Personal View site logo
Make sure to join PV Telegram channel! Perfect to keep up with community on your smartphone.
Official Low GOP topic, series 2
  • 1022 Replies sorted by
  • Reserved for compilation and best hacks
  • @driftwood
    I am sorry for my repeated continuation post.
    However, furthermore, since it improved, I post.
    Since I am poor at English, I cannot explain well, but this setting is stable rather than any old of mine, and frame size does not become extremely small at a dark place.

    First seth.ini, the multiple of "1080p24 Frame Limit" is set to 1.2.
    In camera play may not be possible.

    Second setj.ini, set to 1.0
    In camera play possible.

    In these settings, I think that it is also good to set the value of AQ to 0.
    Both of screen shots are the results of recording Pappas death chart vibrated.

    However, since I may not understand one, why does frame size turn into the almost same size that the bit rate is different by 24H-88Mbps and 24L-66Mbps?
    How do you consider this?

    Thank you for your advice or hint.
    seth.zip
    552B
    setj.zip
    551B
    88_66m3gopaq1fbx3flx1.2tbx8_24h (2).JPG
    1297 x 633 - 395K
    88_66m3gopaq1fbx3flx1tbx8_24h.JPG
    1297 x 633 - 400K
  • @bkmcwd There does seem to be a fixed upper limit on I-frame size of 1,180,000 at the moment, maybe a subsequent version of ptools could break this limit but for now it's the limit, no matter what the settings are it seems. http://personal-view.com/talks/discussion/comment/17168#Comment_17168 but I don't know if that means there is a different upper limit for 24L.
  • Thanks @Stray! :)
    Regarding limit, I got it. But why also 24L...
  • Sorry, wasn't clear, Chris mentioned the limit specifically with 24H. I've no idea if its the same limit for 24L, or if it has a different lower limit. About your results being about the same for both, dunno, I guess its down to the scene and how the codec works (vbr).
  • @Stray
    Thank you carefully! :-)
    Furthermore, when improving, I encountered the scene where 24L differed from 24H.
    Since a setup of Bitrate is different, although I think that it is a natural result...
    However, sometimes, this difference may usually be unable to be recognized.

    @someone
    Although it is persistent as usual;-) the setting which can use AQ=2 is stuck.
    In the result of my test, both 24H and 24L were able to be satisfactorily used also in this setting.
    The stuck screen shot is tested with shaked Pappas Chart, respectively.
    AQ3 may be able to be used if it is not stress test but usual recording.
    setf.zip
    547B
    88_66m3gopaq2fbx4.8flx1tbx8_24h_2.JPG
    1297 x 634 - 176K
    88_66m3gopaq2fbx4.8flx1tbx8_24l_2.JPG
    1295 x 633 - 176K
  • @bkmcwd Here's your buffer charts on pappas death chart you requested for your latest 3 GOP 88M/66M. Looking good. :-)

    Chart 1 1080p24 H 88M
    Chart 2 1080p24 L 66M
    *Shot on Panny's 14-140 lens with ois off on a tripod focussed to 25(50mm) length.

    Good work bkmcwd. Notice your B frames are resolving better prediction on the 88M.
    buffer test of setf - bkmcwd 88MB H 3GOP AQ=2 FBx4.8 FLx1.0 TBx8.0.png
    1481 x 884 - 228K
    buffer test of setf - bkmcwd 66MB L 3GOP AQ=2 FBx4.8 FLx1.0 TBx8.0.png
    1481 x 883 - 242K
  • @bkmcwd

    What do you mean by "Although it is persistent as usual;-) the setting which can use AQ=2 is stuck" ?
  • @driftwood
    Thanks a lot! :-)
    For now, I am glad for it to turn out that this set is not bad.
    BTW, how can it be recognized as the meaning of "B frames are resolving better prediction" looking at where? Sorry!
  • @bkmcwd look at the B frames using more buffer memory in the 88M. Youve probably got a bit more room to play with in 88M too. Keep testing :-)
  • Still working on this Short GOP 1 1080p25H Native Hack :-)
    GOP 1 25p Native Hack.png
    1297 x 684 - 88K
  • @proaudio4
    My English was strange? sorry!
    The meaning is as follows.
    Since it is continuing sticking many settings persistent repeatedly these days, although someone may seem "Not again!", I stick this setting which can use AQ2.
    Although there is no telling whether a meaning gets across to you now, a great meaning cannot be found from the first...
  • @driftwood
    Thanks!
    Of course, I try many things again.
  • Going higher... ;-)
    GOP 1 25p Native Hack - Frames going up! more tests.png
    1297 x 673 - 87K
  • @bkmcwd I think you've got the width to run that 88M at AQ3 if you wanted to.
  • @Stray
    Of course, I want!
    I will try!
  • @driftwood
    Could you explain that it means "25p native hack" please? 25p? really?
  • @paglez Its probably a false positive in GOP1 - test my earlier setting on the PAL thread 720p50.
  • @Driftwood, big problem to play and convert correctly the 220M GOP1, stuttering, on mac and also on PC.
    Any idea?
  • @Ulisse I know it stutters - now change it to 3 GOP or 12 GOP.
  • Seems 25p is all about the 30000000 buffer or just plain GOP1 initialising something as it only shows the option of Native 25p hack in GOP1. Defo looks like a false positive.
  • Three tests:

    1. BKMCWD's 24H 88M AQ2 3GOP
    2. STRAY's 24H 88M AQ2 MAX VARIATION
    3. CBRANDIN's 66M AQ2 (original August 29th)

    NOTE: These tests were done under:
    A, Fixed lighting daylight lamps.
    B. Same camera settings, f-stop, ISO, etc...
    C. Two tests; (1. pappas chart framed) (2. graph paper framed)
    I chose to include the "graph paper" test because it does not max out the i-frames
    and allows us to compare differences easier.
    D. Repeatable motion test (Motorized base fixed on tripod)
    E. Static tests (2 minutes); Motion tests- motorized pan movement (~30 seconds)

    _________________________________________________

    1. BKMCWD's 24H 88M AQ2 3GOP (pappas chart)
    Settings:
    f-stop: 6.3
    Shutter sp. 1/1600
    ISO 3200

    Pappas Death Chart
    Results:
    Passed during static check.
    Failed for Motion error (stopped recording) at start of motorized pan.
    Two Streamparser images below labeled:
    bkmcwd 88M AQ2 3GOP.png
    bkmcwd 88M AQ2 3GOP futher in video.png
    NOTE: These streamparser images are from the same video (pappas chart). I just wanted to show that it
    displays the issue earlier, then further in the video before it drops recording.

    NOTE: Since this setting failed on the motion test (pappas chart), I did not run the graph chart test.
    ___________________________________________________


    2a. STRAY's 24H 88M AQ2 MAX VARIATION (pappas chart)
    Settings:
    f-stop: 6.3
    Shutter sp. 1/1600
    ISO 3200

    Pappas Death Chart
    Results:
    Passed static and motion tests.
    Streamparser image below labeled:
    88M AQ2 MAX VAR PAPPAS STRAY.png


    Test notes (I omitted first i-frames from data results):
    i-frames 1.05M smallest i-frame during motorized base pan
    p-frames during static ~682K
    b-frames during static ~320K



    2b. STRAY's 24H 88M AQ2 MAX VARIATION (Less dense Graph chart)
    Settings:
    f-stop: 8.0
    Shutter sp. 1/1600
    ISO 3200

    Graph Chart
    Results:
    Passed static and motion tests.
    Streamparser image below labeled:
    88M AQ2 MAX VAR GRAPH STRAY.png

    Test notes (I omitted first i-frames from data results):
    i-frames 905K smallest i-frame during motorized base pan
    p-frames during static ~710K
    P-frames during motion max ~800K
    b-frames during static ~300K

    ________________________________________________

    3a. CBRANDIN's 66M AQ2 (original August 29th) (pappas chart)
    Settings:
    f-stop: 6.3
    Shutter sp. 1/1600
    ISO 3200

    Pappas Chart
    Results:
    Passed static and motion tests.
    Streamparser image below labeled:
    Chris B 66M AQ2 AUG29 PAPPAS.png

    Test notes (I omitted first i-frames from data results):
    i-frames 909K smallest i-frame during motorized base pan
    p-frames during static ~220K
    b-frames during static ~510K
    Streamparser image below labeled:



    3b. CBRANDIN's 66M AQ2 (original August 29th) (Less dense Graph chart)
    Settings:
    f-stop: 8.0
    Shutter sp. 1/1600
    ISO 3200

    Graph Chart
    Results:
    Passed static and motion tests.
    Streamparser image below labeled:
    Chris B 66M AQ2 AUG29 STATIC GRAPH.png

    Test notes (I omitted first i-frames from data results):
    i-frames 794K smallest i-frame during motorized base pan
    p-frames during static ~560K
    P-frames during motion max ~630K
    b-frames during static ~205K

    __________________________________________________________

    SUMMARY (& LESSONS LEARNED ;) ):

    1. BKMCWD's 24H 88M AQ2 3GOP really looks good on the charts (very solid).
    It just needs a bit more tweaking so it does not fail. (See images below)

    2. STRAY's 88M AQ2 MAX VARIATION looks great and passes these tests.
    If you directly compare STRAY's 88M (graph paper) tests with CHRIS's 66M (graph paper)
    you can see that it offers a bit stronger i-frames (also motion vector frames are a bit stronger):
    88M i-frames: 905K
    66M i-frames 794K
    88M p-frames (max motion): 800K
    66M p-frames (max motion): 630K
    88M: b-frames: 300K
    66M: b-frames: 205K

    PLEASE NOTE: These tests were performed under the SAME conditions. The camera was never moved.
    The camera is fixed on a motorized base on a tripod.
    During the motion test, the motorized base speed control was set and left the same for
    all motion tests.

    3. CBRANDIN's 66M AQ2 (August 29th) really perfoms well and has NEVER failed for me.

    4. LESSONS LEARNED: Don't upload the same fucking files twice! LOL There are duplicate STRAY images below.
    bkmcwd 88M AQ2 3GOP.png
    1920 x 1080 - 108K
    bkmcwd 88M AQ2 3GOP futher in video.png
    1920 x 1080 - 118K
    88M AQ2 MAX VAR PAPPAS STRAY.png
    1920 x 1080 - 476K
    88M AQ2 MAX VAR GRAPH STRAY.png
    1920 x 1080 - 497K
    88M AQ2 MAX VAR PAPPAS STRAY.png
    1920 x 1080 - 476K
    88M AQ2 MAX VAR GRAPH STRAY.png
    1920 x 1080 - 497K
    Chris B 66M AQ2 AUG29 PAPPAS.png
    1920 x 1080 - 148K
    Chris B 66M AQ2 AUG29 STATIC GRAPH.png
    1920 x 1080 - 153K
  • @proaudio4 Many, many thanks for these tests, very interesting results. I honestly didn't expect it to perform that well against the original 66M AQ2. Once my current work here is out of the way I hope to be able to get back to contributing to testing more than I've been able to recently.
  • Thanks Stray.
    I'm really impressed with your 88M AQ2 MAX VARIATION settings. Nice work!
    I'm willing to bet bkmcwd will be able to fine tune his settings for stability.
    I will compare his again with your 88M once he tunes it.
  • Yep, I expect bkmcwd to get it stable and I look forward to trying it out when he does.
    I do expect that in real world shooting that the original 66M AQ2 will give better shadow detail than my 88M AQ2 max variation settings with some scenes. I will check in a few days to see if this 88M is really stable at AQ3 too.
This topic is closed.
← All Discussions