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
  • 1003 Replies sorted by
  • @driftwood
    well done man! Footage looks and 'feels' stunning, going to test some more today. Thanks VK and Chris too.
  • @Stray
    "Looks like the cadence problem happened around frame 580 or so, I frames drop a lot in size there in some of it. Did the focus change or was the shot mostly static in that section ?"

    Uh oh - it looks like that was right where the focus got nice and sharp. (Take a look at the vimeo link. Frame 580 is right at about 24.2 seconds - it's actually the frame that's displayed in @bkmcwd's streamparser screen shot.)
  • @Stray
    I also think that the problem is caused as you say.

    The scene in the first valleys up to the third is seen suitably to the leaf that is reflected on a full screen of focus momentarily though there are some valleys in B frame.
    On the other hand, I think that anything has not occurred momentarily of the valley afterwards.
  • @bdegazio
    You are correct.
    However, I think that several places afterwards are different though focus seems to be suitable momentarily up to the first three places.
    Does it continue for a while when the cadence falls into disorder once?
  • Yep, it happened when you paused, slowed down your motion. I have seen the same thing on AQ settings as low as 2. I was going to properly test the IQ today to see if the cadence hiccup effects it really. But it looks like I won't have time. But honestly I don't think I have to, because if it happens on static shots too (which it does), and your I frame size drops to half or less than it was consistently before the uneven cadence started then obviously IQ is dropping as it's compressing more.
  • I tweaked @driftwood's setting. It passed my bush of death test for 2 mins.

    132Mbps
    AQ3
    1.33x on FBs and EBL

    Why is it 1.33x? I don't know. I randomly picked lower than 1.5x. I'm sure curious ones can test out other multipliers like 1.42, 1.46, etc.
    setb.zip
    392B
    110M-3GOP-AQ3-1.33X.jpg
    1280 x 683 - 256K
  • @Stray
    Thank you for your information.
    Other files were examined now.
    Certainly, it seems not to be only that though the size of I frame tends to get depressed momentarily of relating to focus.
    There ahead ・・・ I'm not sure.
    And, I cannot test more than this today because the weather is also bad ・・・
  • @bkmcwd
    "Does it continue for a while when the cadence falls into disorder once? "

    Yes it does, it does get back to an even cadence but it takes a second or two before it does. On a static shot it doesn't get back at all but stays at an uneven cadence.
  • Now trying Chris's idea about QP adjustments - unchecking AQ. All tests conducted on Pappas Death Chart.

    1st test is QP = 28.
    (ie Short GOP 3, 140mbps, AQ disabled, Initial QP = 20, QP =28, 1080p24 T4 = T1, 2 x multiplier on FB1 FB2 and FBL)
    Edit: Notice in Streamparser the min frame rate has now improved.

    2nd test is QP = 26. (with the other settings as above)
    Edit: A massive drop in bitrate all round.

    3rd Test is QP = 24. (with the other settings as above)

    4th Test is QP = 22. (with the other settings as above)

    5th Test is QP = 20. (with the other settings as above)

    6th Test is QP = 18. (with the other settings as above)

    7th Test is QP = 16. (with the other settings as above)

    8th Test is QP = 14. (with the other settings as above)
    Edit: Decent rates are on the climb again. @cbrandin Chris you are right about QP=14, your calculations from last night match this chart.
    140mbps GOP3 motion test with QP=28 - FB1FB2FBLimit = 2 x multiplier - 2 mins on pappas death test chart hi iso.png
    1348 x 733 - 493K
    140mbps GOP3 motion test with QP=26 - FB1FB2FBLimit = 2 x multiplier - 2 mins on pappas death test chart hi iso.png
    1323 x 707 - 127K
    140mbps GOP3 motion test with QP=24 - FB1FB2FBLimit = 2 x multiplier - 2 mins on pappas death test chart hi iso.png
    1339 x 700 - 96K
    140mbps GOP3 motion test with QP=22 - FB1FB2FBLimit = 2 x multiplier - 2 mins on pappas death test chart hi iso.png
    1342 x 731 - 214K
    140mbps GOP3 motion test with QP=20 - FB1FB2FBLimit = 2 x multiplier - 2 mins on pappas death test chart hi iso.png
    1364 x 766 - 286K
    140mbps GOP3 motion test with QP=18 - FB1FB2FBLimit = 2 x multiplier - 2 mins on pappas death test chart hi iso.png
    1369 x 776 - 306K
    140mbps GOP3 motion test with QP=16 - FB1FB2FBLimit = 2 x multiplier - 2 mins on pappas death test chart hi iso.png
    1440 x 758 - 348K
    140mbps GOP3 motion test with QP=14 - FB1FB2FBLimit = 2 x multiplier - 2 mins on pappas death test chart hi iso.png
    1363 x 745 - 249K
  • @lenuisible non of my recording tests over the 4gig fat32 limit have been able to span.
    @cbrandin I get the strange feeling that a fixed 32QP will be extremely effecient... gonna test! EDIT: its effecient in that the bitrates drop down hugely to close to factory 24mbps settings, massively compressed images. Dont think Im gonna bother with any more further QP tests. In Summary, its (for me) safe to say the sweet spots with my current setings are QP28, QP14 and QP20. @cbrandin I did do a QP of 12 (attached) but its not that much better than 14.
    140mbps GOP3 motion test with QP=12 - FB1FB2FBLimit = 2 x multiplier - 2 mins on pappas death test chart hi iso.png
    1386 x 764 - 305K
  • @driftwood

    Given the stable QP values you listed, QP=14 is the best because the lower the value, the better the individual frame image quality.

    By the way, I think the settings you posted for the GOP 3 at 100Mb aren't the right ones - they don't contain the FBx and frame limit patches.

    Chris
  • @cbrandin Thats why VK had a go at me - I put them in but forgot to checkbox tick em, I resent the corrected ones. want them again?
    Driftwood - 100mbps GOP 3 AQ=2 T4=T1 FB1FB2FBLx1.5 - setg.zip
    453B
  • @driftwood

    Thanks, these make more sense. Why did you set video buffer to 3600000? Did that setting fix something?

    Chris
  • @cbrandin Long story/short version: Process of elimination after doing some reading on video buffers. It all started when I thought Short GOP wasnt gong to happen after seing the new 3.62 and so I messed with some long GOP stuff (I did a GOP36 stable test and it looked very pretty after reaching 360000000!!) I methodically then upped the bitrate with no probs... so I thought f*ck it lets apply the settings to extreme short GOP (GOP = 1) and it went through without probs... got it upto 100mbps then I thought try it on 3GOP... blah blah blah... My conclusions ? I reckon all the GOPs will find stabilisation with a video buffer of 36000000 under 1080p24. Has anyone applied this buffer setting to 720 modes yet? Basically, I followed VKs rule of testing just one change at a time. Spot the trends and go for it!

    Edit: If anyone wants my latest fav settings here they are (bitrate=140mbps, QP=14, T4=T1, Vidbuff 360000000, FB1FB2FBL=x2 multiplier)
    Driftwood GOP 3 140mbps QP=14 T4=T1 FB1FB2FBLx2 - seta.zip
    435B
  • Love what youre guys are doing with the GH2. Any way we can get some of these tweaks available for the GH1. I have been working on a GOP 1 24p patch and its almost there. I think there is only so much we can do with the current version of ptools.
  • @driftwood This all is your reward for keeping a cool head . . . KISS method is the best thing ever! change ONE variable only. too many people cut corners and this mucks up things. Thanks for taking the time to test properly by process of elimination/trial and error. It is slower to test this way, but clearly it pays off. I certainly have found it so in many circumstances.
  • @cbrandin Is QP=14 close to AQ2 or AQ3?

    @driftwood Any chance to try QP on 100Mbp?

    140 is great... but the SDHC card write speed is kinda concern. I didn't have any issue at 130Mbps write speed... but it seems pushing the limit.
  • for 100Mb it would be 2
  • You are saying QP=14 is equivalent to AQ2 for 100Mbps. What would be QP value equivalent to AQ3 for 100Mbps?

    If FB and FL multipliers are below 1.5x for 100Mbps, will it degrade the image quality? e.g. 1.4x with AQ3. 1.33x worked fine with AQ3.
  • Just little more tweaking... to squeeze tiny bit more juice :)
  • @driftwood I noticed in your latest fav setting you checked Quantizer for 1080 modes to 14. Can you tell me why? Just trying to understand it a little bit better. Thanks.
  • @stonebat

    At 100Mb AQ0=QP=20. AQ4 = about QP=7. The values are evenly distributed across AQ0-AQ4

    Chris
  • Thanks @cbrandin

    Tweaked @driftwood's 100Mbps settings. QP=12 didn't work. Raised the overall bitrate to 110Mbps. Then it worked fine. I guess QP12 equivalent to AQ2.5.
  • @cbrandin can you confirm if AQ is adaptive quantise or auto Quant settings for users but set to constants?
  • It calculates the manual minimum quantize settings for you based on bitrate. Otherwise the math gets a little complicated. The codec does some adapting, but these values are fixed.

    Chris
This topic is closed.
← All Discussions