@mpgxsvcd
"However, I simply don't believe it produces a more film like look."

Here's where we'll have to agree to disagree in a friendly way. I can't prove it because it's a 'look' that we're discussing. I saw "Another Planet" in the theater the other day, it was shot on video with some long GOP interframe codec (I'll find out which cam) and they did a 35MM blowup. It was fine in static shots but the second anything moved the motion was different and there was substantially more blurring than there would have been with 35mm because the same frame by frame info just simply isn't there to transfer frame by frame to 35mm. My theory is that at 3GOP in our case (or best case AVC INTRA one of my favrorite acquisition codecs) there are more I frames per second that get you closer to the same frame by frame representation of the motion that you have in film. I think the longer the GOP the less chance you have at properly representing the motion, forget detail for now. I think an audience has a much firmer grasp of what motion looks like to them on film than they do of detail. Afterall we watched films for years on VHS with a 240 line resolution so there was no detail but motion representation was consistent. Now is there a huge difference between 3 and 6 GOP? No. But to my eye 3 'looks' more filmic to me. But that's entirely subjective. And that's what great about Vitaliy's patch, we can set the GOP (within limits) to what we want.

EDIT: From interview with "Another Planet" director -- "Mike Cahill: Sure. I shot with a Sony EX3 – which is an HD camera that shoots in 1080p."
The EX3 uses an Mpeg-2 long gop codec. Go see the film projected if you can and see how crappy the motion looks - I swear it's due to the long GOP. Then watch Danny Boyle's "28 Days Later" on Blu-Ray. That was shot on a Canon XL1 in DV 25P - an intraframe codec. The detail is less but the motion is a helluva lot more filmic. I shot some second unit with that pal XL1 for a couple years and the motion on the 35mm blowup of those shots (though less detail) always looked great and filmic.

I really think this is the "scientific test" you're looking for. Hell, even watch "Paranormal Activity" (long GOP codec) on Blu-ray and then watch "28 days Later" Forget resolution, just focus on motion. You will SEE the difference, I guarantee it.

GOP 3 won't be more precise, just different. I frames are much less efficient than P and B frames, so the overall effective amount of information per frame is reduced with low GOP. Now, there are many good things that come with low GOP that are desirable, increased resolution of fine detail is not one of them though, in fact that's pretty much what you trade for the other good things.

Chris

PS I mean resolution in the frequency domain, which is somewhat different that resolution in the strictly visual domain.

@mpgxsvcd
"I haven't been able to do a proper test of this yet. However, if that is the case then a GOP of 3 would in fact offer a more precise representation of all scenes than higher GOP values would."

So if this is in fact true, which visually it always seemed to me that it was, are you prepared to reverse yourself on 3 GOP and the very negative way you first reacted to the concept of it? ;-)

@Stray

That's exactly what I'm saying. If a bunch of noise is encoded, that detracts from bandwidth that could be devoted to other things. The encoded noise is of no value. You want to be sure not to overdo it, though, because NR can end up removing actual detail. Detail that is equal to the noise floor is virtually indistinguishable from noise - it's basic signal to noise ratio stuff.

@mpgxsvcd

That's true until you max out bitrate - then the codec will start compromising macroblock quality to avoid exceeding the allotted bitrate. It looks like in your tests the I frames are meeting target macroblock quality without maxing out allotted bitrate. The bitrate goes up with extra I frames because they are simply less efficiently encoded than P and B frames. Actually, it all makes sense to me. With a GOP of three, P frames are eliminated - that's fairly efficient. However, B frames are typically encoded with lower quality quantizations than P frames, so some detail may be lacking in B frames.

B frames are smaller than P frames for two reasons: They can use previous and future frames as references, which is more efficient that just using previous frames - I think everybody knows about this. B frames are typically encoded with a higher QP value (higher QP=lower quality) than B frames - maybe not so well known.

Chris

@cbrandin I just tried turning up the cameras noise reduction to +1 in Nostalgia film mode at ISO 640 (-2 -2 0 +1) on a GOP3 60MB hack and the most obvious thing I noticed in Stream Parser is the B frames are considerably larger. However the stream was only averaging about 35MB for what I was shooting, I will try it with a more detailed shot in a bit.

Edit : Ah no, thats related to the lowlight and movement in the shot, just did a mostly static shot (averaging a 55mb rate) with and without NR and the streams look pretty identical in the parser (the max GOP size in the NR version is a smidge larger, but that could just be happenstance). Which is to be expected I guess. Is what your saying that turning NR on is worth doing, as we'll still get detail and the codec is spending its bandwidth catching detail and not noise ?

I think the 24L isn't working right - I've seen this when all I did was change the bitrate to 22000000.

Chris

When it comes to shadow noise, I think all the low GOP does is starve I frames (because there are more of them). The effect of that will be that the codec will use higher QP values on macroblocks. The higher the QP value, the less high frequency resolution you get, and noise is high frequency content. It does exactly the same thing as lowering the bitrate does in that context. I think you might want to try turning up the camera's noise reduction. Noise reduction occurs before encoding, so the codec can use bandwidth for other things instead of consuming itself with reproducing noise. It's the opposite of sharpening where extra, fake detail is produced causing the codec to expend bandwidth on something that isn't really there and can be better done in post anyway. I can certainly buy the idea that low GOP can improve temporal response, but not so much if the codec is maxed out because of noise.

Chris

@cbrandin

I agree that low ISO static shots will not benefit much for different GOP values. However, noise plays a huge part in determining the bit rate the camera uses. You can shoot a static scene at ISO 3200 or greater and easily max the bit rate out. All of the tests I have done so far where at ISO 3200.

>Static improvements require adjustments to quantization tables... coming soon...

Wow, I wonder what can be further improved, I thought static, well lit shots are the situation were GH2 is at it's best already..can't wait
Thank you for your work Chris!

I don't know what happen with the above movie!!! Which I like a lot but got attached in my previous post.

@mpgxsvcd / Prove that with a scientific 1 to 1 comparison and I will believe it. Otherwise I haven't seen any indication that extreme bit rates are producing a visual difference. /

Here's a "comparision" … don't know whether or not this is "scientific" enough … but that is what I am also seeing in actual footage (but only in dark tonal values).

I've shot my monitor displaying Kae's torture chart.
At the bottom edge of the monitor I've put up a grey card - lit with very low tungsten light (see jpg attachment). Camera set to daylight.

I've shot 24p at 24H (set to 52Mbits; real outcome in this scene is 43Mbits) and 24p at 24L (set to 32Mbits; real outcome = 32Mbits).
1080i50 and 1080p24 GOP Size=6.

Imported into Avid MC to DNxHD 10bit.
Added a curve (quite steep) to open up shadows (so a worst case scenario for tungsten). Rendered at 16bit.
Output "same as source" rec709 (hence to a DNxHD 10bit QT), cropped the grey card (the zipped video attached shows a 100% crop), transcoded to DVCPRO50 NTSC.

First 2 seconds represent 52 (actually 43) Mbits, second 2 seconds 32Mbits.
When you look at the left part of the video you will notice that the noise pattern changes after the first 2 seconds.

The difference is marginal and absolutely doesn't matter in most real life situatuions (IMHO) but it might matter under certain circumstances ...

You can download the 4'' video here: http://www.2shared.com/file/yvQYuU3K/noise_52vs32Mbit.html (6.9MB)

@stonebat
>are you a robot?<
yes ;)