@Stray
66M AQ4 is stable on my camera so far (in 24H mode). I do have the feeling that IQ is somewhat better compared to 44M with fast motion and high shutter speeds... but if so, it's negligible, that's probably right.
The problem is you can't do real side by side comparisions... as long as you can't control the motion in the scene :-)
However I have not done a lot of tests with the 66M settings by now ... as I am actually really happy with Chris' 44M setting. 44M is plenty (IMO), the camera is stable, in-camer playback is fine, files do span at 24L... and finally 50i in the 44M setting looks better than 50i in the 66M setting. So actually I am not yet sure whether or not I will use the 66M at all ... I will however use a 24H GOP3 setting for its motion look. But here again I haven't had the time to look deeper into the GOP3 settings ... other than some preliminary tests.

I too feel we're glossing over a bit the dramatic difference in Ralph's tests re movement. That should be the baseline for now right? I mean, that's the only 'real world' A/B comparison we've got. Here's where I'm confused: it seems clear the 66 AQ4 doesn't perform close to the HDMI Ralph capture with movement, and so the move to AQ2; but is AQ 2 giving a considerably lower performance on static shots? Ralph, are you planning more tests? Did you, at the end of your testing, have a strong feeling about the 'correct' setting in cam for general use, for a happy medium?

@cosimo_bullo and proaudio4

I did those tests with a completely open mind, to find out for myself what was going on with the various settings. The results were pretty dramatic. The 66M AQ4 completely falls apart with motion. But then, that shouldn't be surprising. As the note says, with AQ4 all the additional bandwidth goes to detail; the implication being that motion is the same as the stock setting. Although I did not do a stock setting test, it certainly looks like that's the case.

On the other hand, 66M AQ2 shows solid improvement for both static and motion, and comes reasonably close to the HDMI for both. So, I've happily settled on that for now. It also works fine on all the variable movie modes.

If anybody comes up with a setting that they feel is superior to Chris's 66M AQ2 for BOTH static and motion, I will be happy to run an HDMI comparison test. And I agree, using the HDMI picture as a reference is the ultimate acid test. It cuts through all the muck and shows what's really going on.

Ralph did test the 66M cbrandin setting at AQ4 and as he says, it falls apart with motion, so I'd imagine 44M would be similar but worse. But I suppose it's worth trying too. Ralph I just went back to look and it appears you didn't test an 88 AQ2 setting. Do you think that's worth a shot re motion? Seems like it might be the sweet spot. I for one have no concern about using the extra bandwidth... really just want this thing redlining as max quality.

OK guys, I'll test 44M AQ4 today. I'll also try an 88M AQ2, but shouldn't the buffers and other settings be readjusted to handle the larger data flow?

Panning at ISO 160

Now I don't know how much sense comparing these framegrabs makes, panning speed obvisouly varied and shutter was 1/50 like I said, so I won't post ISO 1600 framegrabs unless someone is interrested.

Ralph, thanks a lot for your time on this!
I do have an 88Mb/s AQ3 setting I'd like you to try.
Maybe, just maybe it's enough to maintain strong static I frames and still offer enough to extra bandwidth for motion.

Here's the basic (untuned) 88MB AQ3 settings:

Video Bitrate 24H=88000000
AQ=3
GOP Size=12
Video buffer 24p=0x3600000
1080p24 T4=1080p24 T1




@stip
I'm not sure if you saw my earlier message regarding high noise levels and how "I believe" the codec responds to it:

As I mentioned earlier, "I believe" In a static scene, higher ISO settings (increased noise) will demand the extra bits to process more motion vectors. In a static high ISO (noisy) shot, the I-frames get smaller as P and B get larger. In static low ISO (low noise) conditions, I-frames get larger and p and b frames get smaller.

@proaudio4

To avoid any confusion, could you post your 88M ini file?

in stips 1600 pics there is a lot of difference. Not that I'm all that clear on what you're doing in after effects, but the colors are markedly different. Compare the left side rooftop. the 132 has a whole lot more yellow splotchy areas.

I'm not sure what that means, if there is more color or detail info that ae is trying to differentiate.
They are different, don't know if that means one's better than the other.

Yeah, if anything that 132Mbps 1600 image looks like it would give you more grief in post, but as you say the artifact size looks identical. As does gradation.

Also - Where are we at with the cadence thing? Are there any high (or high-ish) bitrate settings that are immune at this point? Has anyone decided the cadence thing just looks bad in streamparser and doesn't matter in the real world?

ALSO - just a thought, but for those of us without the ability to capture HDMI cleanly, what about A/Bing AVCHD against very high 1080 MJPEG. I know the MJPEG starts with less resolution, but might it hold up better than AVCHD and AQ4 with lots of movement? Just a thought, and if given a chance (family...), I'll try to test later today.

AND finally... if this AQ thing turns out to be unresolvable (finding the perfect mix), I wonder about just keeping two bodies around all the time, each tuned for a different thing (66M AQ4 for interviews, 88 AQ2 for b-roll?).

Here's a new round of HDMI vs AVCHD testing. Pictures are all lossless.

The contestants -

Chris's 44M AQ4
Chris's 66M AQ2
88M AQ2 based on Chris's 66M.

This time I increased the shutter speed to 1/2000 sec to freeze the action and throw as much information at the encoder as possible.

OBSERVATIONS

All three static pictures are so close to each other, I can't make any definitive statement about them. The 88M might be a teeny weeny closer to the HDMI, but in a blind test, I'm not sure I would pick it out. It's also possible the 44M picture might not have quite as fine noise, but boy, it's subtle.

On to the motion...

44M AQ4
Motion is suprisingly good. There is a very slight loss of detail compared to the HDMI (but not much).

66M AQ2
Very, very good. This is the best motion of the three and comes closest to the HDMI.

88M AQ2
Suprisingly, motion is not as good as 66M AQ2 (look at the little orange flowers). This may be because this setting hasn't been fine tuned. I only changed the data rate of the 66M setting.

First pair of pictures is 44M AQ4 Static.

66M AQ2 Static