D-20 v. Red Filmout Test

[LEON]

Last Week i worked with testmaterial from the "RED Testing Team - Workshop" in Munich - Germany.
We had 5 REDs
For us Final Cut is only the Quick and Dirty viewing tool.

MR mradeck, did you try new mushrooms recently ?
Some infos are not properly going to your brain, i am afraid for your health !!

 

[Dj Joofa]

Yes, it is good test to compare CMOS with CCD (D-20 has single CCD sensor)
CCD is slower and has more consitent column/row read-out but it is lacking the speed and sensor cropping capability.

One has to understand that a phrase such as CCD Vs. CMOS by itself is not very meaningful if the number of "equivalent" pixels on both were more or less same. If that fact is a given, then internal implementation of the whole camera pipeline from signal acquisition to by the time you as a user have your hands on the raw data (or what ever format) is more important. There are so many internal stages involved, that a lousy implementation in a single one can degrade image quality. Lousy implementations happen on a more frequent basis than one would imagine.

Therefore, a phrase such as D20 Vs. Red is more meaningful because that compares the actual implementations. CMOS Vs. CCD with similar "equivalent" resolutions is not very helpful.

 

[chuckt]

The D-20 uses a single CMOS sensor the width of a Super 35 film gate aperture. Effectively the D-20, when used with current 35mm PL mount motion picture lenses, yields the same field of view and depth of field as that of Super 35 mm film motion picture cameras.

The D-20 captures images in two main modes.

In Data mode the sensor uses 2880x2160 active pixels generating RAW Bayer-data at 12 bit in a 4:3 aspect ratio. The RAW data then needs to be processed outboard to generate a full color image. A delivery aspect ratio for theatrical release, commonly 1.85:1, is achieved by cropping from the original image, similar to the cropping necessary when shooting 35mm film. In Data mode the sensor size also allows for the use of anamorphic lenses, producing the 2.35:1 widescreen aspect ratio from the full sensor height. Captured data is recorded to either flash or hard drives. Arri hasn't completed the development of the data mode of the D-20 and it is not presently available, different from Weinberg, SI, red, Dalsa and others, who offer data recording.
It is predicted that by mid 2008 Arri recording technology will develop sufficiently to accommodate the flow of data generated by the D-20.

In HD Video mode the sensor uses 2880x1620 active pixels to generate an image with a 16:9 aspect ratio. This output is 1920x1080 pixels in either YUV 4:2:2 10 bit (via single link HD-SDI) or RGB 4:4:4 10 bit (via dual link HD-SDI). Typically, the D-20 is tethered to a Sony HDCamSR recorder. Other recording options exist, including the Grass Valley flash mag which provides untethered recording of up to 15 minutes per "mag" in RGB 4:4:4. There are no recording option for RAW data as of yet.
The D-20 has a mechanical shutter, variable from 11.2° to 180° or an electronic shutter that simulates a 270° mechanical shutter at 24fps. The camera is capable of running at speeds from 1 to 48fps, though this shall be improved in the future. Numerous components of the camera were borrowed from Arri film camera models (most notably the 435ES), assuring compatibility with Arri film camera accessories and support equipment.

Though the D-20 system is capable of variable speeds from 1 to 60 frames per second in Data mode, these are currently limited to 23.976 fps, 24 fps, 25 fps, 29.97 fps & 30 fps in the video mode.

The sensitivity of the D-20 in video mode is regulated by the application of LUTs (Look Up Tables) prior to output. With sensitivity settings ranging from ISO 100 to ISO 320 with linear responses, the D-20 also offers log curve options designed to mimic the response of film negative. Unlike some other digital cameras, the D-20 does not offer gain boost, instead relying on the advantages of adding gain in the post production process. The sensitivity of the camera in Data mode is regulated by applying LUTs in the outboard processing of the image.

(The above quotation from Wiki.)

There is no need to beat up mradeck for his post. He was only telling what he found out from the tests.
D20 is well built and accepted by Big studios.
RED is still working on issues but has the price advantage, for now.

Talking about the price advantage: has anybody checked out the picture quality of Sony EX1 ($6500).
It has a native16:9 1/2" CMOS sensor with 1080x1920 pixels and a fixed lens and 4:2:2 through HD-SDI.

 

[Jeff Brue]

Working on 4 different movies right now using Red, Viper, F23, and D20. They all have real world issues. The only reason you don't hear as much about them is because there are so fewer cameras, and the people who can shoot them normally don't post on forums, they're too busy shooting movies. BTW Huzzah the strikes going to end.

 

[ChrisLyon]

Huzzah indeed- but it ain't over 'till it's over. We will see how the vote turns out (if there is one).

 

[Petr Dvorak]

I did some very simple tests yesterday at CSC in New York with my Red vs. a D-20, for a project that will film out to IMAX. Next week they will look at the Genesis and the F23. Short clips of each will be filmed out and we'll screen them in IMAX.
...

Harry, at which quality are filmouts made?
Are they all printed with 2K or with 4K laserprinter? And from which data - 2K, 4K or other resolution?

 

[StephenMurphy]

Arri hasn't completed the development of the data mode of the D-20

Thats not entirely correct. Arri now offer the D20 with the sTwo for Data capture - they gave a big demo of the system at Camerimage. As i understand it they've also solved a few of the problems they were having with noise in shadow areas.

 

[David Nardini]

any samples ...

any samples ...

Last Week i worked with testmaterial from the "RED Testing Team - Workshop" in Munich - Germany. We had 5 REDs and many different lenses... one week with 40 students. We finished 4k, 2k & HD 1920x1080. We printed out with arrilaser in 2k, compared this with 2k digital projektion

In 2006 i finished & graded also the first Shortfilm ever shot on ARRI D20 (1920x1080 YUV Fullrange). This D20 material is almost better then most of the red material. Resolution, noise, compression artefacts & sharpness.... You pay for what you get.

Focus pulling is a big issue with the RED!!! If you are not in focus, you have only produced many data for nothing. The RED has a little bit more resolution then the D20 ore Viper, if you are in focus, but also only when you finish in 4k and project in 4k. Well in 2k or HD you have a little chance to reach the quality of a D20.
Another problem is that the quicktime ref "proxies" have no colormatrix applied which would reproduce the correct color. All pictures are very low saturated and colorshifted. I can not imagine and all other people here, that someone like to make an offline edit with those material. So you have to render quicktimes with colormatrix and gamma applied. Rendertimes are very long: 1:20 - 1:60.
Another issue is, that Final Cut cannot deal correct with RGB-FullRange. Also theres no FullQuality Fullframerate, Fullcolor - Realtime Playback on FCP also not in HD 1920x1080. Also all onboard Effects on FinalCut are only 8 bit Effects. I recommend to finish RED on AVID DS Nitris, which produce much higher Quality, Oflline on Mediacomposer in HD.

For us Final Cut is only the Quick and Dirty viewing tool.

Can you post some links, FTP, etc to the sample outputs so we can digest and make up our own opinions ... (rather than just read your verbal one ?)

Cheers

 

[Patrick Tresch]

Thats not entirely correct. Arri now offer the D20 with the sTwo for Data capture - they gave a big demo of the system at Camerimage. As i understand it they've also solved a few of the problems they were having with noise in shadow areas.

What system is that? What recording times? Is it on board? What compression?

Thanks.

Patrick

 

[StephenMurphy]

What system is that? What recording times? Is it on board? What compression?
Thanks.
Patrick

www.stwo-corp.com

 

[Patrick Tresch]

www.stwo-corp.com

Thanks.

So you have 32 min of uncompresse 4k on a "first truely portable" HDCAM-SR type of deck/diskrecorder of 30-40kg?.

I prefer REDCODE raw on a 900$ harddrive of apporximative 400grams with about 2 hours of 4k.

Different needs different tools...


Patrick

 

[laguun]

Last Week i worked with testmaterial from the "RED Testing Team - Workshop" in Munich - Germany. We had 5 REDs and many different lenses... one week with 40 students. We finished 4k, 2k & HD 1920x1080. We printed out with arrilaser in 2k, compared this with 2k digital projektion

In 2006 i finished & graded also the first Shortfilm ever shot on ARRI D20 (1920x1080 YUV Fullrange). This D20 material is almost better then most of the red material. Resolution, noise, compression artefacts & sharpness....

Sorry, but i have to say that the 40 students still need to learn quite abit.

The resolution of red is almost double the resolution of the D20.
Its interesting that you think that the 2880 x 2160 of the D20 (in 4:3, in 16:9 that would be 2880 x 1620) would be more than 4520 x 2540.

Btw: You can measure the much higher resolution. We did so not only in the tests we shot in nov/oct - but with any decent chart. But you should be in focus.

Noise was and is an issue with the D20 in lowlight, reds cameras behaved much more silent in the same light. How high did you set the ASAs in redcine?

If you are talking about Sharpness btw - if your focus puller didn´t get it right, that isn´t exactly a camera misdesign. If you are talking about the the debayer strategy, which of the many available shrpening strategies in RC oder RA did you use?

You pay for what you get.

So, do you recommend the Sony F23 then as maximum quality? And does the image gets better if you attach a codex?

Focus pulling is a big issue with the RED!!!

*sarcasm*
And even worse on 65mm!!!
*sarcasm ends*

If you are not in focus, you have only produced many data for nothing.

Maybe an autofocus camera might be better for you? As alternative you can always reduce the active sensor area, then the nasty 35mm DOF goes away and focus is much easier with S16 sensorsize. Also you produce less "many data", which it seems is something you also would prefer.

The RED has a little bit more resolution then the D20 ore Viper, if you are in focus, but also only when you finish in 4k and project in 4k.

How surprising! You really found out that 4K is downconverted to 2K if you downconvert 4K to 2K? astonishing.

Well in 2k or HD you have a little chance to reach the quality of a D20.

I believe at once that -you- didn´t reach D20 level - Steven Soderbergh and Peter Jackson however came to different findings, and even a modest smaller company as ours is able to get much more out of a red than out of a maxed D20. In resoluion, sharpness, noisefreeness etc.

Another problem is that the quicktime ref "proxies" have no colormatrix applied which would reproduce the correct color. All pictures are very low saturated and colorshifted. I can not imagine and all other people here, that someone like to make an offline edit with those material.

While (for the first in your whole post) i agree, i have to tell you that 35mm -film- mostly worked like that... where you working in the industrie when everyone was using Avid AVR3 btw?

Another issue is, that Final Cut cannot deal correct with RGB-FullRange.
Also theres no FullQuality Fullframerate, Fullcolor - Realtime Playback on FCP also not in HD 1920x1080. Also all onboard Effects on FinalCut are only 8 bit Effects. I recommend to finish RED on AVID DS Nitris, which produce much higher Quality, Oflline on Mediacomposer in HD.
For us Final Cut is only the Quick and Dirty viewing tool.

FCP isnt even 4K, but hey, in which NLE would we edit the D20 in -data- mode? Iridas Speedgrade?

 

[Daniel Reichenbach]

Comparing ARRI D-20 with RED makes not much sense in my eyes: RED started two years ago and delivers a picture and philosophy which is much better than everything I have seen from a digital camera before (haven't really tested ARRI D-20. But: ARRI D-20 lives since 5 years (?) now, is not affordable for a man like me and should be hopefully a bit better than the RED image at the moment. 10%? 20%? Maybe. That's not enough to think about this more than a minute. And I don't really wanna take 30 to 40 Kilos and a bunch of technicians with me if I wanna shoot spontaneous tomorrow morning in the best light on a glacier.

 

[Keith Neff]

We have yet to work with/test a camera that doesn't have some sort of issues. The difference seems to be that Red is aggressive in banging out solutions where many of the other cameras you either 'get what you get' or wait for the next model which is really what you wanted/were sold in the first place.
(Please note the new Red Lab Avitar)

 

[Harry Clark]

Daniel,
The producers, taking advice from DPs, post workflow experts, and filmout lab technicians, found many opinions and (I believe) much misinformation. I suggested that we do a comparative filmout because their project had very specific parameters (digital acquisition in front of a greenscreen, effects work, filmed out to IMAX) Also, unless you shoot a test and view it with your own eyes, it's all just theory and opinion.
CSC has the D-20, and I have a Red, so that seemed like a logical place to start. CSC also has the right personnel, facilities, and equipment to diagnose issues and host a test like this.
Next week they are going to test Genesis and F-23, but I think I'm booked most of the week and may not be able to participate. I agree that all of these cameras are different in design and philosophy, but each project that we shoot has its own unique requirements. I don't want to close my mind and career to a vast set of tools just because I happen to own one of them.

Stephen,
I was really excited to try the S.two, and I felt that using the full chip would be a more fair test in terms of resolution, but Andreas could not get the new firmware update working to his satisfaction. It will be cool when they do, and they're also working on a fiber optic signal path like Dalsa, so just one slim cable for everything. We did try Log C as one of our ISO settings to at least offer up the whole dynamic range of the chip, if not the full res.

Hawaj,
I'm not sure what post workflow they're doing yet. I know that the house they are working with asked for Cineon files. So far, we're just giving them R3D files from my camera and the HDCAM tape from the D-20. HDCAM from the F-23 I guess, and (another guess) files from the SSR-1 for the Genesis, which I think is still just HD res.
I would assume that the material will all be uprezzed to 4K for the filmout. I don't think anything less would be correct for IMAX.

I'll keep everyone posted.
Cheers,
Harry

 

[David Mullen ASC]

There were some circuit board changes in the D20 recently that reduced noise and allowed them to rate it a little faster, plus make their "LOG C" mode more usable. Although if you are recording RAW data, you wouldn't use a LOG mode.

 

[Andrew M.]

I think we were discussing D-20 in length one year ago in some other thread.

How you can even compare 12MP Misterium with 6MP D-20?!

Also just the fact that ‘mradec’ mentioned that

“Focus pulling is a big issue with the RED!!! If you are not in focus, you have only produced many data for nothing.”

proves that 4K resolution is not 2K of D-20. To focus 4 times smaller pixels needs some practice.

Remember that RED has 4 pixels in the place where D-20 has one. So the first indication that camera resolving power is much better is complaining of the operators that it is difficult to set perfect focus. (don’t mix it up with the statement “impossible to focus”)

Also the sharpness of lenses is the big factor here. IMHO anything below Master Primes will not show the real resolution of the Misterium. (and Master Primes above 40mm FL)

I remember HD camera operators complaining in the past, that it is so difficult to set the focus on HD cameras as compared to SD.

 

[Harry Clark]

Hi David.
CSC had those changes on the cameras in house, and we tested up to 500 ISO, a setting that I don't think was available before, and that I would think too noisy anyway... but it looked pretty good on a Cinetal, though I'd wait to hear from the experts on your coast about how it would work for the project we're doing.
The Log C was Andreas' idea. I figured tape is cheap, why not blast off some there too?
Cheers,
Harry

 

[StephenMurphy]

Daniel,

I was really excited to try the S.two, and I felt that using the full chip would be a more fair test in terms of resolution, but Andreas could not get the new firmware update working to his satisfaction. It will be cool when they do, and they're also working on a fiber optic signal path like Dalsa, so just one slim cable for everything. We did try Log C as one of our ISO settings to at least offer up the whole dynamic range of the chip, if not the full res.

The Fiber optic path should make a big difference to its on set workflow - ill look forward to that. As i understand it the d-20's new upgrades have completely removed any of the problems it had with shadow noise, but i was also told that the native asa of the sensor in RAW mode is 250 asa, which i still feel is a bit slow. Even still, Id be happy to deal with that if it meant i could shoot true anamorphic, and have an optical viewfinder
Look forward to hearing what you thought of the results.

 

[M Most]

www.stwo-corp.com

The s.two is an HD video recorder. It can take in dual link or single link HD video, and it writes it to disk as a DPX sequence. It is incorrect to call it a data recorder for the D20, as that implies it is recording data directly from the sensor, which it is not. It is recording the same information as an SRW1 tape deck, only it's writing it as a DPX sequence. This is considerably different than recording RAW bayer data, even if the word "raw" is often misused these days when referring to the dual link video output.

Is there a new version of the S.two that is a straight data recorder?