Bad sensor or bug in pixel mapping?

[Guest roey]

Advertisement (gone after registration)

I have noticed some artifacts on some of my M9 images. It usually shows up when I have thin, close-to-vertical lines (in portrait orientation) in a specific area of the frame. Here is an example:

 

 

 

This is a 200% screenshot (top) and the same screenshot modified to highlight the artifacts: vertical line segments of cyan, magenta, or some light, background-like color. I have ruled out that this is a lens defect, it is not specific to a specific ISO value, and it is not a problem with the Raw converter -- the line shows up in the raw DNG data as a one-pixel line.

 

The lines always have the same distance from the right edge.

 

When I had a sensor defect on my M8 it showed up as a solid line staring in the frame and extending down to the edge of the frame. This is different.

 

What do you think: is this a defective sensor or a bug in the pixel-mapping code, that is responsible for mapping out defective pixels?

[jaapv]

How do you look at RAW data without using a converter?. I would try another converter than your current one anyway, if only to rule out the possibility of an interpolation error.

[ckthual]

Can you post a 100% crop (a 200% crop can be a lie) please ?

[vanhulsenbeek]

A sensor flaw ( as bad pixels etc) would have to show up in the exact same spot every time.You say: specific area: on a 18Mb pixels sensor that would be many, many pixels!

This looks like an interpolation error indeed: try another converter.

[jaapv]

Otoh, the problem in this thread looked very much like a interpolation problem as well, but it did turn out to be the sensor. It never harms to contact Leica in cases like this.

 

http://www.l-camera-forum.com/leica-forum/leica-m8-forum/112841-strangest-artifacts-i-v-seen-2.html#post1196472

[adan]

I had something similar that showed up in about 1 in 500-1000 M8 shots. Rare enough that I can't even find an example right now.

 

A thin green-cyan "slit" of maybe a mm in length, like the bottom half of your largest example. I would have thought "interpolation error" except that it was too consistent in color (no barberpoling between green and magenta as it cut across pixel rows - just green) and location.

 

My conclusion was that it was a minor sensor flaw - silicon, or a tiny scratch in the cover glass - that only showed up when there was a potential for aliasing anyway. Ulltimately I decided it was not worth being without the camera for 6 weeks since it showed up so rarely.

[Guest roey]

Advertisement (gone after registration)

How do you look at RAW data without using a converter?. I would try another converter than your current one anyway, if only to rule out the possibility of an interpolation error.

 

You ask somebody who develops raw converters for a living. And, BTW, I have tried other converters and like I said, this isn't a conversion problem.

 

Can you post a 100% crop (a 200% crop can be a lie) please ?

 

Here you go:

 

 

 

A sensor flaw ( as bad pixels etc) would have to show up in the exact same spot every time.You say: specific area: on a 18Mb pixels sensor that would be many, many pixels!

 

I also said: The lines always have the same distance from the right edge. But I guess I should have been more precise in the statement you quoted. I think there are three columns in the same area of the sensor that can show this problem (depending on the picture).

 

I had something similar that showed up in about 1 in 500-1000 M8 shots. Rare enough that I can't even find an example right now.

 

Now that I know what I am looking for (and where) I can find it in many more files. E.g.:

 

 

a tiny scratch in the cover glass - that only showed up when there was a potential for aliasing anyway.

 

That's an interesting theory. But I think in my case the lines are too straight and vertical.

[ggits]

I was playing in my yard with M9 & Summilux 50, and I experience something like this.

 

I was shooting at a table covered with snow and the pictures show a tiny purple line between the snow and the background (light but darker than snow).

 

Just playing. I was shooting a series with different aperture and speed setting - all with same ISO. I don't have the pictures anymore, but I think to recall that it was seen with f 1,4 - 2.0 and slow closing time. Picture was overexposed.

 

My conclusion on that time was that shooting white in white one have to be very careful with exposure time.

 

Never seen it since.

[adan]

Roey, you might want to look at this S2 thread: http://www.l-camera-forum.com/leica-forum/leica-s2-forum/112897-s2-vacation-pointnshoot-gallery.html

 

Skip to the 12th post or so.

 

I'm beginning to agree with the idea that this is a place where Leica mapped out a bad pixel line. I'm not sure it necessarily means a bug in the code, though.

 

More likely it is due to the fact that Leica doesn't use anti-aliasing filters over their sensors, which makes for a sharper image, but means the data from adjoining pixel lines is more "different" and thus can't be used to replace a mapped-out line as cleanly.

 

An AA filter would smear detail a little into the surrounding pixels, making it easier to "borrow" it back without noticeable effect in mapping out a bad pixel or line of pixels.

[jaapv]

It sounds like a reasonable supposition, Andy, but it also means that it is possibly solvable in software ( from a layman's point of view )

[Guest roey]

Roey, you might want to look at this S2 thread: http://www.l-camera-forum.com/leica-forum/leica-s2-forum/112897-s2-vacation-pointnshoot-gallery.html

 

Thanks, Andy! It looks remarkably similar to what I am seeing. I was actually warming up to your idea about scratches on the sensor cover glass ;-). I really need to contact Leica CS about this -- so far I have been putting that off, because I was afraid to hear their answer.

[Guest roey]

A quick update: I have since contacted Leica and have been told that they need to replace the sensor and circuit board. They are currently waiting for parts.

 

I have found images with the defect dating back to November which is about 2 weeks after I got the camera. So I am assuming that it had this problem from the start.

 

I have also verified that my sensor has four specific columns of pixels that show the defect whenever there is a strong contrast edge: I shot a pattern of vertical stripes displayed on a computer display at a slight angle. Here is a 100% crop of the result:

 

 

[Guest roey]

Today I got the M9 back from Leica NJ with a new sensor -- it no longer exhibits the problem I reported above. Once Leica had received the necessary parts and I had gotten a slot in their repair queue everything went very quick. I was without the camera for a total of 10 days and it came back in perfect working condition and with a clean sensor.

 

I know Leica NJ has lately been criticized on this forum for the quality and speed of their service -- and I don't mean to diminish those problems -- but I figured it doesn't hurt to post about good experiences as well.

[MvMaltese]

Interesting, I just got my M9 from BH yesterday and have a similar line/slit issue (posted pics here: http://www.l-camera-forum.com/leica-forum/leica-m9-forum/122633-line-artifact-m9.html). Thinking about sending back to BH for exchange. Very disappointing considering the wait time to get the thing. Well, back to the M8.