Red Rebuttal

What follows is a rebuttal to Chris Kenny's own rebuttal of our Red Facts article. His original rebuttal can be found on Indie4k's site, part 1 here and part 2 here.

Chris's original text is indented and italicized, our answers are between.

Johnson argues that the Red One is over-optimized for resolution at the expense of other criteria, like dynamic range. As part of this argument he makes the case that 2K is enough resolution for a feature. This is largely true, but there are several mistakes in the evidence Johnson presents for this. For instance, Johnson uses the fact that Kodak defined 2K as a standard resolution for DPX/Cineon files as evidence that Kodak considers 2K sufficient. He says:

"Note that Kodak could have made any specs they felt were required here to fully achieve the quality of 35mm film: for example, they could have prescribed a file that has more resolution and less color information. Here’s one they could have used: 4K pixels, 3 channels-per-pixel, and 8-bits per channel — that would have actually been the exact same file size as the spec they did choose, and they even could have said the word, “4K.”

There are two problems with this. First, 8-bit 4K frames aren’t the same size as 10-bit 2K frames. If we assume a 16*9 frame, 10 bit-per-channel 2K is 30 bits per pixel * 2,359,296 pixels, or 8.44 MB/frame, while 8-bit 4K would be 24 bits per pixel * 9,437,184 pixels, or 27 MB/frame, over three times the size. Oops.

Thanks for the correction, that was indeed a math error on our part. However, it was a math error on a HYPOTHETICAL spec that doesn't exist anyway and the math of which has no bearing on our argument. If you check our site, you'll see that we've already changed the math to be correct, and it doesn't change the argument one iota.

Second, Kodak didn’t define 2K as the standard resolution for digital film scans. They defined it a standard resolution, alongside, you guessed it, 4K. It’s pretty widely acknowledged that a 35mm original camera negative captures more than 2K worth of resolution. Certainly 2K acquisition and workflow can provide a perfectly acceptable level of quality. It’s more resolution than ends up making it into a theatrical release print. But Kodak didn’t declare that 2K is the end of the road, and nothing beyond it is worthwhile. It’s true that most features are currently finished at 2K rather than 4K, but in large part this is simply because it’s good enough and technically much less of a challenge… not because nobody should ever want anything better.

If you read our article, we never said Kodak described 2K pixel-count as "THE standard" -- in fact we go out of our way to say that what Kodak DID define, Cineon (DPX) files which are 30-bit-per-pixel, can be ANY pixel dimension. What we do say is that these files are 2K in ACTUAL APPLICATION for most films today. We agree with you that exceeding the minimum can only be an advantage. The problem is that Red doesn't even get up to the minimum. We think you need to first meet the minimum: 1.8K and 30-bits-per-pixel (or there-abouts) and THEN increase. Sacrificing one for the other is disingenuous and it sacrifices the image.

If you disagree with Kodak about the minimums, then that's fine (then it's just a matter of disagreement, and you can say Red doesn't meet Kodak's spec but doesn't have to because you think that the spec is bad), but you can't say that Red is exceeding or "better" than the DPX spec. The established standards include color depth AND resolution: Red is flagrantly shirking one minimum spec (color depth) to an extreme while exceeding the other (resolution) incrementally.

The resolution discussion continues, with an argument that because the Genesis and F23 capture 2K (or something close enough to it), they capture enough resolution that resolution doesn’t matter, and therefore Red’s higher resolution is not a meaningful advantage.

There a couple problems with this argument as well. The first is that it assumes (at least in the case of the F23) that three 1920*1080 sensors capture full 1920*1080 resolution. This just isn’t so. All general-purpose cameras have optical low-pass filters to avoid aliasing, which results in a loss of resolution. So The F23 necessarily resolves something less than 1920*1080. Johnson is right, of course, that the Red One doesn’t actually resolve 4096*2304, of course. Bayer-pattern sensors like the Red’s do resolve a fair bit less than their raw photosite counts might imply. But several tests have now shown the Red One does resolve about 3.2K, or about 1800 horizontal line.

This argument is so flawed that it's hard to know which way to refute it. Any of the following will do on its own without the others:

-The IDEA of an optical low-pass filter is to fuzz or blur resolution information that is SHARPER than the sensor can even see -- not to blur information that it CAN see. So the mere fact that there is a low-pass filter does not in itself prove that resolution is lost (an underperforming low-pass may not fuzz ENOUGH and create aliasing without reducing resolution at all). So, even though a REAL low-pass filter as opposed to an IDEALIZED one can reduce resolution, the mere fact of a low-pass filter does not in itself prove that resolution is reduced (at least not measurably).

-Our article points out that 1.8K is MORE THAN ENOUGH for theatrical resolution, so even if the low pass filter fuzzed the 1.9K image to 1.8K (which is extreme), it'd still be sharp in excess of the standard. There is no reason to think that Sony has gaffed the design even the amount mentioned. Additionally, it is not mentioned in your refutation that Red also has a low-pass filter and that it is much more difficult to optimize a low-pass filter for a single-chip (like Red has) than it is for 3-chips like F23 has.

-More important than either of the two points above: the entire point of our article is that microscopic measurements of resolution are not what's important to cinema. You need to get ENOUGH resolution, but you also need ENOUGH color data. Once you have enough of both, then you can start in with getting extra. Red doesn't even have ENOUGH color data. F23 (and Genesis) have more than enough of both -- no matter what the microscopic measurement of resolution are.

This argument also seems to take it for granted that nobody would ever need more than 2K resolution even for an original camera negative (or digital equivalent). While it’s true that, as noted above, 2K is higher resolution than a release print, 4K projection is likely to be quite common in the years to come.

As mentioned above and throughout our article, we are not trying to impose an upper-limit cap or maximum resolution. But, we are saying that you NEED COLOR DATA to get THE MOST INFORMATION THAT IS IMPORTANT FOR CINEMA. Again, once you have ENOUGH data, then you can start outperforming. Red is vastly underperforming in color data. If your car doesn't even have wheels on it, then you can't start boasting about the horse-power. If 2K looks great across a giant screen on a film projector or a 2K projector, it'll look great on the exact same screen with a 4K projector -- let's get to the MINIMUMS before exceeding them. Once you're up to scratch, then you can start making claims about outperforming. Cinema has a lot of standards for quality -- not just counting dots on the screen.

And consider: most still photo photographers shoot everything at the maximum resolution allowed by their cameras, even when their photos are going to end up being displayed at less than a megapixel on a web page. Few would be happy if you suddenly told them they couldn’t do this anymore. It gives them the flexibility to reframe later, to repurpose photos for higher resolution media later, to do things like masking on a large image before scaling down to the appropriate size for the deliverable. The only reason moving images aren’t regularly shot this way is because it’s technically difficult to build cameras that can do it, not because it wouldn’t be beneficial.

This is another one that is so full of holes that it's hard to even know how to answer it:

-Our article discusses perceptual resolution being different in images that are far away and moving than in images that are close and not moving (the difference between stills and motion photography) so most any comparison of perceptual sharpness between stills and motion is meaningless. Since we know how motion imaging ACTUALLY works, we don't need a bad analogy to stills to try to understand it. When you understand the thing itself, a vaguely workable analogy is useless.

-Still photographers often do extreme reframing by cropping. Motion pictures almost never do this (at least not more than a tiny bit that has no practical effect on resolution). Motion images ALWAYS color correct and need rich image information and almost NEVER crop in any meaningful way -- the analogy to stills is meaningless, because, in motion, color information is constantly used to adjust the image with color grading, while resolution is almost never used to do "cropping."

-Yes, we agree, there is nothing wrong with shooting "extra" resolution if you can get it. But, if you're going to claim to have the best system, you've got to at least get into the ballpark of being up-to-snuff on color before moving on in superfluous resolution that can't be seen or taken advantage of in any meaningful way. We think that F23 and Genesis are better than Red because they meet and/or exceed ALL the cinema standard specs. Red incrementally outperforms in one area and VASTLY underperforms in another.

The discussion moves on to bit depth, noting that high-end HD cameras generally have 14-bit analog to digital converters (ADCs) and record 10-bit data (with a log curve applied, though Johnson doesn’t mention this). Moving on to the Red:

The Red [...] (still just talking about the image before it’s recorded), is so busy trying to outperform the other cameras in just the one area of superfluous resolution that it hasn’t got to the full cinema quality of the other requisite aspects. They’re using up all their photosites on resolution (so they can say 3.2K or 4K is a bigger number than 1.9K) and are truncating color information. The Red sensor only has one pixel per channel instead of three, which means it is 3-times color subsampled. The Red sensor samples at 12-bit (compared to 14-bit on the other cameras) for conversion to digital (in a seemingly unpublished bit-depth).

This seems to show a bit of confusion. Photosites are analog components. They don’t sample at a specific bit depth; bit depth is a function of digital encoding. A photosite simply generates an analog electrical signal that varies in strength according to how many photons strike that site. This analog electrical signal then goes into an ADC, and comes out as digital data with a specific bit depth. In the case of the Red One, the ADC outputs 12-bit linear data, which is recorded as… 12-bit linear data. Applying a log curve to image data puts more of your pixel luminance values “in the right place” (it distributes them across the tonal range of the image in a more useful way), so Red’s 12-bit linear doesn’t really provide an advantage over the 10-bit log of cameras like the F23. But the Red One isn’t really at a disadvantage here either. And keep in mind that the Red One is recording raw bayer data. My understanding (and I admit I’m not an expert on de-bayering algorithms) is that when generating an RGB image from a 12-bit bayer pattern, one can actually end up with a bit more than 12 bits of color data.

So, does that initial 14-bit ADC conversion benefit cameras like the F23 and the Genesis in any substantial way? Given that 12 bit encoding allows the Red One to record everything from the sensor’s noise floor (the camera doesn’t artificially clip to black below a certain point like traditional video cameras) up through its sensor-clipped highlights without any noticeable banding in images (even when you push them around in post, within reason), it seems unlikely that a 14-bit ADC would make any noticeable difference.

-Thanks for the correction, though once again you got us again on minutiae, but NOT ON CONTENT. This time you called us out on an issue of wording that doesn't effect our argument at all. If you check our site, we've now revamped the article to not use the word "analog" in the discussion of original quantization of the camera. We are clear now that we're talking about quantization of the camera (not necessarily of the photosites themselves) -- and you may notice that our argument isn't diminished at all when we clean up our wording to conform to your stringent standards. Sorry that we worded it too loosely the first time, but the reality is that F23 and Genesis gather 42-bits of REAL DATA ABOUT THE IMAGE per pixel (before mapping to 10-bit-per channel or 30-bits-per-pixel) whereas the Red gets 12-bits-per pixel. 30-bits represents over a billion unique colors/luminances whereas 12-bits represents 4096. That's a hell of a difference.

-You are dead-wrong when you say that you get "a bit more than 12-bits" out of 12-bits. That is self-evidently absurd, as you've just stated almost verbatim that 12-bits is more than 12-bits. Red may PERCEPTUALLY hide lack of data or PERCEPTUALLY appear to have more data than it actually does, but the lack is still real. You never recorded it. You never got it. It's not real data about the original image that was out there in front of the camera: it's Red's processing smoothing over the output image and making it LOOK LIKE you got more data than you did. Again, we agree that Bayer is optimized for hiding it's lack of data. But the other cameras mentioned have no lack to hide.

-In your last paragraph above, you are comparing F23 and Genesis's "14-bit" to Red's "12-bit." We know that F23 and Genesis map to 10-bit (per channel, not per pixel), so even if we cripple our own argument and say 10-bit instead of 14-bit, here's what we get: F23 and Genesis have 30-bits of REAL DATA per pixel. Red has 12-bits. Remember, in calculating the number of unique color-combinations that the camera can record about the image (the real color depth in number of possible values), bits are exponential (the calculation is 2-to-the-bit-depth). And 2-to-the-30th is over a billion while 2-to-the-12th is just over four thousand -- that's "substantial" if we've ever heard of it. One billion as compared to four thousand!? Not really sure how you arrived at, "unlikely... [to] make any noticeable difference"; that's a hell of a lot more data.

Compression is easy to address. I already posted on this subject a couple of months ago. If you don’t feel like reading that entire post, the short version is that capturing a much larger image, even with a substantially higher compression ratio, is going to preserve more useful image data than capturing a lower resolution image.

This is absurd. You're saying that the higher the resolution and the higher the compression, the better the image, no matter what. You're saying there's no give and take in a trade off: that it just keeps getting better no matter how much information you throw away. Instead of a 4K image in 1 megabyte, why not put an 8K image into just a few bytes? Or better yet, a 12K image into a few bits? The best would be a 100K image in 1-bit. Obviously, it doesn't just keep getting better by tipping the scales. We have discussed the requisite balance in detail and won't do so again here. But, the argument you've made for resolution at the sacrifice of everything no matter what the proportions is obviously untenable (that's why we discussed the actual proportions in our article).

As a really obvious proof-of-concept thought experiment, consider the fact that the data rate of 4K shot in Redcode 28 on the Red One is about the same as the data rate of uncompressed standard definition video. Throw both of them up on a movie screen, or even a moderately sized HDTV, and there’s absolutely no contest. Similarly, a 4K image, recorded at a high compression ratio, can look substantially better than a 2K image recorded uncompressed or at a lower compression ratio.

We are suggesting that the best format is the one that meets and/or exceeds the minimum agreed-upon standards. The example of SD video on a movie screen does not remotely meet agreed-upon standards of resolution. Just as Red camera doesn't meet agreed-upon standards of color depth. F23 and Genesis meet and exceed minimum standards for both. Your "thought experiment" is another very long way of saying that you only care about resolution and not color/luminance information. As always, we disagree: an image is made of both resolution information and color/luminance information.

It won’t necessarily, of course. It depends on the details. But one can’t argue from principles that it will worse. One has to actually evaluate the results side-by-side. Subjectively, Redcode holds up extremely well.

We agree that RedCode holds up extremely well. All we ever said is that It just not BETTER than HDCAMSR or than uncompressed. We have actually looked at both and believe that HDCAMSR and uncompressed are the best we've seen. Remember, we never said that Red is bad: just that it's not true when people say that it's better than F23 and Genesis at getting more information for cinema.

HDCAM SR is a DCT-based compression algorithm, like JPEG. Redcode is wavelet algorithm, like (and, in fact, based on) JPEG 2000. These are well established facts. And the fact that wavelet-based algorithms are more efficient than DCT-based algorithms — and have less objectionable artifacts for for image compression — is not particularly controversial. Wavelet algorithms require more computational power, which is why they’re often not used in real-time systems. But they do provide better results, i.e. higher image quality and/or smaller file size.

HDCAMSR may or may not be "like" JPEG, but it is NOT JPEG -- that's a likening that tells you nothing about HDCAMSR. The few sentences above, on DCT vs. wavelet do not compare the two technologies at all: there are many details of both compression schemes besides those two aspects -- you can't tell a thing about which is better by that comparison. Our paper was mostly about comparing specs rather than subjective results, and we used numerical quantification when possible, but some of what is discussed is only perceptual and not mathematical. Unfortunately, how good a compression scheme looks perceptually cannot be quantified (as you're trying to do here) just by naming one of many technologies that went in to it. Which compression scheme is "better" is mostly perceptual -- and WE happened to think that HDCAMS is better perceptually. As we pointed out, the thing that you CAN quantify is how hard the scheme has to work -- and you say wavelet is more efficient (that's in general, not in the specific case) -- but RedCode had better be more efficient, because Red is a LOT more compressed than HDCAMSR and it has to be a LOT more efficient just to look AS good.

Also, we just don't think Red has really trumped everyone with such a jaw-dropper as you do. If you're trying to say which compression scheme you subjectively think "looks" better -- that's fine. We agree that you think Red looks better (and we think HDCAMSR looks better). You can't say which compression is better by the name of one technology used in it. For example, Sony has a long whitepaper on HDCAMSR that speaks of a lot of fancy technological names besides DCT and it's full of long technical words about why they think it's incredibly efficient and high quality. Red's description of its Redcode on their FAQ is simply that "it's the magic codec." Now, we're not saying this proves anything, we're just showing that you can't prove the perceptual quality of a codec by throwing out fancy technical words in this way.

Johnson misses the mark by the widest margin, perhaps, when discussing raw workflow. He says:

Some people claim that the processing of F23 and Genesis “bakes in” a look that you can’t undo in post-production and therefore Red has more information about the image available in post, but this “baking in” problem is only true if you have bad settings in the camera (like, if you crush blacks down to zero in the camera settings). Actually, when the equipment is operated correctly, you have MORE information about the image in post from F23 and Genesis (which is our whole point through this article), because these cameras record MORE color (and luminance) data about the image — you get more breadth and depth, more range for color-grading from the richer HDCAMSR file. Also, the exact same “baking in” problem applies to Red’s processing software as it does to F23 and Genesis — if you have bad settings, you’ll truncate data. So, if trained professionals are handling the equipment (whether it’s F23, Genesis, Red Camera, or Red Software), then there will be no “baked in” data truncation.

I barely need to write a response to this, do I? The advantages of capturing raw data are, by now, probably quite obvious to the readers of this blog. Johnson tries to set up an equivalency here that doesn’t exist (”if you have bad settings, you’ll truncate data”). There are two problems with this. The first is that a camera recording the output of a 14-bit analog-to-digital converter to a 10-bit tape format will always throw away data. If you have everything set up correctly this shouldn’t be a big deal, but it does remain a mathematical fact.

All due respect Chris, but no, actually, this is where you've missed by the widest margin. Raw is indeed a good way for Red to design their camera: it's an advantage that Red has over an imaginary Red camera that's designed a different way, but it's not an advantage of Red over HDCAMSR. Raw is a description of WHEN YOU DEBAYER. Bayer is a subsampled kind of sensor. F23 and Genesis don's subsample, so there's no question about WHEN to debayer.

As to 14-bit to 10-bit, you're only mischaracterizing our argument, not refuting it. Firstly, we explain clearly that the F23 and Genesis are 14-bit to 10-bit in a very transparent way and then later we say that no data is truncated FROM THE WAY WE DESCRIBED IT. Oversampling is not truncation. The spec -- the number we're holding everything up to -- is 10-bit per channel (30-bit-per-pixel) -- so 14-bit is just oversampling, not truncation.

More important, though, the bit-depth doesn't even address the issue at hand in the paragraph that you're quoting. We're talking about whether or not you truncate color/luminance information at the top and bottom of the image (like crushing a dark part of the image all the way down to black or boosting the bright part to 100% white). In our paragraph that you've quoted, when we speak of truncation, we're not speaking about bit depth. Bit-depth does not describe whether you're recording a an image that's crushed and clipped or an image that's not crushed and clipped -- it just tells you how many bits you're using to record that image. You could have a 4-bit image that's not truncated (not crushed and clipped) -- it would be missing information for OTHER REASONS, but it wouldn't be chopping off information in the way that crushing and clipping do.

Secondly, there’s a very large difference between baking in what might turn out to be the wrong data on set, and baking it in using Red’s post tools. In latter case, fixing things requires reprocessing some files. In the former case, it requires a reshoot. Dismissing this by saying the article is “about technical quality of motion imaging, not time and place of motion imaging”, as Johnson does in the next paragraph, is simply bizarre. This is a factor which, in the real world, can have substantial impact on the technical quality of motion imaging.

I guess you're saying here that the camera is going to be setup in prep and/or operated in a completely incompetent way, and that with Red camera, the incompetent operator gets 2 chances to screw up, and in the second chance, he can re-do when he gets it wrong. We're talking about professional equipment operated by professionals. When operated correctly, the F23 and Genesis get MORE information than RedCode -- that's real information about the image -- real information to use in post.

That's information that you can even use to "fix" mistakes made by your hypothetical incompetent operators -- like misexposure, which can be fixed easier if MORE color/luminance information is available to you. More color/luma information about the image (which F23 and Genesis get) gives more choices in post, not fewer choices. You talk about re-shoots: using such a compressed and subsampled image as Red supplies means that a greatly misexposed shot from F23 or Genesis is MUCH more salvageable in post than one from Red would be. As is the whole point of our article: F23 and Genesis get more actual color/luma information (which comes along to post) than Red does about the image.

Dismissing this by saying the article is “about technical quality of motion imaging, not time and place of motion imaging”, as Johnson does in the next paragraph, is simply bizarre. This is a factor which, in the real world, can have substantial impact on the technical quality of motion imaging.

"Sony has been able to build image processing software and native hardware that can fit inside the camera and do full-quality lossless processing in real time. Red didn’t build it on-board and they can’t do it in real time. That’s not an advantage for Red. Red’s image processing software has to work much harder than Sony’s because the camera only gathers one piece of data per pixel instead of three, and the processing software has to work hard to turn that subsampled data into an intelligible color image."

It is true that in some workflows, it’s beneficial to have a high-quality video image available for playback immediately, rather than after substantial processing time. Fortunately, you can do this today with Red footage via SCRATCH CINE. Sometime next year, you should, according to current announcements anyway, be able to do it with the dedicated hardware in the RedRay player, at up to 4K.

Once again, you're refuting our argument only by misconstruing it. Our article never calls for on-set playback of the full-quality image (which we agree is of negligible advantage). We are just saying that it's NOT AN ADVANTAGE of Red that you have to wait to render images in post. Obviously, having to wait is not an advantage. And, for professional jobs where the post house charges money (as opposed to home-made jobs where your own time may be free of charge), render time is expensive. F23 and Genesis get all the information and there's no render time in post.

We need to keep things in perspective. The cameras Johnson is comparing the Red One to cost ten times as much. Or more. The fact that the Red One is even being compared to them shows just how much Red has achieved. The fact that the Red One wins some of those comparisons is absolutely stunning. The Red One’s more significant feature, when all is said and done, is that it puts a high-end imaging tool — a tool which shoots at sufficient quality that its footage is suitable for any deliverable, right up through theatrical projection — into the hands of thousands of people who would otherwise never have access to such equipment.

We absolutely agree with you here. Our article says nothing to the contrary, and actually says some similar things to what you're saying here. We do indeed think Red is a fantastic camera for the price. We agree that it's very impressive that something so inexpensive is even being compared to higher end cameras. Our article doesn't say Red is bad -- it only says that some people are mis-representing technical data to try to show (incorrectly) that there is technical evidence that Red is better. If you say, "Red is great -- very good for the price -- amazingly close to high-end cameras for something so cheap," then we agree with you. If you say, "hey, I personally think Red LOOKS better to my taste than Genesis even though some its specs are lower than Genesis," then we can't disagree that you think so. The only thing we're refuting is the phenomenon where people try to quote specs to show that Red is a technically better CINEMA camera. F23 and Genesis see and record MORE INFORMATION ABOUT THE IMAGE THAT IS IMPORTANT FOR THEATRICAL CINEMA.