Notes on RED and Arri D21
- The following is a compilation of a thread from the TIG message archives for September 2009
From the TIG archives http://tig.colorist.org/pipermail/tig/2009-September/016590.html
Sat, Sep 12 2009:
Can anyone update us on the status of Redcode RAW- has it been improved; and by now, we know that RED has - what is the term, Certified Partners, who are able to use software for various Certified Platforms, or is it Certified Platners, which would cover both; how effective has this been in making it easy to work with the Redcode RAW?
On Jul 5, 2008, at 4:45 AM, Robin Rowe wrote: > RED uses an undocumented proprietary format called Redcode RAW, a > lossy > variable bitrate wavelet codec something like JPEG2000. At this time, > only proprietary software from RED is available and only for Mac Intel > hardware. Reportedly, the layout of this format is still evolving.
-- Rob Lingelbach rob at colorist.org
From the TIG archives http://tig.colorist.org/pipermail/tig/2009-September/016592.html
Sun, Sep 13 2009:
I do work with a TON of Red footage, have been since the very first week the camera became available. Here is a little of my experience and what I know so far:
Support for Redcode has become widely available, however some platforms/partners are still better at this support than others.
There are two forms of Redcode support for now - native, and SDK. The native support has the speed advantage in not having to go through an SDK 'firewall' to retrieve RGB images for further processing. Red has stated that the SDK will be continually developed until it is equal to 'native' debayer support, but that has not yet occurred.
RedRocket just came out, and according to Red, all systems using the Red SDK will be able to use RedRocket, which should significantly level the playing field. Also has output capabilities: 4k DVI via optional breakout box, 2k via built-in SDI. Currently very much an alpha/beta product software-wise, with the only way to use the board being RocketCine-X software - available from red.com. The coming months should see some action on this device as it is a real game-changer for Red post production.
Here's a list of systems I know to have support for the Redcode file format:
Assimilate Scratch - native support, RedRocket support shown at IBC, but not yet available Apple Color - native support Baselight - Red SDK transcode (Baselight Transfer Station) Filmmaster - Red SDK transcode DaVinci - Red SDK transcode Iridas Speedgrade - Red SDK support DVS Clipster - native support via hardware debayer on proprietary card
I'm sure I left some out, but since I use mostly Scratch & Baselight, I'm pretty happy to have remembered what I did!
I like Scratch for grading Red footage, due to it's native support - 1/2 debayer is real-time, and I don't have to transcode to DPX for fluid, reliable workflow, which saves some time. There are also clear advantages to working in a floating point environment from the 12-bit linear chip data without having to transcode to a 10-bit file format, though working in 10-bit log does help with perserving the data. Still, nice to work in linear when not going to film.
Grading Red footage (as with most digital cameras at this stage) requires perhaps a bit more effort, as shadow noise can creep in from the practice of exposing 'to the left' to perserve highlights. With proper exposure & careful grading it's possible to get consistent 10-11 stops of DR out of the Red.
-- john tissavary | colorist | hi ground
From the TIG archives: http://tig.colorist.org/pipermail/tig/2009-September/016594.html
On Sep 12, 2009, at 8:46 PM, John Tissavary wrote: > I do work with a TON of Red footage, have been since the very first > week the > camera became available. Here is a little of my experience and what > I know > so far:
great information to have John, thanks for posting it.
> to working in a floating point environment from the 12-bit linear > chip data > without having to transcode to a 10-bit file format, though working in > 10-bit log does help with perserving the data. Still, nice to work in > linear when not going to film.
As a colorist, I have found working in linear to be a challenge.
there is a good overview of the situation at
It doesn't seem that there has been any progress in Redcode RAW, an inherently lossy format. It is good to know however that more manufacturers are embracing this proprietary format, at a price. 12:1 and 9:1.
> Grading Red footage (as with most digital cameras at this stage) > requires > perhaps a bit more effort, as shadow noise can creep in from the > practice of > exposing 'to the left' to perserve highlights. With proper exposure & > careful grading it's possible to get consistent 10-11 stops of DR > out of the > Red.
The Wikipedia entry cited above has this paragraph:
"Red specifies the sensor's signal to noise ratio at greater than 66 dB, with 11.3 stops of total dynamic range.However, the dynamic range reported from many cinematographers place the camera's dynamic range significantly lower, at about 8 stops. This difference has been attributed to the use of different mechanisms for measuring dynamic range. Red rates the sensor at 320 ISO."
I've done side-by-side comparisons between Arri D21 and Red One, and there's no comparison- D21 is extraordinarily better. However, as clients have pointed out, it's more expensive.
-- Rob Lingelbach TIG founder/admin rob at colorist.org
From the TIG archives: http://tig.colorist.org/pipermail/tig/2009-September/016595.html
Sun, Sep 13, 2009
In Scratch I work directly on the 12 bit RAW files, and since the Mysterium sensor is natively linear (as are all digital cams) I find the best results in a linear workflow as long as I'm working for a linear deliverable. But if I'm going to film I'll work in Raw Log with print emulation LUTs. I've never been much for working in Log when delivering to a linear format... The D21 has a very nice log workflow, but that comes courtesy of Arri's matrices.
There has been considerable progress to Redcode, though because it's a raw file format, the work has largely been done on the debayer algorithm and color matrices. The camera firmware & post debayer process have improved quite a lot from build to build, with build 20 showing the greatest leap forward, easily getting 1 more stop in the shadows compared with earlier builds, and a couple more compared to build 15 & prior. So much happens to the image in the debayer process that it is impossible to consider the file format without considering the debayer as a dynamic, adaptive, and ultimately integral part of the resulting images.
Build 21 of the Red One firmware, which just went into beta, features 4.5k recording (not sure yet, but I believe it's @ 2.33:1), and Redcode 46, a higher bitrate variant. This is probably as high a bitrate as will come from the Red One, as it's reaching the limits of it's processing power. With Redcode 46 one can record all current formats, as well as the new 4.5k
I've compared D21 & Red footage side by side as well, and while I think the D21 provides somewhat nicer images, I didn't see what I would categorize as a huge difference. What was immediately apparent is how much easier it is to grade the D21 footage... but I attribute that to Arri's matrices being rock solid for existing grading pipelines, whereas Red's are... trickier to put it nicely.
Red footage requires more work to get looking good in any given colorspace, but with care it's possible to give any other digital camera a serious run for its money.
-- john tissavary | colorist | hi ground
From the TIG archives: http://tig.colorist.org/pipermail/tig/2009-September/016596.html
On Sep 13, 2009, at 1:44 AM, John Tissavary wrote: > I've compared D21 & Red footage side by side as well, and while I > think the D21 provides somewhat nicer images, I didn't see what I > would categorize as a huge difference.
I wouldn't berate the point except to say that I found the difference substantial.
Rob disclaimer: I am not compensated in any way by RED or ARRI. -- Rob Lingelbach rob at colorist.org
From the TIG archives: http://tig.colorist.org/pipermail/tig/2009-September/016597.html
On 9/12/09 9:51 PM, "Rob Lingelbach" <rob at colorist.org> wrote: > I wouldn't berate the point except to say that I found the > difference substantial. Rob disclaimer: I am not compensated > in any way by RED or ARRI. >------------------------------<snip>------------------------------<
My experience is exactly the same to Rob's, based on doing final color on two large-budget D21 projects this year, and quite a few Red projects over the last couple of years. I think the blacks and highlights hold up better on the Arri, and I also think it looks sharper overall.
Four good references I would recommend:
RED: The Ultimate Guide to Using the Revolutionary Camera by Noah Kadner published by Peachpit Press [ISBN #0321617681]
(more a general guide to producers and post people using the Red)
plus John Galt's technical white paper report:
"The Truth About 2K, 4K, and the Future of Pixels"
as well as Rian Johnson's report
"Red Facts: Straight Talk on the Technical Realities of the Red Camera"
and Daron Keet's comparison
"The Red One Camera -- Through the Crosshairs of my Eyepiece"
I think the bottom line is that the Red camera is definitely "good enough" for a lot of productions (as witnessed by recent theatrical films like KNOWING, DISTRICT 9, and THE ), and most audiences can't tell the difference. But film-savvy DPs and colorists will definitely see the limitations of the pickup. No way is it the equivalent of 35mm film.
I take exception to the opinion that the Red can capture 11 stops of latitude. I think it might be 5 or 6 stops -- at best -- and even then, I think DPs have to use a little more fill and watch the highlights much more carefully than they would with film. I believe that digital cameras are harder to light for than film, because they're much less forgiving with exposure problems.
Again, I'm not slamming the camera, because I think in the hands of a good DP, the Red is capable of making good pictures. Like any medium, it's all about the quality of the lighting and lenses, and the skill of the person using them.
--Marc Wielage Cinesound/LA
Rob Lingelbach 19:15, 13 September 2009 (UTC)