Personal View site logo
AVCHD maximum image quality settings and testing
  • 1002 Replies sorted by
  • @Stray I'm eagerly awaiting your take on the GOP1 versus GOP6, as I've tried the GOP1 and the GOP3, and preferred the GOP1 immensely. (GOP3 had the jitters real bad, and still had some funny blocking for me as well as banding. GOP1 every frame was a beautiful, organic, gritty, sharp, wonderful thing @ISO800. I'm no scientific tester, but I liked what I saw when I got it on the computer.) I guess I'm curious to know what GOP6 looks like in comparison, as the filesize on GOP1 is ginormous ;-)
  • @driftwood

    do you have the .ini file i can download for these settings?
  • @mbowen84 6 gop - its below next to the chart (but here it is again)

    http://www.personal-view.com/talks/uploads/FileUpload/d2/ad3a719246732829d3aee862fea792.zip

    GOP1 is over on the low gop thread
  • @B3Guy Well it's GOP 1 so the motion rendering is outstanding. However I can't see the point of it for myself, I can't think of when I'd use it. I've taken a good series of shots and the average bitrates ranged from 120Mbps-177Mbps. It's not that stable for me I'd get write errors when I walked around with a shoulder mount in about 20-40 seconds, sometimes a lot less than that in lowlight with high ISO (like under 10 seconds). On static shots it does behave marginally better, but theres no point shooting static shots with GOP1 really imo. Also, only VLC can play it back, Premiere CS5.5 just can't play it back sensibly at all, and 5DtoRGB can't transcode it (exhibits the same frame skippery as Premiere). Sooo.. I haven't gone into any proper comparisons with other settings, as I only do that if I think it's going to be of use to me.

    Also from an IQ standpoint I don't see, or think there will be technically, any difference between this and the 132M GOP 3. That is if you consider the size of I-Frames and how B & P Frames work, and their ratios when using the 132M GOP 3. What I'm saying is, you're going to get the same IQ at much lower bandwidths so the only reason to use this is for the motion rendering of GOP 1 (which, definitely is incredible, if you can actually use it/play it/edit it). I'm not planning to test the 220M GOP6, I have a good GOP6 from bkmcwd. I do expect the 220M GOP6 version to give better IQ, and larger I-Frames in each GOP, but I don't think it'll be such a colossal improvement that it will make the massive increase in file sizes worth it. YMMV.

    Edit : 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. I think this 220M setting is overkill, but I think if you can run it under say 150-170M then IQ is going to be lower than much lower bandwidth GOP3-GOP12 settings. With this in mind I think a 220M GOP6 is massively over the top, and you'ld probably get the same results IQ wise running GOP6 at 150M.
  • @Stray GOP1/3/6 - Mac: Movist or that other one beginning with 'M' plays it!, Quicktime with the AVCCAM panasonic plugin, try cineform neoscene too or NLE: or fcp 7 & convert to prores on the export tab. PC - win7 WMP plays it. NLE: Vegas, and Panny AVCCAM may transcode it.
    They work for me :-)

    High bitrate: The whole point of the exercise was to see how far the bitrate can go to stable video buffer graphs. So in theory these high settings are providing perfect sustained buffers. Its down to other settings now to harness it down methinks ... I'm looking at tuned settings in frame buffers etc... I'm also looking at timings to see if there's a way of getting round the spanning.

    GOP1 provides the filmmaker with an interesting hi resolution effect & they work tremendously in darker shots - I can't just discard it ;-)
  • Incidently, Cineform shows Panny GH2 stuff as source: AVC Lite ;-)
  • I don't have, and don't plan to buy neoscene, particularly when I don't need to tbh. I do have a PC an i7, these are the first files that haven't played in WMP without issues (GOP 3 and 6 are fine it's just GOP 1 thats giving grief). I definitely don't want to use vegas, I'd miss the mercury engine and tools of CS5.5 too much. As I said if they work for you and you need GOP1 at 220M then thats fine, I don't though and it defintely does not work for me. If I desperately needed it then maybe I would completely rip up my workflow and change all the software I use.

    Yep, I understand the reasons behind your high bitrate tests, I think. It is good to know exactly what the camera can achieve in regards to bitrate. But at these bitrates it becomes more than just an issue of diminishing returns, I think it becomes an issue of no returns at all when you're up at these bitrate levels due to how the AVCHD codec works. In the case of GOP 1, AVC-Intra which is the nearest comparison isn't just AVCHD running at GOP 1. There is a lot of filtering that P2 cameras do before it goes near compression (and are, IIRC, part of the AVC-Intra spec), and AVC-Intra doesn't run at these kind of bitrates because it doesn't have to (it tops out at 100Mbps, but yeah 4:2:2 and 10bit makes a massive difference there too). These filters and other processes are not available to us, so GOP1 is always likely to be of a higher bitrate than AVC-Intra to be worth doing, but still it shouldn't be anywhere near 220M and if it has to be then imo we shouldn't be trying. Of course the next Ptool version may provide us with much easier to get, lower bitrate short GOP settings. Although this is all an important exercise in finding the cameras limits, it's not going to produce any workable or worthwhile settings imo. What would though is to slowly drag the bitrate back down again that produce the same GOP and the same IQ, and most importantly, doesn't need a very specific software recipe to be able to edit/transcode/play the footage (and very specific SD cards).

    Edit : Don't get me wrong, its all good stuff, I'm really glad to have the 132M GOP3 setting you did available to me. Its not going to be my everyday setting, but it's good to have it up my sleeve if I ever need it. Good luck with the timing/spanning research, I guess you mean timing as in the buffer(s) state(s) when it comes to the point when it needs to span ?
  • @Stray Start testing then. Start to bring it down. :-)
  • Mate, I'd love to, but I have an ill baby and missus here. Will hopefully be able to contribute sometime more over this weekend. I want to get a GOP6 that looks good in a wide variety of shooting conditions, the current 66Mbps thats stable does lose a lot of quality in lowlight as the AQ is pushed right down. I'm figuring, from your tests so far, that it may be doable somewhere around 100-140Mbps, at AQ2 hopefully, or at least AQ1 or Q equivalent. Arguably the lower the GOP the more we should be able to push towards detail rendering as the lower GOP will improve the motion rendering.
  • @Ulisse Nice tele grabs there of the moon.
  • @driftwood
    Your 132 settings span for me just fine. Just filled up two 16GB Transcend Class 10 bought from Amazon UK. As this was a big issue for me, I will be keeping this setting for now.
    Thanks so much!
  • @Stray tend to the missus :-)
  • @allenswrench Good news! Thought 132M would... But I'd like to know what you filmed (in terms of maximising the bandwidth..) :-)
  • @driftwood....this has cheerd me up!!!, I just got one of the memory cards someone sujusted off youtube. It mayof been you dont knot but.....its a 32GB class 10 200x and its gone past 10 mins. It may of writen maltiple files but I dont care its not stopping like it used too. A celerbration is in order!!
  • More great news......I just tested the video in Premire CS5 its 14 mins long and has been split into 2 files. I have checked whether there is any pause or slight glitch in the part of the video that splits up into the next file and guess what.....no glitch at all its practically seemless so thats it I ca now shoot for 1 hour and 10 mins on a 32GB card, fantastic.
  • Can you point to the specific ini file you are having luck with, to avoid possible confusion? Why has everyone believed that 32 was the limit for spanning? What do you think is the factor that brought this 132 formula into success? Thanks
  • @Jspatz Probably due to the video content. I only checked spanning of death chart which if I remember 2 weeks ago I had a problem on...can't recall, but I havent run many tests on spanning - thought others would have done that earlier. Too busy on other tests :-)
  • Incidentally everyone, max settings for 12 GOP AQ2 1080p24 is 88M in H and 66M in L. Keep em together.
    And 66M/44M for AQ4.
  • Could someone point to the specific ini file. I will also test.
  • @Jspatz
    On the 13th page of this thread towards the end of the page Driftwood posted the 132Mbps ini file.

    ini files would be easier to find if we all posted them to the the thread below I started yesterday.
    http://personal-view.com/talks/discussion/953/gh2-patch-vault#Item_8
  • We seem to be getting a more consistent high bitrate with the 132 gop3 compared to the 220 gop6. Does that seem correct, driftwood?
  • I havent even looked at a top setting for 3 GOP! But 132 I know is the sweet spot. The GOP 6 setting was all about complete stability.
  • @driftwood "Incidentally everyone, max settings for 12 GOP AQ2 1080p24 is 88M in H and 66M in L. Keep em together.
    And 66M/44M for AQ4."

    Thanks driftwood, this is exactly what I was wondering.
  • @driftwood....one thing I carn't quite get when I record video and transfer it to my Laptop. It says 25p in windows and premire CS5 but when I play it back in Media Play Classic and observe the Stats it says 50 fps roughly. Can you confirm its 50 fps or 24p only.
  • So the 132M 3 Gop's span? Is it possible then to mix 132M with CBrandin's 44M 12Gops...so I'm using Chris's settings for 720p and Driftwood's settings for all 1080p stuff? I was thinking maybe a marriage between the 2 would be good.I'd rather not have to keep swapping hacks...
This topic is closed.
← All Discussions