Leica M9 CAL and M9 FIX Software for out-of-calibration M9 Sensors

[Rob-F]

This thread should be a sticky (do we still have that?). It looks to be over my head, but still, I have an M9 and an M9M, so you never know.

 

[Sonnar Brian]

This thread should be a sticky (do we still have that?). It looks to be over my head, but still, I have an M9 and an M9M, so you never know.

Done. Hard to believe it's been over 3 years since doing this.
If the M9 and M Monochrom have the new CCD that uses BG-55 glass, the calibration was done when the CCD was replaced.
Buying a camera with a broken sensor, or one that is corroded- then this code comes into play. We used to call this "Non-Uniformity Correction". I was writing code to do that in 1982, one of the first Digital IR sensors to fly.

 

[jwewj]

The Kodak CCD is comprised of two smaller CCDs which are abutted. The calibration for the two halves brings the image into agreement.
The type of filter glass was changed, which upsets the calibration.

Keep us posted on results.

After testing, the program appears to brighten the image on the left

And my photo is bright on the left and dark on the right

 

[Sonnar Brian]

I will need the .dng files that you used to produce the calibration. I would also like a .dng file of a scene showing the problem.

C FIVE FILES WITH UNIFORM LIGHTING MUST BE TAKEN USING THE CAMERA.
C IT IS BEST TO USE AN 18% GREY CARD FOR THE CALIBRATION:
C M2.DNG: UNDEREXPOSE 2 STOPS
C M1.DNG: UNDEREXPOSE 1 STOP
C P0.DNG: CORRECT EXPOSURE
C P1.DNG: OVEREXPOSE 1 STOP
C P2.DNG: OVEREXPOSE 2 STOPS

 

[Sonnar Brian]

After testing, the program appears to brighten the image on the left

And my photo is bright on the left and dark on the right

I would suggest uploading the DNG files using dropbox, and either post links or send me a private message.

 

[jwewj]

I would suggest uploading the DNG files using dropbox, and either post links or send me a private message.

Dropbox

 

[Sonnar Brian]

I downloaded the files: what we learned 3.5 years ago is that the calibration changes based on the ISO setting. This has to do with Analog Gain circuits in the camera.

Please shoot another set of calibration images with the camera set to ISO 320. The supplied calibration files are for ISO 640.
Also- I'd like an image shot at ISO640 to test the calibration code.

 

[Sonnar Brian]

The imbalance on your camera is much larger than the previous two cameras. I think this means the calibration will only be valid for the ISO actually in use.
I need to look for the Serial Number for the M Monochrom- I thought I had the correct location using my camera, but "M9CAL" did not find it in yours. I probably have the incorrect offset, will check this weekend.

These are the calibration points generated for your( @jwewj ) Monochrom:
CALIBRATION POINT, LEFT vs RIGHT
1071.0 , 1059.0
1981.0 , 1959.0
3010.0 , 2968.0
5179.0 , 5080.0
11269. , 9865.0

This is the 2021 case:
CALIBRATION POINT, LEFT vs RIGHT
563.00 , 550.00
1080.0 , 1065.0
2085.0 , 2046.0
4112.0 , 4058.0
8145.0 , 8015.0

 

[jwewj]

Thank you very much for your help

In fact, recently I have also used chat-gpt to write a new program to deal with this problem.

Now we have a prototype, but we still haven’t been able to successfully complete the calibration.

The imbalance on your camera is much larger than the previous two cameras. I think this means the calibration will only be valid for the ISO actually in use.
I need to look for the Serial Number for the M Monochrom- I thought I had the correct location using my camera, but "M9CAL" did not find it in yours. I probably have the incorrect offset, will check this weekend.

These are the calibration points generated for your( @jwewj ) Monochrom:
CALIBRATION POINT, LEFT vs RIGHT
1071.0 , 1059.0
1981.0 , 1959.0
3010.0 , 2968.0
5179.0 , 5080.0
11269. , 9865.0

This is the 2021 case:
CALIBRATION POINT, LEFT vs RIGHT
563.00 , 550.00
1080.0 , 1065.0
2085.0 , 2046.0
4112.0 , 4058.0
8145.0 , 8015.0

 

[Sonnar Brian]

The location of the ISO setting is at offset '02C2'x in the DNG file. The location of the copy of the Serial number is at '0694'x, different from the location with the new CCD and sensor board. I wonder if ChatGPT can figure that out by looking at a HEX dump of the DNG file. Just look for the value '0280'x in little endian format, that is "640" decimal.

 

[jwewj]

During these days of shooting tests, I found that the imbalance between the left and right is the most serious when the ISO is at 640, and it rarely occurs when the ISO is 1000. After 1250, this problem disappears at all.
So currently I am shooting at iso 1000.

 

[Sonnar Brian]

The first set of calibration files was shot at ISO 320, so that is interesting.

You can shoot another set at ISO160 and ISO1250, may make a difference.