Red - it's grotesquely incompetent

[Thomas Worth]

the results still aren't surprising considering the RED uses a Bayer-filtered 4K sensor, and the Canon a 12K sensor (~12K / 3 for R,G,B = ~4K).

I just read this and realize that it's confusing. What I meant was "12 megapixel," not 12K. With 12 mp, you can get just under 2.7K of real RGB if you dedicate 4,000,000 pixels to each color channel:

 

12,000,000 / 3 = 4,000,000

 

4,000,000 / 2,670 = 1,498

 

2670x1498 is roughly 1.78:1 (the same ratio as RED's CMOS)

 

This isn't how Bayer does it (Bayer subsamples red and blue), but interesting nonetheless.

[Sam Wells]

Are you saying the Canon 5D does NOT use a Bayer stripe ? I think it does.

 

-Sam

[Thomas Worth]

Are you saying the Canon 5D does NOT use a Bayer stripe ? I think it does.

Oh, I'm pretty sure it's a Bayer pattern, like the rest of Canon's DSLRs. I'm just saying if you had 12 million pixels and wanted true RGB out of it (as opposed to subsampled chroma), the most you'd get out of it is about 2.7K. I believe the Panavision Genesis does this (and not surprisingly, outputs only a 1920h image).

 

Bayer favors luminance over chroma, so there is a little more "real" resolution at the expense of color accuracy.

[John Sprung]

I believe the Panavision Genesis does this (and not surprisingly, outputs only a 1920h image).

 

Bayer favors luminance over chroma, so there is a little more "real" resolution at the expense of color accuracy.

Genesis isn't Bayer. The chip has 5760 photosites across by 2160 vertically. I know that the RGB filtration is in the form of vertical stripes. That gives them 5760/3 = 1920 across. I've also heard that they have alternating high and low sensitivity rows to increase dynamic range, which means 2160/2 = 1080 sets vertically. Clearly this is optimized for 1920 x 1080 HDTV.

 

As to color accuracy, it would be spatial rather than chromatic that would suffer. Given limited bandwidth/bitrates, there's an excellent reason to trade off some color resolution for more brightness resolution. The human eye/brain combination resolves luma better than chroma, and actually "corrects" small spatial errors in color to match bright edges. It's only a problem when you want to pull a chroma key.

 

 

 

-- J.S.

[Thomas Worth]

Genesis isn't Bayer.

Yes, I meant that the Genesis' chip uses an equal number of photosites for RGB (from what I've read), in contrast to Bayer filtered sensors. I agree with everything else you said.

[Luke Haywood]

In short it's a very cheap, very nasty cellphone type CMOS camera that happens to go at 24fps.

 

 

Phil

 

Oh come on Phil, it's not THAT bad!

 

OK the RED lenses are somewhat of Handycam quality, and most of the RED fan base clearly have no experience of anything bUT Handycams, and RED need to learn some lessons about the ruggedness requirements for on-set equipment and so on, but it's still a damned good body for the price.

 

To me the RED's biggest flaw is not making any provision for live 1920 x 1080 output. There should be a connector of some sort where the raw (as in direct off-chip) data could be taken off and processsed by **third-party** software and hardware, and recorded on industry standard formats. This Redcode nonsense smacks far too much of the software house "tail" trying to wag the industry "dog". There are plenty of other JPEG2000 re-jigs out there waiting int he wings.

 

I also think the strain might be starting to show on Jim Jannard as the realities of mass production start to sink in.

 

It can't have been a very easy year for him. First he lost control of the corporate empire he founded with his bare hands all those ago, and then watched as it was immediately sold out from under him to their biggest competitor.

Edited February 23, 2008 by Luke Haywood

[Phil Rhodes]

Hi,

 

Actually I have to disagree that it's a good body. For a start, it's a hugely unfinished body - presumably this will change, but I'd say they're actually being very slow on that. Too much effort put into rushing the thing out the door so they can crow that it's finished in record time, when in fact it's nothing like.

 

But the pictures are just not nice. It is, indeed, like a very high res cellphone camera. It's all resolution and nothing else - and, as we've seen, not even that much resolution.

 

Now it is very low noise, or that could just be the codec taking it out, but it appears to be very low noise. I suspect that with quite a bit of underexposure it could be made to look OK, but I don't think that many of the people who are currently using it are really capable of figuring that out.

 

Phil

[Nate Downes]

Oh, I'm pretty sure it's a Bayer pattern, like the rest of Canon's DSLRs. I'm just saying if you had 12 million pixels and wanted true RGB out of it (as opposed to subsampled chroma), the most you'd get out of it is about 2.7K. I believe the Panavision Genesis does this (and not surprisingly, outputs only a 1920h image).

 

Bayer favors luminance over chroma, so there is a little more "real" resolution at the expense of color accuracy.

Actually one unconfirmed rumor I'd heard is that Canon is supposedly not using a Bayer with their newest cameras. What is known is that they have licensed some of the Foveon technology and that combined with the rumors I keep hearing suggest to me that they have been slowly introducing it into their top of the line DSLR's It's an interesting move, but a necessary one to get the data transfer rate they needed to support the Mark III and beyond.

[Luke Haywood]

Hi,

 

I suspect that with quite a bit of underexposure it could be made to look OK, but I don't think that many of the people who are currently using it are really capable of figuring that out.

 

Phil

But it only has 12 bit analog-to-digital converters on the chip. Unless there is some sort of analog gamma compression before the digitization process (which I don't believe there is), if you stop the iris down to preserve the highlights, every stop knocks another bit off the ADC output for the bulk of the signal. To get four more stops of overhead would effectively turn it into an 8-bit ADC. You'd then be trying to capture an 11 stop dynamic range with an 8-bit ADC!

 

This particular technological brick wall has been around for many years.

 

One of the main reasons the film/telecine system is still popular is that film capture pre-compresses the dynamic range down to something that an electronic imager can cope with.

 

OK I don't think much of the pissy little mini-BNC connectors and other bits that are sure to snap off at the most undesirable time, and some of the other ergonomic features appear to have been designed by people with limited experience of both professional film and video cameras, but given time, I'm sure this camera will mature into a worthwhile tool.

[Thomas Worth]

What is known is that they have licensed some of the Foveon technology and that combined with the rumors I keep hearing suggest to me that they have been slowly introducing it into their top of the line DSLR's

Wow, I just looked this up and found this page:

 

http://www.ddisoftware.com/reviews/sd9-v-bayer/

 

This is some interesting stuff!

[Max Jacoby]

Actually one unconfirmed rumor I'd heard is that Canon is supposedly not using a Bayer with their newest cameras. What is known is that they have licensed some of the Foveon technology and that combined with the rumors I keep hearing suggest to me that they have been slowly introducing it into their top of the line DSLR's It's an interesting move, but a necessary one to get the data transfer rate they needed to support the Mark III and beyond.

I have a feeling that film camera manufacturers are also looking at options beyond CCD and CMOS because those 2 chips designs have such inherent limitations, among others the fact that resolution is so heavily dependent on the color information present.

[Phil Rhodes]

> But it only has 12 bit analog-to-digital converters on the chip.

 

Y...yes, that's why it's fatally flawed.

 

Apart from nothing else, excepting everything, overlooking the huge compression, the low effective resolution, the unfinishedness... apart from all that, it quite simply doesn't produce very nice pictures. I've avoided saying it like that because it's inherently a matter of opinion, but I think it's well enough supported by the obviously fairly low dynamic range. Even if they are capable of really finishing it, I don't think it'll ever be a truly good camera.

 

P

[Sam Wells]

I'm sure the Canon MK III's are not using Foveon sensors. We'd sure know about it.

 

I've seen some exceptionally nice images from the Foveon Sigmas, and some not-so-nice also. I _am_ curious about their forthcoming DP-1 "point & shoot" with RAW capability ---- Well it could be a way to play without big $ commitment...

 

-Sam

[Luke Haywood]

> But it only has 12 bit analog-to-digital converters on the chip.

 

Y...yes, that's why it's fatally flawed.

 

Apart from nothing else, excepting everything, overlooking the huge compression, the low effective resolution, the unfinishedness... apart from all that, it quite simply doesn't produce very nice pictures. I've avoided saying it like that because it's inherently a matter of opinion, but I think it's well enough supported by the obviously fairly low dynamic range. Even if they are capable of really finishing it, I don't think it'll ever be a truly good camera.

 

P

Well all current model video cameras has got that limitation. Only the Dalsa Origin has 16 bit conversion, but the first four bits appear to be quantitized noise! Poor, but honest I guess!

 

As far as noise goes, we don't never get the chance to see what the true noise performance of the RED is, because the digital data goes straight off the chip into Wavelet, so the RED RAW books are always going to be at least slightly cooked.

 

RED say they don't use noise reduction, but is that the same as just not recording the noise, since recording noise with wavelets requires extra effort.

 

I have to say though, I have seen some very nice pictures taken with the RED. I wish my cellphone camera was that good!

 

But I guess they aren't going to post any awful ones, are they? :lol:

Edited February 23, 2008 by Luke Haywood

[Chris Kenny]

OK, I haven't posted here for months precisely because this forum is so nuts, but the sheer amount of nonsense and outright disinformation in this thread is just too much for me. All of this Bayer hatred doesn't make sense.

 

- The Foveon vs. simulated Bayer comparisons linked above are useful for gaining an understanding of how the technologies work, but have essentially no real-world value. A sensor technology that gives you full RGB at every photo site, but with 1/3 of the number of photo sites (which appears to be what Foveon is in the real world) is not a clear winner at all. A 4 megapixel Foveon camera vs. a 10 or 12 megapixel Bayer camera should deliver higher chroma resolution, but lower luma resolution for typical subjects. This is a very poor tradeoff for non-specialty imaging (i.e. not chroma key), given the characteristics of the human visual system.

 

- The "Is it really 4K?" posts in this thread almost universally ignore the fact that "Is it really [native sensor resolution]?" questions can be asked about any camera, not just Bayer-sensor cameras, and the answer is virtually always "no". Even with a 3-chip system, Nyquist sampling issues and the need for optical low-pass filtering to avoid aliasing prevent the output image from actually having the full resolution of the sensors. The only way to get around this is substantial oversampling. I've seen tests that show 1800+ lines/picture height for Red, which would imply it's a 3.2K camera. Of course with a Bayer sensor it depends on the subject, but this is a lot of resolution, and in real-world shooting conditions, Red is going to produce an image with substantially more resolution than will fit into a 1080p image, and it's certainly going to produce an image with substantially more resolution than a 3-chip camera with 1920x1080 sensors.

 

- Comparing a large-sensor Bayer camera to three-chip systems is largely meaningless anyway, if you value having a 35mm-format camera. Nobody seems to think it's practical to make 3-chip systems with sensors that large, and even if you could make one, it would probably need special lenses.

 

- The 4K -> 2K -> 4K vs. original 4K image that has been discussed in this thread, which supposedly demonstrates that Red is really 2K or less, actually doesn't demonstrate that, for at least two reasons. The first is simply that it's far from the sharpest Red footage I've see. The second is that a 2K image created by downsampling a higher resolution image will have more detail than a 2K image shot at 2K, for various reasons discussed above. In other words, even if the images did look very similar in a well-conducted test, that still wouldn't demonstrate that Red was producing an image that might actually be produced by a 2K camera.

 

The most obvious way to get more resolution into 35mm-format digital cinematography cameras, at this point, is simply to wait for higher resolution Bayer (or striped) single sensor cameras to emerge. I strongly suspect we'll get something that basically amounts to "real 4K" through this mechanism long before we get it with a Foveon or 3-chip system.

[Gunleik Groven]

I must admit I find these threads quite entertaining....

 

Gunleik

[Phil Rhodes]

You're overlooking a couple of things there, not least among which is the need for vastly more aggressive OLPF on Bayer arrays to alleviate the aliasing inherent to such a poor effective fill factor. Red does exhibit these problems, from what I've seen. This sort of filtering makes a mockery of the higher luma and even chroma resolution that's so frequently claimed, even when you start getting very clever.

 

I'm not particularly trying to claim better results for 3-chip systems; I would expect a Red to outresolve something like an F23 or Viper, not that it's nearly as nice as them in any other way.

 

I'm equally sure that Red and co. are sitting back in full knowledge of the shortcomings of their product and the claims they make about it and watching the cash mountain pile up. Why should they care - as I mentioned in another thread, nobody's distracted by facts these days.

 

P

[Chris Kenny]

I'm equally sure that Red and co. are sitting back in full knowledge of the shortcomings of their product and the claims they make about it and watching the cash mountain pile up. Why should they care - as I mentioned in another thread, nobody's distracted by facts these days.

 

A $17.5K camera that's even remotely in the same league as an F23 is something that people who can't afford a $150K camera can justifiably get excited about, even if they're in full possession of the all facts. A $17.5K camera that's at least arguably better for some use cases is even more amazing. Now, I understand the price doesn't matter much to people who work as freelance DPs, using other people's equipment, but for individuals or companies operating on some other models, this is a really big deal. People aren't, for the most part, getting excited because they've been duped. They're getting excited because Red, despite its flaws, is vastly superior to the tools they'd otherwise have access to.

[Phil Rhodes]

Yeah, but it's:

 

a) Shitty

B) Much more expensive to use than anyone seems to realise

c) Built by some quite unpleasant people

 

This is enough to put me off. You of course are free to do what you like.

 

And no, it's not in the same league as an F23

[Walter Graff]

nobody's distracted by facts these days.

 

 

Are you saying RED is the Barack Hussein Obama of cameras? :)

[Walter Graff]

" A $17.5K camera that's at least arguably better for some use cases is even more amazing."

 

I totally agree but can we be honest, you can't take this camera out for less than $30 to 40k so it's not a $17.5k camera. Sort of like saying a car only cost $800 because that's how much the tires cost.

[Gunleik Groven]

" A $17.5K camera that's at least arguably better for some use cases is even more amazing."

 

I totally agree but can we be honest, you can't take this camera out for less than $30 to 40k so it's not a $17.5k camera. Sort of like saying a car only cost $800 because that's how much the tires cost.

 

 

Sorta, but that goes for all cameras. One price for the body, and then you add whatever you want.

 

If you really need just "an image" out of it, you could have it rolling for 19200 with a crappy 2nd hand ebay lense, cf module, 1 card and, monitoring on a cheapo 24" lcd and hard-wired to electricity adapter.

 

I don't suggest that this is a good solution, I just say it's possible.

 

Last time I checked, that's less than one of the Pro35 adapters people use on their HDX-900 or F-900's aroud here. (Norway)

Incidentially it's substantially less than those cams costs as well.

 

I also find the "is it really 4k" discussion quite fun. It seems most agree that the luma rez = quite high, and the chroma rez is in the 2,5-3,5 range @ 12 bit.

 

Most/all DI is done @2k from 10-bit DPX's here (in Norway).

This gives a nice startingpoint for a standard film post-workflow.

 

I've seen filmouts done in 2k/4k done by highly Red Critical shops, and what we saw (both on film-outs and digitally projectet) left no room fr doubt on the subject that 4k from Red @ least has "much more" rez than 2k.

 

The exact numbers are for others to debate, but the difference was tere for all who wanted to see.

 

 

Monitoring the image is happening in 720p from a quick and dirty debayer and placed in a REC 709 colorspace, and this solution has it weaknesses, as the image you monitor is quite different from the image you record.

 

The RED images are low contrast low saturation and need a one-light like process in RedCine to stand up through post.

If You screw up in this process, you also lose most of the image and the post possibilities.

As all image-modifications in RedCine is done by applying metadata, this is a lossless process, but making decissions bases on a 12 bit unprocessed image when watched as 8-bit jpegs on a computermonitor, takes away most of the fun.

 

To me the real fun of Red happens in post. To make any qualified judgement on the images, without having RedCine/one light as part of the equation - is rather silly.

(I don't say you'll like the images then, but it's sorta counterproductive to talk of the camera, without looking @ the workflow)

 

I've seen tons of images that have looked like they've been screwed up in the development phase, and when I had my first test with the cam, I tried to emulate as many of the problems I've seen as possible, to try to figure out what has happened in-camera and what has happened in post.

 

From what I've seen, quite a lot of crap happens in post.

 

The most obvious problem is matching the ISO/exposure data in one-light with the on-camera settings when shooting.

This effectly diminishes the DR, adds noise and makes the images blow early.

 

The on-camera iso setting is best viewed as "preview gain"

The ISO/exposure settings in RedCine, should usually be set @ 160-320 ASA, and not be pushed beyond that if you don't really intend to grain your image and lower DR.

 

So:If you judge the image from the preview contrasty image - I see all of your points.

 

If you've had some fun with Red images in post...

 

Let's say:

I think these images are quite useable.

 

Gunleik

[Radoslav Karapetkov]

Had the Romans let Christ live, the rest of the world would have never heard of him and we would have been spared. It was by making him a martyr that the LEGEND was born.

 

Actually, the Romans did leave Christ to live.

[Walter Graff]

<br />Actually, the Romans <i>did</i> leave Christ to live.<br /><br />

<br /><br /><br />

 

Too many folks watched The Passion of the Christ and believed it as Gospel. It was just a Road Warriors view of the death of Christ.

[Radoslav Karapetkov]

<br /><br /><br />

 

Too many folks watched The Passion of the Christ and believed it as Gospel. It was just a Road Warriors view of the death of Christ.

 

(sorry for the continued off-topic).

 

Well, all the "evidence" we have on this actually is in the Gospels. The film is based on these scriptures.

 

But let's look at things from a different POV. Why would the Romans want to kill a carpenter who preaches eternal life? The guy wasn't instigating any political or military form of anti-Roman revolt and thus wasn't confronting Roman imperial interest. Nevertheless it's quite likely that he was confronting some other interests...

 

Of course all this is speculation, because it's based on the Gospels, which are human account, and thus requires the brackets around evidence.

 

(I propose to close the Christ off-topic here).

 

...

 

But I'd wanna ask something else.

 

It's obvious that the development of the RED Camera has cost enormous investment - money, time, energy, commitment.

 

But in the end, the creators of RED are offering their camera for 17 500 $ [25-30 000 $ for working kit].

 

This is MIGHTY cheap.

 

Even with 3000+ reservations, the profit from such a huge undertaking is questionable with these prices.

 

(Although I hope they'll have more reservations\profit in the future).

 

So my question is simple:

 

Why would they wanna do that?