why 5k? go after uncompressed instead please

[Brook Willard]

Well at ~415MB/s uncompressed RAW 5K at 24 frames per second, you're looking at roughly 24.3GB/min or 1,459GB per hour. A 20,000 foot day would be roughly 5.22TB. Say you're talking about a million feet on a feature, you're up to about 262TB of storage.

Unless my math is way off... which it very well may be.

 

[reality]

For 3 copies, you'd be looking at 700TB or 900 LTO4 tapes.
900 tapes would cost $70,000 or $24k for 1 copy.

The equivalent RED feature would cost $3k in tapes for 1 copy/1Mfeet.

That not including operator and equipment time to backup all that data which would change the 70k to 150-200k through a post house.

 

[Bing Bailey]

I think less compression options rather than uncompressed. it wasn't so long ago everyone was happy with hvx 200 and that footage is compressed like crazy. are we getting spoiled or what.

redcode raw in whatever improved permutations with a better sensor is the way to go and will probably solve even the most difficult issues

 

[Vincent Thomas]

I don't know what is the position or experience of people here, but as someone working as both a photographer and in post production FX, i'm not especially attracted by being able to shoot in 5K.

More, i dont even think it's quite possible to handle for most places!
For example i'm leading the FX of the first visual sequence of this last Terry Gilliam's movie ( http://www.imdb.com/title/tt1054606/ ), it's a 10 minutes sequence, the movie is going to be deliver in 2k only and the server for the image is closed to 30 terra gigabytes already... So make the calcul by yourself, not to mention that working with 5k image is more difficult than in 2k off course

Now, i'm just wondering.. Is it reasonable to imagine Red make a larger sensor than 35mm or/and with black silicon? It would be the bomb indeed!

vincent*




------------------------------------------------
Artist / Photograph / Author & Director
http://www.linkedin.com/in/vincentthomas
http://blog.myspace.com/bout_de_lune

 

[Sidney L. Plaut]

I am with Brook,

I kindda fear uncompressed too - if its there i have to use it.. goodbye finishing at home... hello posthouses


BUT that said, if it can be done ...do it. As Frederik said the compression is an issue, as compression always is.

That said, that said, Graeme has done a fantastic job on the redcode and i am sure redcode100 will be even more fantastic

 

[Stuart English]

5K

5K

I don't know what is the position or experience of people here, but as someone working as both a photographer and in post production FX, i'm not especially attracted by being able to shoot in 5K.

More, i dont even think it's quite possible to handle for most places!

The point of shooting 5K RAW though, is to generate a high MTF ("full resolution") 4K DPX digital intermediate (or other file type)

Similar principle as scanning film at 3K for a 2K DI or 6K for a 4K DI.

So it shouldn't be any extra burden on post production / VFX.

 

[I Bloom]

Just to weigh in:

COMPRESSION IS GOOD

Try shooting on the Phantom a few times and you'll realize how problematic uncompressed workflow can be. And that's just 2K.

If you preserve bit depth, then a higher resolution more highly compressed image will always look better than an uncompressed, low resolution image that takes up the same amount of storage space. Compression gives you more. Not less.

IBloom

 

[Gunleik Groven]

Compression to a certain extent is good. And then it turns bad...

Still remembering the milkgirls, I think they are technically the best RED images I've ever seen - and they were recorded tettered to a fridge...

Not that I want that fridge, but RED ONE @ 100 MB/sec would have been fun to see...

For a long while I guessed that we'd see a 72 or 84 MB/sec mode, untill it was stated that RED internally cannot move more than 36 MB/sec

Not that I am in any way unhappy with the current state of things, just rememberig the milk girls...

 

[Adam Glick]

I would be interested to know how many VFX supervisors or editors have actually stated a desire or need for more data (less compression artifacts) in their R3D plates in order to pull off a tricky shot in post.

My bet is that except for the most demanding effects shots, RC36 is probably "good enough".

 

[Rudi Herbert]

Still remembering the milkgirls, I think they are technically the best RED images I've ever seen - and they were recorded tettered to a fridge... Not that I am in any way unhappy with the current state of things, just rememberig the milk girls...

Yes sir, indeed. That footage was simply astonishing, and what really sold me on the idea that RED was "quad HD", so much cleaner it looked than anything else. Anything since has looked great, but never had that intangible feel the milk girls had. Which is why, however adequate current onboard compression is, I was left with the desire of an uncompressed output option, which could be tethered to the proverbial refrigerator for those anal enough, or by fibre cable to a computer on set capable of recording direct to drives at 100, 150 or 200 mb/s, which are still quite doable. That, would have been amazing, and perhaps not all that difficult to do...

 

[Michael Lindsay]

Cintel used to have a white paper on there website that claimed (if I remember correctly) the advantages of compressed very high resolution images over lower resolution uncompressed images.

Anyone got a copy?

Michael

 

[Dan Hudgins]

The times...

The times...

I would be interested to know how many VFX supervisors or editors have actually stated a desire or need for more data (less compression artifacts) in their R3D plates in order to pull off a tricky shot in post.

My bet is that except for the most demanding effects shots, RC36 is probably "good enough".

I would like to see 24fps uncompressed (or at least single frame mode uncompressed) because the compression/redcode burns in a "look" to the images due to the filtering required in the de-Bayer used to overcome issues caused by the compression.

People are confusing the use of compression on graded images at video gamma or DPX for the film recorder and such vs. the use of compression for the sensor data.

Files of Sensor data needs to hold all signal so that later you can process the linear sensor data with S and other curves to get a film-like result that has higher contrast in the midtones and less contrast in the highlights and shadows (like in a 4th generation print).

If you use compression that was intended for use on final graded images on the sensor data the artifacts of the compression degrade the result. Although if you light the well for the RED ONE (tm) you should be able to get good results, if you had the full sensor data you could get better results under more real world conditions where the lighting may not be optimal. Even under optimal conditions the chroma resolution could be higher if you de-Bayer true raw data rather than compressed data.

The workflow issue is also not just black or white.

If you keep uncompressed true raw sensor data packed using lossless "compression" (relative rather than absolute pixel values) you can store the sensor data without issues that are beond what a flash or DRAM module can hold, like the A-cam dII, those files can be dumped to cheep harddrives for later de-Bayer.

You do not need to stay uncompressed through the whole workflow since after grading the compression artifacts would have less impact on the final viewing.

Then you need to ask, in the theatre can you see any difference, I would say that if you compress before grading you might, if you compress after grading then it would matter if double compression artifacts will amplify the compression so that it is noticeable.

Once film prints are no longer made in a few years, there will probably no longer be any way to view an uncompressed image in a public avenue, so since all images viewed will be degraded by lossy compression, how much can the source images be degraded so as to not be greater than the distribution compression degradation.

If you do compress for archive, then all future versions will show double compression artifacts. It is common now to see archive footage that has been transfered to some low bit rate compression used in TV shows so that very bad double compression artifacts take away much of the "brilliance" of the original film/video and make squirmy muck.

For a camera to make 4K images you need real brightness values for the pixels. In fact if you make a compressed camera with OLPF the brightness of the pixels is off, edges move off position and get soft, so maybe 25% or less(?) of the pixels are close to 10bit data equal to true raw bayer but the others are less close in brightness and color. To say it is easer to compress 4K images [because the viewer only sees 900lines on the screen], seems odd to me. To make something that is 4K in all parts of the image all the time may not be a need, but if it is not a goal, then 4K is just a pixel count and not an image property.

I would not say that compressed 5K Bayer images would always look better than uncompressed 4K Bayer images. Maybe the compression would blur the sensor noise in some parts of the image, and make tone banding and noise blobs in the other parts, someone might say that looks better, but it is burned into your images since you cannot trun the compresson off.

If you record uncompressed true raw sensor data you can run the images through filters later to reduce noise if you like, and since the result is not burned into the images you can make adjustments and re-do the filtering if you did not like the result, if you compress before de-Bayer you are stuck with burned in artifacts.

I have been working on an uncompressed DI system option for post, it is not for everyone, but it would be good if uncompressed sensor data was available for those who would like to use it to help insure the best result under difficult lighting.

How bad good enough can get is a slippery slope, 70mm prints were being made from bad 35mm dupe negatives, so people did not see how good 65mm/70mm was when done right, the 70mm was just a way to market "bad" films, not so much a way to improve the image on the screen as much as it might have been. Will 4K projection of low grade compressed and re-compressed re-resed source images be just another 70mm fan fare for marketing?

 

[Antoine Baumann]

I would be interested to know how many VFX supervisors or editors have actually stated a desire or need for more data (less compression artifacts) in their R3D plates in order to pull off a tricky shot in post.

not a supervisor, but compositor, and YES I desire less compression on future RED's cameras.

Thanks,
antoine.

 

[Stuart English]

Compression

Compression

If you use compression that was intended for use on final graded images on the sensor data the artifacts of the compression degrade the result.

If you do compress for archive, then all future versions will show double compression artifacts. It is common now to see archive footage that has been transfered to some low bit rate compression used in TV shows so that very bad double compression artifacts take away much of the "brilliance" of the original film/video and make squirmy muck.

To clarify - REDCODE RAW is optimized to compress sensor data, and its also the suggested archive format.

 

[M Most]

To clarify - REDCODE RAW is optimized to compress sensor data, and its also the suggested archive format.

It would be a much better archive format if Red would release tools to allow for trimmed R3d files.

 

[doondoon]

You had me at 2K : )

You had me at 2K : )

I've recently seen uncompressed footage shot on the Arri D-21 and the VIPER... captured both the the Codex box and the S2.

Neither one of these cameras shoot at 4K (they're both lower resolution than the RED). Nonetheless, the images I saw projected were way beyond RED. So I'm not sure why everyone is so excited about 1K more of resolution?

When I think of many HD feature films I've seen on the big screen I've NEVER been bothered by the resolution (even at 1920 X 1080). Im usually bothered by the noise, or the plastic skin look, or the limited dynamic range!
MORE resolution is not going to fix these problems, more bandwidth will.

My list in order of importance.

1. Latitude
2. Bandwidth
3. Resolution

 

[reality]

I've recently seen uncompressed footage shot on the Arri D-21 and the VIPER... captured both the the Codex box and the S2.

Ha. Every side by side I've seen show Red as superior.

 

[Graeme Nattress]

Every dpx I've seen from a Viper is riddled with artifacts...

Graeme

 

[Charles Angus]

I've recently seen uncompressed footage shot on the Arri D-21 and the VIPER... captured both the the Codex box and the S2.

Neither one of these cameras shoot at 4K (they're both lower resolution than the RED). Nonetheless, the images I saw projected were way beyond RED. So I'm not sure why everyone is so excited about 1K more of resolution?

When I think of many HD feature films I've seen on the big screen I've NEVER been bothered by the resolution (even at 1920 X 1080). Im usually bothered by the noise, or the plastic skin look, or the limited dynamic range!
MORE resolution is not going to fix these problems, more bandwidth will.

My list in order of importance.

1. Latitude
2. Bandwidth

3. Resolution

I would just throw out there one more time that, to a certain degree, resolution is latitude for a digital sensor. More resolution means less noise, and more useable space in the shadows.

 

[Dan Hudgins]

Film like vs. Video like... Film is RAW video is signal processed...

Film like vs. Video like... Film is RAW video is signal processed...

My list in order of importance.

1. Latitude
2. Bandwidth
3. Resolution

1A) Film has more Latitude, or can when processed and printed film to film, sometime film is clipped when scanned or graded on a DI, but if you pull process Film you can get more highlight range in negatives.

2A) Film has variable Bandwidth with exposure so hides some blemishes in light or dark skin tones. Midtones have the most information, especially when film is printed: camera neg -> master pos -> dupe neg -> print pos, the curves are upsidedown every other generation, that blurs and grains the highlights and shadows so makes the images look smoother while still having fine detail in the target areas exposed for mid-tones.

3A) Resolution of film is distributed over several frames when projected from film-to-film all uncompressed. Although compressed Digital Images may read higher on test charts than single frame film images, when viewed on a screen moving film images from film printed to film, not scanned to a DI, can resolve small details at lower contrast without aliasing artifacts.

In a movie theatre the film projector might produce a image on the screen that has 1280x720 at 40%MTF (the "pixels" are not fixed in place but relate to the spot size on the screen). To reduce the aliasing to the same level from a digital projection (fixed pixel addresses) at that resolution you might need a Digital camera with 2.5x or 3200x1800 or about what the RED ONE (tm) is now putting out with its OLPF limiting the low contrast resolution.

It would be hard at this point for the RED company to switch to lossless true RAW recording, but it would be wise maybe to offer it as an option to provide competitive results since other companies will offer lossless compression of the sensor data.

As for the bandwidth issues, if the RED ONE is 10:1 compression and Lossless compression can reach 2.5:1 with the same OLPF blur (OLPF reduces the pixe to pixel delta values), then you would just need 4 CF cards rather than one to record on, and I do not see that as a big issue given the improvement in image quality that true RAW Bayer data could give, and the benefits to long term archive with an open RAW file type.

If you archive as true RAW Bayer data in a open documented file format, then anyone in the future can open and process the true RAW data to pull out more detail using better software in the future, you can also trim the shots since each frame can be recovered to full "uncompressed lossless" true data and edited without any loss then recompressed with lossless compression for archive.

As things stand, the R3D files have burned in losses from their compression, you cannot open the files then re-compress them without additional losses. Since the R3D filetype is not open, in the future after PCs are history and RED is gone, you will not be able to open the R3D files with new software you write yourself since you will not have the open documentation on the details of the files bits.

With true uncompressed RAW data there are no "builds" or revisions to the file type, that data format of the files can be fixed in something simpler than DNG, something that anyone in the future can understand and get fine images from. Also the quality of the images from true RAW Bayer data is dependent only on the sensor and exposure, not the proprietary de-Bayer program that mates with some compressed format, since true RAW data can be formated to be de-Bayered with any type of software you can use now or in the future.

The RED ONE is a unique device in that it compresses the image before de-Bayer with lossy compression, most of the other cameras that record compressed do the de-Bayer, gamma, white balance, then lossy compression, that way the shadow detail can be pulled up to overcome the "compression noise" that the compression causes in shadow areas. The eye is not linear and most compression trys to use the eyes weaknesses to reduce bandwidth in the dark areas. Since the RED ONE trys to record linear data compressed, it would seem that too much information is recorded for the bright areas and too little for the shadow areas, although what exactly is being done is a "trade secret".

You can minipulate images more that are recorded uncompressed true RAW sensor data to compensate for exposure issues.

So the big questions is: does the RED company want to lock their future to RED CODE while other players use newer and faster memory to record true RAW, or do they want to change coarse and stay competitive.

In a few years the extra cost for recording true RAW sensor data will be much less of an issue, if RED stays with RED CODE then the "look" of their future cameras will stay with the RED CODE artifacts while other cameras will be true RAW and free of such artifacts.

Given the costs of the true RAW cameras will come down, people may in the future make their purchase decision more on fine aspects of image quality than price of the camera as RED now has the advantage.

The RED ONE is a great achievement that has forever changed the motion picture camera market and film production choices. But if RED decide not to offer a true RAW option, that will not stop others from making such an offer, and with uncompressed DI getting easer and lower cost everyday, then where in the quality Hierarchy will the RED cameras fit in from HD to Cinema quality?