Chris's 66M AQ3 Static

Eiji's 132M GOP3 AQ2

@Ralph_B

>>But the hands down winner is 132M AQ2 GOP3.>>

Maybe I need an eye doctor. The only difference I can perceive is more noise. Please, can you tell me at what area of the image I have to look to see a difference between the 44mb, AQ4 and the 132mb AQ2?

@stonebat

Well, um,... interesting. Thanks for enlightening me!

14-800 Ultrazoom

http://dl.dropbox.com/u/39764940/I_love_my_gh2.avi

;)

Ralph,
These tests are excellent man!

1. 66Mb/s AQ4 Static Night (ISO 640) comparison:
Notice how the noise is blurred by the codec. I believe this is probably due to the AQ4 setting.
The higher noise levels are seen as motion to the codec. Since AQ4 steals the bandwidth for
static shot, the noise motion has less definiton.

2. 66Mb/s AQ2 Static Night (ISO 640) comparison: The AQ2 setting helps render the noise motion
better ending up with a closer look to the original. There is still some loss here, but it does a bit better of
a job than AQ4 under these noiser conditions.

3. 132Mb/s AQ2 GOP3 Night (ISO 640) comparison: I'm with Ralph on this one. Very impressive.
There's VERY little loss in detail and maintains damn near the same rendering of noise. Impressive!

4. 44Mb/s AQ4 Static Night (ISO 640) comparison: Seems to be close to 66Mb/s AQ4. There is some detail loss, but not bad when comparing to 66Mb/s AQ4.

5. 66Mb/s AQ3 Static Night (ISO 640) comparison: Seems to be close to 66Mb/s AQ4. I believe the 66Mb/s AQ2 setting does a bit better job on the noise motion rendering.

I'm wondering how well 88Mb/s AQ2 would of faired in this test? Any thoughts?

@proaudio4

They were posted yesterday over on the Low GOP thread: http://www.personal-view.com/talks/discussion/comment/14616#Comment_14616

@reimaginecinema I'd go with AQ4 and also use these settings http://personal-view.com/talks/discussion/comment/13692#Comment_13692 (Set J) which are stable. Tests indicate that those 44M AQ4 are pretty close to the maximum image quality. If you just change bit rate and set AQ it more than likely won't be stable as you also need to change frame buffers and limit which the settings in the link do. The 66M AQ2 settings also in that zip file are very, very good. Maybe shoot steadicam with AQ2 for better motion rendering.

@Ralph_B Thanks for these tests and thanks for sharing them. I may have missed it but what where the camera settings: Film mode and all the others, especially NR? Al

@stonebat - "The difference is very subtle. Not enough for me to switch from 44 to 132. But it's good to know the subtlety."

I want to say that for me, the difference is anything but subtle. The holy grail is a small, lightweight camera that not only shoots great details but that holds shadow detail as well as much larger cameras. This is the hard part. A LOT of small cameras give a highly detailed settings, but most fall apart in shadows/out-of-focus areas. This is what we love so much about our red when 4k is rendered to 1080. Every pixel has value, there's no throwaway areas, no slop (or very little). That (of close to that) is where we're heading with the GH2 hack and that should be the goal. Chris's 66M setting, while awesome at 160 without too much movement, is not, I don't believe, going to get us there at higher ISO's. The 132 setting is giving a file nearly twice the size, and - at least at higher ISOs - a file with more info in the hardest-to-reach parts of the image: the shadow/out-of-focus areas.

@Ralph_B - you've done so much already, but if you can test the 132 vs 66 at 160 ISO and make sure to leave areas with shadow, maybe even underexposed a bit, I think we'll learn a lot.

@cosimo_bullo

That's the plan. :)

@cosimo_bullo

I agree with you here.
If there's a visual gain even slightly, this certainly can equate to better image corrections (color grading, level changes, green screen work) in post.

Also, the comparisons you mention should be at AQ2. 66 AQ4 just gives up to much with motion.

@cosimo_bullo. I'm with stonebat the difference is subtle, and in terms of PP stips test shows little difference, nor do my own experiences so far. But yeah, this is all gut feelings and instincts really. You can't though judge quality with this codec purely by the relative file size (as in relating to the compression amount), its more complex than that, it's not like MJPEG.

Edit : I figure the small improvement just isn't worth the dramtic increase in file size, and much shorter shot length for most of us.