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.
GH2 v1.1 new firmware with 25p discussion topic
  • 637 Replies sorted by
  • @Ishvar

    So The NR improvement is effective only with 25p? Am i getting this right?
    ...It's urgent to be patient IM(f)O.
  • Just done a convert using 5D2RGB and it works very well. Bit of a shag you can't drag the native files directly into FCPX but 5D2RGB also handles the colour much better especially with reds etc. In the original mts file there is distinct jaggedness around high contrast red.

    See the attached images. I transcoded the native file using 5d2rgb and placed both in an After Effects comp on on top of the other. I then exported stills of each. In the colour frames if you look at the orange book there is a jagged tearing at the edges. It is much easier to see the reason for this when you look at the red channel which I have also included. These are the black and white images. 5d2RGB really does a fantastic job of clearing these artefacts

    Sorry I realised I hadn't turned off the upper field interpretation in After Effects so I redid the red channel for the mts file. Can't delete the other one for some reason
    5d2rgb.jpg
    1920 x 1080 - 429K
    5d2rgb-redchannel.jpg
    1920 x 1080 - 357K
    mts.jpg
    1920 x 1080 - 419K
    mtsnative-redchannel.jpg
    1920 x 1080 - 345K
    mts-redchannel.jpg
    1920 x 1080 - 387K
  • @driftwood

    As much as I like the native 25p, FCPX doesn't and your hack settings are leagues better than the "HBR" panasonic claims.

    I got some very strange artefacts in one shot and I have attached a crop of the full size frame. Just very weird macro blocking with interlace in there for good measure

    I was thrilled initially but now I think my ardour has been quelled by the poor way FCPX handles the files and by mud I can see. I think we had all got used to mega high bitrate images at 24p. I had at least

    Roll on the new hack I say, and can someone suggest a way for FCPX to work nicely with these new files

    Screen shot 2011-12-08 at 5.21.32 PM.png
    925 x 815 - 1013K
  • I have only played around with it for an hour or so. Maybe I am imagining it but the AWB seems a bit better. Grain at ISO 800 seems marginally better, with contrast and NR both at -1, I am seeing plenty of detail. Will try it a bit outside to check the foliage tomorrow.
    Tried the HBR, and the output *on playback* of the HDMI cable is recognized by my TX as interlaced, and I see interlaced stuff in the output--that is, my cam is not deinterlacing to the TV, but maybe there is a setting for that. Mediainfo reports 29.970 on the file on the card. Live HDMI monitoring is reported as 60i for both. 24p outputs the same seemingly progressive image to the TV with no jaggies. Focusing for camera (not video) seems marginally slower, but maybe I am imagining that as well. Flash exposure in the few shots I snapped seemed less bleachy and with truer colors in just a few samples.
  • Exactly the same problem with 1080i50 /60 INTRA. But can be sorted with the field dominance set to none. The HBR seems not that great a change.

    Pic: HBR mode (Left) v 'Quantum' 1080i50 25p in a wrapper (Right).

    HBR:
    max qp : 23
    min qp : 21

    Quantum:
    max qp: 24
    min qp : 1
    HBR 25p (00004.mts) in a wrapper v Quantum 1080i 25p (00001.mts) in a wrapper.png
    1920 x 1200 - 1M
  • cool! yeah - on closer inspection it seems these artifacts are only on things are a just out of focus. areas in crisp focus and those totally out of focus seem to not display them - which makes me think that maybe it is compression, as the areas that are very sharp would get most of the bitrate and there would not be much left over for the almost-sharp areas. The totally out of focus areas - well - maybe i just don't notice it there :) i dunno.

    Yeah - "High Bit Rate" my rear. Quicktime is reporting ~21Mbps when i turn off the soundtrack. Well - at least its a starting point for Vitaliy, and already looks useful.
  • @arvidtp, @kavadni was able to capture the HDMI output of 25p (while recording) without any interlacing issues. So hopefully your discovery is bit rate related. 24mb/s is almost insulting to call HBR. :-P
  • @alcomposer they are in every file, even the .MTS files directly from the camera if you look at 1 to 1 resolution. Don't get me wrong - they are not that bad. Just sub-optimal.
  • @arvidtp are you getting the artifacts with JES or any other software?
  • With the new 1080p30 HBR footage, it is definitely progressive, but seems to have interlaced-like, every-other-horizontal-line artifacts in soft areas in some frames - see attached screenshot (1 pixel in image = 1 pixel in video). They are weird - as if they are different levels of compression or something acting on the different fields - i.e. since it's interlaced, the codec is making every other line self-similar so that frame compression looks right when shown on an interlacing screen at 60fps- ? not sure - just a hypothesis.

    I'm hoping maybe this will be able to be mitigated if VK gets this mode to run at a higher bitrate, if it actually is a compression artifact.

    Edit - Transcoding comments moved here: http://www.personal-view.com/talks/discussion/comment/29773#Comment_29773

    Edit - @edwardm yeah - why didn't they eliminate interlacing when they have the chance in the HDTV spec??? it is a blight on all video makers, especially those who are just learning. I made so many avoidable mistakes with interlacing in projects before i understood it :)
    Screen Shot 2011-12-07 at 7.47.02 PM.png
    930 x 709 - 702K
  • Packing a progressive video stream inside an interlaced stream has been done for a long time, and seen by those who shoot HDV video. The Canon HV30/HV40 tape-based HDV cameras all shoot "30p" mode (NTSC version) that splits the progressive image into two interlaced fields and writes that to tape as 60i. The two fake interlaced fields are creating by slicing and dicing the original progressive image. Each progressive frame is split and stored as two interlaced half frames (there is no loss of data). When played back, what you are seeing is the original progressive image - its just been wrapped inside a 60i stream because that is what the format supported.

    The Canon XH A1 also records 30p inside a 60i stream. In some cameras, 24p gets recorded in strange modes like 2:3:3:2 "pulldown" (I am not going into that here) which needs to be handled by the editing software or converted with something like NeoScene or JES Deinterlacer (Mac).

    This is just what happens when better stuff comes along (progressive) but must be made to work with legacy systems (interlaced standards).

    Depending on your editing software, you might be able to choose either 30p or 60i (25p or 50i) to avoid having to do re-rendering of the video.

    Some day, interlaced will be gone for good!
  • toxotis70 That's a pretty impressive noisefree image for ISO 3200. Good tip about overexposing
  • @paglez - "In my opinion, new 25p stream is a malignant PsF. "

    I think PsF would more accurately be characterized as "benign". By packaging the 25p video stream in a 50i wrapper, Panasonic insured compatibility with Blu-ray-spec'ed HDTV's and players. It's not technically challenging to reconstruct the 25p stream from a PsF video. If FCPX doesn't detect and do it automatically, it should provide manual settings that allow you to specify exactly how it should be done. If the problem lies in Panasonic's native AVCHD importer, it's a bug they should fix ASAP.
  • @naavt ... i shoot with fast lens (pany 20mm f/1.7) and exposure to the right.
    Try to shoot manual with shutter speed 1/25
  • HBR hibitrate probably also stands for hybrid mode :-)
  • probably shot with fast lens
  • I just realized that after updating fw, a footage @ 25p is brighter than 24p, also exposure meters shows around 3-4 bars more. I kept GH2 settings and light the same, Manual, 1/25sec. 3200 iso, Standard Film Mode all=0, noise is slightly thinner/less visible. Also after pressing record color shifting seems a tad smaller. MediaInfo (ver 0.7.51) say it's 25fps interlaced, SplashPro 25fps, VLC says 50fps.
  • @toxotis70 and other fellow members.

    How do you end up with such a clean image with high ISO?

    I've finally picked my GH2 from the box now that it offers me 25p and made a couple of shots in my bedroom with awful noise pattern at 3200ISO. Completely unusable! (Standard Film Mode, HBR)

    Any thoughts on this will be very aappreciated
  • the problem is when you have footage that is Real 25p (not wrapped in 50i).
    I have both footage from sony nex and sony a65 that is 25p and when i put them both in the same timeline , it needs render!
    So ... we have a problem.
  • if you open the mts in VLC an go for codec details you will find out frame rate is 50.
    Imported to FCP X it's 1080i / 25fps - what does that mean??
  • Question - I updated the firmware and then shot some footage using the new 'High bitrate" option.

    I then used Log and Transfer in Final Cut Pro and the footage was automatically 25p.

    Did I do it correctly to shoot in 25p? I heard about all this 50i wrapper talk, thought there would be a more complicated procedure to conforming to 25p.
  • Updated an NTSC (U.S.) GH-2 to 1.1. For the NTSC camera, the new HBR mode delivers 1080/30p wrapped in 60i/29.97 stream. Works fine imported to Sony Vegas. Looks and behaves just like progressive images.

    I did a brief and poorly executed before and after upgrade shot to look at the video noise at ISO 2500. Subjectively, I'd say the video noise is about 40% to 50% reduced.

    I just ordered a 2nd GH-2 after doing my tests, although that was in the planning stages before the new firmware update.
  • In my opinion, new 25p stream is a malignant PsF.
  • @paglez ... you are right
    It reports as interlace.....!
    So, no NLE sees the footage as progressive?
  • @Driftwood:
    In my test with HBR in Streamparser, the stream is GOP 12, not 25