AVCHD maximum image quality settings and testing

driftwood

Chris - Not disputing the tuning. All I'm trying to ascertain if all the fine tuning is neccesary on the VBR!

chrimsbroome

driftwood - ok cool

cbrandin

@driftwood

I hadn't assumed any dispute:-)

I was just trying to explain the difference. My goal for my settings really come down to one issue. I assume that static IQ is limited by maximum I frame size. The purpose of the tuning is to allow I frame size to go above the 1MB limit imposed by some default settings. Other parameters were changed to stabilize the result.

Tuning certainly is not necessary - unless you want I frames to be allowed to get bigger. The difference between my tuned settings vs. untuned settings is that you get bigger I frames with the tuned settings - that's all.

Chris

driftwood

Like I said before, I think I speak for everyone, we all need to grasp which and what tunings will work together with each and all of the settings that vitaliy has provided us with. Your 44M is impeccable. I believe my 132M GOP3 is too. Its the middle ground now I am working on. Spanning & cadence is the main issue for a lot of people no matter what the setting. I think there's been some interesting testing recently by all here. We're all getting there.

PS Can't wait for the picture settings you're working on.

proaudio4

driftwood,

What am I missing here I see the elecard invalid values on the tuned 44M settings?
Why the errors and none in the larger 24 buffer (0x3600000) untuned?

driftwood

They all have invalid errors - even the Panny default :-)

proaudio4

thanks drift.

I did not realize this.

driftwood

Ive posted to elecard about these errors. No response so far.

proaudio4

word

Stray

@driftwood thanks for the heads up on the buffers with the 132M GOP6, will push it up to 36000000.

@proaudio Yes, I've seen exactly the same drop in I-Frame size when using the T4 to T1 cadence fix, if IIRC it would always happen around the same time in every stream. Worked through this a bit with Vitaliy (can't remember the thread its in sorry, will dig it out later if needed), changed some timing setting which should have resulted in moving the drop in I-Frame size further on into the stream but not remove it. It actually only moved the drop further on by a few frames, nothing substantial.

I'm liking the idea of an 88M AQ2, I agree with the argument for it, just to avoid bitrates from dropping too much.

daimon

@Stray Thanks for your comment. So GOP3 is ideal if you want to keep the P en B frames. Just shot two cards with the 220M GOP1 AQ2 and 220M GOP6 AQ2 at the seaside. All GOP6 shots are 217Mbits according to Neoscene. Will try to show something later. Thanks again everyone for the good work.

driftwood

@cbrandin what's the chances of you including a video/picture buffer analysis within streamparser? too much work probs?

pingpongcomputer

I shot a short last weekend with cbrandin's 44/32 settings, I used mostly 24L and 720p, no errors, no spanning problems on 24L, no drops in IQ. We shot in a broad range of locations, interior and exterior, and the settings held up great everywhere. Working with the footage now transcoded to prores 422 and everything looks great. I'll post a link when editing is done. Big thanks @cbrandin

driftwood

720p60 55M/48M AQ4 should work perfectly. Change AQ =2 if you do get high shutter /iso problems. This is rock solid on the chart and is still brilliant IQ.

Stray

@proaudio http://personal-view.com/talks/discussion/comment/12058#Comment_12058 and there is more talk about it a bit on page 4 of this thread. This was with an untuned 44Mbps setting IIRC. Altering timevalue as you can see on that thread did shunt the drop about 5 seconds further on. Maybe the I-Frame drop when setting T4 to T1 is related to bitrate as I don't remember seeing it happen on the early 132Mbps GOP3 driftwood posted that also used it (or was it eiji). Also, as it doesn't happen with a tuned 44Mbps maybe that could solve the problem for you too and you can skip using T4 to T1.

proaudio4

Thanks Stray.
Somehow I missed that entire thread. lol
Yes, this is what I'm seeing with the replacement T4 to T1.

driftwood

I havent seen cadence on 132M gop3 with or without t4=t1

neveraholiday

@driftwood
those chris's 44M-settings give me onle one time spanning (2 clips with app. 4.29 Gigs). Then it stops. Does anybody get more minutes out of it? Tried with 32gig extreme 45MB and 16Gig 30MB Sandisk.

Butt

@Stray
>Obviously to get a GOP1 with really good IQ is always going to require a large bitrate, as GOP1 can't take advantage of B & P frames to maintain IQ.<<br />?????
nonsense - GOP1 is AVCHD Intra: No one here needs b and P frames

stoney

> 720p60 55M/48M AQ4 should work perfectly. Change AQ =2 if you do get high shutter /iso problems. This is rock solid on the chart and is still brilliant IQ.

Thanks @driftwood!!!

Stray

@Butt It is not AVC-intra it is AVCHD-intra though ;), it is AVCHD running at GOP1. The Intra codec produces a 10 bit stream and is usually created (as in within P2 cameras) using different intra prediction methods (allegedly, whatever, I'd imagine more time is spent on this as it is only going to deliver Intra frames). It isn't the same. The GH2, and the AVCHD codec it uses is designed to produce streams containing B and P frames. Sure you can run it without them, but I don't think it is in anyway the same as shooting AVC-intra on a P2 for example.

Edit: What we're getting is AVC-intra 8 bit (which isn't AVC-intra). Also IIRC the size of each I-Frame in AVC-Intra is fixed and we're definitely not getting that from GOP 1, now that could be better or not, depending on the shot. Also an Intra only stream is going to require a higher bitrate than a long GOP stream, thats obvious.

I think maybe that the judder and odd results some of us are getting playing back GOP 1 could be related to the software expecting at least a single B-Frame after an I-Frame. Maybe if we could wrap a GOP1 stream into an MXF file flagged as an AVC-Intra stream it would work.

Stray

@proaudio Been trying an 88M patch, the settings attached (88M AQ2) have passed all the usual shots I do that create cadence problems and it all looks good. Deathchart was no problem running for over 2 minutes at shutter speeds of 1/50 and 1/125. With detailed shots outdoors I'm getting I-Frame sizes of 1.17M which is about the same as the 66M AQ2. Lowest averages (indoor, lowlight) was around 52-54M. Built this following the cbrandin 66M patch (just FBx4, FL, Top/Bottom etc), so it has settings for other shooting modes from that patch. However the FSH settings come from the 44M patch, I could never get the FSH settings in the 66M patch to work.

As I-Frame size seems to top out at 1.17 arguably the buffers could probably be run lower. The 24L runs at 64M but I havent tested it properly yet, will do later. @driftwod, can you look at this with elecard if you have time ? Its a work in progress, have you had any more luck trying an 88M setting ?

chrimsbroome

@everyone.....I found! that setting NTSC mode from PAL solves "recording limitation" on 720p mode. It Generally fixs most of the problems that I face when trying to record. I dont think this willl restrict mixing videos together and exporting it back to PAL.

You will get a message when switching output " cannot record due to Pal/NTSC fomrat"............all you have to do is format the card in the camera. This will convert the card to NTSC mode. It makes things far more stable for me. I am going to test the spanning to see if it spans.

proaudio4

Thanks Stray!
I will test your 88M AQ2 settings today.
I was confined to working on music yesterday, but should find time today.

paglez

@chrimsbroome
If you switch to NTSC mode, you get a 720 p 60 stream, not a 720p 50.

Howdy, Stranger!

Categories

Tags in Topic

Top Posters