@vladnik tried loads of tests to get stable AQ3 or AQ4 at 100mbps on pappas death chart, most failed after about 2 - 10 seconds. @danyyyel try out either my 100mbps or 140 mbps short GOP 3 settings on the LOW GOP thread with 3.62 - they are stable and preferable to most people over the old best GOP3 settings of 3.61.
Yeah, I like @driftwood's work on short GOP very much. It's sort of the opposite of what I'm doing with the the patches I posted. Depending on what I'm shooting, I'll be using different patches. I guess I should have said that for short GOP work, I'll be using @driftwood's patches (or a derivation thereof). I plan to, however, make use of standard GOP patches as well; and I'm working on making them as good as possible.
Actually, your 100Mb at AQ 2 should have the same frame quality (approximately) as my 44M at AQ 4. AQ is not an absolute quality level - it is scaled according to the bitrate increase. Now, at 66M, AQ4 will produce significantly better individual frame quality.
I want to max out quality and stability at 44M first. At some point lowering QP won't do anything any more. Eventually I plan to start increasing bitrate in order to determine when there is a diminishing return. My guess is that will happen by the time 66M is reached and going higher will not produce better results.
For me, the occasions of gradient banding in areas of subtle tonal and chroma shifts over large areas like blue sky and blank walls are the Achilles' heel of the GH2 video image. I am finding @cbrandin's 44M settings are producing less banding. The banding has not been eliminated entirely but it appears to be noticeably reduced.
I attach two PNG images of typical images that would produce a more pronounced banding effect with the default firmware as well as some of the high Mbps/low GOP settings being tested, at least so far as I can tell. I have an allergy to that banding, so I am very aware when and to what degree I see it!
I guess a simple tactic would be keeping I-frame high enough and QP low enough. Once we reach I-frame large enough, we wouldn't gain much. Then shortening GOP makes sense for higher bitrate settings as lower GOP demands more redundancies which may not be a bad thing. The more I-frames per second, the less motion aliasing... although not proven yet.
44Mbps GOP12 66Mbps GOP12 or GOP6 88Mbps GOP6 or GOP3 110Mbps GOP3
All those settings might have almost same image quality on I-frames, but the higher bitrate gets more I-frames. So the diminishing gain in image quality might become non-issue for those who prefer shorter GOP. I think each setting would be perfectly fine, and we have more choice.
BTW it would be nice if 24L and 24H can have different GOP value so that I can have 44Mbps GOP12 and 110Mbps GOP3.
@cbrandin I am using your setj.zip setting posted from 27 August. On 24H mode, I have no issues with file spanning past the first 4Gig size. I am using PQI class10 16GB SDHC card.
I wanted to share these settings. GOP is unchecked 24H -88 24L -66. AQ4 The 66 setting seemed to use more of it's bit rate but have to shoot more at 88 to really make that call. I've posted a earlier 24H@66 but these settings use more of the bit rate. I've been using the 66 at 24H and just changed to the higher bit rate. The footage looks great. I've shot bushes trees bees dog and cats and a number of humans. Load to FCP no problems. I'm wanting to try the new low GOP for the motion and to see the difference if any. thanks to everyone Luke
Here are some updated standard GOP settings. One is at 2x(44M), and the other is 3x(66M). As I expected, 4x (88M) does not improve on 66M significantly, so I am not bothering with those. As before, there is still the spanning issue sometimes.
@cbrandin do these settings apply to all modes? I used your last set on some test shots for an upcoming project and noticed some pulsing at regular intervals on 720p60, otherwise they looked great
@cbrandin I could not produce an uneven cadence with your settings, both 44Mb version looks great in stream parser and on screen, lucky we have you, thanks ;)
Though I do not firmly understand the theory, I changed a little and tested Chris's 66Mbps setting.
It was possible to record continuously during 30 minutes in 32Mbps(24L) though it was possible to take only up to 4GB with this setting in 66Mbps(24h). Both, there was occasionally what was not able to be played in the camera. However, it could play after turning off and on camera's power.
Mixing up these settings won't work. The 44M settings do have 32M at the 24L position. Also, 32M works fine with just the user settings bitrate and AQ.
When I first used 66m setting, 24L was not able to be played in the camera. Then, I combined with 32m though I understood the difference from what you intended because it was inconvenient. Is it still useless?