I know you don't want to hear it but the low GOP doesn't work right - Vitaliy is working on it. We know what the problem is - it will just require a little time for him to work it out. Then, and only then will you get what you want. This is a little like trying to drive a car with a flat tire - it just doesn't work very well - no matter how much you want it to.
Just got back in town. 3.62d potential looks great! Huge thanks to V and C!!! I haven't found a way to get stable high data rate death charts yet with 3.62 and 3 GOP 24H. I tried playing with 66mb and triple fb but just produced unusable low data rate footage. I tried the old 65mb extreme settings and set aq to 2. Death chart footage was horrible. So am sticking with old 3.61 65mb settings until someone can point me to stable 60mb or above 3 GOP FSH settings.
Also, I guess I don't understand the benefit of big B frames in 3 GOP. If the B frames refer to differences between it and the previous and next I frame and that I frame is just one B frame away in 3 GOP's case why would the B frames need to be huge? It seems to me in 3 GOP's case the I frame is the most important since it comes 8 times a second so the bigger the I frame the better. Or am I missing something technically here?
I guess I'm looking for someone to get as high of stable data rates as those shown below (death chart on 65 extreme settings with 3.61). I'd even settle for a 60mb constant stream with 3 GOP - Is there a way with 3.62 yet? And please don't answer 'set GOP to 12 or 6', I'm interested in 3.
I wanna keep going... but I have like only 30 mins each day during daytime. 15 mins in the morning. 15 mins in the evening. I'd be busy pushing and pulling my kids bikes. I got time at night but can't test this at night. That's why I didn't volunteer for AVCHD ST team.
It will take me weeks. Hopefully some people with more time can pick up the slack.
The problem with not setting a high AQ value is that the additional bandwidth does not contribute to individual frame quality. I'm not sure adding a lot of bitrate without also setting AQ to a high value results in improvement that is very visible. High AQ values result in image improvements that are clearly visible. I suspect you would be better off setting AQ to three and then finding the highest bitrate it can support, rather than finding the highest bitrate and then the highest AQ value that still works.
Higher I-frame size means more detail, but that implies B-frame has harder time to catch up with change between frames.
Taming I-frame size gives more allocation on B-frame. Lower I-frame size gives more room for shorter GOP. But high AQ value might trigger frame size going wild causing camera failure. But lower I-frame size might give lower low lighting performance.
That makes sense. I think the codec sets up a budget for I frames per second. A higher proportion of I frames would naturally result in smaller individual I frames. The B frames are simply allowed to get big enough to use up the additional bandwidth.
In low lighting, GOP3 produced two B-frames that were slightly bigger than one I-frame where each I-frame was about 5Mb.
As GOP size increases, it increased I-frame size where B-frame size remained about same with spike in P frame size. So that increased I:B ratio. But I think short GOP people prefer lower I:B ratio. If we find a way to increase B-frame size, that would demand high overall bitrate which would increase the chance of failing. Since GOP3 gives smaller I-frame, it may actually stabilize the system. In that case, AQ3 might be too high. It failed my highly detailed scene test. I will give one more try on 88Mbps GOP3 AQ1 24H.
3.62d GOP3 is producing very big B frames. It seems logical to stretch GOP longer. Added benefit would be better blending between 24p GOP6 and 60p GOP6.
Since the release of ptool 3.62 this thread has gone very quiet and I have seen a lot of talk towards moving from 3 to 6 gop. Is it not defeating the idea of the low gop look and does 3.62 work well with 3 gop and lower. No news from Kae and Butt about some high and stable settings.
@cosimo_bullo So you are able to have stable and high bit rate (how much) 1 gop recording with the ptool 3.62 ??? So if this is the case, Vitaliy has invented the first intraframe avchd codec :-)
I know that this is low GOP topic. But you can now also try double GOP. I mean 24 instead of 12, etc. And PTool also will recalculate things inside. Hints for GOP patches have all supported values listed.
Clipwrap however is chocking in the files and creating a strange on output not matter if we rewrap or convert to prores. - however Final Cut Pro 7.0.3 was able to ingest the files fine and they play back correctly.
GOP 1 is a revelation! Now seems to leave mjpeg in the dust grainwise, and for what it's worth, we've yet to have any crashes this evening even shooting detailed scenes at 2500iso. We'll try to do a detailed vs. test tomorrow.