Personal View site logo
Official Low GOP topic, series 2
  • 1022 Replies sorted by
  • yeah I did, just enough so I could hack and shoot last night, I had cooking, and animals, and a gf to take care off after my 9 hour work day... will get more on it this weekend
  • @PerryWilson
    The data rate on this clip is only 36 mbs. I guess it's because most of the frame is black.

    @driftwood
    Your GOP1 settings seem to be highly variable in datarate, unlike cbrandin's GOP12 settings, which remain quite constant at the specified rate - 44 or 66 mbs. Is this the expected behaviour? I'm still trying to figure out if it's an advantage or a problem.


  • @bdegazio depends what youre filming. Film the same scene or static with each setting and test.
  • @JDN its probably the controlled bitrate which is making it span. Take a look at Streamparser and see how big the data rate and bitrate is - it could be low enough to span. More detailed shots involving higher rates dont appear to span.
  • But motion shouldn't make a difference any more, since it's GOP1, right?

    I can partially confirm the theory of bitrate, my shots of the evening sky from yesterday have much lower bitrates and spanned for a while. This morning's shot of trees has far higher bitrate and doesn't span.
  • Chris already told, I think many times, that span barrier is around 32Mbit and it is important that have such bitrate in the moment of spanning. It is fairly easy to understand as otherwise internal buffer overflows, because file operations necessary to make new file, add infor to AVCHD structure take some time.
  • interesting perhaps @driftwood could let us know why its only 36mbs... is it because of what I shot or was it because I put it into H.264?
  • Yep @PerryWilson cus its mainly dark with unchanging areas and not much detail outside the face - its the VBR encoder which changes the bitrate accordingly, and your test shoot bitrate used shows there's not that much changing.
  • cool, another question dont know if you could answer... If I were to use cineform to change this footage to 10bit 4:2:2 would the bitrate remain? I guess what im saying is once the clip is shot in a bitrate, does it hold that through the post process?
  • All depends on the codec being used but yes, so long as the codec offers equal or above, like you suggest, it should translate across. Try it!
  • @bkmcwd Ok, wow, spent all this morning and most of the afternoon shooting with your 66M GOP3 AQ3, everything shot handheld. I never thought I'd like a GOP3 patch, but this, I like a lot. I'm also more than a bit confused by the results, particularly the average bitrate figures I'm getting from streamparser on high detail shots. I'm not getting the half framerate notification either. In windows I'm getting er.. believable averages (62-65M averages) on the shots that streamparser is getting odd results out of.

    The killer thing for me about this patch is its maintaining a high bitrate in lowlight (relative to its 66M maximum) probably because of AQ3 and maybe because there was still a lot of detail in the the shadow areas. Plus, motion is still very good (I was shooting water) due to it running at GOP3. Anyway, check these graphs (106M average from a 66M patch..er really ? ;)). Now, are you going to try and increase it's maximum I-Frame size or am I ? I think this patch is my new general, default, shoot everything patch. Many thanks for it. Oh and Image quality is great too, I shot everything pretty flat really and will do some post work on some of the better shots later and probably vimeo them.

    Edit : One other thing that first half a second which we know is never right, is pretty odd looking motion wise with this patch and unusable. Nothing to worry about though.
    bkmcwd 66M AQ3 GOP3 detail.png
    1297 x 683 - 439K
    bkmcwd 66M AQ3 GOP3 lowlight.png
    1297 x 683 - 414K
    bkmcwd 66M AQ3 GOP3 lowlight2.png
    1297 x 683 - 372K
    bkmcwd 66M AQ3 GOP3 lowdetail.png
    1297 x 683 - 351K
  • @driftwood and Vitaliy - Thank you so much for your efforts along w/everyone else in the GH2 hacking community. Yesterday I shot a video in New Mexico canyons with the IntraPure (version 1) ini hack and it was unbelievable. No muddiness, judder and exceptional dynamic range for 1080p/24fps.

    The 20mm Panny lens yielded ~120Mbit playback while my Voigtlander 35mm kept around ~60Mbit. Of course the Panasonic 20mm was razor sharp and had many auto parameters enabled, but literally every frame looked like an uncompressed PNG file. I can't believe this is a 700 dollar stills camera!
  • Holy shit! bkmcwd, Stay's info and post on your 66M AQ3 look great! I'm looking forward to testing this tomorrow. Great work.

    Stray, I always look forward to reading your posts -always methodical and a great read.

    Question: I see there's a "p" at the start. Is a frame reorder required, or is it ok?
  • Its fine, it just looks pretty odd for the first half second or so, the rest of the stream plays back okay.

    Edit : Also I'm guessing those weird averages from streamparser are something odd happening in the way the stream is being interpreted (related to that half frame rate message maybe), I don't think that the stream is actually running at those rates. I could be wrong either way though.
  • Stray, OK, good to know.

    It will be interesting to see how large i-frames remain under the same conditions comparing your 88 AQ2 12GOP and bkmcwd's 66M AQ3 3GOP.

    I was really happy to see someone made progress with a moderately hight bitrate (66M) with AQ3. Like bkmcwd mentioned, he has not had any luck with 88M AQ3. I have not been able too myself. I imagine it's possible, but will take many tests to find it.

    A good test would be to compare motion with Chris's 66M AQ2 (Aug 29th) to bkmcwd's 66M AQ3 3GOP.
    66M AQ3 3GOP may just be the ticket! If motion is very close to 66M AQ2, the added detail from AQ3 is well worth it. I was surpised to see how high the video bitrates remained on your tests for 66M AQ3.
  • Agreed, also need to do some better, more accurate tests with fast motion particularly with this patch. On some shots I see somethings that give me some concern, but this could just be my subjectivity kicking in (as said before myself and GOP 3 have never actually got on before now). Maybe it could end up with us running this 66M GOP3 at AQ2, or finding some way to widen its QP range slightly.

    Still this is my subjective take, and I only go 'hmm...' during extremely fast sudden motion (probably because I'm so used to blur/smear of longer GOPs rather than the jumps of shorter GOPs), it really needs the eyes of some diehard GOP 3 fan to try it out. Don't get me wrong, I really like the way the motion looks, even the sudden motion, but it's definitely a suck it and see thing for anyone who wants to use it in my opinion.

    Edit : More likely, it could end up with me runing this patch at AQ2 ;) Or I'll learn to stop worrying and love the GOP 3. Fundamentally there is I think absolutely nothing wrong with it, its just my old school pair of eyes being totally freaked out by this motion and detail.
  • Stock GH2 versus 176M GOP 1 Intra. See charts (chart 1 =Interior ISO 1600/shutter 100 same WB, chart 2 = Exterior ISO320/Shutter 100 same WB, all shot on 14-140 lens).

    Ive been experimenting with FBs and IQ and Q to increase the quality of the macroblocks in the latest setting and balancing for movement. A massive thing Ive noticed in the Intra patch is how much more colour detail there is too, and much better in the shadows. The stock settings are almost a joke. The QP is 28 on the first frame! The Intra - at the moment 16. Image quality as smooth as chocolate!
    GH2 Stock 22M Default Versus GOP1 Intra 176M.png
    1902 x 1055 - 1M
    GH2 Stock 22M Default Versus GOP1 Intra 176M - Exterior.png
    1913 x 1052 - 2M
    GOP1 Intra 176M - Every frame a png!.png
    1913 x 1057 - 2M
    GOP1 Intra 176M - Smoother than chocolate.png
    1292 x 677 - 330K
  • I am noticing an obvious visual improvement in 176 over 244 and everything else, even without specialized examination. Spanning works consistently on Sandisk extreme pro, unless I happen to be on an extremely detailed seen at the moment of switchover.
  • @driftwood
    I think that there is nothing to compare the original settings for your GOP1 174Mbits. Only once I had the suspension of the camera with your latest settings at 80% slowmo mode.
  • Nick, One word: whatthefuckover
    Well, maybe it wasn't one word. ;)

    You didn't white balance, right? :)

    The difference is simply amazing!
    You are right about the word "massive".



  • @proaudio4 I had to double check the WB - exactly the same. The Intra brings out the true colour of my curtains. This is a massive plus for colour rendition.
  • @Stray @proaudio4 "Question: I see there's a "p" at the start. Is a frame reorder required, or is it ok?"

    Try putting Encoder Setting 1 1080p to = 2 or 1 to get rid of the P frame
  • ah cant wait for more intra... is the screen grab of the woman close up from the same 176 i messed with last night?
  • @driftwood,
    I can't help ask: Do you have a new 1GOP ini file or is the last one you posted still current?
  • Still working on improving QP.
This topic is closed.
← All Discussions