Official Low GOP topic, series 3

proaudio4

LOL
@mrsquirrell
There's no way that was ISO160.
If it was, you'd better send it back.
That's at least ISO1600 !

Check your settings!' Turn all that auto shit off.

Ralph_B

@driftwood
I just want you to know that the mosquito noise is real and is an issue you need to be aware of. I shot the same 60i scene with Quantum 5 and Sanity, and the mosquito noise is only present in the Quantum version.

balazer

The extra noise in ETC mode is inherent in the way it is generating output pixels from a smaller number of input pixels (that capture less light, since you are using an aperture that's too small for that focal length and resolution). Encoder patch settings won't change that. At best, better encoder settings will make the encoder represent that noise more accurately. It is a good thing to represent noise accurately, because that noise looks better than compression noise, and you can model the noise more accurately and remove it more precisely with something like Neat Video.

kholi

@driftwood

Sorry to be a pain... being an ETC fan I'd like to ask: do the gains and advancements in your patches trickle down to ETC or am I only seeing things that I want to see? It's hard to really test that, but I feel like between seaQuake and Quantum V2 I did see a noise improvement with ETC mode on.

Right now the threshold is like 800 ISO in lowlit situations. 400 can even be too noisy, and 200 is a severely low level to work at even with fast Super 16mm lenses. Do you think your latest patch will help with the noise levels at 400 ISO?

Is there anything that can be done to tame or adjust the noise for specifically working with ETC mode? Like a patch that JUST helps with the noise levels?

My earlier question didn't pertain to noise, just ETC mode crop function and custom sizes.

driftwood

@mrsquirrell Check your ISOs - be wary of the sometime occuring iso bug which you can read about in other threads. Try and perfect your exposure.

csync

@mrsquirrell, are you sure that isn't ISO 1600? I get the same noise at the higher ISOs, with the same horizontal bands that you see in that video. It seems that not everyone has this problem, which makes me think some sensors perform significantly better than others. Also, I've noticed it more at extreme temperatures, very cold, or after the camera has been running a while. I'm going to try to acquire another GH2 soon to see if some of my problems aren't just due to the specific camera I have.

driftwood

@Ralph_B Thanks for the report. A new version of both are in the pipeline. I'm liking the higher bitrate/short GOP 720p60 and 60i for another reason... It's got nothing to do with your excellent Long Gop versions. However, I have to remind you mosquito noise can surface not just because its short gop.
Mosquito noise generally happens around high frequency parts of the image like hard to softer edging.
This happens when the DCT transformation attempts to rebuild the image after compression - and it can't rebuild discarded bits. Its here that I'm looking more into Quantisation factors to improve things.

The link http://pro-av.panasonic.net/en/technology/technology.pdf explains Panasonic's theory why "Intra prediction predicts within the limits of a single frame, and has an advantage over inter-frame prediction in preventing the deterioration of prediction accuracy even for volatile movement".

Ralph_B

@driftwood
I tested Quantum 5 with 60i and 60P and had problems with both. 60P failed after 5 seconds or so on a 16GB Sandisk 30M card. 1080 60i suffers from mosquito noise. 24H and 24L both ran fine.

mrsquirrell

hello firstly thank you very much for all this work, I'm a bit of a noob so i apologise if i'm doing something silly. I have been playing around with quantum beta 5. i did a film in the snow which I'll post later and at least to my eyes was wonderful. I just shot some stuff today and three shots in particular seemed to have a lot of noise, i was hoping to get some advice, maybe I am doing something wrong.

settings were 160iso smooth -2-2 0 +2 and looking at camera histogram exposure looked ok..

thanks again and I'll post the other film in the next few days

driftwood

I keep getting asked so here is why. For detailed description about AVC INTRA look at this link from Panasonic where they outline their case for INTRA v Long GOP. Quantum is INTRA on 1080p24 and 1080i50 plus 720p50.
ftp://ftp.panasonic.com/pub/Panasonic/Drivers/PBTS/papers/WP_AVC-Intra.pdf

Swiss_Boy

@stanlymanly2 if aquamotion V2 works fine with your card you should use it till something new comes out. I switched from Quantum v5 to Aquamotion V2 because it wouldn't span on my cards and i couldn't playback on my camera. Now Aquamotion is great and so far i don't have any problems with it. Maybe just try Quantum and you will see if it works for you or not.

stanlymanly2

I need a lil help. So the Aquamotion V2 has been perfect for the cards i use. I use Class 10 200x, i dont know if it is as good as the Sandisk cards and what not but Aquamotion works perfect on them. Question is, will the Quantum work with my cards or is it just too much? Aquamotion H mode is perfect so I think i can draw back to that if Quantum's H mode is too much.

sanzadez

@svart
I just started trying out Quantum this past weekend but think I might be able to answer your question, at least to some extent. Quantum is currently a beta patch that @Driftwood is working on. It is mean to have the best all around performance. The new additions to it are that it has a lower "L" setting for spanning and longer record times. Also, a lot of time has been put in to developing MJPEG settings. These MJPEG settings are being used for anamorphic adapters so that no stretching is needed in post. Hoped that helped. If I am mistaken in any of my info or left anything out, someone please do correct me.

svart

Ok, so I've been trying like hell to catch up on all these patches and variations. Even though the patch vault explains most of the patches pretty well, I still don't know what Quantum is and why it's any different than any of the others.. Can someone explain this please?

PaPa

Hey. on Quantam V2, is it normal that i can't access 60p mode at 720? I tried in both cinema mode and manual movie mode and it's always greyed out.

*edit*
found the culprit. Can't have anything going in through hdmi... very strange.

driftwood

The only thing you have to watch out for when changing the L setting to a low value is that Quant matrix for INTRA changes to a poorer secondary matrix in which case the low and hi frequencies will experience quite a substantial change. This is why the H setting is high bitrate/highest possible working frame buffer limit and match checked with the buffer analysis that it maintains peak and then QP checked using streameye. Always check your settings 'work' on different scenes and movement. What might look good initially to the eye, used in other scenes could look poor in hi to low frequencies.

sohus

@Zaven13 @balazer
My words were not meant for the experts. Just some words of caution that bitrate isn't the only paramater that Driftwood changed. You guys know what you are doing!

bkmcwd

@soru
I made a test patch new again. :-)
The maximum frame size is set to 1M as well as my usual patch.
Please try.
Although you can also change AQ value for 1080p into 3 from 2, I think that AQ3 is not so practical with 1M max frame size.

UPDATE(2011.Dec.26th):
Second patch is improved version.
It can be spanning with "95MB/s Sandisk Extreme Pro (64GB SDXC)".

NOTE:
In this patch, it is ignored except 24p(24H) as usual.
Please combine with your favorite patch except 1080p.

balazer

Longer GOPs are more efficient, yes. But you can't have a high and low setting of different GOP lengths in the same firmware. Go long or go 1. There's not much sense in using a GOP length of 3. At 66 Mbps, GOP1 is still pretty darn good looking. I wouldn't worry about it. If you go much lower than 66, that's when GOP1 won't be good quality anymore.

PaPa

Hey all,

I'm currently using Quantam V2 for 24H but am wondering something about 24L. Because 24L is something around 65 or 66 MBps, and still has a GOP of 1, doesn't that mean that we would basically have a lower quality video than something that runs at 65 or 66 MBps with a GOP of 3?

I'm just looking for something to pair up with my 24H setting that will let me record longer videos and be more space efficient on my cards. And i do understand the difference between motion in the GOP numbers but i figure that the difference between 3 and 1 might be very very very very subtle that at a certain point i would just be motion peeping too hard.

What is a good balance for a lower quality setting that perhaps uses a GOP of 3 to increase the compression but use the same data rate more efficiently?

Muchos appreciation!

Zaven13

@sohus
Thanks for your candid comments and concerns. I agree with you that if you change the bit rates and settings you need to fully test. And I have. I have spent hours to test various bit rates and settings under different conditions and different options including fast motion, slow motion, EXTC, different lenses, different SD cards, etc... And yes, I use Streamparser.
Driftwood and others have provided us with a foundation to be able to do our own testing with the understanding of the impact of the setting individually and in relation to each other as also explained above by balazer. I am an IT Architect by trade and systems programmer for many years. Testing has been part of my life.
Cheers.

balazer

If you want to mess with bit rates, yes, you should do your own testing. But just because Driftwood has painfully balanced his settings doesn't mean other settings don't also work well. A lot of the time it is as simple as changing the bit rate or frame limit. In fact, I'll give you a nice little recipe: start with Driftwood Reaquainted or Aquamotion v1. Set the 24p frame limit to whatever you want, but not higher than what's in Reaquainted. Set 24H to the frame limit times 24 plus 10 Mbps. That will work to give you intra-frame coding with any bit rate you want from 50 to 150 Mbps.

sohus

@zaven13
Everyone here is messing with bitrates but these patches are painfully balanced. You can't just reduce the bitrate and expect everything to work fine. I wonder if you guys all run your customized @driftwood settings trough streamanalyzer and see if your settings work. I wouldn't mess with them as I know that Nick tests them thoroughfully. It is not as easy as changing the bitrate... buffers, frame limits all influence stability and quality.

Zaven13

@danyyyel, @johnnyblank,
I personally like SeaQuake, so I kept the H settings at high but reduced the L to 100MBs. As balazer said, if you keep the H settings and just play with the L settings you should be ok. If 100mbs is too high and 60 is too low for L then pick something between 60 - 100 like 80 or 88 and try some test shots. If you like the resolution and storage size then that's your new L number.
Good luck and happy holidays.

balazer

@danyyyel, try 60 Mbps for 24L. That's what driftwood had in an earlier version of Quantum. It doesn't need to be a multiple of anything. The reports I saw were that 60 was working, so I'm not sure why he raised it to 100. But I'm sure he had a good reason to raise it.

My understanding is that the H and L settings are independent, but the H setting Nick uses and the frame limit are chosen in concert to give the stream some nice properties and very predictable performance. When you choose a much lower value for L but keep the frame limit corresponding to a higher H, you lose those properties, but it still works and it's still good enough. In fact that's how the stock settings work. Nick may have a different understanding. ;)

Personally, I use Aquamotion v1 with 24L = 66 for the few times I use 24L.

Howdy, Stranger!

Categories

Tags in Topic

Top Posters