ARRIRAW vs REDCODE RAW

[Chris Hocking]

I don't want to start any D-21 vs RED discussions, but I am curious as to the difference between ARRIRAW vs REDCODE RAW. I've read bits and pieces about both formats, but just curious to learn all the nerdy nitty-gritty details....

[Brian Drysdale]

Originally Posted by Chris Hocking

I don't want to start any D-21 vs RED discussions, but I am curious as to the difference between ARRIRAW vs REDCODE RAW. I've read bits and pieces about both formats, but just curious to learn all the nerdy nitty-gritty details....

ARRIRAW is uncompressed out of the camera and REDCODE RAW uses wavelet compression. You record the ARRIRAW on certified data recorder such as the S Two or the Codex - I believe the latter can record using JPEG 2000 or as uncompressed data. JPEG 2000 is similar to REDCODE in that it uses wavelet compression.

[Robert Monaghan]

To add a bit more, ARRIRAW files are similar in concept to DPX. Its a sequence of files representing a single frame per file. The ARRIRAW files are 12-bit (the maximum gamut that the human eye can see), and are largely untouched. ie: no lossy compression with a log curve applied. (Redcode does this to bring the file size down.) I am not sure what the acquisition bit depth is, for a Red One, but I do believe that the file is 10-bit.

Red does a fantastic amount of processing on the footage to get it out of the camera, ARRIFLEX D-21 does some bias/gain tweaks (for light sensitivity), and then sends the image out of the camera.

Doing a comparison of Red to other digital cinema cameras isn't straight forward, unfortunately. Many of Red's competition are $100,000 - $250,000. The difference when that sort of money is involved, is that the "big-iron" cameras are built for maximum uncompressed fidelity, while Red One is built to be inexpensive (trading camera cost for image quality, and post production ease of use.)

Red workflows requires a fair bit of noodling with the footage to get the desired image. Raw formats aren't as much work. (Naturally, the skill of the DP has much to do with this, too, to be fair..)

Big-Iron cameras have an "easier" workflow in many respects, than Red, but are far and away more expensive. Red setups are much less expensive, but require more time to make adjustments. (time == money in many cases, so arguably this is a wash, as far as the accountant is concerned.)

But, the phrase "you get what you pay for" does comes to mind.

bob.

[Kevin Duffey]

I am still not sure what $100K+ buys you in terms of quality over a $20K RED? You might get a fraction better image quality.. I am guessing the glass and such is worth a lot more. But the video footage I've seen from RED that is going in to movies looks every bit as good as any other movie I've seen. I am no pro like many of you with regards to color correction, looking at footage and knowing which is better.. but I do have somewhat of an eye compared to say 95% of movie goers who are there to see a movie and not judge the quality. With that in mind, I can't imagine why anyone today would want them big expensive cameras with those like RED, EX-3 and so forth available. They shoot stunning video footage that is easily put into movies and looks outstanding.

[Perrone Ford]

Well,

A RED at $20k won't record anything. You're going to be at closer to $50k before you're ready to shoot anything seriously. And the Arri and the RED are using the same, or very similar lenses. You can take the lens off the Arri and put them on the RED.

The margin in look between these high end cameras is subtle, but I have no doubt it's there. When you see the extent to which directors go for the VFX in modern movies you KNOW there has to be a difference for them to get approval from the studio for some of the gear they have to use.

But in terms of sheer size, an outfitted RED is quite a bit larger than the EX3. A fair bit heavier to. But to be honest, something like the D21 is just large to make things "easier" for a certain group of users. James Cameron talked about this a year ago in HDVideoPro. (HDvideoPro Feb. 2008, p38) Specifically saying how he felt it was silly that some people (DPs) wanted these monstrous digital cameras when it wasn't necessary at all.

[Brian Drysdale]

The D21 has an optical viewfinder, which adds considerably to the costs. The reason this feature is requseted is that quite a few film camera operators like to be able to to tells if the the image is sharp, which is a problem with many of the electronic V/Fs shooting HD. The RED is around the same size as a 35mm mute film camera rig.

The size of the camera will depend on the sensor size requirement for the productions and how the camera output is being recorded. These are often dictated by the requirements for the market the film is being made for and for many HD broadcasters and the EX3 isn't up to spec as the main production camera. The fashion for 35MM DOF is another factor which will add to the camera rig's size.

There are other issues besides the camera costs that people have to consider on a production including the post and the tight turnaround of large quantities of rushes requiring a lot of post production work. The camera is just one element amongst many to be considered and even the D21 can be one of the cheaper items in the budget.

[Craig Parkes]

Originally Posted by Robert Monaghan

To add a bit more, ARRIRAW files are similar in concept to DPX. Its a sequence of files representing a single frame per file. The ARRIRAW files are 12-bit (the maximum gamut that the human eye can see), and are largely untouched. ie: no lossy compression with a log curve applied. (Redcode does this to bring the file size down.) I am not sure what the acquisition bit depth is, for a Red One, but I do believe that the file is 10-bit.

Robert - RED ONE is 12 Bit Acquisition, and the file format is 12 Bit as well.

Red does a fantastic amount of processing on the footage to get it out of the camera, ARRIFLEX D-21 does some bias/gain tweaks (for light sensitivity), and then sends the image out of the camera.

RED does next to NO processing on the footage in camera - it outputs RAW with a Gamma Curve (which you can turn off to get straight back to very close to the original LINEAR file if you want). It does do quite a lot of compression - but this is very clever compression and is 'visually lossless' (you do not see compression artifacts because the compression is on the RAW data, before the demosaic.)

Doing a comparison of Red to other digital cinema cameras isn't straight forward, unfortunately. Many of Red's competition are $100,000 - $250,000. The difference when that sort of money is involved, is that the "big-iron" cameras are built for maximum uncompressed fidelity, while Red One is built to be inexpensive (trading camera cost for image quality, and post production ease of use.)

I'd argue that Red is built for maximum compressed fidelity, which means the compression is visually lossless, so just smarter than an uncompressed solution (same quality, smaller hd space.) RED One offers post production flexibility - this flexibility means more work in post to a degree, but with the increasing number of tools available this work isn't necessarily any more complicated - but it does require keeping up to date and developing new tools (while other digital cameras at the high end were designed to work within existing infrastructure and workflows, generally developed for film, RED ONE was more revolutionary in it's outlook and as such offers more workflow options that run the gamut from inexpensive to full film scale production. As such it didn't fit into any existing workflow instantly but is rapidly being accepted and developed for.

Red workflows requires a fair bit of noodling with the footage to get the desired image. Raw formats aren't as much work. (Naturally, the skill of the DP has much to do with this, too, to be fair..)

This is just confusing - Red IS RAW. RAW is Sensor Data, both require correct processing to get the desired image, both require someone on set who understands the camera and monitoring path to give a wysiwyg preview (DIT)

Big-Iron cameras have an "easier" workflow in many respects, than Red, but are far and away more expensive. Red setups are much less expensive, but require more time to make adjustments. (time == money in many cases, so arguably this is a wash, as far as the accountant is concerned.)

But, the phrase "you get what you pay for" does comes to mind.

bob.

You get what you pay for in terms of both Camera and Crew - given a knowledgeable crew, the D-21 and RED are both Tools with different workflows that result in mainstream cinema release suitable results. The cost advantage of RED obviously isn't so little that it's a wash otherwise it wouldn't be gaining traction in so many different applications - including studio pictures.

[Robert Monaghan]

Originally Posted by Craig Parkes

Robert - RED ONE is 12 Bit Acquisition, and the file format is 12 Bit as well.



RED does next to NO processing on the footage in camera - it outputs RAW with a Gamma Curve (which you can turn off to get straight back to very close to the original LINEAR file if you want). It does do quite a lot of compression - but this is very clever compression and is 'visually lossless' (you do not see compression artifacts because the compression is on the RAW data, before the demosaic.)

Umm, so which is it? Compression or no compression?

As for "visually lossless" vs. "lossless", there is a difference. "Visually Lossless" means that estimations are made on the original image. Lossless means that there isn't. Thats just a fact. In some cases this is important, and many other cases, it isn't. As with any tool, it depends upon the job you have.

Secondly, R3D files are 10-bit. Not 12. So while you may capture at a higher bit depth, processing is being done in order to save your footage to a file. Another way to say this, is that the data you are acquiring, isn't the data that is being saved.

That is the basic difference between R3D and other uncompressed raw solutions out there.
This also true of other file formats that save imagery, too. Its not just a "Red" thing.

Lastly, I am not interested in starting a flame war. My comments are based upon feedback I have gotten from a lot of my customers. In a nutshell, they say that Red footage is cumbersome to work with. This is due to its nature. I have *tonnes* of requests to create a replacement R3D quicktime component, so that people can work with the footage real time. (Currently, my Phantom quicktime component is real time, as it is the raw sensor data. Not a compressed/curved/processed movie file.) Unfortunately, there isn't too much I can do to make R3D footage decompress any faster. My customers that use both Phantom and R3D files, find that working with the Phantom files is substantially easier than R3D.

ARI files are very similar to ARRIRAW files. Just in a slightly different container. My QuickTime component for ARI files works at about the same speed as the Phantom files.

Another comment my customers have mentioned, is that in order to improve turn around times on their footage, they have to buy extra gear to process the R3D footage. This has been the case for a lot of studios, both big and small. They've also commented on the fact that they've had to set up and maintain a special "red" processing pipeline that isn't nearly as efficient as other pipelines that deal with DPX. (for example.)

So, while you may want to pound the red drum, and say that red is the best cheapest fastest, greatest, etc., etc., etc., the facts and details from my customers that are a better indicator as to how the formats stack up.

Red == Low Cost Ownership of Camera/10-bit Lossy Compressed Bayer Sensor data/Time Consuming
ARRI == expensive to operate/12-bit Uncompressed Bayer Sensor data/(fairly) instant use of imagery
(and for those that are interested)
Phantom == Moderate expense/14-bit Uncompressed Bayer Sensor Data/(fairly) instant use of imagery and High Speed.

Regards,

bob.

[Craig Parkes]

Originally Posted by Robert Monaghan

Umm, so which is it? Compression or no compression?

It's compression - generally 'processing' when it comes to image manipulation can be interpreted as something else. That may well just be a difference of usage, so I would say we agree.

As for "visually lossless" vs. "lossless", there is a difference. "Visually Lossless" means that estimations are made on the original image. Lossless means that there isn't. Thats just a fact. In some cases this is important, and many other cases, it isn't. As with any tool, it depends upon the job you have.

Agreed.

Secondly, R3D files are 10-bit. Not 12. So while you may capture at a higher bit depth, processing is being done in order to save your footage to a file. Another way to say this, is that the data you are acquiring, isn't the data that is being saved.

This I need further explanation on - documentation and posts by Graeme Nattress who designed REDCODE describe REDCODE as containing 12 bits of colour data. If you export 16bit TIFFs from a R3D would you not end up with 12 bits of accuracy in the TIFF?

Lastly, I am not interested in starting a flame war. My comments are based upon feedback I have gotten from a lot of my customers. In a nutshell, they say that Red footage is cumbersome to work with. This is due to its nature. I have *tonnes* of requests to create a replacement R3D quicktime component, so that people can work with the footage real time. (Currently, my Phantom quicktime component is real time, as it is the raw sensor data. Not a compressed/curved/processed movie file.) Unfortunately, there isn't too much I can do to make R3D footage decompress any faster. My customers that use both Phantom and R3D files, find that working with the Phantom files is substantially easier than R3D.

Number of clients that use R3D versus number of clients that use Phantom files, and the amount and type of hardware that is required for uncompressed workflows versus compressed workflows are very different. Access to the technology that allows the compressed workflow of RED is increasing rapidly, especially now that RED Rocket is shipping.

The difference in methodology is one of choice - the choice RED one offers in my mind an unparrallelled compressed workflow. This does not negate the value of uncompressed workflows - but it does offer greater flexibility and, because of their pricing, accessibility.

Red == Low Cost Ownership of Camera/10-bit Lossy Compressed Bayer Sensor data/Time Consuming
ARRI == expensive to operate/12-bit Uncompressed Bayer Sensor data/(fairly) instant use of imagery
(and for those that are interested)
Phantom == Moderate expense/14-bit Uncompressed Bayer Sensor Data/(fairly) instant use of imagery and High Speed.

Regards,

bob.

Agreed, except for my question on the 10 bit thing which hopefully you can explain. Additionally, hopefully soon to see changes on the Time Consuming thing as RED Rocket becomes available for accelerating the debayer to realtime.

[Graeme Nattress]

Originally Posted by Robert Monaghan

Secondly, R3D files are 10-bit. Not 12

12bit actually.

Graeme

[Peter Moretti]

Robert,

You have to realize that the post support for Red has been improving. So what may have been a problem in the past probably isn't one if you are using the latest system. Also, the camera does have a somewhat different learning curve, both to operate and more so in post, that people have to acquaint themselves with.

As for compression losses, you are barking up the wrong tree. Yes, there are losses; and no, you can't see them. I imagine some crazy scenario could be concocted where the compression losses would matter, but in anything mildly resembling the real world, its simply not an issue.

I don't know why you believe .R3D is 10-bit when it's 12-bit.

In short your clients are feeding you a hodge podge of information, some of which is true, but much of which is dated or incorrect.


HTH.

[Tim Kolb]

The support for RED files in real time, and including the ability to maintain RAW file structure through the edit (as opposed to having to 'flatten' the file for editing...effectively putting primary, or even the 'best light' color correction step before the edit) is increasing rapidly.

At a recent RED workflow seminar in New York, I showed Premiere Pro RT on the timeline and Sony showed similar Vegas capability. I would have to assume Avid and FCP are at least equal....(right?)

As far as the D21 is concerned...the optical viewfinder and simplified menu structure are really nice features that make handling the camera very tactile and quite frankly increase the comfort level of many DPs who (rightly) have trouble completely trusting a completely electronic evaluation viewing pipeline. I really like the pictures that camera makes...it's that simple.

I'd also say that there is something to be said for shooting a bayer sensor that oversamples. (the D21 has a 3K Bayer sensor and most usually outputs 2K or HD to HDcamSR, whereas the RED has a 4K Bayer sensor and outputs 4K...the downconvert needs to happen in post for RED...not the end of the world, but a small point that could make the difference for some situations).

It's been said many times...it's money vs. time, and depending on what situation you're in, either one could be the winner in your decision making.

I think that too many conversations on various production forums degenerate into a politics/religion sort of discussion that clarifies nothing for anyone. I see it on post production forums as well.

Our industry isn't 35mm vs BetacamSP vs VHS anymore. Specs are tossed around by many who really don't even understand them. Anything a manufacturer offers in the way of information is typically designed to make a case for their product. It may be completely accurate, well-meaning, and honest... But with so many different types of productions, budgets, and professionals out there, each product's benefits may not carry the same weight for every situation.

...and just because a particular product isn't right for your situation, that doesn't make it the wrong choice for everyone.

RED is good...D21 is good...

(Question: Why aren't we all concerned that the art of being a really competent DP is fading away on us instead of worrying about what tool we put into their hands?)

Sorry...a bit of a rant there. I guess I've had my fill of these types of exchanges over the last couple days.

[Chris Hurd]

Originally Posted by Tim Kolb

I think that too many conversations on various production forums degenerate into a politics/religion sort of discussion that clarifies nothing for anyone. I see it on post production forums as well.

You don't see it here, though -- and that's the DV Info Net difference.

Originally Posted by Tim Kolb

Sorry...a bit of a rant there. I guess I've had my fill of these types of exchanges over the last couple days.

Well, this is a two and a half year old thread and I'm not sure why Graeme choose to resurrect
it. I suppose I should simply auto-close discussions which have been dead for more than 365
days... although his input and yours are of course always welcome here.

Gently laying this thing back down in the ground so it can continue its dirt nap. Thanks all,