Personal View site logo
AVCHD: Stable settings
  • 288 Replies sorted by
  • If you set a buffer limit, you are trying to keep it occupied at its most safest /fullness /maximum stable picture to bitrate settings. The two set video buffers have to be able to cope with a variety of settings. Anything over creates write failure because youve gone over the limits of the buffers - hence buffer overrun - the opposite is buffer underrun where we are not utilising its full capabilities - if it reaches empty there will also be write problems. The frame buffer sizes should be kept as close as possible to the expected size of frames under extremely busy conditions again without maxing out its buffers. This is why a lot of people are testing multipliers to push the limits of the GH2. Very high bitrate settings as proven by Chris are not neccesary but its good to work out the maximiums and then work backwards until every setting is working. The spanning issue may well have something to do with buffers too. Obviously a good fast SD card is required to keep up with the data write size and speed. We just keep testing... ;-)
    A good AVC read on buffers albeit a tad old is found here;-

    http://www.google.co.uk/url?sa=t&source=web&cd=19&ved=0CGsQFjAIOAo&url=http%3A%2F%2Fciteseerx.ist.psu.edu%2Fviewdoc%2Fdownload%3Fdoi%3D10.1.1.130.5026%26rep%3Drep1%26type%3Dpdf&rct=j&q=frame%20buffers%20and%20video%20buffers%20avc&ei=DS9cTsj7Ooig8QOO4v3wDQ&usg=AFQjCNGWVl9YskP5NjZ_ItkPRylvf2JquA&sig2=udyEeqMHOqQkY_IDv9mq4Q

    Edit: I'm now thinking 88M is the top workable setting for the GH2 looking at the buffers and available memory with two differing multipliers employed for Video Buffer and Video Buffer 24p.
  • Edit: I had a spanning failure on 24H 80% with these settings - so don't trust them.

    @sdbest At the moment I've got settings using ptool 3.61 that, as best as I can test them shooting intense scenes (trees blowing in the wind) span and never crash or error out while improving on the stock image quality. Like you, I need reliability - in my case for videos of performances. I'm pretty sure these settings are fully reliable. They are conservative - 32mbit 24H and 28mbit SH/FSH, about about 1.5x and 1.75x the original bitrates, respectively. A modest but useful improvement I think. Only problem is the 'blip' that was fixed in 3.62.
    arvidtp conservative spanning settings for ptool 3.61.zip
    382B
  • @mrmukke @namtot I'm in your camp, too. As a working "film" maker, I welcome reliable settings that bridge the 4GB gap and produce better pictures. So far that goal seems yet to be achieved. Meanwhile, I can, at least, trust Panasonic's default settings to give me excellent usable images. But these are early days, and I expect there will be a Ptools version coming that will provide better images, stability, and no 4GB file limit. Patience is today's watchword, I think.
  • unfortunately im not a tester, just a user trying to get the best IQ with stability. im not sure what an ini is and ive never used stream parser.
  • @namlot could you post an ini of your 24H 66mbs Aq3 setting, I am searching for such a setting as the low gop topic is now only 100+ mbit, I think it is a bit high for me. Have you tested with stream passer that you are not getting cadence problem.
  • i'm confused. this thread became an experimental settings topic in the meantime, isnt it?

    i'm still leaving my settings to:
    24H 44
    24L 32
    30 min removal
    (sandisk extreme 32gb card)

    thats it for now. stable.
    but i would appreciate some serious advice for GOP and quantizer settings.

    we're working on an documentary at the moment and i really need to count on this camera and my 5D ;)
  • after much testing today and trying many different setting, ive settled on 24H at 66mb, gop3 and AQ3. the only problem im having is occasionally the clip wont play back and ill have to turn the camera off and then back on and it will play fine. with a 16gb card im getting right around 31min worth of footage per card which is good for me. whats even better is that these settings so far have been stable for me on both a silicon power class 10 and a transcend class 6.
  • @cbrandin

    Thanks for the reply, I was hoping that was the answer.

    I tried 34mb for H and it still spanned and was recognized. I will do some more informal testing and try to find the tipping point. Last time I tried 36mb H it didn't work out.
  • Over the top Long GOP 36 test. Passes the Pappas death chart with ease at 50mbps and LOng GOP at 36 with a few adjustments to buffers. AQ = 2.
    36 GOP 50bitrate test.png
    1311 x 704 - 89K
    driftwoods stable long gop 36 - 50mpbs bitrate test - seta.zip
    427B
  • @parkgt

    24H at 32M with AQ set to 4 can produce quite good results. I have noticed in my tests that the difference between factory settings and 44M is greater than the difference between, say, 44M and 66M. I would say go for it, even if you can only go to 32M - there will be a noticeable improvement. In fact, I would say that the improvement would be about the same as the difference between factory 24L and factory 24H.

    Chris
  • @schrigg

    I would recommend buying a few cheap 64mb SD card from somewhere (drawer, ebay) and put the stock firmware (with PAL/NTSC switch and 29 minute removal if you are in PAL countries) on one of them and your experimental, high-end settings on the other. Just keep a full original battery at hand and I think the risk you take is well covered with a good back-up plan.
  • 3.62 will span and is recognized by Photofun at 32mb H and 24 mb L. 3.61 spans and is recognized at higher bitrate setting.

    Is there an inherent advantage to using 3.62 with lower bitrates rather than 3.61 with higher? I can adjust the Quantizer in 3.62 but it has been suggested that the limited bitrates of a 32H setting may not show much gain from those adjustments.
  • shrigg..
    I like your thinking bro!
    I also like your DSLR cage which I bought a couple weeks ago.
  • @sohus

    Looks like a sensible approach to have 24H at 66 and 24L at 44, that way if the camera freezes you can reboot and switch to 24L with no reflash necessary. Aside from the benefit of spanning, that 44mbps 24L setting gives me confidence that I won't be caught with a non-functional camera on a shoot for hire. Clients just wouldn't understand, lol
  • @sohus, yes that looks good. I think you'll find though that if you shoot for over a minute, maybe two that the pattern in streamparser may start to get uneven. Try shooting something else other than the death chart for that time length too, a mostly static shot. Theory is that the uneven cadence doesn't actually effect image quality, but we've not heard anything definite about it from Vitaliy or Chris yet AFAIK. I can't see how a large drop in I frame size isn't going to affect the IQ myself but I'm no expert on this at all, so maybe I just don't get it.
  • My mistake, I uploaded my 44mbit version. Here is the right one.

    I uploaded a typical graph I see with these settings. This particular one is Kae's Death Chart on my 1080p Panasonic Plasma TV. Is this normal behavior? I don't know what to expect from the B- and P-frames.

    File: F:\PRIVATE\AVCHD\BDMV\STREAM\00008.MTS
    Size = 1.071.194.112 Bytes (5.579.136 Packets)
    Stream: Max Speed = 70.530.620 bps
    Mode = 1080/24pN
    PCR Timing Interval = 0,094 seconds
    Clip Time = 0:02:38,20
    Average Total Bitrate = 54.170.109
    Average Video Bitrate = 52.675.256
    Average Audio Bitrate = 235.016
    Average Other Bitrate = 1.259.837
    Max GOP Size = 3.928.320 Bytes (31.426.560 bits)
    Video Frame Size (Min/Ave/Max)= 24.960 / 280.430 / 1.001.280
    setc.zip
    350B
    3.JPG
    1182 x 220 - 48K
  • You have change from 66mbit to 44mbit, is there a reason. I was particularly interested in a good 66mbit 24h mode.
  • These are the settings I'm testing now. Some observations:
    - 24H does not span
    - 24L does span (in my limited testing)
    - EX TELE works in 24H/24L
    - Camera profile: nostalgic 0,-2,-1,0

    I hope someone who has been testing longer can also test this setting as I have just started testing and don't have much experience with streamparser so I am not sure how to analyze what I am seeing. Can someone point me at a 'proper' avchd streamparser graph. How should it look?
  • @Sohus you should post your ini and details like what is your AQ settings.
  • real mess is going on with these settings, maybe we should make the contents and the winner settings to put in one sticky topic with the chart explaining the pros and cons???

    or maybe is too early for something like that....
  • Maybe the spanning has to do with the blip that has been removed. 3.61 spans but has blip, 3.62 doesn't span and doesn't blip. I have faith that Vitaily can fix the spanning in the next Ptool release.

    For now, I'm using the following settings:
    - 66mbit 24H
    - 32mbit 24L (files span with 3.62, need to test different scenes/motion though)
    - GOP12

    For interviews, I will shoot 24L and B-roll in 24H. No need to flash firmware each time.
  • Like I said I no longer use 3.61. They were Kae's settings, but set to 60Mb/s 6GOP.
    Look at the first post at the top of page 5 here. Kae mentioned he had no issue spanning with 3.61 at 65MB/s.
  • huh. and your resulting average bitrates are definitely above 32mbit? Can you post your settings?
  • arvidtp >"Anyhow, Test results are in - 24H @ 36mbits/sec with ptool 3.61 did not span for me. 32mbit seems to be about the max, at least for my SD card."

    Hmm..
    You're making me think SD card data rate are also tied to spanning issues more than I thought.
    Maybe it's a function of how fast the SDHC card random writes, not just contiguous data write.
    Using Sandisk Extreme 30MB/s SDHC I had no problem spanning using 3.61 @ 60Mb/s.
    You had issues at 36Mb/s..... odd. Maybe other parameter differences?

    I no longer use 3.61 and have no intention on going back. 3.62 with QA4 looks too damn good to go back to 3.61 LOL!