GH1 firmware research volunteers required

Also impossible to understand, just strange.
How they can be hardcoded then we change them and it works?
Try to read my posts in last pages.

GOP is an arbitrary number. It may define a length of I/P frames in a group but it does NOT define the quality, type or actual quantity of the frames in that group.

Therefore some other process MUST be determining the quality/quantity and makeup of the GOP.

Being able to define how many I or P frames and their quality/compression would be more beneficial than defining how long the GOP is and letting the subroutines do their best to try to accommodate.

That's why I think people are having issues with writing to the cards, because adjusting the GOP might bring any routine that defines the quality/size and or number of I/P frames outside of it's coded boundaries and therefor confuse any other routines that might happen later in the processing chain.

EDIT:

I suppose what I'm saying is that more research needs to be done on the processes involved and we need to focus less on just changing GOP numbers and seeing what happens.. We are on to something with changing the GOP but we need to investigate the lower level stuff first due to the fact that we are seeing the GOP changes cause issues and not actually "fixing" anything.
 
GOP is an arbitrary number. It may define a length of I/P frames in a group but it does NOT define the quality, type or actual quantity of the frames in that group.

Therefore some other process MUST be determining the quality/quantity and makeup of the GOP.

Being able to define how many I or P frames and their quality/compression would be more beneficial than defining how long the GOP is and letting the subroutines do their best to try to accommodate.

That's why I think people are having issues with writing to the cards, because adjusting the GOP might bring any routine that defines the quality/size and or number of I/P frames outside of it's coded boundaries and therefor confuse any other routines that might happen later in the processing chain.

Sounds smart.
Main problem is that even having code it is not that easy to do.
And looking at pure assembler it is almost impossible.
So, we are doing that we can.
Of course fantasies are good, but hardly archivable.
 
PTool 3.12 released

Native patch now in officialy added (same as in late v3.11)
Japan->English interface patch added (must be popular for many people) - it DO NOT change your region or any other settings, only interface.
 
I suppose what I'm saying is that more research needs to be done on the processes involved and we need to focus less on just changing GOP numbers and seeing what happens.. We are on to something with changing the GOP but we need to investigate the lower level stuff first due to the fact that we are seeing the GOP changes cause issues and not actually "fixing" anything.

Yes, we need.
But we better look at various encoder sources to understand that we can expect. It is pointless to research thousands of low level assembler functions related to AVCHD encoding.
You could have some result around 2040 :-)
 
My all players show this video how 25p progresive video ;-)

PS: Maybe someone can repeat this test and check it on NLE or some better software for that. ;)

Regards
NURBS
 
Last edited:
My all players show this video how 25p progresive video ;-)

Does this mean the dreaded pulldown is now gone?
NURBS, could you please specify which patches you've applied/with which settings? Any specific camera settings?

If this is real, we are looking at a first major breakthrough in this project ... I'll hold off on congratulations until confirmed :2vrolijk_08:
 
Sounds smart.
Main problem is that even having code it is not that easy to do.
And looking at pure assembler it is almost impossible.
So, we are doing that we can.
Of course fantasies are good, but hardly archivable.

Don't be too rough on people offereing suggestions.
Not everyone here has an understanding what a disassembled listing of a few MB of binary code looks like.
So here is some simple math.

The 1.22 bin file is 8.9 MB in size.
Assuming one line of assembly code for 4 bytes of binary code brings us to 2 million lines of assembly code.
Throwing away the sections dealing with constants, empty blocks, etc. still results in hundreds of thousands of lines of code to look through.
Being disassembly code this one has no comments whatsoever.

This is just to underscore my amazement at the rapid progress achieved thus far. Good luck to you. You are our hero.
 
You need turn on this settings: AV Native 24/25p attempt on Ptool3 and nothing more.

And one sore point to tester13

Possible in the Future this changes in camera menu:

Recording Mode: ( 50p SH: 17 Mbps, 50p H: 13 Mbps, 50p L: 9 Mbps)
change to (50p SH: 20 Mbps, H: 25p 20 Mbps, 24p L: 17 Mbps) ? :-)
 
You need turn on this settings: AV Native 24/25p attempt on Ptool3 and nothing more.

And one sore point to tester13

Possible in the Future this changes in camera menu:

Recording Mode: ( 50p SH: 17 Mbps, 50p H: 13 Mbps, 50p L: 9 Mbps)
change to (50p SH: 20 Mbps, H: 25p 20 Mbps, 24p L: 17 Mbps) ? :-)

I tested the AV Native 24/25p feature on an NTSC model. I only checked it and the 'version change' checkbox. I recorded in FHD mode which looked like it was working. When I tried to play back in camera, the camera froze and I had to remove the battery to shut it off. Looking at the file on my computer, it is very jerky, recording every other frame and repeating frames. The file size itself is extremely small: 7.5 MB for a 20 second clip.
 
Don't be too rough on people offereing suggestions.
Not everyone here has an understanding what a disassembled listing of a few MB of binary code looks like.
So here is some simple math.

The 1.22 bin file is 8.9 MB in size.
Assuming one line of assembly code for 4 bytes of binary code brings us to 2 million lines of assembly code.
Throwing away the sections dealing with constants, empty blocks, etc. still results in hundreds of thousands of lines of code to look through.
Being disassembly code this one has no comments whatsoever.

This is just to underscore my amazement at the rapid progress achieved thus far. Good luck to you. You are our hero.


I am familiar with it. I am suggesting that, while using the GOP changes as a start is good, it seems that it creates issues that are not well understood and therefor should not be inferred as being the actual fix we are looking for. I believe there are underlying functions in the code that would be much better for changing to suit our needs but we should attempt to understand the whole picture before simply just changing parts without method.


I'm not saying that Tester13 is not doing a good job, because he is doing a GREAT job. I just think there seems to be a rush to get some kind of result yet we don't really know what the code is doing. I understand that it's a lot of code, that's why I'm not understanding why there is such a rush. Rushing makes errors happen.

It's my belief that the GOP length setting is very likely independent of the I/P frame attributes in the AVCHD encoding engine(at least it SHOULD be if someone wrote the code with full adjustment in mind), and that the GOP changes we are making in the firmware are only a piece of what we should be looking at. If this is not the case then we would not be getting errors when changing GOP lengths.
 
Users who bought Japanese cameras rejoice.

Users who bought Japanese cameras rejoice.

I have a Japanese model of GH1 (Cheaper and I love the color Red), will the hack firmware work? If so, I'm planning to donate $$. Oh, how difficult to convert from Japanese menu to English?
Got your wish. Congratulations.
Check the first page.
 
Suggestion for version control.

Suggestion for version control.

Tester13.

I have a suggestion regarding version control.
Can you replace the "compare for greater than current" by "compare if not equal to current" version?
This is how it works on ZS3 and LX3 cameras.
Once we have this feature users will be able to load the original 1.22 version.
This will produce more testers, once people know they can always go back to the official firmware version.

Thanks and good luck.
 
Very interesting moment!

Wen turn on: AV Native 24/25p attempt
We can do setting: VA 1080p25 GOP Size: 3 (no freeze or text message). ;-)

But if I want to play this video on camera "Play mode" Camera go to freeze..
 
Rolling shutter

Rolling shutter

Hi,

I don't know if what I am proposing is possible or not, but may be worth a look: the GH1 has the infamous rolling shutter problem. However, it is not that bad with higher frame rates. E.g. 720p 50fps rolling shutter is much less visible than 1080i 25fps. This means two things:

1. Has anybody checked the effect of 720p 24/25fps modifications on the rolling shutter issue? 50fps was acceptable, but changing to 24/25fps may reintroduce the issue.

2. It may be worth a look if it is possible to get 50fps from the sensor in 1080 mode, and skip every second frame in the generated AVCHD file. This would mean that the rolling shutter would be reduced while having a nice 1080 file.

I am not saying that these are easy modifications, maybe they are impossible without the source code. I am just mentioning to keep them in mind :cheesy:
 
This was addressed to Tester13.

This was addressed to Tester13.

I am familiar with it. I am suggesting that, while using the GOP changes as a start is good, it seems that it creates issues that are not well understood and therefor should not be inferred as being the actual fix we are looking for. I believe there are underlying functions in the code that would be much better for changing to suit our needs but we should attempt to understand the whole picture before simply just changing parts without method.


I'm not saying that Tester13 is not doing a good job, because he is doing a GREAT job. I just think there seems to be a rush to get some kind of result yet we don't really know what the code is doing. I understand that it's a lot of code, that's why I'm not understanding why there is such a rush. Rushing makes errors happen.

It's my belief that the GOP length setting is very likely independent of the I/P frame attributes in the AVCHD encoding engine(at least it SHOULD be if someone wrote the code with full adjustment in mind), and that the GOP changes we are making in the firmware are only a piece of what we should be looking at. If this is not the case then we would not be getting errors when changing GOP lengths.

I think we need to help in organizing better analysis of test results.
Many times the results are not very clear.
 
Hi,

I don't know if what I am proposing is possible or not, but may be worth a look: the GH1 has the infamous rolling shutter problem. However, it is not that bad with higher frame rates. E.g. 720p 50fps rolling shutter is much less visible than 1080i 25fps. This means two things:

1. Has anybody checked the effect of 720p 24/25fps modifications on the rolling shutter issue? 50fps was acceptable, but changing to 24/25fps may reintroduce the issue.

2. It may be worth a look if it is possible to get 50fps from the sensor in 1080 mode, and skip every second frame in the generated AVCHD file. This would mean that the rolling shutter would be reduced while having a nice 1080 file.

I am not saying that these are easy modifications, maybe they are impossible without the source code. I am just mentioning to keep them in mind :cheesy:

AFAIK the rolling shutter effect is caused by relatively long exposures. As the sensor data is being read sequentually, by the time the lower part of the frame is being processed the subject may have moved a bit causing visible distortion of vertical lines. That is why with shorter exposures the effect is minimized. Lowering the frame rates allowes slower shutter speeds, which cause the effect. But some users like the 24/25fps mode as it produces more of a "cinematic film look", whatever that is.
 
Got your wish. Congratulations.
Check the first page.

What? Already? Wow, didn't expect this quick, awesome. Will try to install it tonight but first, I need a PC, I am MAC. Guest I'm going to get a copy of Parallels Desktop.

Again, thanks a bunch. Keep it up!
 
AFAIK the rolling shutter effect is caused by relatively long exposures. As the sensor data is being read sequentually, by the time the lower part of the frame is being processed the subject may have moved a bit causing visible distortion of vertical lines. That is why with shorter exposures the effect is minimized. Lowering the frame rates allowes slower shutter speeds, which cause the effect. But some users like the 24/25fps mode as it produces more of a "cinematic film look", whatever that is.

Rolling Shutter is more of an artifact of the CMOS sensor used and less of the camera itself or codec. But your right, since you have longer exposures the more noticeable the effect is. CCDs and Uniform Synchronous Shutter CMOS sensors aren't afflicted by Rolling shutter.

The Foundry, the company that makes the compositing software Nuke, has a plugin for After Effects and Nuke called "Rolling Shutter" that is designed to take out the artifacts from a line-scan CMOS sensor camera.
 
What? Already? Wow, didn't expect this quick, awesome. Will try to install it tonight but first, I need a PC, I am MAC. Guest I'm going to get a copy of Parallels Desktop.

Again, thanks a bunch. Keep it up!

It works! English instead of Japan language. Thanks!! Going to test MJPEG 24fps now.
 
Back
Top