Best setting for Mjpeg and Class 6 Card

C

Chryso

Member
Hi, today i tried out several settings for my GH1 in Mjpeg-Mode. First i set the resolution to 1920x1080/30 but i've got big problems recording outdoor-sets with much details, due to my class6 card. I don't want to mess around with AVC-Bitrate, because i still need the high compression rates.

Now i am using 1280x720/30 with

Q: 256, 220, 200, 184
T: 33, 36, 36, 36

Settings and it seems to work fine.

Does Somebody here have better Settings with higher resolution for mjpeg-mode and class6 cards? Is the low 720p Resolution really a drawback against 1080p?

Would be nice if i could only patch the Bitrate of FHD-AVC-Mode, but the firmware change seems to affect all four AVC-recording modes, right?
 
Chryso said:
Now i am using 1280x720/30 with

Q: 256, 220, 200, 184
T: 33, 36, 36, 36

Settings and it seems to work fine.

Does Somebody here have better Settings with higher resolution for mjpeg-mode and class6 cards? Is the low 720p Resolution really a drawback against 1080p?

Would be nice if i could only patch the Bitrate of FHD-AVC-Mode, but the firmware change seems to affect all four AVC-recording modes, right?
Click to expand...

Does this setting play back in the camera OK?
 
I am using now:

Q: 205, 176, 160, 147
T: 83, 91, 94, 97

as Settings and all is stable, i can even playback the clips with my cam.
 
Class 6 Maxflash 8gb Card does NOT work with these settings:
Q: 256, 256, 256, 256
T: 5, 5, 5, 5
High audio bitrate
Recording stops after 3 seconds (GH1, Zuiko 14-54mm Lens).
 
Depends from the scene, with the 20mm Pancake from 20Mbps up to 48Mbps on high Detail outdoor Shots, the average is around 35Mbps.

I noticed now, that Playback isn't working in high detailed scenes, looks like playback is capped by the camera with bitrates around 38Mps.
 
bauhausler said:
Class 6 Maxflash 8gb Card does NOT work with these settings:
Q: 256, 256, 256, 256
T: 5, 5, 5, 5
High audio bitrate
Recording stops after 3 seconds (GH1, Zuiko 14-54mm Lens).
Click to expand...
Leave the original Q-setting and alter only the T settings to 5!
 
Chryso said:
Depends from the scene, with the 20mm Pancake from 20Mbps up to 48Mbps on high Detail outdoor Shots, the average is around 35Mbps.

I noticed now, that Playback isn't working in high detailed scenes, looks like playback is capped by the camera with bitrates around 38Mps.
Click to expand...

Yes. I agree. At 720p, the camera seems to have an internal bottleneck for MJPEG playback. I tested the read speed for four different cards Class 4 through 10. They ALL had a read speed of 19MBs or slightly lower. The 19MBs might be a ceiling on my card reader. I'm not sure. But either way, this is way more speed than is needed to play back an 80Mbs MJPEG clip. So the card isn't the issue with MJPEG playback. It is either the camera's internal read capabilities or some other bottleneck.

It is interesting that the camera can read the sensor, demosaic and resize the video frames, correct for distortion and chromatic aberrations, compress them, pack them into a .MOV format and finally write them faster than it can read and decompress the MJPEG from the .MOV container. I'm not a MJPEG expert, but I'd expect reading and displaying to be easier/faster than recording /creating. If for no other reason than because of the whole sensor read/demosaic bit. Of course, there could easily be important things I'm not considering.

Anyway, it sure looks like MJPEG at 720p has permanent playback issues with the GH1 that seems unlikely to be addressable. I will play around with the other sizes and see if they hit the same "about 38Mbs" limit.
 
I've been using these MJPEG settings on both GH1 and GF1 at 1280x720p with 422 color depth:

Q1-4: 256, 220, 200, 184
T1-4: 56, 56, 56, 56

Q1-4 are exactly twice the default values displayed in PTool. I have no idea how the T1-4 settings work - can anyone shed some light on this factor? I get about 55Mbps bitrate, and of course the movies don't playback in camera.
 
lpowell said:
I've been using these MJPEG settings on both GH1 and GF1 at 1280x720p with 422 color depth:

Q1-4: 256, 220, 200, 184
T1-4: 56, 56, 56, 56

Q1-4 are exactly twice the default values displayed in PTool. I have no idea how the T1-4 settings work - can anyone shed some light on this factor? I get about 55Mbps bitrate, and of course the movies don't playback in camera.
Click to expand...

This is just my current working theory, so take it with a grain of salt. But the table settings seem to be something like the arbitrary JPEG quality setting on an image editing program. 0 is the least compression and largest file. 1024 is the most compression and smallest file. I think the Q setting might be numbers that tell the camera what to expect for file length so that it can predict clip sizes for you. I'm not sure they change the encoding at all.

I haven't carefully tested this yet. So it could easily be wrong. I frankly have no idea why there are four settings. I wonder if different settings are used depending on scene complexity. Maybe the MJPEG is somewhat adaptive or maybe it analyzes the first frame and then picks a scheme based upon the initial scene complexity. I thought that perhaps the four settings might be there for the four different MJPEG movie sizes, but I've tested and confirmed that they only affect the HD MJPEG movies.

Note that this inverse relationship jibes with the default values.

Quality Table
128 133
110 146
100 150
92 155

A higher table value results in a smaller sizer per frame.
 
I have been using the following settings without a glitch and find the results quite good. I use a very cheap Transcend 16gb Class 6 card. No playback in camera though...

Quality settings (E1 to E4) - 352, 220, 200, 184
Table settings (E1 to E4) - 24, 24, 24, 24

Here is a video test: http://vimeo.com/12515307
 
wturber said:
This is just my current working theory, so take it with a grain of salt. But the table settings seem to be something like the arbitrary JPEG quality setting on an image editing program. 0 is the least compression and largest file. 1024 is the most compression and smallest file. I think the Q setting might be numbers that tell the camera what to expect for file length so that it can predict clip sizes for you. I'm not sure they change the encoding at all.

I haven't carefully tested this yet. So it could easily be wrong. I frankly have no idea why there are four settings. <snip>
Click to expand...

OK, did some testing tonight. Changing the Quality values has a direct and apparently proportional affect on the camera's estimate for the maximum clip length. I halved all the Quality values while leaving the Table values alone and my estimated clip length doubled.

However, halving the table values also reduced the actual data rate. It didn't cut it in half, but reduced it by about 33% going from 31.4Mbs for my test scene to 22.9 Mbs.

When I subsequently halved the table values, I got a very minor bitrate change to 21.2Mbs. That is so small that the real difference could have been a slight change in the scene I was recording due to small camera position shifts.

However, when I leave the table values at their modified (halved settings) and put the Quality setting back to where I started, the actual data rate jumps to 47.4 Mbs. So clearly the two values work together to determine the actual data rate.

Now with the Quality settings back to where I started I set the table values to twice the value where I started. The result was no change in the predicted time remaing - once again confirming that only the Quality settings affect this prediction. But the actual measured bit rate dropped from 47.4 Mbs to 20.0 Mbs. Again, the two values seem to act to work in conjunction or perhaps limit each other.

How they interact is unclear, but smaller Table values tend to create higher bit rates while lower Quality values tend to create lower bit rates.
Though it seems like the two tend to limit the influence of each other in some as yet undetermined (by me at least) way. I'll do some snooping on the web and see what I can dig up. Based on a thread on DVXUSER between tester13 and others, the Table values probably point to quantization tables 0-1023 that are stored as permanent values in software or firmware. My guess is that certain quantization tables are more efficient/effective given a particular degree of compression. But then, we can see how poor my previous guessing was so I wouldn't put a lot in my current guess. :^)
 
Last edited:
Chryso said:
I am using now:

Q: 205, 176, 160, 147
T: 83, 91, 94, 97

as Settings and all is stable, i can even playback the clips with my cam.
Click to expand...

For me
It records but not playable on my cam. It plays about 1/2 a second and comes up card read error.
Class 6 Transcend, and in benchmarkings is constantly over 8mb/sec

On the computer Windows media player, has no audio about 1 sec in, VLC player has audio all the time, but about every 5 secs pauses..
 
You must log in or register to reply here.
Back
Top