@towi

Unfortunately, the AQ settings are common for all 1080 modes (high, low, progressive, and interlaced). To preserve a good balance you would have to raise all 1080 modes in the same proportion. That is, if you double 24H, then you should double 24L, FSH and FH as well.

Chris

Well, you could get some instability with those settings. AQ = 2 would avoid that. Now I wonder which bitrate setting controls the AQ parameters for 1080? Vitaliy?

Chris

@film_noir try GOP12 instead of GOP24.

@cbrandin

"I haven't tested my settings under enough different environments to be confident that I haven't just been lucky."
same here ... the settings need much more testing. Unfortunately at the moment I don't have the spare time to run multiple tests ...

I am looking forward to hearing your findings with my settings!

' = minutes
'' = seconds

for instance 01'30'' is one minute, thirty seconds ... or simply 90''

:-)

@cbrandin

Can you post your current settings as a .ini file? I am more than happy to join your testing.

@proaudio4

"I see the improvement in overall detail and low light detail (shadows)."

Can you show us a before/after comparison?
I would like to quote cbrandin:

"Improvements to image quality are readily visible. Absolutely not true."

Maybe a placebo-effect?

@cbrandin "First, I want to test my simple settings under more circumstances and see if I can make the uneven cadence happen reliably."

Weird, I can make it happen consistently with a static indoor shot that has a reasonable amount of detail (a big pile of my kids toys, cuddly animals etc) at AQ 3 or AQ 4 at 66Mbits and at 44Mbits and 48Mbits. I have only doubled framebuffer settings now, tripling them for 66Mbits didn't make much of a difference, nor did lowering QP. Its predictable in the sense that it will happen eventually within 3 minutes. The shot is lowlit, I don't have NR on and I'm using ISOs in the range of 400 to 800. Writing to a Sandisk Extreme Pro 45MBits card. If it is a buffer fragmentation problem, as I asked ealier, would making the video buffer actually shorter help, as in it would flush it more frequently ?

@sohus Maybe the camera gets bored of our horribly dull shots and starts dicking about with its encoding strategy to stop itself from falling asleep.

@cbrandin

Cheers, Interesting, yeah I get your theory it makes a lot of sense to me. Right then, what I'll do tomorrow is go back to an AQ4 set up and shoot both static and moving shots, then I'll try some heavy post work on them. If there is a change in IQ I'll defintely notice it where the cadence change happens. I'll also output the files as a sequence of TIFF files from Premiere CS5.5 and take them into Nuke where I can push them further. If I get the cadence problem during a shot with motion then (as I have done once) then it'll be interesting. However, I do remember the shot in question that did that was a camera movement, and not an actually motion within the subject. Sooo.. I guess it's possible that as the camera is recording a hell of a lot more data than it does normally it really doesn't need the frames to remain as high as they were to maintain IQ consistency. I like this theory, maybe it also explains how a lower GOP causes it more consistently, as the I frames themselves (containing much more data in themselves) are more frequent. But then again I frames are encoded stand alone without reference to any B or P frames right ?

@Angry_C>"Can you show us a before/after comparison?
@Angry_C>"Maybe a placebo-effect?"

Yes I can.
It's high ISO 2500. This test really shows how well 3.62d 66Mb/s AQ4 can really resolve fine detail such as noise.
Also, the flower pedals show more detail.


And Yes, Chris is right. It's not always easily visible, especially in static shots. But throw it on your timeline and
start grading and you will see who causes the most issues.

Also, let's face it, the GH2 stock firmware looks great! Afterall, this excellent image quality is what sold us in the first place.

Thank you! Great example of the huge improvement in the codec.

@proaudio4 so x3 with no gop changes? this is interesting... trying it out right now. I have been in the middle of a shoot this week and last but with rain delays, ill even switch up mid shoot if its worth it!! The png is very promising indeed, thank you! i actually went stock due to so many inconsistency in my tests and reports here =(