GOP vs. Bitrate

I just did a bunch of tests today where the GOP was 1/4 the frame rate (720 60p = GOP 15) (1080 24p = GOP 6). Now that I have a few settings that are close to stable I will try GOPs where it is half the frame rate (720 60p = GOP 15) (1080 24p = GOP 12) and see the difference tomorrow.
 
Bitrate will make a bigger difference, in theory, but it depends on the application

For higher motion situations, a smaller GoP will result in more I frames that contain all the information, so it means that the codec will be better at resolving the large differences between frames. However, a small GoP with a low bitrate is basically useless, because that means that EVERYTHING loses information to the video down to your specified bitrate. But a larger GoP (like 12), and a high bitrate should be good for most applications.

Oh, and when I say "high movement" I basically mean where every frame is drastically different from the last. Even in myt high motion tests, a higher bitrate produced more detail and better image quality than lowering the GoP and raising the bitrate. C settings with a god of 12 seems to work best, or so I think so.
 
I just ran these:

No native 24p
"C" settings
GOP=10
Camera gave card speed error on very fast panning. bitrates avg around 30mb/s

No native 24p
"C" settings
GOP=12
Camera worked ok. Bitrates avg around 20mb/s

No native 24p
"B" settings
GOP=10
Camera recorded ok, would not play file back in camera. Bitrates avg around 20mb/s
 
Try to stay with multiples of 24, so 12, 8, 6, 4, 2.

I don't think anything under 6 has been stable so far.
 
Tester said it first I believe, some other people have mentioned that since it's doing I-frames within the stream, that it can get weird if it doesn't fit into 24.

That's why when it's in the 60i wrapper the GOP is set to 15 which is a multiple of 60, while the 24P patch automatically adjusts it to 12 when it's activated.
 
Tester said it first I believe, some other people have mentioned that since it's doing I-frames within the stream, that it can get weird if it doesn't fit into 24.

That's why when it's in the 60i wrapper the GOP is set to 15 which is a multiple of 60, while the 24P patch automatically adjusts it to 12 when it's activated.

I thought I read something from Vitaliy a while back saying we SHOULD change it from 15 to 12 if we go with the native patch. I never read anything from him saying that it automatically changes it. I can't find proof tho, it was a while ago and I just spent the last 30 min searching for it. If you can find proof to back your claim that would be great. I've been checking these threads daily for weeks, but I guess it could have slipped thru.
 
Last edited:
I thought I read something from Vitaliy a while back saying we SHOULD change it from 15 to 12 if we go with the native patch. I never read anything from him saying that it automatically changes it. I can't find proof tho, it was a while ago and I just spent the last 30 min searching for it. If you can find proof to back your claim that would be great. I've been checking these threads daily for weeks, but I guess it could have slipped thru.


It was the other way around, he said the 24p patch changed the GOP to 12 and we should put it back to 15. However, this was in reply to the failures seen when using the Native 24p patch before it was learned that it is a bug and is not related to GOP length changes.

anyway, I tried "C" settings and GOP=12 today with the morning sun on the trees and the MTS files bitrate was much higher than usual, 45mb/s where the GOP=15 bitrate is usually around 25-30mb/s. This was tested on a Patriot LX Class 10 and I never once got the write speed failures.

I couldn't tell much difference in the picture by just peeping pixels but the difference I do see seems to be in the form of more detail in the shadows.
 
OK so I just compared the "C" settings with GOP=15 and GOP=12 in the woods.

GOP=12 is better than GOP=15 by a good stretch. It's not the detail in the foreground, it's the details in the background that are the difference for sure.

GOP=15 still has some blocking and mud while GOP=12 has less blocking and mud. The bitrates are definately higher by about 15mb/s on average. I had no card speed errors on a Patriot LX either on the tripod or handheld(normal and samurai).

I have to go do some test shots of a garden for an upcoming movie and I'll definitely be needing to peep these shots to make sure the codec holds up for production.

I still need to test with a Class 6 card as the highest bitrate I saw was 47mb/s which is still theoretically below class 6 speed.
 
The camera just locks for me when I try GOP 6, 60.66.70.. I'm about 85-90% stable at those settings with GOP 12
 
Last edited:
The camera just locks for me when I try GOP 6, 60.66.70.. I'm about 85-90% stable at those settings with GOP 12

GOP 12 with 60-66-70. Is that your formula? I used the 60-62-70 after it was said to be good.

In one experiment I tried one that was higher, however the image was worse then way lower numbers. So what's in the numbers is the question..... Were all trying to find the sweet spot. The best numbers are those that are going to perform excellent in low light level conditions- if it compresses well in those conditions, then were good to go.

Nothing wrong with GOP 15 either. Many systems use 15 GOP- Now with the 24PN- I believe it was said that it becomes 12GOP when 24PN is selected, so maybe we should set if for 17 GOP, and then it will be 15 again in 24PN...?

Pappas
 
OK so I just compared the "C" settings with GOP=15 and GOP=12 in the woods.

GOP=12 is better than GOP=15 by a good stretch. It's not the detail in the foreground, it's the details in the background that are the difference for sure.

GOP=15 still has some blocking and mud while GOP=12 has less blocking and mud. The bitrates are definately higher by about 15mb/s on average. I had no card speed errors on a Patriot LX either on the tripod or handheld(normal and samurai).

I have to go do some test shots of a garden for an upcoming movie and I'll definitely be needing to peep these shots to make sure the codec holds up for production.

I still need to test with a Class 6 card as the highest bitrate I saw was 47mb/s which is still theoretically below class 6 speed.

Did you shoot in 720 or 1080?
What would be the best GOP for 720?
 
Back
Top