For those who have used them: How does Red compare to the Genesis and CineAlta?

[Jeff Brue]

Eh, guess I'll have to respectfully disagree.
In my mind something that has a pixel difference ratio of more than 7% is not a valid format professionally anymore, and in cases like the Red comparing a bayer pattern data rate with wavlet to an YCrCb data rate with a hackneyd version of jpeg isn't cool and just confuses people, as with bayer pattern the way the sensor works means you're really only recording a b/w image so instantaneously you're getting 1/3rd the data rate. Call me a purist I just know too many vfx guys that have had to pull nightmare keys out of hdcam, because it was marketed by sony as something valid.

Sr on the other hand even at its 440 rate is the bee's knees (and you're talking to someone who hates compression). I just do want to point out that mathmatically shooting HDCAM means that you're only going to have a NEG with about 5 stops of lat (hence the origination of the DIT) the person that says yes this image will survive the horrible compression.

 

[laguun]

Eh, guess I'll have to respectfully disagree.
In my mind something that has a pixel difference ratio of more than 7% is not a valid format professionally anymore...

Hello Jeff,

HDCAM is the most used professional HD format. Worldwide.

and in cases like the Red comparing a bayer pattern data rate with wavlet to an YCrCb data rate with a hackneyd version of jpeg isn't cool and just confuses people, as with bayer pattern the way the sensor works means you're really only recording a b/w image so instantaneously you're getting 1/3rd the data rate.

Jeff, its pretty easy.

datarates @ 24P
HDCAM 144Mbit/18Mbyte datarate for 1080p.
REDCODE 224/28 and 288/36 for 4K. At 2K24P, redcode was ~ 72Mbit/9 Mbyte IIRC.
HDCAM SR 440/55 and 880/110 for 1080p.

bayern pattern raw in itself is a form of (clever) datareduction, as is YUV chroma subsampling.

Call me a purist I just know too many vfx guys that have had to pull nightmare keys out of hdcam, because it was marketed by sony as something valid.

If you need excellent and complex keys from HDCAM cameras, recording the cameras 10bit 422 uncompressed and fully avoiding the datareduction on the tape is recommendable.

Almost any *realtime* live key in 1080p you see is done that way, be it with sony, thomson GV, panasonic or ikegami camcorders or camera heads.
Our hdcam often are rented out and return after days of use with 0 additional drum hours due to this.

I think many people who dont use hdcam on a daily basis miss that the 750/900R/950/1500/23 etc deliver uncompressed via HD-SDI *and* compressed via tape simultanesly.

I just do want to point out that mathmatically shooting HDCAM means that you're only going to have a NEG with about 5 stops of lat (hence the origination of the DIT) the person that says yes this image will survive the horrible compression.

Sorry Jeff, but you are wrong.
Its no problem to get 10 stops out of HDCAM.
I dont know which Sony camera you own or use, i just can speak for the the 750, 900 and 23.

 

[Jason Sinclair]

What is the current standard and agreed upon testing for dynamic range? Can someone post image results in uncompressed form of the different formats, shot in real life situation. (a scene with high contrast range) of the same scene under same conditions. Then we are in a position to actually see exactly the trade off and we know there are always compromises. We are also in a position to take those images into post and see what the latitude can extend to. For me until testing becomes a standard that we can all share and experience with real footage then it's all talk, numbers and may i even suggest rhetoric. If people are going to make statements on their personal opinion that are in conflict with manufacturers statements then they need to post their comparative results. And also give the community those images so the testing can be verified. It doesn't matter who you are or what you have done, it's basic respect to explain your methods and standards and to post your results. Then we can let results speak for themselves and people can make their own informed decisions on the given results and their own particular circumstance and also allows the manufacturer to reply. That way we all know where we stand and comments can be agreed upon and unified.

 

[Evin Grant]

What is the current standard and agreed upon testing for dynamic range?

That's really the problem, there isn't one. Every filmmaker has a different opinion about what constitutes DR. I've shot HDCAM, Varicam, and was involved in one of the F23 Vs. Red tests, which Jeff was also part of.
All I can do is give you my opinion based on my experience and tests, which is exactly what you should do when deciding a camera to shoot on. My Experience leads me to peg the F900 at about 7.5-8 stops, the Varicam at about 8 and the Red at a solid 9 if exposed at ISO 400 Redcode 36. The F23 seemed to be 10 maybe 10.5 in the direct comparison but it seemed like mostly highlight detail and while very important I was unable to really test it in a way that I feel complete. Jeff should be able to shed more light here. This test was recording to the S2 system uncompressed not the HDCAM-SR tape deck. One interesting thing is how much noisier the 1/4 and 1/3d tones looked on the F23, but again this was before any processing. In the end your tolerance for noise is really the determining factor, there is no doubt that the Red has all 12 bits being recorded, if you don't mind the shadow noise you could have 12 stops, but most agree, as of build 15, 9 is what's really useable.

 

[Stephen Williams]

Hi Evin,

What gamma curves were you using? The digital praxis curves give way more than 8 on an F900R. The BBC recommended set up for the Varicam gives way more than 8.

Are you getting a Solid 9 stops from Red using tungsten light? With my tests I only got 9 stops with daylight.

Stephen

That's really the problem, there isn't one. Every filmmaker has a different opinion about what constitutes DR. I've shot HDCAM, Varicam, and was involved in one of the F23 Vs. Red tests, which Jeff was also part of.
All I can do is give you my opinion based on my experience and tests, which is exactly what you should do when deciding a camera to shoot on. My Experience leads me to peg the F900 at about 7.5-8 stops, the Varicam at about 8 and the Red at a solid 9 if exposed at ISO 400 Redcode 36. The F23 seemed to be 10 maybe 10.5 in the direct comparison but it seemed like mostly highlight detail and while very important I was unable to really test it in a way that I feel complete. Jeff should be able to shed more light here. This test was recording to the S2 system uncompressed not the HDCAM-SR tape deck. One interesting thing is how much noisier the 1/4 and 1/3d tones looked on the F23, but again this was before any processing. In the end your tolerance for noise is really the determining factor, there is no doubt that the Red has all 12 bits being recorded, if you don't mind the shadow noise you could have 12 stops, but most agree, as of build 15, 9 is what's really useable.

 

[Jason Sinclair]

That's really the problem, there isn't one. Every filmmaker has a different opinion about what constitutes DR. I've shot HDCAM, Varicam, and was involved in one of the F23 Vs. Red tests, which Jeff was also part of.
All I can do is give you my opinion based on my experience and tests, which is exactly what you should do when deciding a camera to shoot on. My Experience leads me to peg the F900 at about 7.5-8 stops, the Varicam at about 8 and the Red at a solid 9 if exposed at ISO 400 Redcode 36. The F23 seemed to be 10 maybe 10.5 in the direct comparison but it seemed like mostly highlight detail and while very important I was unable to really test it in a way that I feel complete. Jeff should be able to shed more light here. This test was recording to the S2 system uncompressed not the HDCAM-SR tape deck. One interesting thing is how much noisier the 1/4 and 1/3d tones looked on the F23, but again this was before any processing. In the end your tolerance for noise is really the determining factor, there is no doubt that the Red has all 12 bits being recorded, if you don't mind the shadow noise you could have 12 stops, but most agree, as of build 15, 9 is what's really useable.

So in essence if proper exposure is used you get 12 stops regardless of build but just not on the set monitor? That is, if you expose the shadows properly. And in essence this is better than all the systems you have noted?

 

[liquidigital]

In the end your tolerance for noise is really the determining factor, there is no doubt that the Red has all 12 bits being recorded, if you don't mind the shadow noise you could have 12 stops, but most agree, as of build 15, 9 is what's really useable.

That's an interesting point, because ultimately, it is subjective. I read this post and immediately thought of Soderbergh. I remember an article a long, long time ago where he talked about pushing the gain all the way up on a Sony. Traditionally, (or at least in a number of digital projects) he appears to not mind residual grain/noise, so I am looking forward to seeing what direction he goes with the RED. What is Soderbergh's tolerance, so to speak?

 

[Jeff Brue]

(sigh)
"HDCAM is an 8-bit, 3:1:1 sampling structure with DCT compression and a limit of 1,440 horizontal pixels to record a signal of 140Mb/s to tape."

In regards to everything Evin's statements are really the most dead on, it comes down to what you and whoever is going to work on the movies tolerance is for noise and compression. Personally I can pick out HDCAM artifacts no matter whether its being pushed any which way in terms of latitude. The 5 stops statement was made by a DP that has a few F900 major hollywood films under his belt, for what he was willing to accept or deal with in post.

I like the F900 and its variants I think that for the time it was a good camera, I just think sony really boffed it with HDCAM and well I wouldn't be repeating it if sony themselves hadn't said it to me. This is why they invented SR.

Lastly I will say Laguun you're talking to someone who has the highest delivery requirements, DCI packs and Film Outs. There's a lot of background processing that has to happen to make those deliverables happen that affects color and latitude. I can't make an effective argument about throwing away established infrastructure (there's really no reason) for using F900 and HDCAM for meeting delivery requirements to things like broadcast. Its definately good enough for that.

I'm just hoping that you won't let the thing be put out on a green screen shoot thats going to mean someone in the vfx department is going to be fired, because they can't roto 200 shots in 2 days. I just heard about that particular scenario happening 2 months ago.

 

[Jason Sinclair]

From what i am gathering it is best to expose at 125asa and you gain an extra 1 1/3 rds stops. Then you can crush the last 1/3 stop and pull 1 1/3rd details from the highlights and you get 9+ 1 2/3 = 10 2/3 stops . So expose for 100asa and then take it down 1/3 stop (or expose 125 asa) giving 1/3 stop of noise. In red cine pull out the highlights in curves and export (if you have scratch then do it all there). Then crush whatever noise you have left and you should get close to 11 stops in your tests.

 

[David Mullen ASC]

You don't gain any more dynamic range by overexposing. Or underexposing. The range that the camera captures in RAW is the range. All you are doing by over or underexposing is sliding that 11 stop range (or whatever) around to either favor highlights or shadows. And if you end up lifting the image in post to make it brighter, you start to get noise but good highlight detail, and if you end up crushing the image to make it darker, you'll have clipped areas that are unrecoverable, but with less noise.

Overexposing by a stop just means that you've lost a stop of bright information but gained a stop of shadow information.

It's really a matter of taste whether you prefer to reduce noise at the risk of clipping of overexposure information, or reduce clipping at the risk of more noise. Or just work in the middle where perhaps you are happy with both the noise level and the highlight latitude.

 

[Evin Grant]

From what i am gathering it is best to expose at 125asa

BAD IDEA!
Use 320 as your base ASA. It's much better IMHO to have more highlight headroom than to blow it out in favor of slightly better noise performance in the 1/4 tones. Mp personal preference is to expose at ISO 400 in daylight and 320 with Tungsten. I'm happy with the noise and the texture under those conditions.

 

[Jason Sinclair]

Edit...

Yes, my thinking is in fact backward here.

Thanks for your patience guys.

Jason

 

[Nova Invicta]

Hungarian Tests

Hungarian Tests

Lonesillver

I agree the tests did not standardise lenses in the 2/3rd format but they could not for Genesis against the D-20 or Red as PV have a proprietary mount. However lenses do not affect the latitude of the cameras and on one simple element of these tests the overexposure of the Red at the window (all HD cameras were worse than film) was worse that the D-20 & Genesis & the F900R.

Graeme

Unfortunately I was not present at the tests so I was not able to question how the Red One was set-up and where I agree with Lonesillver is that the tests appeared to be rushed somewhat. The Red as I stated before appeared fine on the color chart it was only going into the underexposure scenes that the blacks appeared crushed & lacking detail. I have since seen this test again and the color chart when the tape is frozen looks well exposed in the openning part of the Red test so the comparison for latitude should stack up and compared to all the other cameras gave the least range in this particular test. As stated I dont know what version of software was used or how they processed the raw data I am trying to find this out.

 

[Graeme Nattress]

Thanks for looking into it. As I said, I could understand seeing many diferent things, but not crushed blacks.

Graeme

 

[Antoine Fabi]

I think people should understand that, the RED camera's output is of highest quality. It is almost infinitely tweakable because of the richness of the output info. You can make it look almost anything you want.

BUT, you have to fully understand or "feel" the camera.
It is NOT a "point and shoot and edit" camera like some others.
It's a lot much more. That's the magic about it i think.

With such a powerfull tool, the best and the worst is in the operator's hands.

 

[BradWright]

Could this test be the result of the wavelet compression dropping detail in the blacks and the peak whites? This is pure speculation, but to really compare these cameras side by side it should be a very tightly controlled situation. The test should have some fine detail in the in all the grey areas of the test to check for lost information and compression artifacts. Solid colored squares on color charts don't typically contain fine detail, and therefore you can't see what may be getting lost due to compression.

 

[N_Villers]

comparison "tests"

comparison "tests"

Lonesillver

I agree the tests did not standardise lenses in the 2/3rd format but they could not for Genesis against the D-20 or Red as PV have a proprietary mount. However lenses do not affect the latitude of the cameras and on one simple element of these tests the overexposure of the Red at the window (all HD cameras were worse than film) was worse that the D-20 & Genesis & the F900R.

Graeme

Unfortunately I was not present at the tests so I was not able to question how the Red One was set-up and where I agree with Lonesillver is that the tests appeared to be rushed somewhat. The Red as I stated before appeared fine on the color chart it was only going into the underexposure scenes that the blacks appeared crushed & lacking detail. I have since seen this test again and the color chart when the tape is frozen looks well exposed in the openning part of the Red test so the comparison for latitude should stack up and compared to all the other cameras gave the least range in this particular test. As stated I dont know what version of software was used or how they processed the raw data I am trying to find this out.

I toned down my original reply concerning the tests done here in Budapest out of consideration for those involved but I will stick to my claim that using the testing methodology in which these tests were carried out does not illuminate any of the real differences between these cameras. I have since had several conversations with others who were also present and they have only reinforced this. With especially the shot by the window, the lighting differences and actor variances were too dramatic to really be useful. Did you notice how differently the natural light varied from shot to shot - look at the differences in direction of the shadows - or that the curtain was completely cropped out in one shot. In fact at the screening that I was at this brought about some quite sharp words and Sz. G. rather defensively, in my opinion, admitted they were shot on different days!! It would be great to see side by side footage from an invariable and well lit scene. Unlike many (as I am patiently waiting for my own camera where I will make my own tests) I can make no claim as to the merits of red over any of the other cameras used. It would not be a complete surprise if indeed film looked best followed by the genesis but I feel that those differences would be far more subtle and involve "feeling" more organic, not the dramatic differences that this test outlined. That the red (not to mention the d20) looked so dramatically wrong I think is one more indicator of a well intentioned but ultimately failed test.