AVCHD maximum image quality settings and testing

proaudio4

Correct, just 12GOP.
I'm still testing it...

proaudio4

Stock is fine unless you want the best out of the GH2 right now and that's 3.62d.
If you convert to an intermediate or stay native in post, higher quality more detailed data will always allow better color grading. Of course, our limitation is 4:2:0, but we'll get the best we can out of it.

OSGondar

@proaudio4 i use the adobe suite, imo there is no beating the Premier and After effects combo / workflow. Especially with red giant all over it =) i looooove that you can import your premier pro project into after effects natively so you can keep your edit and cuts to the mts untouched no need for an intermediate render, i have always had the idea that there simply is need for an intermediate codec EVER thanks to that amazing feature. It saves both time and space, Just do the finalizing in after effects ie special effects and color correction after importing the primere project immediately after you are done with the editing, no waiting needed.

maybe that's off topic but i rarely get to talk about how much i love the workflow =) you brought it up! and then theres cinema4d if you need to get fancy, and that too plays nice with AE.... just saying...

ok so...i did x3 with no gop changes, that parameter is unchecked. If you get results with setting 12 as well i would love to know about it especially if its before Monday =) unfortunately this version of ptools i've been too busy to stay as involved as last time around.

So in the meantime i will do a proper test shoot / edit tomorrow with current set up of just the x3 and 66mbs ill test gop 12 as well after im done with that if i can get around to it, please keep me updated. I'm very excited to step up the quality even if its mid shoot i don't care. I was afraid it was too buggy in contrast to last release when i was not nervous about anything.

vladnik

low qp importance .

qscale -16 is about qp 6.

nic2011

@proaudio4

>>Yes I can.
It's high ISO 2500. This test really shows how well 3.62d 66Mb/s AQ4 can really resolve fine detail such as noise.>>

Ok - but I don´t like noise. I would use Neat Video anyway. So, to clutter up the sd-cards with a lot of datarate only for getting more detailled noise isn´t worth the hassle imho. But to each his own...

sohus

@nic2011

Better noise is easier to remove in post. As simple as that.

Stray

@nic2011 I feel pretty much the same way about noise, and avoid high ISO shots when I can (except when testing where its a good way of forcing high data rates) and I really don't like the in camera NR much either. For me one of the main benefits of a high bitrate is that the noise is more consistent and a lot easier to clean up with Neat or whatever. However, this benefit is only really there with a low GOP as well as it does still smear the noise with high bitrates. Don't get me wrong, I'm not jumping on the low GOP *gasp* 'its like film' thing. I have too much knowledge of the experimental film scene, particularly the structuralist film movement, to get all hot and bothered about the 'film like' motion aesthetics relating to low GOP. At the end of the day the flicker of film contributes just as much to the motion as its inherent method of capturing light in consecutive stills. 'Flicker' is not something we can ever do, and as most cinemas use DLPs now anyway its gone from the aesthetic of film anyway. We are all heading towards a new aesthetic of motion which is a bastard hybrid which is coming from trying to make digital video ape film, and we're ending up with something that is neither one thing or the other. I can see nowt wrong with celebrating the digital aesthetic for what it is, as it's own unique and flexible look. Ha, and yeah, that includes noise, but that's not something I want as part of my aesthetic.

nic2011

The problem is that you get more visible noise even with lower iso values like 160. So, you always need to use neat video. And, as everybody knows, this could be a pain in the ..neck because it´s time consuming.

Stray

Yeah agreed you do see it thats very true. The scene itself does determine how much you notice it at a lower ISO (in brighter scenes its much less visible). But for me thats a very small price to pay for the latitude and detail I now have when working with the footage. But thats my workflow, I always process before I edit anyway (I'm an experimental film maker, video artist etc), unless I'm just making 'home movies'. Also, for me anyway, the quality of the noise at any ISO < 400 is acceptable. I must say I do find peoples comparison of it to 'film grain' to be annoying, grain has a wholy different quality to digital noise, no matter how consistent it's modulation may be. Saying all this though I do plan to try and make a short this week exploring digital noise as a result of this discussion and other statements I've read here (I did a piece about 20 years ago made from processed digital noise), as it does after all create some truly beautiful looks sometimes.

edit : If I was shooting a documentary or working on a comission for some sort of traditional piece (music video, advert etc) then I agree that always having to have a denoise process in my workflow would probably get on my tits.

proaudio4

@nic2011>"The problem is that you get more visible noise even with lower iso values like 160"

So, why do you think you may see more noise?

It's because the noise is there along with fine image detail and due to the improved codec, we can now actually see it.

Neat Video does a better job removing noise opposed to trying to deal with smeared dancing blocks that turn detail to plastic.

I'm not sure what you would expect.
If you like the original look, stay at the original bitrate and settings. You could lower them and rob even more detail including noise. Of course, at the expense of a lousy looking image.

Also, I never shoot at high ISO settings like this. I find ISO 160 to be magic with 3.62d. Looks very good.
If you're saying you would always need to use Neat at ISO 160, then there's something else going on with your camera. The same shots I did above, I have also at ISO160. I see no noise in either ones shadows or the back gray board behind the flowers.

Also, remember this is a static shot. This is only one aspect of improvement, another is motion.
Shoot the sea and compare.

cbrandin

Please try the attached settings. They basically double the bitrate for all modes. GOP was not changed. I did rather a deep dive into what many settings do. I think these maximize quality and efficiency. One thing I noticed was that P&B frames were too big in static scenes - they should be very small. I made some adjustments so that bitrate is more appropriately allocated now. I think the quality of these 44M settings will outperform the 66Mb settings that have been posted - maybe even the 88Mb settings. I tested a lot with different lenses, subjects, lighting, static, motion - and everything looks very good. There are no cadence issues except a very tiny bit (insignificant) with 720p.

I studied the streams pretty carefully and everything looks like it should. Even though B frames don't get very big unless you do a whip pan, motion rendering seems to be perfect. The only issue I know of with these settings is that files still don't span.

If these settings hold up well, I'll do another set at 66Mb.

Note that these are experimental settings, don't rely on them yet. There may be problems with some editors, etc... I haven't tested anything but Vegas.

Chris

oscillian

@cbrandin

So that means I get roughly 10minutes in 1080 or 15minutes in 720 with the 4Gb span limit?
Since I do weddings and need to roll for at least 45 minutes with multiple GH2s it's a bummer :(
Up until now I've only checked the 30 min limit on my cams.

How low do you need to tweak the bit rates for the files to span?
This probably should go into the "stable settings" thread, on second thought :)

cbrandin

@oscillian

From what I hear it's 32M

Stray

@cbrandin

I just tried those settings on some shots, its after midnight here so I can't do too much, I will do more tomorrow. However, wow, it's very impressive. I took the static shot I've been using to check for cadence problems and there were no issues, perfect. I shot it 3 times, first with the stock 14-42, then with a Helios 44 and finally with a 28mm wide paragon. I then took a fast motion shot walking around from low light to bright light at 3200 ISO with the paragon around F8. I paused motion a few times to focus on things (another thing guaranteed to cause hiccups), and no problems at all (there are some fat B frame sections as you said when you whip it around, but its difficult to gauge really. tThey are mostly around 30-50% of I frame size in this lighting and with this ISO). Then I filmed myself from a static position throwing things and generally moving about, the motion looks amazing and again the stream is great . Also so does shadow detail, incredibly so, more than I expected at this bitrate. I shot the death chart just so I could see what the max I Frame size might be, beautiful cadence pattern on this and I Frames consistently at 1.166MB. As I said will try more outdoor and motion shots tomorrow.

I'm wondering though how much of a balancing act this all is, in that if you do create a 66Mbps version if it could coexist with a 44Mbps 24L setting? That is what I've been trying to get myself, well that and also with GOP 6. Saying that, with what I've seen a 66Mbps has got to be a huge jump upwards in quality to make it worth not just sticking with this 44Mbps. Many thanks for these settings.

cbrandin

@Stray

As to mixing 66M and 44M the simple answer is no. These settings only work when all bitrates are raised in the same ratio. Now you could possibly get 66M and 50M, or 44M and 57M, etc...

Chris

paglez

Excellent work cbrandin!! The next step would be to fix the span files issue. Thanks again

towi

@cbrandin

you rock! the settings work perfectly in 24H, 24L and FSH (50i here).
Too, they DO span... at 24L.
Files also import in to Avid MC just fine (which has been an issue with my 88Mbps settings).
These will be my standard settings for the time beeing... plus a second set for 24H only (GOP3 / 88Mbps would be cool).

Thanks!!

Tomaso

@cbrandin
you relly rock! the settings work perfectly in 24H, 24L, FSH and FH (50i also here).
Both on SanDisk Extreme and SanDisk Ultra!!
These will also be my standard settings for the time beeing...

nic2011

@cbrandin

>>Please try the attached settings. They basically double the bitrate for all modes. GOP was not changed>>

Sounds promising. Is it possible to get a patch with these settings only for 720p60 (no change in 720p50, 1080p24p...)?

danyyyel

@cbrandin I saw that you were very enthusiastic about the drifwood low gop settings. What is your take on yours here and the one of drifwood. Because I don't know if it was you that said that it would be his shooting setting.

driftwood

@danyyyel See for yourself - here's my 12 GOP default settings at 100mbps. Now test it against Chris's. His should be better image quality. Conduct a test with this against chris's. AQ=2 = best stability at 100mbps hence inferior image quality. You could bring this setting down to 80mbps and try & improve AQ - as it never gets to 100% use of bitrate setting - see if image improves. Note: this setting is only for testing 24p H.

HillTop1

@oscillian or any person, so with the hack and settings, you can't do continuous recording on say a large sd card? You can only do it in 10 to 15 intervals based on if it's 1080 or 720?

danyyyel

@driftwood I was thinking more about your 3 gop setting (I prefer the short gop) against the latest Chris has posted here. I am still on 3.61 (lots of work lately) was going to go on the 12 gop because of what was reported about the 3 gop cadence problem, before yours came out. Now I am unsure about where to start (I would be trying a lower bitrate and higher AQ) of your setting if it makes a difference. I don't know if a 100+ megabit is not too high for me?

vladnik

@driftwood
did you try to up AQ ?
maybe it can go to AQ4.

driftwood

@HillTop1 it seems most of the new 3.62 tests are having probs with spanning after the first 4gig is reached so yes. Still testing why this is happening...

Howdy, Stranger!

Categories

Tags in Topic

Top Posters