Personal View site logo
Official Low GOP topic, series 3
  • 1008 Replies sorted by
  • @balazer
    There are definately differences, especially over the 244M patch.
    It's definately due to compression levels. I can also see that Quantum v2 appears to offer less compression especially looking at the back wall.

    Thanks @FGCU for the excellent test!
  • driftwood, as for 2x squeeze anamorphic adapters, the sensor mode is still locked to 16:9? 4:3 recording is still not possible right?
  • thank you FGCU , do you also have playback on your gh2 with all those patches? or playback works only on lower bitrates hacks? i would like to buy gh2 as soon as i can and i would like to know if all those patches are playable or i must buy also an external monitor to check my footage.
  • @driftwood. I'm going to be testing Quantum Beta 2 today and tomorrow. Only change I'd like to make is on the mjpeg for timelapse.

    I set the following
    mjpeg fps to 2
    mjpeg size adjustment for 720p 30 width to 1920
    mjpeg size adjustment for 720p 30 height to 1080
    and I put setting on the 720p compression for E1-E4 Quality & Table

    In you Quantum mjpeg settings, are there any other settings you make that I can go put back to default unless there is a good reason to change those?

    Thanks again for your efforts. Have a good day.
  • Hi! can i ask somebody, do all those stunning gh2 patches allow in camera playback? (i´m gh1 user only) if not, which of them have problems with playback? would an external hdmi monitor solve the playback problem?
  • Thanks FGCU for the tests. They are indeed small improvements as we go along now - but who knows - older INTRA versions could be better - even though streameye tests for me show better overal macroblocking than some of the earlier patches I tested against. I don't mind if I get proved wrong.
    I am looking forward to @balazer tests as this guy is a walking talking font of information and holds some excellent opinions and theories - indeed he reminds me of cbrandin. :-)

    Forgot to mention: MJPEG testers (for those who don't know) can test their images using jpegsnoop. And VK has provided excellent stuff here. http://www.personal-view.com/talks/discussion/766/jpeg-estimator-and-quantization#Item_1
  • @FGCU, what differences are you seeing between in those screen shots between Quantum V2 and Seaquake v2? To be honest I can only spot some tiny differences, none of which I would necessarily attribute to compression.

    I still need to do my own tests for Driftwood.
  • @driftwood, how does the anamorphic shooting work in mgpeg mode? I would assume, ideally you'd shoot a 1.33 (4:3) at as higher res as possible with a 2x anamorphic and stretch.
  • @FGCU doubt it but you never know its gonna be interesting to see how far we can push mjpeg.
  • @killagram / @CraftyClown I'll be trying out your suggestions soon.
  • No I already know seaquake is better at L. Test against H if you like for me :-)

    Please understand, Quantum L setting ( a much lower bitrate but based on a high limiter) is like the AQuamotion/SpanMyGopUp patches' H bitrate setting - ie lower i frame quality (but still good quality) - we're trying to combine the best of both worlds - High quality setting (which won't span on most cards) because of the high bitrate / plus a decent quality L setting for spanning/long recording work.
  • I think there's been talk that the fps ><30fps on mjpeg is simply duplicating frames or dropping them so if anyone else could answer this for definite that'd be cool.
  • Or possibly a 1920x810 patch for everyone who has a 1.33X anamorphic lens. I purchased a Century Optics adapter from LPowell a month ago and love it.

    I'm sure a seaQuake MJPEG deviant would do the widescreen users justice (of course if those settings are possible outside of AVCHD)!
  • @Driftwood Nick if we can get 1920 x 540 in the MJPEG for anamorphic lenses, that would be amazing! The 2 x anamorphic lenses are definitely the most commonly available. I have a 2 x Kowa, I'd love to test this out with. Are we still stuck with 30p for MJPEG though? As with avchd I presume 25p is impossible? Cheers mate
  • For now, in my opinion, GH2 Mjpeg codec -in HD size- is in a very preliminary phase ("MJPEG is in its infancy " Sir Driftwood said)
  • @FGCU seAQuake wont work with AQ4 - it will fail after a number of seconds. The new encoder adjustments and AQ4 are in Quantum v2 beta. Test it against seAQuake and give me YOUR thoughts.
    Still testing how good/clean we can get the quality for MJPEG, still some way to go...
  • @Driftwood
    Do you think that you'll be able to get similar clean images in Mjpeg than Avchd? I guess that MJPEG codec efficiency never will get avchd quality .
  • @Oedipax Ive been testing a variety of anamorphic settings, the 3840x1080 could be difficult as its fairy high on the res, and Vitaliy has already told me that these may not work as the encoder cant handle it. But we can try...
    1920x540 should be a breeze.
    This reminds me - I need to get some anamorphic lenses & adapters next week...!

    I have found that there is a definite effect on chrominance on E1/F1 settings in ptools and I'm guessing Luminance is affected more on E4/F4. Needs testing more :-)
  • @pchristoph Hi mate, try 'Quantum' BETA 2 and test the L setting - should be very close to spanmygopup/AQuamotion2. I'm still fine tuning the setting and will probably tune it till it reaches slightly better quality than AQuamotion v2 and at AQ4!
    The H setting should in theory be better than seAQuake but needs testing. Looks better to me on streameye macroblocks, but all situations need testing.
    Hope youre well & good, Nick :-)
  • @driftwood, for 2x anamorphics I think we need either 3840x1080 or 1920x540 in MJPEG. Unless the milder containers you listed are cropping in-cam (I haven't played with MJPEG much), then 2.39 would be great.
  • @jfro I'll try and answer both your thread comments:
    1/ 1080p24 H and L are both INTRA GOP1 - sorry (better than GOP3).
    2/ 56000000 should work without underflows / overflows on the L setting only. Keep the H setting high if you try a very low bitrate on the H it will fail / produce poor i frames.
    3/ extended times: Yes, go ahead and try changing GOP1 to GOP3 but you'll get very high P frames - the saving will be neglible on H setting, but better on L setting. ;-)
    4/ P frames predict to the next i frame, B frames can predict backwards and forwards and are lower than in size than P frames. To achieve B frames, turn encoder setting 1 = default (not 2). Driftwood Hi quality GOP3 patches have already been made in the past using P and B frames and can be found on one of the low gop threads (some time ago!) you could always adjust those!
    5/ Test all settings and check graphs with Streamparser (look for cadence issues where the graph goes strange) and test on the various death charts like Pappas/sStray's colour chart discussed throughout these threads (use search)

    Hope that helps you. :-)

    Nick
  • @Driftwood, I've had great success using AQuamotion 2 spanning across my Panny Gold 16Gb cards.
    I am stoked that you're working on Quantum 24H & 24L patch that lets us have the best of high quality, and spanning together. For run-n-gun events where you're grabbing b-roll and interviews - this combo is fantastic with Panny Gold 16gb as I can get 10 seconds of footage for b-roll on 24H. If I need longer record times in 24H I just pop in my San Disk Ex.Pro 45 16gb. No reflash! Awesome. Many thanks.
    I have no idea how to make patches but my Christmas wish from you would be :
    24H seAQuake & 24L AQuamotion 2 and the best 720p50/60 patches. "seAQuabake"? cheers.
  • @FGCU
    1/ No - the 'L' Cinema 1080p24 setting is lowered bitrate GOP1 to around the quality of the AQuamotion/SpanMyGopUp - to enable spanning and long recording. This was confirmed as a working possibility by @balazer and so far it looks good without underruns / overflows. So thanks to him.
    2/ The H setting is still the hi quality of seaquake adjusted (with new encoder settings) to run AQ4 (seaquake was AQ3). We need to see which one is best so report back your findings.
    3/ To access MJPEG mode, goto 'Creative Mode' and 'OK' then go to REC MODE and 'OK' the bottom setting 'Motion JPEG'. Then goto 'REC QUALITY' where you'll find 4 modes. HD mode = Reference 2K (2048 x 1152) and VGA Mode = 2K container (2048 x 1080). The other two settings in 'REC MODE' can't be touched by ptools so are irrelevant.

    Hope this helps. :-)
  • 60i on Quantum is reported by elecard buffer analysis as iPiPiP...etc... all other software inc. streamparser, elecard streameye etc... report it as 60i in a 24N wrap ( it looks on some software as all INTRA - this is not true). So you need a method of extrapolating the i frames. AVIsynth have very good soultions. Maybe @Ralph_B could answer this better for PC users?
  • @driftwood, I'm a PC, but I looked at that software. It seems like it would only convert the container for AVCHD to Quicktime. How would this affect the clip frame rate (60i -> 30p)?
This topic is closed.
← All Discussions