GH1 firmware research volunteers required

Sensor m4 Adjustment 7 Test

Sensor m4 Adjustment 7 Test

MJPEG Quality Patched -> 384, 330, 300, 276
MJPEG Tables Patched -> 24, 24, 24, 24

Each adjustment tested in 4 conditions:

1) On-Camera Light On - ISO 320
2) Light On - ISO 1600
3) Light Off - ISO 1600
4) Light Off - ISO 100
-------------------------------------------------------------------------------------------------

Sensor m4 Adjustment 7 Tested with below settings

Set 1) 2200 (50% of normal)
Set 2) 6600 (150% of normal)
Set 3) 17600 (4x normal)

Results:

http://www.jerseyforged.com/sampleclips/m4_adjust7.zip

Set 1's settings seems to squeeze (not crop) the image to the upper portion of the screen... making the image roughly 442 pixels high and leaving a stripped pattern in the remainder of the screen. Where the pattern begins isn't at a constant pixel height as the border seems to bend upwards several pixels in the middle. The pattern is relatively static, with the exception of a slight side to size shake. Set 3's setting gives nearly identically results as set 1, except there's slightly less shake at the bottom.

Here's a reformatted video clip from set 1 -> http://www.jerseyforged.com/sampleclips/adjust7_clip.zip

Set 2's setting has an entirely different effect. The image is shifted (not squeezed) to the left by roughly about 311 pixels. Once again the image border has a curve. Furthemore, no diagonal stripes are present, but there's a white stripe at the border with the rest of image being black. The left side of the image is pushed off the side of the screen. The video shows these artifacts to be entirely static.. with no subtle shake as preset in 1 and 3.

In both cases you seem to see a bit more of the image at the border then you normally would. In the shifted image you can see much more of a stack of VHS in the room (on the upper right next to the can and yogurt). On the squeezed image you can see make out more of my nerdhood on the box below the test chart.

adjust7_pic.jpg
 
Sensor m4 Adjustment 8 Test

Sensor m4 Adjustment 8 Test

MJPEG Quality Patched -> 384, 330, 300, 276
MJPEG Tables Patched -> 24, 24, 24, 24

Each adjustment tested in 4 conditions:

1) On-Camera Light On - ISO 320
2) Light On - ISO 1600
3) Light Off - ISO 1600
4) Light Off - ISO 100
-------------------------------------------------------------------------------------------------

Sensor m4 Adjustment 8 Tested with below settings

Set 1) 1239 (50% of normal)
Set 2) 3717 (150% of normal)

Results:

http://www.jerseyforged.com/sampleclips/m4_adjust8.zip

I did not include Set 1 results because Set 1's 50% results were identical to the 50% value on the adjustment 7 test.. same squeeze, stripes, and subtle stripe shake.

Set 2's results initiated a shift like adjustment 7's second test, although in this case it was a 164-176 pixel shift upwards resulting in both a blank area and striped lines in the remaining bottom area.

Furthermore, the blank area in set 2 seems to flicker depending on the ISO setting. Here's a clip showing the patch at 320 and 1600 ISO....

http://www.jerseyforged.com/sampleclips/adjust8_videos.zip

Flickering at 320 is alot more noticable when I play it in Quicktime where the brightness/gamma/something seems to be a little high on my computer.
 
Those bends would be due to the automatic correction of the lens distortion, most likely if you'd be using an adapted lens you would not get this bending, but a straight line.
 
Each adjustment tested in 4 conditions..

Thanks again.
But due to my fault this results are not very useful.
Thing is that I believe that last two parameters (sw and sh) are directly related to sensor width and sensor height (hence the abbreviations).
So, such huge changes have little use.
It is advisable to use small adjustments and do not cross 4400 for width.
And same thing for height.

My understanding is as foollows - this settings together with other two(I'll add them to ptool soon) define actual window.
But RAW data transfer have other setup parameters (also will be added).
 
Yes, it is still useful.
I need reprted resolution and mode (like 1080i60) by TV for all playback modes.

Done those earlier, but probably got lost in the craziness. http://www.dvxuser.com/V6/showpost.php?p=2016680&postcount=2476

a) Regular HDMI playback mode.
1) Pictures, all aspects - 720p.
2) AVCHD (unpatched) FHD - 1080i.
3) AVCHD (native) FHD - 1080i.
4) AVCHD SHD - 720p
b) Viera-link on -- everything shows as 720p.

The two TVs I tried this on, a Panasonic and Sony only report resolution in HDMI mode. Refresh rates only reported when using VGA cable.
 
PTool 3.37d new release

www.pentax-hack.info/firmware/pan/ptool3d.zip

Removed some sensor m4 adjustment patches.
Four new sensor m4 patches added (promized to GrgurMG).

Note. This release is intended to experienced testers only.

to GrgurMG. I hope you could do something useful with this new patches.
Note that it is better to keep (raw w)*(raw h)=constant. But experiments also ok.
 
Last edited:
I know about this.
Looks pretty strange.
As no one of guys who tried this firmware found any difference.
I bet it addresses the blue shift bug.

People who want to try it, should google for "blue shift" in the GH1 forum. It happens with skies and certain shades of blue, on any gamma other than "standard."
 
I bet it addresses the blue shift bug.

People who want to try it, should google for "blue shift" in the GH1 forum. It happens with skies and certain shades of blue, on any gamma other than "standard."

OK, we'll see after testing.
Until this fix will be officially available at Panasonic site (not under password protected section) all I can do is wait for this version to become official.
 
Last edited:
Sensor m4 Adjustment 9 & 10

Sensor m4 Adjustment 9 & 10

MJPEG Quality Patched -> 384, 330, 300, 276
MJPEG Tables Patched -> 24, 24, 24, 24

Each adjustment tested in 4 conditions:

1) On-Camera Light On - ISO 320
2) Light On - ISO 1600
3) Light Off - ISO 1600
4) Light Off - ISO 100
-------------------------------------------------------------------------------------------------

Sensor m4 Adjustment 9 Tested with below settings

Set 1) 240
Set 2) 960
Set 3) 1920
Set 4) 0

Results:

http://www.jerseyforged.com/sampleclips/m4_adjust9.zip

Although the task at hand is pretty obvious, I've shown these to illustrate how this adjustment seems to function and how under or overshooting the saught after value looks. Set 1 and 2 are obviously undershot... Set 3 is overshot. FYI, a setting of 0 (Set 4) completely and immediately freezes the camera once you start recording.

Now, for the goal... to find standard value that you would plugin for this adjustment to operate as if it wasn't patched. Sounds easy enough... but not quite. As illustrates above I tried 1920... then I tried 1280... then upon doing some pixel math on how far 1280 seemed to undershoot it seems like it was going to be 1440... which was close but no cigar.

Closer I got to the value, the harder it was to compare. Shooting across at something with a tripod was no good.. cuz every little touch of the camera askewed the results.. so I wound up tapeing something across the lens hood to compare against. Also, you reach a point where the difference between the live view and the m4 while recording isn't noticable, due to the LCDs low resolution. But alas, long story short.. the number I came up with was 1452.. although due to various factor's.. I won't say I'm 100% on it. It seems simply selecting Adjustment 9 (which effects the horizontal) nudges the vertical up two pixels.. without having Adjustment 10 even selected. Furthermore, just playing with Adjustment 10 seemed to leave the horizontal nudged one pixel to the left (or was it right). Someone with a more stable testing setup may wish to retest for accuracy.

In any case, I moved to the Adjustment 10 setting which I will glance over because it simply works the same way as 9 in terms of overshooting, undershooting, freezing at 0, and so forth.... just with height instead of width. The number I've come up with so far for it is 820. That leaves the Adjustment 9 & 10 combo at...

1452 x 820

This however, may be slightly off as used together the image seems to be off by a pixel or two vertically... but when adjusting Adjustment 10 by two.. I got no difference. There's alot of factors that I'm sure are just toying with me.. like the scaling on encode. I tested with unpached size as well as patched to 1920.

I attempted to set the encoder width and height to 1452x820... but as with many irregular sizes it seems to instigate some error. Do we know much in regards to what allows a MJPEG size setting to work or not? Was there any possible benefit to having the MJPEG size and encoder size patched non-integrated in terms of resolving sizing issues?

Anyway.. it doesn't seem like I came up with much.. but just getting those numbers required ALOT of firmware reflashing. I'm very glad we got the version compare patch when we did. Lets assume 1452x820 is the correct number... what exactly does this mean? It seems to sensor scales down to this number.. then scales/crops to 720... or at least I'm guessing. Perhaps this number gives us insight into the cameras actual effective video resultion. Then again, perhaps I'm entirely off base... in which case feel free to chime in.

I was hoping to get to adjustments 11 and 12... but they'll have to wait a short bit.
 
Thanks for testing,
But it is slightly hard to clearly understands results.

Could you edit your post to include following things:
Pictures inserted in post (use Member Upload Center and some screenshot making utility).
Clear captions for each picture containing "Adjustment 9 sensor w cut" and "Adjustment 10 sensor h cut" values.
Thanks in advance.

It'll save some time for me and other people looking at this thread.

One very interesting thing is that:
1452 x 820 = 1190640
and 1119 x 984 = 1101096 (this is default values for raw adjustments 11 and 12)
pretty close.
So, you can try 1119 and 984.
 
Last edited:
Well for now I'll recap.

In the pics I posted, the first 9 pictures are the initial results of setting a value that's too low or too high.

Beyond that most of my pictures looked like the last two... a blurry card in the foreground (taped to my lens hood).. with all the pictures being just a pixel or two off. Problem I was having is not having a test rig/area near enough to my computer that I could easily work with the camera without moving anything even a fraction of a milimeter... which is what I'd need to get it pixel perfect. So I had to judge it off the blurry movie stub in the foreground.

Essentially 1452x820 seems to be the w (adjust 9) and h (adjust 10) settings that most closely look like unpatched 720 mode.. being only a pixel off if they aren't correct. Overshooting those values slightly clearly shows pixels passing the point in which patched and unpatched would be the same. Overshooting more then slightly actually produces video with a frozen image... in other words it doesn't freeze but there's no movement in the image.

Just to reformat... here's the comparison....

sensorcut_unpatched.jpg


sensorcut_patched.jpg


As you can see.. the test is not ideal. Preferably I'd have both the camera and test chart tethered down so they wouldn't move at all while operating.. then I'd have something in focus and detailed to compare the patch with... but alas' this is what was available.

As I mentioned... the patched version seems to be a pixel or two off in height... but when I tried to compensate with the sensor h cut setting.. it didn't change the results.

I'll see what else I can come up with.. in terms of a clearer test.
 
Last edited:
Better use tripod for next test and scale pictures to something like 720x405.
And printed chart.
Plus try to use values from my previous post.
 
Last edited:
Better use tripod for next test and scale pictures to something like 720x405.
And printed chart.
Plus try to use values from my previous post.

Might of misunderstood me in regards to the shot. Those shots are on a tripod.. there's just something taped to the lense hood to shoot. Essentially my tripod and work area aren't rigid enough to operate the camera without moving it ever so very slightly... in which case would throw off the results.

However, I'll see about maybe taking my GH1 into the garage and carefully using a table vice to hold it perfectly still while I operate the camera and subsequentally run back into the house to reload firmware for the second shot. Also, by scaling I think you mean to say cropping. Scaling down the images would completely hide the results we're trying to see.

As for the settings (1119 x 984)... here's what that looks like when plugged into adjustment 9 & 10....

P1000825.JPG


... like I said.. the settings to match up to unpatch 720 are about or exactly 1452 x 820. If those numbers are off at all it's by a pixel or two at the most.
 
Oh, also.. I'm not sure if there's already been notation on this or not, but these newer m4 adjustments effect all 720p (MJPEG/AVCHD) modes in addition to the 848 x 480 MJPEG mode. So to break it out....

Effected by new m4 adjustments: HD, WVGA, SH, H, L
Uneffected by new m4 adjustments: FHD, VGA, QVGA

Interesting that WVGA is included. In this case it's simply HD mode scaled whereas the other VGA modes function differently and aren't just the HD mode scaled and cropped.
 
M
... like I said.. the settings to match up to unpatch 720 are about or exactly 1452 x 820. If those numbers are off at all it's by a pixel or two at the most.

OK, thanks.
Before storing, routine take this numbers and multiply them by 3.
So, in reality it is 4356x2160.

As for affected 30p and 60p modes, yes, I am aware of this.
 
Quick Test

Quick Test

Tested these settings...

m4 Adjust 7: 4400 (default)
m4 Adjust 8: 2478 (default)
m4 Adjust 9: 1452 (found default?)
m4 Adjust 10: 820 (found default?)
m4 Adjust 11: 1119
m4 Adjust 12: 984

Results: Everything operated normally. Just wanted to verify having all these settings patched at default didn't pose any problems.

Then I tested this....

Same as above except
m4 Adjust 11: 1678 (50% increase)
m4 Adjust12: 1476 (50% increase)

Results: I got this picture, with the corresponding video doing alot flashing and flickering...

P1000830.jpg


What's also interesting is that upon stopping the recording, the camera is forced into a brief focus shift.. even though it's on manual the whole time. The increase of values seem to just crop the bottom of the image. Must test these settings individually in the future.

But for now lower values.....

m4 Adjustment 11: 560 (50% decrease)
m4 Adjustment 12: 492 (50% decrease)

Results: Instant camera freeze on record.

One battery pull later....

I decided to play with the sensor w/h settings instead. After all that systematic testing.. I just wanted to try some things....so....

m4 Adjustment 7: 4356 (reduced to x3 value you mentioned)
m4 Adjustment 8: 2160 (same as above)
m4 Adjustment 9: 1452
m4 Adjustment10: 820
m4 Adjustment 11: 1119
m4 Adjustment 12: 984


Result: Frozen footage squeezed to the top half of the screen... moving on....

m4 Adjustment 7: 4440 (default)
m4 Adjustment 8: 2478 (default)
m4 Adjustment 9: 1452
m4 Adjustment10: 820
m4 Adjustment11: 1319 (-200 from default)
m4 Adjustment12: 984 (default)

Results: Forced the camera out of focus with this image and flickering. Video did not freeze or become a static image however.

P1000835.JPG



Now......

m4 Adjustment 7: 4840 (+10%)
m4 Adjustment 8: 2726 (+10%)
m4 Adjustment 9: 1597 (+10%)
m4 Adjustment10: 902 (+10%)
m4 Adjustment11: 1408 (+10%)
m4 Adjustment12: 792 (+10%)

MJPEG Width/Encoder: 1584 (+10%)
MPEG Height/Encoder: 1188 (+10%)

Results:

http://www.jerseyforged.com/sampleclips/P1000842.JPG

Much to my amazement I got something close to proper footage. Aside from the bendy crop on the bottom of the screen.. I got actual 1584x1188 footage that didn't freeze or flicker.

I also tried that same test over, but with everything at +50% (1920x1080)... the clip started recording, but very glitched out and card error occured after a few seconds.
 
Get a wired remote.

Get a wired remote.

Might of misunderstood me in regards to the shot. Those shots are on a tripod.. there's just something taped to the lense hood to shoot. Essentially my tripod and work area aren't rigid enough to operate the camera without moving it ever so very slightly... in which case would throw off the results.
....

Get a cheap wired remoted control on ebay. Search for Panasonic DMW-RS1

This way you won't cause vibrations when pressing the shutter button during testing. You will need to set the mode dial to video for this.

Thanks for your efforts and good luck.
 
Back
Top