M9P sensor fixed by MAXMAX - big problem!

[Sonnar Brian]

I've seen worse problems. The sensor that I worked on in 1983: we used 15-bit A/D convertors, the company that built the hardware did not properly isolate the digital and analog circuitry. 9 bits of noise on a 15 bit pixel. $1M to fix.

 

[andyturk]

[...] a decent programmer should be able to fairly easily dig into the firmware and remap the sensor.

What????

You mean reverse-engineer the firmware binary and then replace pieces of it? For one camera? Yikes.

That's a lotta work. Better to see if you can pay Leica to re-run the factory calibration procedure.

 

[Freakscene]

What????

You mean reverse-engineer the firmware binary and then replace pieces of it? For one camera? Yikes.

That's a lotta work. Better to see if you can pay Leica to re-run the factory calibration procedure.

No, essentially once you get into and understand the firmware you can run the calibration procedure. It is a reasonable amount of work, but most of the documentation is available on the web from others who have hacked the camera firmware. That’s how I did it, and the first one took about half a day. There are a series of calibration images in .jfif format of dark, middle and light grey and the same for each of r, g and b embedded in the firmware, and these map the pixel values across the sensor and feed them back in to the secondary processor (the Blackfin 547) for management of the file after capture and before recording. This is what the calibration procedure does.

Marty

 

[raid]

Brian knows very well what he is talking about here. Very impressive stuff! I hope that Luuca can get his camera(s) fixed.

 

[Sonnar Brian]

Marty- wow. That is an eloquent solution. If possible, posting the links to the information you found would be useful and interesting. Posting your work would also be interesting.

I'm "old-school" quick and dirty. I had my Fortran code fixing the bad scan line on the M Monochrom in a couple of hours. Always remember- "It's just typing".

 

[luuca]

leica answered!!

guess what they said???

it's in italian so I will resume the mail in english:

"no problem Luuca, we've the solution!
you can buy a brand new m10!!!!" 🙄

they say there is a problem of "dual input" probably caused by Dan job that caused the "loss of calibration", not fixable by leica.

but they have the solution.
A fantastic trade up with new leica models.

 

[luuca]

No, essentially once you get into and understand the firmware you can run the calibration procedure. It is a reasonable amount of work, but most of the documentation is available on the web from others who have hacked the camera firmware. That’s how I did it, and the first one took about half a day. There are a series of calibration images in .jfif format of dark, middle and light grey and the same for each of r, g and b embedded in the firmware, and these map the pixel values across the sensor and feed them back in to the secondary processor (the Blackfin 547) for management of the file after capture and before recording. This is what the calibration procedure does.

Marty

that would be awesome...

I really love my M9 and although I usually shoot film, I really want it back fully working.
I don't need at all a brand new camera

 

[Freakscene]

leica answered!!

guess what they said???

it's in italian so I will resume the mail in english:

"no problem Luuca, we've the solution!
you can buy a brand new m10!!!!" 🙄

they say there is a problem of "dual input" probably caused by Dan job that caused the "loss of calibration", not fixable by leica.

but they have the solution.
A fantastic trade up with new leica models.

I am sorry. This is one circumstance where you really should have said ‘my M9 has started doing this’. Maybe Leica would think about doing something. But ‘I tried to get a non-Leica repair and now this is happening!???’ We have laws in Australia about third party repairs and manufacturers needing to respect them but most places don’t. Of course Leica might have stopped fixing M9s at all, but I don’t know.

Marty

 

[Freakscene]

Marty- wow. That is an eloquent solution. If possible, posting the links to the information you found would be useful and interesting. Posting your work would also be interesting.

I'm "old-school" quick and dirty. I had my Fortran code fixing the bad scan line on the M Monochrom in a couple of hours. Always remember- "It's just typing".

This: https://alexhude.github.io/2019/01/24/hacking-leica-m240.html provides 90% of what you need.

The .jfif files need to be properly formatted and documented; you cannot just rename .jpgs.

Calibration is in the subprocessor firmware and refers to the localization file. If you can find someone who can get you this far you shouldn’t have too much trouble getting to the calibration procedure. But you might brick your camera.

It has just occurred to me - have you tried reinstalling the firmware? I am not sure if it would make any difference, but it might.

Marty

 

[Pál_K]

I told him there's plenty of images on my pc that can proof that the problem wasn't there before his work... he continues to say that I didn't notice the defect :rolleeyes:
yes, in 10 years I never noticed it...

Send him a few of the earlier photos with the EXIF information as proof as to the date and the camera being used. Then he cannot dispute what you say.

...in the meantime I opened a dispute with the card company

Good.

 

[luuca]

I am sorry. This is one circumstance where you really should have said ‘my M9 has started doing this’. Maybe Leica would think about doing something. But ‘I tried to get a non-Leica repair and now this is happening!???’ We have laws in Australia about third party repairs and manufacturers needing to respect them but most places don’t. Of course Leica might have stopped fixing M9s at all, but I don’t know.

Marty

I agree with you if leica still would repair sensor corrosion, but they don't... I could only tell them the thruth.

 

[jawarden]

I agree with you if leica still would repair sensor corrosion, but they don't... I could only tell them the thruth.

I think you did the right thing and I’m sorry to hear how bad it has gone so far. I hope MaxMax does right by you and if they don’t you should send them a link to this thread.

 

[jsrockit]

Leica’s response was predictable unfortunately. Stay on top of your credit card company about the payment. Keep it simple. For those of us who don’t know about programming at all, I feel for you.

 

[Sonnar Brian]

This: https://alexhude.github.io/2019/01/24/hacking-leica-m240.html provides 90% of what you need.

The .jfif files need to be properly formatted and documented; you cannot just rename .jpgs.

Calibration is in the subprocessor firmware and refers to the localization file. If you can find someone who can get you this far you shouldn’t have too much trouble getting to the calibration procedure. But you might brick your camera.

It has just occurred to me - have you tried reinstalling the firmware? I am not sure if it would make any difference, but it might.

Marty

Interesting that Photoshop does not natively recognize "JFIF"- as it is the standard for the Internet to pass jpeg files. There are programs to convert it. The harder part would be to generate the calibration files and get them into the camera. Calibration data cannot be included in the firmware that is downloaded, but must be retained in the camera.

My "quick-and-dirty" solution is all out-of-camera. I need to see some files from at least two levels of grey from the camera to see if it works. My code for the M Monochrom takes the bad column number as input, figures out the defect by itself. I think a similar "input file" for "gain and offset" to bring the two halves together would work. Very simple compared with what Arvid wrote for the M8 Raw to DNG program.

 

[luuca]

here you are the link for the dng from the cameras:

https://www.dropbox.com/s/vj1gv4lpkav64eo/M9GREY_L1066669.DNG?dl=0

https://www.dropbox.com/s/fuccsnyp9dywfe2/M9PBLACK_L1066664.DNG?dl=0

they are uncompressed dng, shoot to a grey target straight to the cameras
f8
iso 160
1/360
manual AWB

as you can see on the grey camera the issue is huge.
on the black one it is hardly noticeable, but it is there...

 

[luuca]

Interesting that Photoshop does not natively recognize "JFIF"- as it is the standard for the Internet to pass jpeg files. There are programs to convert it. The harder part would be to generate the calibration files and get them into the camera. Calibration data cannot be included in the firmware that is downloaded, but must be retained in the camera.

My "quick-and-dirty" solution is all out-of-camera. I need to see some files from at least two levels of grey from the camera to see if it works. My code for the M Monochrom takes the bad column number as input, figures out the defect by itself. I think a similar "input file" for "gain and offset" to bring the two halves together would work. Very simple compared with what Arvid wrote for the M8 Raw to DNG program.

I didn't see the "at least two levels of gray" part... :rolleeyes:
I will shoot other 2 pics

 

[Sonnar Brian]

I looked at the grey image- the difference is almost 10% between right and left half.

For the Black image- the picture is much brighter than the grey image. It is almost 1/2 of the total scale for a white image, like 50% grey. I would expect "Black" tp be much darker, did you use the same exposure?

It would also help to get a picture of the grey card on a table or something, so I can get a feel for what you used.

 

[Sonnar Brian]

It is interesting that the problem is evident for the image at ~30% scale, and almost gone at 50% scale, using Photoshop to look at values- scaled from 0:255. It's like the calibration has changed, and is applying the wrong slope.

SO: easy measurement using the existing greyscale card: measure the correct exposure. Shoot successive images using Manual Exposure to Bracket from -4 stops to +4 stops in 1 stop increments. Keep the aperture the same for all as to not introduce changes from vignetting. This will cover the dynamic range fairly well. We should get an idea of what is going on at each brightness level. I'm hoping we can measure the slope of the differences between the two halves and just do a simple multiply.

 

[luuca]

beware, they are 2 dng from 2 different cameras!

the one with the bigger difference is done by a grey M9P
the oother is from a black m9p of my friend
both of them were sent to MaxMax

the grey one returned with a bigger issue, the black one with a slighty noticeable issue.

tomorrow evening I will post other dng, as you asked

 

[luuca]

It is interesting that the problem is evident for the image at ~30% scale, and almost gone at 50% scale, using Photoshop to look at values- scaled from 0:255. It's like the calibration has changed, and is applying the wrong slope.

SO: easy measurement using the existing greyscale card: measure the correct exposure. Shoot successive images using Manual Exposure to Bracket from -4 stops to +4 stops in 1 stop increments. Keep the aperture the same for all as to not introduce changes from vignetting. This will cover the dynamic range fairly well. We should get an idea of what is going on at each brightness level. I'm hoping we can measure the slope of the differences between the two halves and just do a simple multiply.

ok, I will follow your instructions