Personal View site logo
Make sure to join PV on Telegram or Facebook! Perfect to keep up with community on your smartphone.
Please, support PV!
It allows to keep PV going, with more focus towards AI, but keeping be one of the few truly independent places.
Official Low GOP topic, series 3
  • 1008 Replies sorted by
  • @driftwood , thanks I remember checking to see it that it was checked but just now I looked and it wasn't... ha my bad. Don't know how I could have unchecked it, thanks for the pointer.
  • @balazer With 3 hacked GH2s I do thousands of tests... testing: Specific numbers - moving small increments at a time, they are close to as possible as I can get without producing buffer underuns/overruns. Go ahead and round them up if you like but dont push it too far or you will get instability (believe me). Spreadsheet your tests, and dont lose track - it saves time.
    You have to keep the L close to the H Ive found. The H settings just give that more headroom for safety reasons. Again, one test at a time. Go ahead and see what happens when you start making greater 'distances'. AQ4 will fail at 6100808. Try it.
    Experimenting with the encoder settings gives us INTRA & largely improved QP - Ive gone down from the defaults a stage at a time - many of the encoder settings will disable recordings (try it ;-)) so BE CAREFUL.
    24p vid buffer at 3600000 is rock solid on Elecard buffer analysis. Frame buffer 2 incresase the margins for the frame sizes.
    There's a certain ceiling for 60fps which no matter what you do refuses to go higher on 720p at low GOPs without a fail.

    Wow, a life times work has just been passed on in seconds!!!!! Happy testing mate.

    PS If youre thinking of testing yourself, search thru the history of the LOW GOP threads. Search for frame buffers, Encoder settings (though theres edits to be made & more to be filled in because we've come along way since then) etc... etc.. ;-) etc... etc... Just remember this, this is a VBR codec, the way it handles bitrates means there really is no set pattern or multipliers - but it can get you close.
  • @driftwood was just reading your last post man it's incredible how you come up with those patches.. Kudos man amazing work. @Vitaly did his thing with the new Ptools, gonna upload your new SeAquake patch now with the new PTools.
  • @driftwood
    About your "thousands of tests", if the number of frame limits is seen, imagination can be done simply. ;-)
    Although I have also done thousands of tests, I cannot do putting finely like you.

    BTW,I stick my newest economical patch "with Driftwood Tune".

    UPDATE:
    http://personal-view.com/talks/discussion/comment/30743#Comment_30743

    Samples are here


    bkmcwd New 3GOP Q18 Test patch with Driftwood Tune.zip
    694B
    Q18_DWTUNE_FOREST.JPG
    1295 x 634 - 208K
    Q18_DWTUNE_BOKEH.JPG
    1299 x 634 - 180K
    Q18_ISO3200.JPG
    1298 x 631 - 188K
  • Medall of Honnor for Vitaly Kiselev and Nick Driftwood

    Thank you very much!!!!
  • Gorgeous settings Driftwood! Thank you! I'm using SeAQuake a lot (sandisk 16gb 30mb/s) and all is great until I start doing quick pans, the bitrate drops significantly, which results in a grey frame...:/

    edit: OK, I get it, it happens in highly over-exposed situations. Stand next to a window, boost iso or open wide so that it's seriously blown out, and start turning around. When I do that, everytime I pass the window I get a dropout and the bitrate goes down.
    The grey frame is a VLC's thing, it inserts such a frame when the dropout occurs, while WMP just holds the previous one.
    vlcsnap-2011-11-17-11h34m52s50.png
    1920 x 1080 - 29K
    1.jpg
    1275 x 658 - 174K
    2.jpg
    1273 x 642 - 189K
  • I tested the spanbitch, recorded 5 files. 4 files had about 64mbit, with 24h. the 5th file had a bitrate of 1700kbit under "details". it was about 40 sec long, but in mediaplayer it showed up 00:38:42, 40 minutes??? whats wrong? also when I select an iso. while im recording it show "auto iso". is that ok?
  • @testor the ISO display bug is well documented in the new PTools- disregard- it has no impact. The other issues could be caused by a slow card?? What card are you using? We should all really be aiming for the Sandisk 32 gig HD Video. (The current benchmark SD Card)
  • I got the transcend class 10 , 32gig
  • @brudney try the same manouevre on stock settings and report. Thanks.
  • @testor youre not very clear about what u did. You recorded five files seperately or u got five spanned files from one recording session?
    Try to give us a bit more detail.
  • five files seperately. what do you mean by spanned files? when i record more than 30minutes they will be spanned or what? i thought there is no limit now? :)
  • @testor
    Can you upload your fifth file? The one that went crazy? dropbox it if you like.
  • @bkmcwd
    Are you saying "with driftwood tune" is better quality than your previous version?
  • u got a pm
  • @testor that file site you use is slow - 2 and a half hours to download this. :-) Use dropbox!!!
  • @Jspatz
    Of course, YES! :-)
    It's quite obvious.
  • @bkmcwd Have you done a recent death chart test on your 1080p24H setting? It drops to an average of 60M after 10 seconds. Look here at this quick test I conducted. bkmcwd v seaquake on pappas death chart test.
    Average frame rate over the seconds for seaquake is constant and reports around 100M on the buffer.

    At the moment bkmcwd your GOP3 is only slightly better than the spanning patch AQuamotion v2 - see below.

    I suggest readjusting some of your Hi Top / Low Top settings and/or frame buffers. Something aint right.
    GOP3 v INTRA GOP1 - bkmcwd GOP3 - test Streamparser.png
    1296 x 683 - 87K
    GOP3 v INTRA GOP1 - 'seaquake' INTRA - test streamparser.png
    1297 x 682 - 85K
    GOP3 v INTRA GOP1 - bkmcwd GOP3 - test Elecard Video Buffer.png
    1376 x 859 - 82K
    GOP3 v INTRA GOP1 - 'seaquake' INTRA - test Elecard Video Buffer.png
    1375 x 801 - 73K
  • AQuamotion v2 - Good quality INTRA - Outdoor 1080p24H test. Holds up very well and spans! :-)
    AQuamotion v2 - Elecard buffer analysis.png
    1368 x 800 - 79K
    AQuamotion v2 - streamparser.png
    1298 x 683 - 406K
  • @driftwood
    Wooooooooow!!!
    It is just like that!
    I was not testing with an old chart recently...
    It will be how when a more intense stress starts with Stray's chart?
    You understand where should be corrected?
    Is this frame size difficult after all...?
  • @driftwood
    Thanks for your advice! ;-)
    I do instantly.

    I think that the frame buffers are not enough...This is my homework!!!
  • @testor I looked at your 'spanmybitchup' file 00004.mts and it shows fine on elecard and streamparser - ignore details on windows file properties - its reporting it wrong. Streamparser defo tells you its 1:05:40 seconds long. Here look for yourself on these graphs. Its a nice recording in all!!!
    testor's 'spanmybitchuptest' on file 00004 mts - streamparser.png
    1296 x 677 - 84K
    testor's 'spanmybitchuptest' on file 00004 mts - elecard buffer analysis.png
    1371 x 785 - 69K
  • @bkmcwd Try utilising AQ2 or AQ3 too.
  • @driftwood -- one of our regular DPs is actually getting a bit ticked off at the quality of the results your producing. We may very well ditch the f800 and do the rest of the doc on gh2s. You and Vitaliy have certainly earned your place in the credits. (oh, and thanks for the dedication -- just in time for interviews I'm doing next week!).
This topic is closed.
← All Discussions