Tracey Lee
Well-known member
I left those unchecked and I am still getting pretty amazing avchd footage. Just thought I would ask in case they make a difference. Thanks!
What's the GOP settings for?
- Thread starter Tracey Lee
- Start date
Tracey Lee
Well-known member
Does setting them lower than stock settings make for less compression/better image?
John Caballero
Veteran
Tester13 said that when you do the 24p Native patch it gets automatically lowered to 12. Some people have experimented with lower settings and got good results. What I understand it does is that it has less frames to compress at a time apparently producing better compression. Of course the only ones that would know better is Barry and tester13.
Barry_Green
Moderator
"lower is better" only if you have the bandwidth to support it, and if each i-frame is capable of delivering adequate quality.
The longer the GoP, the more efficient the codec *can* be. I won't say "will be" because with this codec, all bets are off, the way it's behaving...
But -- you could potentially get much higher quality footage at the same file size, if you made the GoP a lot bigger. And hey, that's part of what testing is for. Set it to 30 and see what happens. Set it to 60 and see what happens.
I wouldn't be surprised if you weren't able to make it much bigger, because there might be some sort of RAM or processor speed restriction. In traditional camcorders, the camera compresses ALL the frames in the group, simultaneously. That's why you'd get a half-second delay on streaming footage out the firewire port, because the GoP was a half-second long, and so it wouldn't do any compression until it had all the frames for each half second group, and then it'd compress them all together.
With the GH1, I don't expect it to work that way, because it doesn't have B frames, so it's more likely a very simplified system of each P-frame being a delta frame from each previous frame. If so, making the GoP longer probably wouldn't really do all that much to help codec efficiency.
But try it. We're all still in the testing phase. Go crazy. Set it to 128. Set it to 2. See what happens with each.
The longer the GoP, the more efficient the codec *can* be. I won't say "will be" because with this codec, all bets are off, the way it's behaving...
But -- you could potentially get much higher quality footage at the same file size, if you made the GoP a lot bigger. And hey, that's part of what testing is for. Set it to 30 and see what happens. Set it to 60 and see what happens.
I wouldn't be surprised if you weren't able to make it much bigger, because there might be some sort of RAM or processor speed restriction. In traditional camcorders, the camera compresses ALL the frames in the group, simultaneously. That's why you'd get a half-second delay on streaming footage out the firewire port, because the GoP was a half-second long, and so it wouldn't do any compression until it had all the frames for each half second group, and then it'd compress them all together.
With the GH1, I don't expect it to work that way, because it doesn't have B frames, so it's more likely a very simplified system of each P-frame being a delta frame from each previous frame. If so, making the GoP longer probably wouldn't really do all that much to help codec efficiency.
But try it. We're all still in the testing phase. Go crazy. Set it to 128. Set it to 2. See what happens with each.
Barry_Green
Moderator
Depends on what you mean by "banding". If you mean 8-bit dithering, that's already been helped a lot by a bigger bitrate, and as Kholi says, juicing the NR down and upping the ISO to 320 will take care of just about any banding that remains.
If you mean the vertical streaking, that's a sensor issue that would have nothing to do with the codec.
Yes, I meant the dithering. I'll have to do some tests at 320 with NR reduced the next time I have a chance to play with the camera. Thanks for answering.
4:2:0 shouldn't be a factor in the banding. 4:2:0 still has an 8-bit bit-depth, just like 4:2:2.
My best guess is that it comes down to the CODEC precision and/or how much noise is in the image. Now maybe shooting at 4:2:2 causes the CODEC to behave differently and indirectly affects the banding. Maybe I'll test that. But I'm not holding my breath.
Counter-intuitively, it may be that shooting at a higher ISO improves overall image quality in some cases by introducing enough noise to minimize the apparent banding.
FWIW, I have some specific shooting I'm doing where I often have blue sky in the background using optics that have very mild vignetting. This reveals the GH1 CODEC's shortcoming all too well. What I've noticed is that shooting 720p at the "c" settings and GOP = 15 does not solve the banding problem. Shooting at 1080p24 with the "c" settings and GOP = 12 does seem to make the banding less apparent. But shooting with MJPEG is the only thing I've found that delivers these shots with no banding. Of course, it uses a different and much less efficient CODEC and while you can bump the quality up quite a bit, you can't go very far before you lose the ability review (play back) the shots in-camera.
On a final note, I haven't thoroughly tested the banding I'm seeing. It is possible that it is an artifact not so much of the recording CODEC, but of the decoding and playback capabilities of the GH1 camera.
PappasArts
Veteran
That's probably why the banding/gradient artifact is even in black & white AVCHD
That's dithering, and I do it all the time in post, to avoid banding which happens on very clean files.
My findings exactly- MJPEG records a very robust image at high bits- The Gamma on Night scenes or shadows can be lifted without having that much banding/gradients artifacts. Also you can convert those to B&W and shots are clean.
Like this lowlight shot- In AVCHD the shadows would have more artifacts compared to the MJPEG.
http://www.dvxuser.com/V6/showpost.php?p=2017000&postcount=93
Low light shots done in MJPEG- I can't get the same smooth tone in the dark's with AVCHD. It could just be the formula- we just haven't tagged it maybe.
I tried GOP 2 with "C" settings- On Freezes, the camera will retain the footage up to that point- and the camera doesn't need to be shut off. Problem is, that 24PN and "c" settings don't do well. And a LOW GOP on 60i/24wrapper is not good.
GOP 4 with "c" 24PN are slightly more stable then using GOP 2. Still high detail stops the shot and this freezes the camera- and then needs to be turned off.
Pappas
Last edited:
x_WOrPIG_x
Well-known member
I shot stuff last night with a GOP of 4. But as soon as I went outside this morning to shoot high detailed trees and bush + sky it froze every time it went to 5 seconds. I'm just sticking with the 50 52 60 24pn setting for now.
PappasArts
Veteran
It's frustrating, that's for sure.. So your sticking with gop15-"C" settings and 24PN?
x_WOrPIG_x
Well-known member
Yeah. It is truly the all around best AVCHD setting. I tried a hell of a lot of variations. And the " C " setting is the one to go with IMO.
PappasArts
Veteran
I'm using 1080p24 (not wrapped in 60i) at a GOP of 12 and the "C" settings. At 720p I'm using a GOP of 10 (thought I had it at 15, but think I typoed the setting on Ptools - I intended 15). I can get the 1080p24 to freeze, but I have to try really really hard to get it to happen. It requires a very high detailed scene (couldn't do it with the star chart and my laptop screen), a very stable tripod based shot with no motion in the scene, low ISO and noise reduction turned up. As a practical matter, it is stable for me. My cards write at 10.5MBs or better.
I'll keep shooting at these settings as they seem pretty darned stable in real world use and see if I can get a real world non-stress tested failure.
I've tried some GOP settings on B and C settings and I get strange macroblocking and very low bitrates for a few GOP settings above and below 15.
I don't think these settings are dynamic in any form. I think there is indeed a formula to them all and changing a few things can/will change the way other routines are run as well, thus the reason we have so many different problems even though we only change a few different things.
I don't think these settings are dynamic in any form. I think there is indeed a formula to them all and changing a few things can/will change the way other routines are run as well, thus the reason we have so many different problems even though we only change a few different things.
PappasArts
Veteran
Totally agree- As so the "C" settings etc reflect on the quality. I tried high numbers,however there was a lower image quality. You would think the opposite. I wish we had a powerful computer that could automatically run routines to seek out the best combination for AVCHD and MJPEG settings.
Pappas