager of several inches in size that they claim may achieve such dynamic range. But presently RED, SI-II, and even that vaunted F23 are all stuck around 10-11 f-stops, maybe just a bit more for the F23. But the equipment is only as good as the operator, and knowing under what conditions to play on the strengths of a given camera. I've gotten amazing footage out of a Canon HV20, under very controlled conditions. It's all in how you set up the shot.
 
Where are you getting 24bit and 30bit from?

The Ex1 is MPEG2 based 4:2:0 and 8bit to its SxS. It can derive 10bit form the HD-SDI I believe but this isn't a perfect implementation and that is live not recorded.

Also lumens has very little to do with contrast, and contrast alone is not the be all and end all. You need good contrast - 2000:1 to be convincing, good greyscale tracking and all the primaries intact. ON/OFF contrast tells you very little. (Not to mention ANSI) - It doesn't even guarantee a good black level, it just means the difference between your low intensities and high intensities is wide - but that could mean a really high lamp output rather than a particularly low black level.
 
24-bit is what, back in the TARGA32 days, was called 'true color'. 30-bit color is what is referred to as 'deep color' in today's standards. HDMI 1.3 supports 30-bit color. My projector processes and displays 30-bit (over a billion colors) color.

8-bit is grayscale, or mapped color, and is found if GIF (graphic interchange format) graphics.

Yes, lumens has little to do with contrast. And in fact, the huge Christie projectors, despite their 30,000 lumens output, have lower contrast ratios. Though in a situation where you have simultaneous bright and dark areas of the picture, 2000:1 seems to be adequate. Room reflected light starts to become a more important factor in dynamic contrast, which is why theaters have dark interiors. Consumer projectors often use trickery like dynamic irises to achieve better measured specs, but that's why their pictures look murky. Achieving both deep black AND bright white on the same frame of a scene is the real deal. Color temperature and flatness of the lamp source' spectral output over the entire visible range is important. And don't forget the screen material! All screens are not equal. A performance of x=.002, y=.004 (red +/-0.0%, green +/-0.0%, blue -2.9%) with a Sencore OTC-1000 using a NASA Calibrated White Reflectance Standard as the reference means a neutral white color that makes for a reference quality projection material.

Accurate fleshtones are such an important part of a good projection system, as is having useful dynamic range into shadows and highlights. I was amazed how much more detail I am seeing on the projection system than what I saw on the LCD panel it replaced. Not to mention it being so much large and subtending so much more of my vision, out to the peripheral. At this size, I readily spot differences in film stock from scene to scene, and those couple of shots in most films where the focus puller 'blew it'. I see chromatic abberation and also internal reflections between lens elements on a number of films. Stanley Kubrick's films suffer a lot of weird lens anomalies. All these flaws escaped me when viewed on a conventional LCD panel.
 
it seems that the ex1 is the only "consumer", use that term lightly, cam to break the 800 line resolution limit, the dslrs, 7d, 5d as well as the hpx, dvx jvc hd250 and sony own z series all max out at just under 800 lines, while the ex1 is pulling in 1100 lines.
 
"Consumer" is a rather inapplicable term here.. the EX series is manufactured by the Sony Broadcast Division, whose products are intended for professional use. Try walking into a museum or other public place with an EX3 strapped to your person and see how far you get before security nabs you for having a professional camera in their facility. :)
 
basspig said:
24-bit is what, back in the TARGA32 days, was called 'true color'. 30-bit color is what is referred to as 'deep color' in today's standards. HDMI 1.3 supports 30-bit color. My projector processes and displays 30-bit (over a billion colors) color.

8-bit is grayscale, or mapped color, and is found if GIF (graphic interchange format) graphics.
Click to expand...

But your Ex1 is not 24bit or 30bit, that's all I'm getting at - neither does it support deep colour - it doesn't even have HDMI I believe (could be wrong on that.)

And we're talking about different compression schemes, colour spaces and conversions.

It's true a regular projector (like many display devices) will derive a full 24bit across the RGB, but your EX1 doesn't store its pictures in RGB (For a start it's trimmed/mapped <16 & >235). There is a translation of course (not forgetting your colour samples are slashed in this process.).

Hence my point being it can come nowhere near the quality of a high quality colour system and display that can access that.

So something seen on HDCAMSR in either 4:4:4 or 4:2:2 will clean it out with the colours - being my original point. And I've seen some nice BAFTA presentations like this. Not to mention the Gamut, Datarate and dynamic range of DCi.

I don't think we can argue the sharpness. Although RED delivers around 3k.
 
futureman said:
it seems that the ex1 is the only "consumer", use that term lightly, cam to break the 800 line resolution limit, the dslrs, 7d, 5d as well as the hpx, dvx jvc hd250 and sony own z series all max out at just under 800 lines, while the ex1 is pulling in 1100 lines.
Click to expand...

Which HPX? - I would've expected the HPX300 and It's more expensive siblings to match this. But I don't know for sure.
 
woops, obviously i did not mean dvx, since thats sd... i meant hvx, i also left out the canon cams xh xl, which also max out under 800 lines... i am not sure exactly why these cams are unable to pull in over 1000 lines, it could be anything, but the sony seems to be the only one to do it...

its particularly curious, since the dslr's have 21mp imagers and still cant break 800 lines!?!?
 
ROne:
The internal processing in the EX is indeed 30-bit color space. Could you imagine if it was only 8-bit? It would be incapable of color! It would be a B&W camera. The dynamic range does indeed exceed 16-235. In fact, it causes problems with my NLE, because it produces 'super-white' and my NLE clips whites at 235 or 100IRE at render time, so I have to use an additional step in my editing preparation to scale back the dynamic range, save that file as a Quicktime file, then import that video-legal file and render to the final format.
At the imagers, everything starts out RGB. Downstream, it gets reinterpreted as YCbCr for encoding. But it still contains the full range of colors. Now at the longGOP MPEG encoder, it gets truncated to 24-bits or 16.7 million colors, whilst the SDI gets the billion or so colors at 30-bit depth.
Obviously, 4:4:4 would enlarge better than 4:2:2 when the picture is displaying SMPTE color bars, or strongly-saturated, sharp-edged objects of opposite primary colors abutt eachother, but otherwise the difference becomes vanishingly small in typical images found in practical shooting. The closer to monochromatic the scene, the less the difference matters. Higher bitrate helps with reducing compression artifacts. HDCam SR will produce a cleaner, more solid looking picture, but so will SDI to the Wafian recorder, uncompressed.


futureman:
Those cams are HDV, which is 1440 pixels. Nyquist's sampling theorem states that the highest sampled frequency is just under half the sampling rate. That's why the HDV cams are limited to 700 lines.
 
@ROne,

I think you're confusing 8 bits per channel (for a 24bit file) with 8 bits overall (one color i.e. black and white) output.

Also confusing 24 bits file output with 30 bits internal (10 bits per channel).

By capturing live through HD-SDI (as with a Nanoflash) you're getting a 10 bits per channel input into a file at 4.2.2 by bypassing the compression. If you were to record it on an SxS card, then capture through HD-SDI you're right that it would be reconstituted from 4.2.0.

The trick is to capture the internal data before its compressed and the data is lost.
 
Last edited:
On the contrary, I come from a graphic design business from 1986 to 1997, where I worked with many flavors of image files, ranging from 1-bit monochrome line art bitmaps to 16-bits per channel (48-bit) deep color drum scanners from Heidelberg and Scitex, as well as TARGA32 graphics systems of the '80s from AT&T.

So you do agree that SxS is 24 bits file output, not 8. That was my point exactly. The SxS should be 16.7 million color output. The SDI out is 30-bit and should be good for about a billion colors. 8-bits only gives 256 colors and if a camera were based on such a crude sampling depth, banding would be so severe that the output would be unusable.

NanoFlash cannot yet record 30-bits. If you want 30-bits, the Wafian HR-1 is the only option that I'm aware of at the present time. And that has to be done during a live shoot. Once it goes to SxS, it gets truncated to 24-bits.
 
Duke M., I need to step away from the computer for a while.. I'm seeing cross-eyed and my brain is confused.
 
I see what your saying... But

When we talk about HD-SDI when do we say 30bit? I've never heard it called 30bit - it's referred to as 10bit .

But as I've pointed out several times irrespective of your cameras internal processing and your projectors final output your camera is limited in by its final output ie. 8bit Mpeg 4:2:0.

I can remember targa boards and 24bit and 30bit etc, and RGB colour derived from that palette - but we are talking compressions schemes and have only ever heard them referred to as 8bit and 10bit.

In this respect it's no different to many cameras at all, which is why I was shocked by you being bowled over by the colours and lack of banding on a inferior colour system.

What would be the point of us all getting excited at 10bit AVC-I and 12bit Si-2k if the EX1 was 30bit!
 
Last edited:
Here you go, the same chap who you quote from who tested the Ex1 for the beeb. I asked him how it handles the bits.

"I don't know to what depth the EX1 samples video at the sensors, it's probably 10-bit RGB, but can't be sure. That doesn't make it 30 bit though, because it's 10-bit RGB in linear light, and we're interested in the bit-depth in non-linear signals. Most consumer cameras are 10-bit RGB these days, while top-end broadcast at 14 or even 16-bits for the film-replacement types. But that doesn't make them 30- or 42- or 48-bit either. As a rule of thumb, the ADCs at the sensor need about 3 bits more depth than the camera is to deliver, because gamma-correction, when done in digital form, needs to apply extra gain near black, and 3 extra bits allows up to 8 times gain.

The internal processing of a 10-bit RGB camera would probably be about 18-bit RGB, in order to preserve accuracy in what should, ideally, be floating point processing, but usually isn't (integer processing is faster and easier, if less accurate). Common practice is to work at 8 bits more than the ADCs deliver, so a 14-bit camera like the Sony HDWF900R (there are lots of others that use 14-bit as well) works internally at 22-bit depth. The reason is to allow enough bits to provide the extra gain for gamma correction, black stretch, linear matrix operation, detail control. Working at lower bit depths would result in premature clipping through data rounding errors. there's a way round that, but it comes later.

Final output for delivery is usually to 8-bit for consumer cameras, 10-bit for broadcast, or up to 16-bit for film-type cameras. The simple use of these bit-depths does not define the camera as being in one of those groups, but the bit-depths are those expected for those groups. These bit-depths apply not to RGB, but to coded YPbPr, where the chroma channels have half the sample rate of luma (4:2:2) or quarter the rate (4:2:0). It's fairly normal for broadcast 10-bit to be 4:2:2 and 8-bit consumer to be 4:2:0. MOPEG2 is usually 4:2:0 but can be 4:2:2. So, at this level of the system, you could describe a 10-bit 4:2:2 camera output as being 20-bit (because for each group of 4 10-bit luma samples there are 2 10-bit chroma samples of each of Pb and Pr, making a total of 8 10-bit samples, spread over 4 picture points), but that would also be very silly. In the same way, a 8-bit consumer camera delivering 4:2:0 could be described as 12-bit (because there are 2 samples of 8-bit chroma for each group of 4 8-bit luma samples). And that would be just as silly.

Have you spotted the pattern yet ? :smile:

The EX1 delivers 10-bit HDSDI, and so could, in extremis, be described as 20-bit, but it doesn't make sense, for all the reasons I've given. Not only that, but the video noise level in the EX1 is about -40.5dB in linear mode (gamma off), and thus the noise floor, the level below which fine detail is lost in the noise, is about -40.5dB, which is a little under 1% of total video level. So, in theory, only 8-bit delivery is needed, and 7-bit wouldn't be amiss either. In practice, delivering to 10-bit means that the noise is preserved in the output and not limited by the output format, so it;s a good idea, it's always a good idea not to throw image content away before you have to. Other cameras exist, HD cameras, which deliver noise levels as low as -59dB (in my very recent measurements), and these really do deserve 10-bit output.

Anyway, back to the question: no, the EX1 is not 30-bit."
 
Last edited:
I know exactly what Bass Pig and ROne are doing. They are getting technicalities confused! You guys are both right, in a sense, but let me try to clarify.

For some reason in the video world we've called things "8-bit video" or "10-bit video", etc. This is truncating (pun intended) a very important part of the phrase out. Just like we seemingly short-hand/short change everything else in this world, here is the correct way of stating video specs:

x-bits per channel color depth.



So the XDCam EX format written to memory cards as MPEG-2 video has a total of 24-bits of information. There are three color channels, Red Green and Blue with 8-bits of data/information per channel. 8 x 3 = 24 bits.

XDCam EX LIVE out of of the camera via HD-SDI supports 10-bits per color channel for a TOTAL of 30-bits.

So, with 8-bits per color channel we have 256 steps from black to white. And with 10-bit there are 1024 steps. These numbers become important when we manipulate this data, especially with color correction.



When discussing 4:2:0, 4:2:2, and 4:4:4, this refers to sampling rate. In audio it would be in kHz, such as 48 kHz (common digital audio sampling rate). With audio this is samples per second; with video it is samples per line.


Color space, known as YUV or Y'CbCr represents brightness, or luminance (Y), and the color difference signals (Cb,Cr).

4:4:4. Y, Cb, and Cr are sampled equally along each field or frame line with one Cb sample and one Cr sample for each Y sample. High-end video editing systems use 4:4:4, often together with uncompressed video.

4:2:2. Y is sampled at every pixel, but Cb and Cr color information is only sampled at every other pixel. While this sampling rate significantly reduces the bandwidth requirements, only a slight loss of color fidelity occurs. Many editing systems process this level of color information.

4:1:1. Y is sampled at every pixel, but the Cb and Cr color information is only sampled at every fourth pixel, saving even more bandwidth. This sampling rate is used in consumer DV cameras.

4:2:0. This sampling rate uses a different technique known as a spatial sampling that takes a 2 x 2 square of pixels.



Sources: http://www.microsoft.com/windows/windowsmedia/howto/articles/UnderstandingHDFormats.aspx

http://help.adobe.com/en_US/AdobeMediaEncoder/4.0/WS8A54CAA7-268E-4af0-B8A4-0EA3EACEC1A3.html



Discussion of DSP/internal processing bits is another thing entirely. This has to do with the number of bits (the smallest unit of information on a machine) that can be processed or transmitted in parallel, or the number of bits used for single element in a data format. Typically with video we are discussing the analog to digital converter processing speed.



Now, with that being said, the important thing is the storage medium. If you shoot with a DSLR camera and use RAW files, you know that this is the ideal way to go for post-processing. JPEG will get roughed up with any extreme post-processing. The higher end cameras enable this flexibility and it's something the EX1 (or any camera that writes non-raw data) can't compete with.
 
You must log in or register to reply here.
Back
Top