How far away are we from RAW processing in an EX style camera?

[Tim Polster]

Just a thought here. So much effort is put into codecs but if we were shooting in RAW we would only care about delivery codecs.

How close or far away are we to getting RAW capture from a Sony or Panasonic model camera?

RAW images out of the 5DMKII are ~23 MB But this is for 22 megapixels. If this was for HD it would be less than 2 megapixels, less than 1/10th the size. By my calculations that is about 125MB per second at 60 fps.

Does not seem too far fetched. It would be way better to stop all of this codec nonsense and really let the cameras and lenses be the focus.

Thanks for your thoughts.

[Brian Drysdale]

One problem with RAW is that you need to do more post production and many productions don't have the schedule to handle it. RED cameras use RAW, but HDCAM SR is popular for television productions because of their tight schedules. Even the Arri Alexa gets used with HDCAM SR rather than ProRes on some productions.

Sony is committed to s-log, so I can't see them going for RAW for those productions that use colour correction in post. However, later in the year the Scarlet should be out, so you can use RAW on that camera.

[Dave Sperling]

Originally Posted by Tim Polster

Just a thought here. So much effort is put into codecs but if we were shooting in RAW we would only care about delivery codecs.

Hello Tim,
I'm personally in no hurry to get a RAW output from the camera.
Codecs are what keep file sizes manageable.
Look at the Convergent Design Gemini or the Black Magic Hyperdeck Shuttle - they use uncompressed recording (ie - no codec) from a 4:2:2 (or up to 4:4:4 signal in the case of the Gemini) - and already we hear complaints about having to deal with data management in the neighborhood of >10GB per minute!
Actually part of the sales pitch of the Gemini is that you can convert to the codec of your choice, and because your initial data is uncompressed, this conversion is really a 'First Generation' compression.
S-Log is essentially a way of logarithmically compressing more exposure latitude coming off the chip into a reasonably sized signal. If you really want to record true uncompressed RAW you will probably also want all the latitude coming off the chip, so think about potentially multiplying your data flow by a factor of 4 or 8, and then redesigning a pipeline to handle all that data.
Now think about RAW a bit more-- and the fact that everything you record must be color corrected in post to become useable. And since you probably want to be doing your color correct from your original uncompressed data, you need to at least keep it around until your online (yes, it's unlikely that a typical edit system will be able to handle all that data, so after converting everything to a useable size for editing, you'll need to conform the master data, then color correct.)
Now all this may be great for post houses and commercials and other very large budget endeavors, but for the rest of us who typically roll 2-5 hours of data a day and do our own transfers afterwards to the producer's hard drive, Codecs are what gives us a chance to finish the day.

[Tim Polster]

Good points. I agree there are some wrinkles and the transition from still to video will require a new set of NLEs and workflow. It is not for everybody just like still RAW workflow. But it seems like quite a dis-service to cameras produced like the EX series when a lot of the community basically thinks the included codec is under tasked for their work. So we go to an external recorder environment.

Thanks for your input and yes, 10GB a minute does seem like quite mountain of data!

[Mitchell Lewis]

Once you use the RAW file to create a perfectly exposed and white balanced master file, you could ditch the larger RAW file.

[Brian Drysdale]

It's still a lot of data and production insurance also insists that you have 3 copies of the rushes. Of course that depends if it's a personal film, but not making at least one back up copy of your camera recording in the tapeless world isn't wise.

Usually you don't know the correct colour correction until you've edited the film and trying to match the shots.

[Larry Kelly]

The Red scarlet that was mentioned in an earlier post shoots 3K raw in various file sizes. The file sizes are very manageable and are no where near the sizes of the uncompressed files stated above. Very workable on a MBP. Working with red's colour management system along with any number of NLE's that work with R3D's natively, it is a piece of cake to apply looks to your footage. You can have a number of "looks pre programmed into the software so you can pick a choose much like Picture profiles on EX cams. The beauty of it is that you have complete control of your image and can undo or change that look if you don't like it later, not the case with other cameras' footage. The Scarlet should be out this fall but there are a number of prototypes that have produced footage for viewing. The Epic user manual is online available for download and it is almost identical to the Scarlet camera operationally.

[Dave Sperling]

So when you say RedRaw isn't nearly as large as uncompressed, that by definition means it is their compressed version of RAW data coming off the chip.
Which of course may not necessarily be a bad thing, since even XDcamSR is a compressed format, it just happpens to use a codec that is virtually indistinguishable from uncompressed. Of course the more compression a codec utilizes, the more chance there seems to be for artifacts down the line...
But going back to the original post, I'm wondering whether the math of 2 megapixels for hd is appropriate, or whether in RAW terms it is 6 megapixels, 2 each for the camera's three sensors? I'm the first to admit I don't know much about the RAW recording system on the 5D. I'm assuming the 5D uses a Bayer pattern, but is the RAW recording of an image pre- or post-Bayering? If post-Bayering, what is the effective pixel count? And does it use compression techniques? And can the RAW image capture more dynamic range than s-log, or RedRaw for that matter?
I suppose the only thing I'm really certain about is that anything that's recorded using a RAW system is going to have to go through color correction (or 'look application') somewhere down the line.

[Alister Chapman]

Looking at RED Raw, consider that for a true uncompressed 4K image you would have 4096 x 2304 pixel locations with 3 sets of data for each, one for Red, one for green and one for blue. A total of 28 Mega Pixels in effect. But what RED Raw records is the output from the bayer sensor so there is only one set of data for each pixel location, 5120 x 2560 (Red MX) or only 13 Mega Pixels, so there is less data to start with and then this is compressed using something not to dissimilar to zip software to virtually losslessly reduce the file size. The RGB is extracted from the bayer data in the computer.