GH1 firmware research volunteers required

I want to clarify some things.

Any questions concerning newly added patches or old patches in testing state go to this thread.
Any general questions to sticky FAQ topic.
And other footage discussion to "high bitrate" topic.
 
Mode 6 is the same resolution than mode 4, and has 50 fps. Only fps changes. If it works, we could have a 720p@50fps MJPEG mode. Also, if then we set FPS for MJPEG encoder to 25 instead of 50fps, we can have a 720p@25fps mode.

And why do you think that it is same resolution?
Yes, after scaler resolution is the same.
But as far as I understand (today) RAW data resolution is different.
This is like 720p60 and new 720p24 AVCHD, second looks like same resolution, but in reality it is not so simple as it is scaled from much higher RAW resolution.
 
And why do you think that it is same resolution?
Yes, after scaler resolution is the same.
But as far as I understand (today) RAW data resolution is different.
This is like 720p60 and new 720p24 AVCHD, second looks like same resolution, but in reality it is not so simple as it is scaled from much higher RAW resolution.

Argh! You're right. I forgot that (many pages ago you told us).
Sorry, maybe isn't a good idea as I though :lipsrseal
 
New 720p24 mode also shows us real mud reason in 1080 modes.

1275932955.jpg


Notice how P frame size (blue bars) increases during movement.
Even here it looks like it hits some wall.

But in 1080p even still P frame size is already near top limit.
So, any movement causes hit to the wall and big overcompression.
Looks like idiotic reason to comply to 17Mbit standard rate caused all this mess.
 
Last edited:
I know not everyone is interested but tried the patch on the gf1, the 30min removal worked as expected :) but in 720 mode the gf1 is different as the gh1, the gf1 does not do 50fps it does 25 (pal) the thing is that it is wrapped in 50 frames, I dont know if it is interlaced or what but my timeline is 25p works perfect.
so I am not clear what the 720p 50 to 25 patch is doing.

I´m buying some batteries to test the patch.

keep up the good work can´t wait for recording in M mode
 
With ptools ver3.34 Tester13 added native 24p in both 1080 and 720 AVCHD and states it is a new patch and is "tested and working" and plays back in cam....

Does this mean that we have true native 24p without skipped frames??? Has this patch been tested by any one else to verify this???
 
Or hardware limit?

They design this LSI :-) So, it does not matter.
Plus I do not believe in such hardware limit as I see P frames size spikes during AVCHD patches testing. And each such spike is times more than current size.
In reality P frames have some constant top size value (and this value is adjusted for lower bitrates, btw).
If you have big bitrate setting, sometimes all remaining bitrate go to last P-frame in GOP sequence.
So, if we could crack this hard nut, it'll be good.
 
Last edited:
Maybe the scaler is our problem. Maybe that function/process/hardware wants a concrete input frame size that we can't change. Maybe if we can skip the scaling process and output a MJPEG with the same size than the sensor mode, it will work...

Or maybe there are some parameters that tell the scaler the input frame size,... :-m
 
Maybe the scaler is our problem. Maybe that function/process/hardware wants a concrete input frame size that we can't change. Maybe if we can skip the scaling process and output a MJPEG with the same size than the sensor mode, it will work...

Please, stop posting fantasies :-)
Current sensor patches are aimed at one thing only - find MJPEG sensor mode, test if it is used for AVCHD 720p mode also.
And changing MJPEG mode to other mode get appropriate movie files (with weird picture) and provide me link to them.
 
They design this LSI :-) So, it does not matter.
Plus I do not believe in such hardware limit as I see P frames size spikes during AVCHD patches testing. And each such spike is times more than current size.
In reality P frames have some constant top size value (and this value is adjusted for lower bitrates, btw).
If you have big bitrate setting, sometimes all remaining bitrate go to last P-frame in GOP sequence.
So, if we could crack this hard nut, it'll be good.

So the hardware has capacity that is wasted due to bad allocation of P frames in GOP sequence, so the code / routines that control P frame allocation are to blame. So far with have parameters related to GOP size, overall bitrate and buffer size. How were these discovered? I take it you are searching for more... :) Great work tester13
 
Hi Great work!
I tried the Mjpeg size/encoder patch it was very excited to see the higher resolution in mjpeg.
Then after shooting for about a sec I had a warning about my card speed...do I need something faster than class 6 sdhc?
 
Okay, a note about card speed, for those who don't know -- first, not all card manufacturers are equal, and second, there are definitely bogus knockoff cards out there. But thirdly, the way the speed ratings work is that the rated speed is a *minimum* but it is not a maximum.

So it's entirely possible that a Class 4 card could actually end up being faster than a Class 10 card!

How? Because you have to recall that the Class designations have grown; there was a time (recently) when there wasn't even a Class 10 designation. So if a card can sustain writing 4 megabytes *or faster* per second, it qualifies as a Class 4 card. If a card can sustain writing 6 megabytes per second (or more) then it is eligible for a Class 6 designation (but could also be classified as Class 4!) In order to earn Class 10 designation, it would have to be able to sustain at least 10 megabytes per second, but could also be classified as Class 6 or even Class 4 (or, heck, even Class 2, if they wanted to.)

So what if you had a really fast card back when Class 4 was the maximum designation? What if you had a card that could sustain 20 megabytes per second? It'd still be rated as Class 4, because that's the highest they had.

So is it possible that a Class 4 card could handle 20 megabytes a second, and a Class 10 card could handle only half that? Yes, that's possible. Not really all that likely, but it's definitely possible. So you can't just assume that a higher class rating will definitely mean a faster card. It SHOULD work that way, but it's not guaranteed. The only guarantee you have is that the Class 4 will support at least 4 megabytes (32 megabits) per second, and the Class 6 will support at least 6 megabytes (48 megabits) per second, and the Class 10 will support at least 10 megabytes (80 megabits) per second.

But any of those classes could be substantially faster. Or, if you're using some ripoff/knockoff card, your card may not meet those speeds at all.

The only way to know is to download a speed-testing program and actually try your cards out.
 
Last edited:
Just ran another quick test. I set sensor mode 4 to 7 only. AVCHD in 720p mode records normally with this setting. I think sensor mode 4 is definitely the MJPEG mode. It is the only one that when I change it, there is a funky picture. All the other modes when changed, this far, produce the same frame skipping picture (like if you don't change sensor modes but turn on 24p MJPEG) .

And on a side note, when I up MJPEG to 1920x1080 and switch sensor mode 4 to 7, there is still a funky picture.
 
Last edited:
Hey Guys'

Testing again'

Setting: AVCHD 1080p25/24 -> 720p25/24

Working but no resize picture ;) it mode crop picture 720p from 1080p ;-) like digital zoom.

Next test:

MJPEG mode sensor test.
(turn on: MJPEG Encoder 30fps -> Xfps: - 25)

Setting: Sensor mode 3: 3 -> 8

Working well, focus assistance too ;-)

Setting: Sensor mode 3: 3 -> 7

Working well, focus assistance too ;-)

Setting: Sensor mode 3: 3 -> 6

Working well, focus assistance too ;-)

Setting: Sensor mode 3: 3 -> 5

Working well, focus assistance too ;-)

Setting: Sensor mode 3: 3 -> 4

Working well, focus assistance too ;-)

(Next modes test and Files will be soon).

Files link: http://rapidshare.com/files/396443657/Sensor_Mode_3_TEST.zip

Regards
NURBS
 
Back
Top