GH1 firmware research volunteers required

Attachments

  • Picture 35.jpg
    Picture 35.jpg
    68.4 KB · Views: 0
  • Picture 36 copy.jpg
    Picture 36 copy.jpg
    178 KB · Views: 0
I really suggest starting new thread and moving all this stuff concerning bitraites, comparisons, etc to it.

I agree. This will cause the thread not to be focused on the research.. If I haven't said it, thank you for what your doing "tester13".



Pappas
 
Last edited:
Hey, been silently following this thread since it was less than 20 pages. Great work so far! If there is a way I can help in the form of contacting folks or shooting specific test footage let me know.

Just gave the latest ptool a try in order to remove the 60i wrapper. Just two questions about it (tried using search, but no dice so sorry if this is a repeat). I noticed that when bringing the new 24p footage into final cut, the format inspector says that:

1. Field is still Upper (Odd) rather than progressive and
2. The bit rate for similar footage was 14.1 rather than 17.2 (I assume this is an average since it's just 1 number in a VBR clip?)

So I wonder, should I deinterlace this footage still? Or is it truly progressive but a bad tag?
 
Hey, been silently following this thread since it was less than 20 pages. Great work so far! If there is a way I can help in the form of contacting folks or shooting specific test footage let me know.

See first page for all possible help (second post).

1. Field is still Upper (Odd) rather than progressive and
2. The bit rate for similar footage was 14.1 rather than 17.2 (I assume this is an average since it's just 1 number in a VBR clip?)

I can see clips using low level tools and they are all progressive if you use native 24p, without doubt.
We can still have some flag set in header, but all players play such footage fine.

To compare bitrate you need very controlled experiments.
 
PTool 3.33 Released
MJPEG Size and fps patches added.

See tester notes on first page-second post , please!

MJPEG Size and fps patches are important to understand how MJPEG part works.
 
Yeah, just shot something again similar with the patch and it turned out 13.8 but looked fine when stacked up next to the first one. I'll try something to max it out tomorrow just to see if the encoder's capability is the same and will report if it's not good. I'll also throw it on a timeline within a progressive sequence with XDCAM EX > ProRes to see if any issues arise with the actual picture.

Thank you so much! You have saved so many future logistical headaches with even just this single feature.
 
Looks like 1280->848 works, but reverse produce heavy artifacts.

P.S. Problem is identified and fixed.

I am able to produce 1920x1080p30 working MJPEG (most probably just unscaled 1280 for now).
Do not see any problems with this.
New patches will be available very soon.
 
Last edited:
PTool 3.33 Updated without version change

MJPEG 1280m Encoder settings patches added.

How-to get working 1920p30:
MJPEG 1280p width - input 1920
MJPEG 1280p height - input 1080
MJPEG 1280p Encoder width - input 2096
MJPEG 1280p Encoder buffer - input 4821120

P.S. Resolution will be like upscaled 1280. But we need more tests on this.
 
Last edited:
I don't understand, are you changing the resolution down and then immediately adjusting it from 848 back to 1280? Is it just to see if you can change it?

It was to understand some encoder parameters and LSI behavious. All I did is change size fo different modes.
And it helped already.
We know that we have "free"(no resources consumed) scaler before encoder.
So, it uppes our chances to have both 1920p24 and 1280p24 if we could switch sensor mode.
And least from MJPEG encoder side I don't see problems.
 
Hey Guys'

Testing again'

MjPEG 30fps - Xfps - input 25
MJPEG 1280p width - input 1920
MJPEG 1280p height - input 1080
MJPEG 1280p Encoder width - input 2096
MJPEG 1280p Encoder buffer - input 4821120
MJPEG 848p width - input 1280
MJPEG 848p height - input 720

1280p mode:

Duration : 10s 80ms
Bit rate : 25.7 Mbps
Width : 1 920 pixels
Height : 1 080 pixels
Display aspect ratio : 16:9
Frame rate : 25.000 fps
Bits/(Pixel*Frame) : 0.496

Working well

848p mode:

Duration : 10s 80ms
Bit rate : 9 454 Kbps
Width : 1 280 pixels
Height : 720 pixels
Display aspect ratio : 16:9
Frame rate : 25.000 fps
Bits/(Pixel*Frame) : 0.410

Picture have a problem :) need hi buffer or something.

Regards
NURBS
 

Attachments

  • P1010059.JPG
    P1010059.JPG
    103.4 KB · Views: 0
  • P1010060.JPG
    P1010060.JPG
    76 KB · Views: 0
Will it be possible, to have MJPEG in 720/50p as well with `the better bitrate?
Or enhanced codec with better bitrate for AVCHD 720 50p?

And?
Something prevents you to use already existing patch to change MJPEG framerate?
Try it. Not 50fps, but 60fps can work.

As for AVCHD - I am working on it.
AVCHD is very complicated thing (plus requires very good hardware), and it looks that it is very tightly programmed for certain bitrates.
Plus it looks that it have some bugs in bitrate allocation resulting in sudden spikes (and this spikes can bre real cause of freezing).
 
Hey Guys'
1280p mode:

Duration : 10s 80ms
Bit rate : 25.7 Mbps
Width : 1 920 pixels
Height : 1 080 pixels

You can use quality adjustments :-)
848p mode:
Picture have a problem :) need hi buffer or something.

Look at tester notes - I updated them.
Use 848p mode changes to test lower resolutions (change to 640x480, 320x240, etc).
As for picture problem in 848m I know how to adjust buffer for this mode also, but it have no use in this moment.
 
And?
Something prevents you to use already existing patch to change MJPEG framerate?
Try it. Not 50fps, but 60fps can work.

OK, I will try, but here in PAL country, 60p is not ideal.
Ayway I give it a try.

As for AVCHD - I am working on it.
AVCHD is very complicated thing (plus requires very good hardware), and it looks that it is very tightly programmed for certain bitrates.
Plus it looks that it have some bugs in bitrate allocation resulting in sudden spikes (and this spikes can bre real cause of freezing).

Good luck for it. If anybody, you can do it.
 
As for 1920 mode.
Even in 1280p upscaled version it have very good use - keying work.
As upscaling is prior encoding, so you have much better color resolution.
I believe someone asked improvements in this field not long time ago.
 
OK, I will try, but here in PAL country, 60p is not ideal.
Ayway I give it a try.

I also think that 60fps may work with current patch.
About PAL countries, 60fps will give us good slow-mo, and 60->25 conversion, although not perfect, gives far better results and motion than 30->25.

So, even not ideal, it would be very useful.
 
Test for fun: same settings and quality settings.

Quality settings (E1 to E4) - 720, 512, 428, 400
Table settings (E1 to E4) - 24, 24, 24, 24


Duration : 2s 880ms
Bit rate mode : Variable
Bit rate : 85.0 Mbps
Width : 1 920 pixels
Height : 1 080 pixels
Display aspect ratio : 16:9
Frame rate mode : Constant
Frame rate : 25.000 fps
Bits/(Pixel*Frame) : 1.639

Working well but camera stop recording after 2 sec. ;-) quality is superb!!!!

Regards
NURBS
 

Attachments

  • P1010062.jpg
    P1010062.jpg
    101.6 KB · Views: 0
Test for fun: same settings and quality settings.

Quality settings (E1 to E4) - 720, 512, 428, 400
Table settings (E1 to E4) - 24, 24, 24, 24

Yes it is too much.
Camera top bitrate is about 75-80Mbit (not average!).
And this settings lead to spikes in more detailed (or noisy) frames.

Try to use 1fps setting :-)
For timelapse like footage.
 
MjPEG 30fps - Xfps - input 60
MJPEG 1280p width - input 1920
MJPEG 1280p height - input 1080
MJPEG 1280p Encoder width - input 2096
MJPEG 1280p Encoder buffer - input 4821120
MJPEG 848p width - input 640
MJPEG 848p height - input 480

1280p mode:

Duration : 6s 0ms
Bit rate : 52.8 Mbps
Width : 1 920 pixels
Height : 1 080 pixels
Display aspect ratio : 16:9
Frame rate : 60.000 fps
Bits/(Pixel*Frame) : 0.424

Working well but stop recording after 6 sec.

848p mode:

Duration : 9s 500ms
Bit rate : 21.0 Mbps
Width : 640 pixels
Height : 480 pixels
Display aspect ratio : 4:3
Frame rate : 60.000 fps
Bits/(Pixel*Frame) : 1.141

Working well ;-) next tes 120fps :) like RED
 

Attachments

  • P1010065.JPG
    P1010065.JPG
    98.5 KB · Views: 0
  • P1010066.JPG
    P1010066.JPG
    73 KB · Views: 0
It is not required to test 120fps, as it is encoder framerate, not sensor.

Test only 1280 mode with 30fps->60fps patch (do not use all else initially).
And check that it is indeed 60fps. Using some motion.

It is also fun to look how scaler work.
It gets full source picture and scales to defined resolution (look at 640x480 squeezed look).
 
Back
Top