Fast action GOP 2 54 mBS settings for GH1

M

Mark Crabtree

Active member
After a lot of experimentation, I seem to be getting fairly stable 24pn at GOP 2 and 54 mBS.
I have not tested the 720p only the 1080p.
It will crash if you are recording very dark almost black scenes or cap on. If there is any kind of bright area, even a small one in the frame is seems to keep recording. I have not had a crash other than recording black and the famous chart of doom which it will record for more than 10 seconds before it crashes. I have let it record for over ten minutes at a time on busy high detail scenes and had no crashes. I hardly ever do longer takes than ten minutes so I have not tested longer. I see no strobe effect and the resolution is very high with the GOP 2. I tried to up the over all rate to better handle bursts of high data rate. I also tried to keep all the numbers multiples of 9000.
I just shot a green screen for a TV spot and it keyed perfectly.

Here are my settings.



Ptool v3.50d

GH1__133.bin

Prevent version compare Check

Battery Third party Check

Interface all languages Check



AVC

Wrapper Native 24p Check

AVC compression
FHD 54000000
H 45000000
L 45000000
Overall bit rate 99000000

AVC GOP
720 60P 3
1080 24P 2

Patches for Testers:


AVC HD Movie Mode:
Preset bitrate 9000
Preset bitrate 2 9000
Overall bitrate 108000000
 
Thanks for your efforts on stabilizing a 2-frame GOP with Native 24p FHD video mode. I made a number of attempts to produce a reliable patch of this type and was unable to achieve reliable results. The problem was that even when the Native 24p videos would appear to record and play back without error, they frequently contained one or more empty, 192-byte P-frames. This would cause the video to appear to pause briefly for the duration of the empty frames. If you haven't done so already, I'd suggest analyzing your test videos using cbrandin's GH13 Stream Parser:

http://www.dvxuser.com/V6/showthread.php?224134-StreamParser-2.0-Beta-Available
 
Checked out with stream parser

Checked out with stream parser

I downloaded the stream parser and everything checks out OK. I get no empty frames and no pauses. LPowel, can you load my settings and see what you come up with. I have made them my standard settings and have several TV spots running right now that look flawless. I have two Red cameras and I am using my GH1 and GH2 as a low budget alternative and as a quality B camera for those shots I just can't do with the Red, like shooting from a medicine cabinet shelf.
 
I duplicated your settings in PTool 3.50d and tested them under artificial LED lighting, producing about 35Mbps of bitrate. The FHD videos played back properly in-camera. Stream Parser reports that the GH1 is indeed recording a series of Native 24p 2-frame GOPs and detected no flaws in the stream.

This patch looks very promising!

Although neither VLC nor DivX desktop players handled the Native 24p videos properly, I was able to import the MTS files into Adobe Premiere Pro CS5 and examine them visually. At 35Mbps the image quality looks very good, and they of course require no deinterlacing. I'll have to wait until morning before I can shoot some high-bitrate stress tests with dense foliage in daylight.

On the technical side, the empty and corrupted frames I've seen in previous Native 24p videos typically occurred in the last few P-frames of a GOP. I don't think I've ever seen a corrupted P-frame at the start of a GOP. It's quite possible that achieving a stable 2-frame GOP might be the key to eliminating corrupted P-frames. With only a single P-frame in each GOP, the codec may not ever get a chance to produce any empty P-frames!
 
The Holy Grail is Real?!

The Holy Grail is Real?!

"It's quite possible that achieving a stable 2-frame GOP might be the key to eliminating corrupted P-frames. With only a single P-frame in each GOP, the codec may not ever get a chance to produce any empty P-frames!"

Very exciting hypothesis, lpowell! Thank you Mark! We appreciate your investigations and anticipate your conclusions.
 
I have been using a simular patch with a 1 gop setting and it has been working very well for the past few weeks. It sure doesn't like a completly black scene, locks up after a few seconds. I tried the 2 gop settings and it works great as well.
 
I can't wait til the sun comes up in the morning, I will try to test these new settings. Thanks!
 
Streamparser reports both NTSC and PAL as error free in FHD and FH modes.

Windows media player/Splash plays them fine. Sony Vegas and Cineform neoscene both process and transcode fine.

The only thing i can see is there seems to be a strobe effect on very slow pans in all the above except PAL FH mode, very strange. Almost like a high shutter but i used 1/50 & 1/60 on 24&25 and 1/100 on 50&60fps.


This is under fluro lights so will need to check this in daylight but it doesn't look like flicker.

FHD mode 36 - 42Mbps with 20mm pancake no failures on record.
FH mode 18 - 26Mbps with 20mm pancake no failures on record.

Tests were done moving the camera in very slow 2ft circles.

Locking the camera down (tripod or on desk) Bitrates jump to over 50Mbps in FHD and FH focused on a bookcase under ceiling lights. First time i have ever seen FH mode 59.94fps this high, high detail and natural light will be interesting.
 
Last edited:
Just for info I tried this using PTool 3.51d and had a "card speed" recording stoppage on a daylight static foliage scene that showed lots of 192 byte P frame errors. There were also some of those on a hand held clip that did complete. I don't have version 3.50d so cannot try that (is it downloadable somewhere?).

Camera is a WE9G using Panasonic Gold cards.

Please bear in mind that, despite taking care, due to my inexperience with all this my info should be regarded as suspect.

It's back to lpowell's reliable settings for now (thank you sir).
 
Mark Crabtree said:
AVC HD Movie Mode:
Preset bitrate 9000
Preset bitrate 2 9000
Overall bitrate 108000000
Click to expand...

Sounds promising.

2 questions:

Will this work for PAL too? In other words:

AVC GOP
720 50P 3
1080 25P 2

And for me, Ptool v3.50d will not accept an Overall bitrate 108000000 in Patches for Testers / AVC HD Movie Mode. Could this be a typo?

Leave all other boxes unchecked? Thanks for a short reply.
 
emory said:
Sounds promising.

2 questions:

Will this work for PAL too? In other words:

AVC GOP
720 50P 3
1080 25P 2

And for me, Ptool v3.50d will not accept an Overall bitrate 108000000 in Patches for Testers / AVC HD Movie Mode. Could this be a typo?

Leave all other boxes unchecked? Thanks for a short reply.
Click to expand...

I have original PAL camera, latest Ptools v3.51d accepted 108000000 for me, but i'm thinking this is too high based on testing done on this in the past. However Ptools has changed alot since i did lot's of testing and i've forgotten what some of the patches actually do.

About to run this thru in daylight.

Any news from you lpowell?
 
Ok, not working for me at all in daylight. instant lockup/unable to write to card errors that plagued all the early testing over the past 6months.

Back to lpowell reliable for me unless someone has found a incorrect parameter that maybe only affecting PAL cameras.

I believe we have done all we can with the current Ptools and hacks in regard to the GH1, lpowell seems to have this thing sussed and all the old testers have abandoned ship. Sorry Mark, thanks for the work on it, good luck with your project.

R
 
Last edited:
rambooc1 said:
Ok, not working for me at all in daylight. instant lockup/unable to write to card errors that plagued all the early testing over the past 6months.

Back to lpowell reliable for me unless someone has found a incorrect parameter that maybe only affecting PAL cameras.

I believe we have done all we can with the current Ptools and hacks in regard to the GH1, lpowell seems to have this thing sussed and all the old testers abandoned ship. Sorry Mark, good luck with your project.

R
Click to expand...

Ramboc and everyone else,

I was having the same problem. On normal, well lit room scenes I was not having any freeze-ups, but the second I began shooting outside and capturing all of the tiny little branches and dead grass my GH13 gave up.

I have spent the last hour or two doing some experiments with the Patch, and I finally found a slight modification that works without fail for me, and gave me a 54mBS readout.

Change the first "overall bitrate" to 90000000
Change the second "overall bitrate" to 99000000

as I realize this can very easily get confusing I am copying Mark Crabtree's initial format below (thanks a ton Mark for your work - so incredibly helpful).


Here are my settings.



Ptool v3.51d

GH1__133.bin

Prevent version compare Check

Battery Third party Check

Interface all languages Check



AVC

Wrapper Native 24p Check

AVC compression
FHD 54000000
H 45000000
L 45000000
Overall bit rate 90000000

AVC GOP
720 60P 3
1080 24P 2

Patches for Testers:


AVC HD Movie Mode:
Preset bitrate 9000
Preset bitrate 2 9000
Overall bitrate 99000000



Hope this helps!
 
Just realized that that highest I've gotten so far is 49.5 mBS. Sorry about the mix-up. I'm still pretty happy with it right now though.

I'm downloading Pappas chart of death right now to see if I can break it.
 
Card speed error and 192 byte empty P frames

Card speed error and 192 byte empty P frames

PJF said:
Just for info I tried this using PTool 3.51d and had a "card speed" recording stoppage on a daylight static foliage scene that showed lots of 192 byte P frame errors. There were also some of those on a hand held clip that did complete. I don't have version 3.50d so cannot try that (is it downloadable somewhere?).
Click to expand...

If the scene is static, doesn't the 192 byte P frame just say "that", i.e. there has been no change in the past 1/24 second?

I have had card speed errors myself when trying to shoot 24p, but I was never left with any file to inspect. (How) do you succeed in having a file written and being able to process or inspect it?

It's kind of funny (but it would be much more fun when this did not happen) that it seems to be a "too little data (to be) written" error that stops the recording. Isn't there anything to be done about that? I guess it should not be that difficult to track down in the firmware code, though I can hardly judge that, not knowing how to look at it.

It never seems to be a "too much data to be written" problem.
 
I just noticed something of interest when testing Pappas Death chart.

Every file that I have transferred into FCP since trying this Patch has had a couple of incredibly blocky frames right at the beginning of the clip. It goes away instantly and does not return.

When I tested the Pappas chart with the camera starting to record on the moving image it instantly had writing errors and froze up, but when I started the recording off of the Pappas chart and moved onto it, I could record for as long as I wanted to.

Take it for what it's worth I suppose. Best of luck
 
Magnus387 said:
I just noticed something of interest when testing Pappas Death chart.

Every file that I have transferred into FCP since trying this Patch has had a couple of incredibly blocky frames right at the beginning of the clip. It goes away instantly and does not return.

When I tested the Pappas chart with the camera starting to record on the moving image it instantly had writing errors and froze up, but when I started the recording off of the Pappas chart and moved onto it, I could record for as long as I wanted to.

Take it for what it's worth I suppose. Best of luck
Click to expand...

Yep i noticed that on all the clips too, you can see that frame in Streamparser, small i and p frame. We each have to decide if it's worth the extra risk or stay with lpowell reliable.

I'm still playing with it for the sake of better 720p 50fp.

R
 
Sorry, I forgot to mention that the first frame looks blocky and streaky. It even transcodes to CineForm that way. But it clears up immediately and stays perfect til recording stop.
 
Fool4UAnyway said:
If the scene is static, doesn't the 192 byte P frame just say "that", i.e. there has been no change in the past 1/24 second?
Click to expand...

I don't know enough to answer. I can say that for most of the clip (presumably similarly static) the problem P frames were not present. There were always three together, and the clip at that point had a weird brief "fuzz" that looked a bit like going out of focus in a grainy way for an instant.

I have had card speed errors myself when trying to shoot 24p, but I was never left with any file to inspect. (How) do you succeed in having a file written and being able to process or inspect it?
Click to expand...

Beginners luck? Honestly, I have no idea. The session ended abruptly with an error message about card speed and I assumed I'd stood in the cold for nothing. When I inspected the card there was a large file so I loaded it into the Streamparser thingie. The clip only lasted as long as it did because I was out of focus to start with (manual lens). Once I was in focus it did its strange intermittent fuzz thing for a while and then fell over. The bitrate follows the focussing nicely.

Note that I just put in the settings as Mark Crabtree listed and set the camera as NTSC rather than PAL. All else was default.
 
You must log in or register to reply here.
Back
Top