40Mbps AVCHD High Reliability Patch

I was not right with the blank wall bit-rate .. but here is the right blank wall test: SEE HERE (full res)

- Blank wall bit-rate is 13,5Mbps for me

.. and here is another tripod test: SEE HERE (full res)


- Nothing special there but bit-rate is 32Mbps

Don't get me wrong, very happy with the 7 - 38Mbps data rate of the hack and these settings, but is the variable codec able to relied upon to throw enough bits at the lower detail stuff. Would a minimum of 17Mbps make the low detail images better, or are they optimal as is??
 
Would a minimum of 17Mbps make the low detail images better, or are they optimal as is?
To evaluate low-detail bitrates, I took two fascinating series of FHD videos of my sidewalk, at 1/50, 1/100, 1/200, and 1/400 second shutter speeds. For the first series, I focused my Nikon lens sharply on the flat concrete details. Then I defocused the lens just enough to blur the details, and shot the series again. Here are thumbnails and GH13 Stream Parser screenshots of the resulting bitrates:

Sharp Focus Series
SharpFocusTN.jpg


1/50 second shutter
SharpFocus050.jpg


1/100 second shutter
SharpFocus100.jpg


1/200 second shutter
SharpFocus200.jpg


1/400 second shutter

SharpFocus400.jpg



Soft Focus Series
SoftFocusTN.jpg


1/50 second shutter
SoftFocus050.jpg


1/100 second shutter
SoftFocus100.jpg


1/200 second shutter
SoftFocus200.jpg


1/400 second shutter

SoftFocus400.jpg


With sharply focused details, bitrates stayed up around 35Mbps for the top three shutter speed brackets. But with an underexposed 1/400 second shutter, the bitrate dropped down to 19Mbps.

With the soft focus image at well-exposed shutter speeds, bitrates were also around 19Mbps, and dropped down to 15Mbps with an underexposed 1/400 second shutter.

On virtually featureless images, I'd expect the bitrate to drop even lower. But for most low-detail subjects, it looks like we can count on at least a 15Mbps bitrate.
 
Last edited:
Thanks for doing these tests ipowell
With all due respect, I wouldn't call concrete texture low detail, obviously it is high detail as evident in the 35Mbps result. However, are you happy with the low detail video of "real world" images at low bitrates, ie a talking head in front of a flat blank wall or talking head in shallow DOF situations?

I'm happier now i've unticked native (to get rid of pulsing), but still think we could benefit from more BR at low detail. Just seems to be something not quite there, maybe i need a kit lens rather than CanonFD's.

R
 
Yes, I do consider the sharp-focus concrete to be a moderately high-detail subject. By defocusing it and bracketing down the illumination, I was using controlled conditions to examine how the encoder's bitrate drops off as image detail is reduced.

I think my blue sky shot of the Lux Mausoleum is a good example of an interesting real-world subject against a low-detail background. It produced a 20Mbps bitrate, and looks very good to me.

http://dvxuser.com/V6/showpost.php?p=2069399&postcount=33
 
Last edited:
If you see a patch for "Limiting Bitrate" then you don't have the latest version of PTool.

Unfortunately, the last few updates to PTool have been marked with the same version number, and the best way to be sure you have the latest version of PTool is to download it again:

http://www.gh1-hack.info/ptool3d.zip

Thankyou! I too had installed your patch and was getting The Big Freeze :-/
will post a 1080/720/MJPEG comparison of your patch later today.

First indication of the stable MJPEG patch is an astounding 40MBS (QUicktime inspector).... Wow...

Thx for your great efforts!

Paul
 
Last edited:
To evaluate low-detail bitrates, I took two fascinating series of FHD videos of my sidewalk, at 1/50, 1/100, 1/200, and 1/400 second shutter speeds. For the first series, I focused my Nikon lens sharply on the flat concrete details. Then I defocused the lens just enough to blur the details, and shot the series again. Here are thumbnails and GH13 Stream Parser screenshots of the resulting bitrates:


With sharply focused details, bitrates stayed up around 35Mbps for the top three shutter speed brackets. But with an underexposed 1/400 second shutter, the bitrate dropped down to 19Mbps.

With the soft focus image at well-exposed shutter speeds, bitrates were also around 19Mbps, and dropped down to 15Mbps with an underexposed 1/400 second shutter.

On virtually featureless images, I'd expect the bitrate to drop even lower. But for most low-detail subjects, it looks like we can count on at least a 15Mbps bitrate.
I don't understand. What's the point of those charts?
 
I don't understand. What's the point of those charts?
what's not to understand? the chart shows what the bitrate is throughout the footage and how it's much higher when the shots are more detailed. seemed pretty straightforward to me.
 
Don't get me wrong, very happy with the 7 - 38Mbps data rate of the hack and these settings, but is the variable codec able to relied upon to throw enough bits at the lower detail stuff. Would a minimum of 17Mbps make the low detail images better, or are they optimal as is??

Well, I think bit-rate like 17-38Mbps would be much better!..
17Mbps in low detail scenes will be much better .. like you said - 17Mbps would make low detail images better.. i think...
 
Last edited:
Using new ptool with Wine

Using new ptool with Wine

Have been doing the hack.........multiple versions, but don't know how to import the latest ptool version into Wine.

Can someone guide me.....I'd like to try these newer settings.
 
Ok, did some testing with lpowell´s settings. :thumbsup:

I have a PAL camera (50i), Sandisk class 10, kit lens and 20 1.7.
Using the Starchart,tripod, autofokus and iA (it is hard for me to push the camera like that in real life).
I got an average of 35,7 using my PS3.
Long story short. Changed Video Bitrate and Overall Bitrate.

Video Bitrate FHD/SH=52000000 (41000000)
Overall Bitrate=54000000 (43000000)
Preset bitrate=180
Preset bitrate 2=180

Now my average is 44,9. It can fail on grass on these settings. Camera stops but video is ok. Photo cropped
Tree.png

http://dl.dropbox.com/u/9863865/Tree52%2C54.MTS
http://dl.dropbox.com/u/9863865/Grass52%2C54Stopped%20.MTS
 
Last edited:
Changed Video Bitrate and Overall Bitrate.

Video Bitrate FHD/SH=52000000 (41000000)
Overall Bitrate=54000000 (43000000)
Preset bitrate=180
Preset bitrate 2=180

Now my average is 44,9. It can fail on grass on these settings. Camera stops but video is ok.
Increasing Video Bitrate and/or Overall Bitrate as you have done will push the AVCHD codec into the range of the PTool "C" settings. While you will see higher average bitrates, the peak bitrate will also range significantly higher, and can provoke the type of recording failures you describe. I suspect these failures are not due to SD card write speed limitations, but are instead caused by internal buffer overflow errors within the AVCHD encoder.
 
And do you have the top limit value of AVCHD stability? The top limit before the overflow errors ..
 
And do you have the top limit value of AVCHD stability? The top limit before the overflow errors ..
Yes, the 41M + 43M settings are optimized to produce the highest average bitrate while limiting the peak bitrate to 44Mbps. According to the GH13 Stream Parser, this also provides an adequate safety margin to avoid internal buffer overflows. When you push the encoder higher than this, you're exploring the lower range of sub-optimal "C" settings.
 
My data (average 44,9Mbps) is coming from my Playstaion 3.
I have never used the GH13 Stream Parser, and I am going on a vacation tomorrow.
Feel free to analyse my clips.

Pal (50i) settings should be higher than NTSC (60i), and ok maybe I am the edge of failure..........But unless you are a iA (Auto) and grass and tripod lover........You are safe..:beer:
 
I redid my test today with SanDisk Extreme 16gb 30MB/s class 10

Fails after 5-10 seconds. The maximum I could record was 50 sec. While my today's attempt with Silicon Power Card was more successful: it recorded 23 min before it hit the 3.99Gb wall. Go figure.

[Settings]
Version increment=5
Prevent version compare=Checked
Third-party battery=Checked
All Interface Languages=Checked
PAL<->NTSC Menu=Checked
Video Bitrate FHD/SH=41000000
Overall Bitrate=43000000
Preset bitrate=180
Preset bitrate 2=180

What is left to blame? Using PAL hack on a native NTSC camera?
Thanks for sending me a copy of your original MTS file, zigizigi. The recording plays back correctly and the GH13 Stream Parser reports no errors in the file. However, when I analyzed the video stream in detail, here's what I found:

StreamParser00001.jpg


This FHD video was recorded in Native 25p progressive mode. I've found that this can cause sporatic recording failures, regardless of the video bitrate. In the Stream Parser Frame Data panel, the selected packet shows the first and only video packet recorded in the last frame of the movie, followed by a number of Null packets. In the Packet Data panel, the selected video packet is shown to be empty of video data.

I suspect that what happened was the encoder abruptly choked for some reason related to the Native 25p setting. The interruption in the stream of encoded video data was detected after a number of Null video packets were written, and the recording was terminated.
 
Last edited:
Yes, Adobe Premiere Pro identifies it as 25p, and when I step through the frames in After Effects, there's no evidence of interlacing.
 
Thank you for all your testing lpowell. Have you uploaded your combined avchd/mjpeg settings as an ini file?
 
Back
Top