Sensor Corrosion Analysis and Fix [Merged]

[jdlaing]

Advertisement (gone after registration)

6 minutes ago, Herr Barnack said:

That was my point in my post #118. 

If that is indeed what happened, I wonder why Leica did not pursue legal action against Kodak to recover the costs incurred by them in resolving Kodak's, uh,  "forgetfulness?"  That seems like a reasonable course of action.

If an agreement was reached to share costs no suit was necessary.

[adan]

I'm just going to add a trio of data points.

1) Living in high-and-dry Colorado, I only experienced one instance of sensor corrosion across two standard M9s. One hairlike line, possibly originating from otherwise invisible surface cleaning damage (faint scratch). And that was after 6.5 years (2016) and was repaired at the very end of the "free replacement" program (8/2017).

2) regarding the thickness of the IR-cut cover glass, we should recall that there was a batch of M9 CCDs prone to stress-cracking across a corner. Which was probably an assembly problem, or possibly simply a batch of glass with pre-assembly stress flaws - I leave that to the glass gurus present here. But may explain why Leica/Kodak did not use the even-thinner 0.6mm glass.

3) If I recall correctly, the M8, where the glass spanned a smaller APS-H sensor, got away with 0.5mm IR-cover glass and had no cracking problem. And, as far as I know (?), also has never had corrosion problems.

___________

And a side question for the glass gurus, that I've always wondered about (but don't put too much time and space into answering here; even a PM will be fine). Is there an air gap between the cover glass and the silicon sensor surface?

And is that sealed around the edges (one presumes so, for dust reasons)?

And can that mean a thin cover glass can act like a barometer/altimeter diaphragm - flexing a bit with a change in external air pressure vs. sealed internal pressure, such as changes in altitude? And what effects might that have on focus (slightly bowed cover glass acting as an additonal "lens element"), tendency to crack, and so on.

I take my cameras, often rapidly (couple of hours) from Denver's 5280-foot elevation down to sea level, and up to 14000+ feet (road-accessible peak of Mt. Evans, CO). And there have been times when it seems the focus calibration goes off slightly (and temporarily) when I'm away from "a mile high" significantly.

Edited September 16, 2020 by adan

[Advertisement]

The Leica Q3 is available for pre-order at B&H Photo Video and Adorama

[dllewellyn]

The cracked sensor coverglass sounds like thermal stress where the coverglass is warming or cooling too fast compared to the ceramic chip package.

Yes, there is an air gap between the coverglass and silicone surface.  The sensor sits inside a small well.

Yes, sensor is completely sealed.

I have rarely seen the coverglass crack because of a change in air pressure though I did see it once for a camera used in a high altitude airplane experiment.  I think the problem is they went from a pressurized cabin environment to suddenly opening a hatch were the camera was located.

Focus will also change with the temperature of the lens.  I have seen lenses that can't focus at infinity when they are cold.

Edited September 16, 2020 by dllewellyn

[martinot]

On 9/15/2020 at 11:46 PM, BrianS said:

So-  you announced that Kodak and Leica were "Stupid, Stupid, Stupid"

Given what actually happened in reality, and how much it has costed them (both financially and reputation wise), I think Leica would agree 100% that is was stupid.

[jaapv]

8 hours ago, adan said:

I'm just going to add a trio of data points.

1) Living in high-and-dry Colorado, I only experienced one instance of sensor corrosion across two standard M9s. One hairlike line, possibly originating from otherwise invisible surface cleaning damage (faint scratch). And that was after 6.5 years (2016) and was repaired at the very end of the "free replacement" program (8/2017).

2) regarding the thickness of the IR-cut cover glass, we should recall that there was a batch of M9 CCDs prone to stress-cracking across a corner. Which was probably an assembly problem, or possibly simply a batch of glass with pre-assembly stress flaws - I leave that to the glass gurus present here. But may explain why Leica/Kodak did not use the even-thinner 0.6mm glass.

3) If I recall correctly, the M8, where the glass spanned a smaller APS-H sensor, got away with 0.5mm IR-cover glass and had no cracking problem. And, as far as I know (?), also has never had corrosion problems.

___________

And a side question for the glass gurus, that I've always wondered about (but don't put too much time and space into answering here; even a PM will be fine). Is there an air gap between the cover glass and the silicon sensor surface?

And is that sealed around the edges (one presumes so, for dust reasons)?

And can that mean a thin cover glass can act like a barometer/altimeter diaphragm - flexing a bit with a change in external air pressure vs. sealed internal pressure, such as changes in altitude? And what effects might that have on focus (slightly bowed cover glass acting as an additonal "lens element"), tendency to crack, and so on.

I take my cameras, often rapidly (couple of hours) from Denver's 5280-foot elevation down to sea level, and up to 14000+ feet (road-accessible peak of Mt. Evans, CO). And there have been times when it seems the focus calibration goes off slightly (and temporarily) when I'm away from "a mile high" significantly.

The most likely cause for the cracking was probably the glue at the edges. The composite material used shrinks when polymerizing If the gap is too wide the amount of shrink will put a high stress on the glass. Add thermal cycling by the heat of the sensor and a crack is a logical result.

[michali]

In early-2015 the original sensors on both my M9 & MM1 showed corrosion. In addition the Monochrom's sensor glass also cracked (see image). These were replaced FOC by Leica at that time. 

Then in late-2019 corrosion re- appeared on both cameras at about the same time.  

After several months of drawn out arm wrestling sessions with Leica they refused to come the party.

Finally in Feb. this year I took the plunge & first sent the Monochrom to Germany for sensor replacement. This ended up costing me in excess of EUR1, 900- after factoring in courier charges & ridiculous customs duties, as a result of Leica's stubbornness & insistence on declaring the camera at a higher value on its return from Germany to South Africa.

Once the Monochrom was returned to me, the plan was to then send the M9 for repair, however Covid & lockdown intervened & I was unable to send the M9 to Germany & of course now they'e no longer replacing these sensors.....

To the point being made about environmental factors, it's interesting that the corrosion on both cameras (MM1 and M9) appeared over exactly the same time period i.e. early-2015 & again late-2019. I suspect that environmental factors may have also played their role in this.  At the time I was travelling extensively with both cameras between Johannesburg alt. 5, 600ft and south of France at sea level. 

Right now I'm sitting with an M9 paper weight, so this new sensor replacement option appeals to me. However I need to work out the customs issues, after being stung by SA customs on the MM1's return last time.

MM1 Sensor:

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

 

 

[BrianS]

Advertisement (gone after registration)

11 hours ago, martinot said:

Given what actually happened in reality, and how much it has costed them (both financially and reputation wise), I think Leica would agree 100% that is was stupid.

From the article, "Leica / Kodak used an uncoated Schott BG type glass!  The coatings seal the glass.  Somebody *really* screwed up. "

Kodak and Leica used sealed cover glass that was MAR coated. The seal did not hold up. Dan announced that Kodak and Leica used unprotected/unsealed glass. If they had, every sensor made would have shown corrosion after a few months. Not just a few spots, but widely spread on the surface. The breakdown of the seal happened, and a lot of factors went into the seal breaking down on individual sensors. To simply announce that Kodak, Leica, and Schott all got their data sheets wrong and that conclusive proof was made by examining one bad sensor is a bit much. 

Using S8612 glass with the thin seal was a mistake, hindsight is 20-20.

 

Even in 2009, Kodak's primary market for CCD's was the scientific/technical market, not consumer oriented products. You could buy a KAF-18500 from Digi-Key. 

 

https://www.photonstophotos.net/Charts/Sensor_Heatmaps.htm#mode=23,camera=Leica M10-R,suffix=14

 

Scroll down the table of sensor performance to the M8, M9, and M Monochrom in the above table. The last two columns are dark signal non-uniformity (DSNU) and Photo Response Non Uniformity (PRNU) is - like Fixed Pattern Noise (FPN) - a way of expressing errors in the output from sensors. The lower the value, the better the raw image. I will never sell my M Monochrom, or M9. My M8 is one of the last in the production line. I'll keep it as well. 35 years ago- would have selected the M Monochrom based on the last two columns alone.

 

 

Edited September 17, 2020 by BrianS

[kristyansen]

vor 20 Stunden schrieb dllewellyn:

I would tend to go with the factory  0.80mm thick glass.  One can second guess the factory and put in 0.60mm glass and change the shims, but there is probably a reason the factory chose 0.80mm.  The factory could have put in 0.50mm glass if they wanted.  I also have doubts about how much improvement your are going to see with 0.60mm glass.  If enough people want it, I can easily make up 0.50mm glass and change the shims accordingly.

RE changing the color profile.  To get a certain answer, I would want to measure camera spectral response of a stock and converted camera because the color response is a combination of the filter and the camera RGB response.  For example, the glass in the M9 I measured goes down to 300nm but the sensor has very little sensitivity down there.  But I think you would notice very little difference and if you set a custom White Balance, then probably no difference at all.

Here is the M9 ICF that I measured versus a typical ICF.

 

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

I would like to study the color response and light transmission, and ideally we would have your opinions, MaxMax and Kolari.

First some anecdotical points showing the importance to me of the color response which is much of the appeal of the Leica M9:  

1. I used to have 5-6 years ago a Leica M9-P which had an official sensor change by Leica. The color response of the sensor had however changed, and was not just a matter of White Balance - the color palette was nice, but different; also in addition the pictures became slightly sharper, luminosity was slightly reduced and the images slightly colder (not as Kodachrome like anymore). As you know the firmware was updated after the official sensor change, which is not the case for the suggested "only glass change" solutions - so same camera RGB response. 

2. Due to this I had bought again a Leica M9 with original sensor without corrosion at the time - I think it was one of the first ones produced - (with the feeling that if something happened I could still have the official sensor changed). The colors are indeed more like Kodachrome. I could not check the camera in time to see the spots that now in retrospect seem to have appeared around July 2020 especially since there was no pre-announcement by Leica that the sensor change program would end in Aug. Either way, I would like to stick to the "old" colours as much as possible if I aim to perform such as sensor glass change.

Based on data from Schott, and as non expert, I am trying to understanding the color response of the two solutions being suggested, original M9, MaxMax and Kolari: 

- On the original M9, the original sensor cover glass was S8612 0.8mm (based on comments on this forum). Schott's official transmission data sheet here: https://shop.schott.com/medias/schott-bandpass-s8612-jun-2017-en.pdf?context=bWFzdGVyfHJvb3R8Nzk3ODUzfGFwcGxpY2F0aW9uL3BkZnxoM2YvaGFlLzg4MjAyMzQ3ODA3MDIucGRmfDUzYzE4ZGI3OWY5NDU4MzY3NDZhZDc1OTk1MGQyZWU0Mzc3N2UyM2EyYWM2ZjI1MzI5ZGMzZTgwZDI2ZjNlOGQ

- MaxMax: replacement cover glass BG38 0.8mm. Schott's official transmission data sheet seems similar to the original S8612, and slightly higher IR sensitivity; still the official graph seems more sensitive than Dan's graph in this post). https://shop.schott.com/medias/schott-bandpass-bg38-jun-2017-en.pdf?context=bWFzdGVyfHJvb3R8Nzk2NTk2fGFwcGxpY2F0aW9uL3BkZnxoMmMvaGQwLzg4MjAyMzM4NjMxOTgucGRmfDA2NGQ4ZmY2MmUzOGEyZmIyNWM4MzA3MGExMGYyNTgxZGJmZDNmYjBmNmQ3NmNmZTRiMTFmYWJlMDAwNTZkYTM

- Kolari: replacement cover glass BG50 0.6mm (with possible need to readjust sensor position). Schott's BG50 glass is mentioned having high humidity resistance; the official transmission data sheet here also shows slightly less transmission than BG38 and higher IR sensitivity: https://shop.schott.com/medias/schott-bandpass-bg60-jun-2017-en.pdf?context=bWFzdGVyfHJvb3R8Nzk4NDMzfGFwcGxpY2F0aW9uL3BkZnxoMGQvaDMzLzg4MjAyMzQzMjE5NTAucGRmfDdjNmQ1NjBiMzY0ODkzMTQ0ZjQzZTRkNzhmYjkyMzg2ODY2ZGRhMDNlNGUwOGRmOTIxMTM5M2FlNjA1YTFhZDU

Questions: 

 a) which solution is more accurate color-wise to the original M9 (I think both increase IR sensitivity similarly) and

b) which solution ensures a similar light transmission? If there is a different transmission, what is the cause? 

[BrianS]

ONSEMI also changed the color dye used in the Bayer mosaic filters of the new sensors, in addition to changing the firmware and using BG55 glass. This was done as an attempt to match for the new sensor to match the color of the original. The response of the color dye and filter both affect the color being recorded.  It is not possible to change the dye used in the old sensors "in a cost effective manner". The BG38 has IR leakage properties similar to the filter used on the Leica M8. The transmission curve for BG55 and S8612 in IR is much lower than BG38. The color being recorded by the sensor is going to change.

Edited September 17, 2020 by BrianS

[dllewellyn]

Hi Brian, Curious, how do you know that ONSEMI changed the dye?  Of course you can't change the CFA on an old sensor.  That would involve removing the microlenses and CFA and then using photolithography to create a new CFA and microlenses.  At that point, your are in the semiconductor chip manufacturing business.  Seeing that current foundries cost in the $5-10 billion dollar range and are designed to work 24/7 making huge volumes of chips, doesn't quite make sense.

[Ko.Fe.]

8 hours ago, michali said:

In early-2015 the original sensors on both my M9 & MM1 showed corrosion. In addition the Monochrom's sensor glass also cracked (see image). These were replaced FOC by Leica at that time. 

Then in late-2019 corrosion re- appeared on both cameras at about the same time.  

After several months of drawn out arm wrestling sessions with Leica they refused to come the party.

Finally in Feb. this year I took the plunge & first sent the Monochrom to Germany for sensor replacement. This ended up costing me in excess of EUR1, 900- after factoring in courier charges & ridiculous customs duties, as a result of Leica's stubbornness & insistence on declaring the camera at a higher value on its return from Germany to South Africa.

Once the Monochrom was returned to me, the plan was to then send the M9 for repair, however Covid & lockdown intervened & I was unable to send the M9 to Germany & of course now they'e no longer replacing these sensors.....

To the point being made about environmental factors, it's interesting that the corrosion on both cameras (MM1 and M9) appeared over exactly the same time period i.e. early-2015 & again late-2019. I suspect that environmental factors may have also played their role in this.  At the time I was travelling extensively with both cameras between Johannesburg alt. 5, 600ft and south of France at sea level. 

Right now I'm sitting with an M9 paper weight, so this new sensor replacement option appeals to me. However I need to work out the customs issues, after being stung by SA customs on the MM1's return last time.

MM1 Sensor:

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

 

 

Mine M-E 220 (purchased new in late 2016) has cover glass corroded for no reasons by 2017.

I have Canon EOS 500D. It was traveling with me frequently, was in all kinds of temperatures and it is still on its original sensor from 2009. I never seen Canon sensors to be reported with similar to M9 sensors covers corrosion. 

It is nice we have two services now taking care of it. 

 

[compman10]

Would it be possible to use S8612 that has a better anti-corrosion coating?  That would minimize the possibility of a change in color rendition.

[BrianS]

6 hours ago, dllewellyn said:

Hi Brian, Curious, how do you know that ONSEMI changed the dye?  Of course you can't change the CFA on an old sensor.  That would involve removing the microlenses and CFA and then using photolithography to create a new CFA and microlenses.  At that point, your are in the semiconductor chip manufacturing business.  Seeing that current foundries cost in the $5-10 billion dollar range and are designed to work 24/7 making huge volumes of chips, doesn't quite make sense.

New Part Number from ONSEMI, kaf-18500-fxa-jh-aa-bg55
Dan-

The part number relates to a lexicon from the ONSEMI website, I found these back in 2015. 

The new part number for the KAF-18500 indicates that a different pigment is used in the RGB mosaic filter, as per the ONSEMI document the "FXA" was "Pigment RGB Bayer, GEN 2". I downloaded the naming convention documents years ago- all this is gone from the website now. I probably have the naming convention PDF file somewhere on one of 12 computers. I have a lot of computers.

It is not possible to change the color dye in the old sensor at this point. It would be worth looking at images taken of color charts before and after the conversion and possibly providing some software to make the new color-space closer to the old one. IR leakage would be a problem for something like this. I ended up writing my own software to process M Monochrom, M9, and M8 DNG files. They read and modify the values without going through the demosaic process. On the M Monochrom, apply a gamma curve and use 16-bits per pixel instead of 14 for the output. On the M8: use an Orange filter to cut out Blue channel, leave it IR only. Then equalize it to get a Visible+IR image. On the M9- Jaap stated that a yellow filter would render better Monochrome. Convert the M9 files to linear DNG with a custom interpolation process. Maybe something similar to scale the new color response back to the old one might work. Image hacking.

The part number of the old sensor is KAF-18500-NXA-JH-AA-08.

This thread will bring back memories- it is the 2015 discussion that went on here when Leica was revising the sensor. There was a noted error in the 2014 Schott data sheet for S8612, seen in the discussion. I Emailed Schott- from work, never got an answer from them. The next Data Sheet from them agreed with all the others collected.

 

Edited September 17, 2020 by BrianS

[kristyansen]

We really appreciate there are two solutions emerging, both with pros and cons. 

The questions in this thread are meant to better understand the final result (and not only the process) since the conversion is new and, still, a considerable investment. 

To increase understanding, would it be possible to create a few test photos (RAW, and JPG Color and JPG) - including skin color / portraits, IR-sensitive surfaces, made with the converted sensors, for both solutions of Kolari and MaxMax? Thank you. (Creating special conversion software is beyond what normal users would be able to do).

 

[AlbertTRAL]

1 hour ago, BrianS said:

New Part Number from ONSEMI, kaf-18500-fxa-jh-aa-bg55
Dan-

The part number relates to a lexicon from the ONSEMI website, I found these back in 2015. 

The new part number for the KAF-18500 indicates that a different pigment is used in the RGB mosaic filter, as per the ONSEMI document the "FXA" was "Pigment RGB Bayer, GEN 2". I downloaded the naming convention documents years ago- all this is gone from the website now. I probably have the naming convention PDF file somewhere on one of 12 computers. I have a lot of computers.

It is not possible to change the color dye in the old sensor at this point. It would be worth looking at images taken of color charts before and after the conversion and possibly providing some software to make the new color-space closer to the old one. IR leakage would be a problem for something like this. I ended up writing my own software to process M Monochrom, M9, and M8 DNG files. They read and modify the values without going through the demosaic process. On the M Monochrom, apply a gamma curve and use 16-bits per pixel instead of 14 for the output. On the M8: use an Orange filter to cut out Blue channel, leave it IR only. Then equalize it to get a Visible+IR image. On the M9- Jaap stated that a yellow filter would render better Monochrome. Convert the M9 files to linear DNG with a custom interpolation process. Maybe something similar to scale the new color response back to the old one might work. Image hacking.

The part number of the old sensor is KAF-18500-NXA-JH-AA-08.

This thread will bring back memories- it is the 2015 discussion that went on here when Leica was revising the sensor. There was a noted error in the 2014 Schott data sheet for S8612, seen in the discussion. I Emailed Schott- from work, never got an answer from them. The next Data Sheet from them agreed with all the others collected.

 

Spot on Brian. The N on the old sensor part number had a shorter red wavelength as per ON SEMI Device nomenclature manual TND310 and F is the Gen 2.

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

Edited September 17, 2020 by AlbertTRAL

[martinot]

14 hours ago, Ko.Fe. said:

Mine M-E 220 (purchased new in late 2016) has cover glass corroded for no reasons by 2017.

I have Canon EOS 500D. It was traveling with me frequently, was in all kinds of temperatures and it is still on its original sensor from 2009. I never seen Canon sensors to be reported with similar to M9 sensors covers corrosion.

Corrosion in less than one year, brand new?

That indeed sounds suspiciously like they totally skipped any coating at all on that one.

(Regarding Canon; I have cameras from Canon, Nikon, Panasonic, Fujifilm, and have never heard of any form of corrosion on any of their camera models.)

Edited September 18, 2020 by martinot

[BrianS]

The M-E 220 purchased late 2016, after production had ended- may have been manufactured well before it was sold. I've had a number of custom boards and chips made for my projects over the last decades. I've seen manufacturing defects bring down some systems, and errors in the design of chips that require revision. For what it is worth- 

 

The project manager that oversaw the glass selection should have caught this. Leica paid dearly for this design decision. Most of us were taken care of my Leica, I have both cameras with new sensors at a fraction of the cost of buying the Sensor from Digikey. Leica ended the program when ONSEMI shutdown the entire CCD production line. They are obsolete. Not even end-of-life, but completely unavailable.

An engineer wants to get the best possible performance with components selected, S8612 is optically the best performing glass from Schott for the job of cutting IR and preserving UV. The latter is important to some people, especially in the technical field. In 2009, Kodak was out of the high-end camera business and the CCD's were aimed at the scientific market. The engineers bet that the "seal" (both sides coated, as per the lexicon on the part number, new and old) selected to protect the filter would stop the corrosion. They bet, and where wrong. The Project Manager should have realized this was "iffy" at best, went against the recommended procedure from Schott for a Class 3 glass, not enough life-testing was performed, and used another glass. BG55 was reintroduced in 2011, too late for the M9 and M Monochrom. Managing engineers can be like herding cats. That's what my supervisor told me when I was new to being a manager 26 years ago. I'm glad to be "Senior Staff" and back in the Lab. Now back to debugging the latest board made for me. The last bug found required capacitors to be changed on the board. Better than herding cats.

[analog-digital]

Hi Guys

 

I have here a M9 (anthrazit, 1200 Shots) with the sensor problem. I got it for free from my leica dealer.

Question: 

Which of the 2 exchange solutions have which empirical values?

Has it done this before and what exactly?

[kristyansen]

In the thread you see the parameters of glass, depth, sensitivity etc. I would at the same time like to see some RAW and JPG test photos of typical scenarios (including people, nature, low, high ISO) to better understand the results. 

[Advertisement]

Hallo,

Look here: Leica M

[analog-digital]

1. Has anyone here in the forum already made this repair?
2. Maybe someone from the two companies is reading along and can help.