@Kae -- hope you get better man! Your 3gop setting was superb on my last shoot.

You beat me to it by a few minutes but can also confirm no problem with spanning, Avid import or playback on mac (with MXplayer).

@Driftwood, have you played around at all with the 720p 60p settings? I have to use that a fair bit and was curious if you had any starting points for experimentation there.

what are record time looking like... IDK the conversion say an 8gb card

Hi @kae, at the moment the bitrate you'll notice is being held back on a limiter... I wanted to 'get there' somehow, this is now achieved, but now its down to fine tuning and it'd be great if everyone could get involved more and help make this work - you know - really push the gh2 to its limits and get quality. If we can't, then its back to good old GOP3, which, because of this test will only improve too. Even great 6 GOP hasnt been truly established yet.

Whether we get it to span like this latest setting is another matter, because with these encoder settings we can now (hopefully) rid ourselves of the dreaded 2nd frame 'P' from GOP 1.

We have to seriously look at what the encoder settings are doing and some of the other ones that Vitaliy has provided. Its taken me a day and half of tests just to get this stable. Ive made some crazy errors - I actually thought I'd found 25p at one point in 1080i50.

Ive got 50i, 72050p stable and 1080i60 (all at GOP1) but indeed, poorer quality than GOP3 upwards. By achieving this, the hardest setting first should in theory help us go beyond and and improve the quality. And no doubt I believe the GH2 has got more in the kitty to offer. Who knows what the next ptools could offer? I know people hate the high bitrates, but GOP3 hi quality and certainly GOP1 needs the bitrate - theyre hungry beasts!

The GOP1 24p is the best of the bunch Ive tested and Ive filmed with it daytime today and it looks very good in daylight. However, I would say in darker settings this will not perform anywhere near as good as 132M GOP3, but if we can get it close - all the better.
In a few weeks we will be beyond this stage, I sincerely believe that (perhaps a few days ;-) who knows?)
It's been fun getting there and you @kae have encouraged me with your original GOP3 to go beyond - sometimes the simple answers are there before you - this is why I suggest to all testers to try everything, don't give up, dont rush and be methodical - and write it down/spreadsheet it. It saves acres of time and you have something to refer to later.
Anyways, Im too tired to finish GOP1 720p60 (it probably wont be worth it - but who knows?!)

What we need to do is establish the limits and then work backwards.

This, I believe, is a great start. PS Yeah congrats on taking on the bigger mission - life!

Thanks driftwood! Give us our marching orders for perfecting this tomorrow. Can't wait!!!

@driftwood
Dude, I go away for a week and you completely rock it! Amazing work! GOP1 like butter... and in every tool, no less. Just bloody fantastic.

Now, how can we get the bit-rate up? :D

I'm excited

Just as an experiment I set my 88M max variation to GOP1 to try and confirm some things, I changed nothing else except the GOP size. It isn't stable on deathcharts unsurprisingly and does have the rogue P-Frame. However, it has confirmed something to me about bitrate usage at GOP1, which is it stays high even on AQ2 in lowlight. On the shots I took it was always running at an average very close to the maximum, so if you are running at GOP1 don't worry too much about pushing AQ. I expected this to be the case, but hadn't got around to seeing if it was true.

On all shots the I-Frames collapse in size at around frame 16 to 20 consistently. so thats the point where it runs out of bitrate and adjusts I-Frame size downwards. this could also be due to it being a max variation patch it allows the codec more possible QP settings. From that point it does remain consistent for the rest of the shot (it being static). Notice in the lowlight graph that if it starts at a lower size it remains consistent. This fits with Driftwoods clamping of the I-Frame size solving the problem. What I feel is that a GOP1 with I-Frame sizes comparable to the 244M is therefore possible at a much lower bitrate than the 240M (as its pretty close to the 244M results in the lowlight shot to the 244M deathchart test). But yeah, the relationship between I-Frame size and bitrate is going to be exponential in some way I guess at GOP 1. Now, I'm saying this without trying to cure the rogue P-Frame issue, which again could also lower the I-Frame size more, but I don't know if thats true yet.

edit : Oh yeah, I didn't move the camera about, that definitely needs doing. Will upload findings in a few. Answer : It gets a bit messy, exhibits cadence problems, however the positive from it is that it is capable of spikes in I-frame size back up to large I-Frames when required (but can't hold them obviously, hence the cadence).

@disneytoy -- oh were it so. Remember a few things: first, bit rate is far from everything -- the alex is an $80,000 german engineered small supercomputer, combined with a far more advanced cmos sensor that is able to extract an incredible 14 stops of latitude. From my testing, the best hacks may have been able to squeeze out another stop of latitude, but that's it. Even at pro res 422 HQ you're getting 10 bit colour, which means millions of colours instead of 256. It's a big difference. And finally, we're stuck at 420 colour.

The work of driftwood and others is amazing but we should be careful about going to far with comparisons. From my experience the gh2 is still struggling for acceptance in the professional market and overstating our case will not help it knock off the 5D :-)

PS: We're also getting off topic. This should only be for testing and posting low GOP hacks, not comparisons to other cameras. If you like, please post something on my thread comparing the gh2 to arri to f3 and I'd be happy to discuss more.

Im interested in seeing this new GOP1 244m footage converted to cineform raw 4:2:2 10 bit...

@Mihuel that should be something AQ does, or setting the QP manually. However at GOP1, from what I can surmise limiting the lowest value shouldn't be necessary (it seems to try the max QP setting for I-Frames as it is, within the range it has been given). Plus if you could it would just increase the chances of instability, and increase them considerably too. You've got to give the codec enough of a range to adapt within after all.

@driftwood Yep its a good strategy of driftwoods, basically go for stability at GOP 1 first (and kill that P-Frame) then tweak it afterwards to get decent image quality and bitrate usage. My own instinct is AQ and QP are not that important when running at GOP1. I also reckon we'll end up with a bloody good GOP 1 at a lower bitrate.

GOP1 footage, dont know who from...

I agree, the 1 GOP looks great. I hope we can find a way to bring up the i-frame size and maintain stability.

ITs there any chance of making GOP patches for GH13 ?