The strangest artifacts i´v seen!

[steinzeug]

Advertisement (gone after registration)

The original images once again:

 

First is a JPEG (100% crop) made from the original DNG file after importing directly into Photoshop

And second a photo of the same file on the LCD screen before it even left the camera.

 

I now understand that different raw converters can react differently either aiding or worsening the effects,

however what is clear is that the issue is occuring first in the camera itself.

 

So the question is, is it the lens or the chip?

 

Mike

Hello guest!
Please register or sign in to view the hidden content.

Hello guest!
Please register or sign in to view the hidden content.

Edited January 19, 2010 by steinzeug

[Advertisement]

Hallo,

Look here: Leica M

[steinzeug]

Another theory that I am hearing is that due to the lens being old (1974) and not having any antireflex coating on the back lens,

that the highlights are bouncing off the sensor and back onto the lens, and in effect exagerating the highlights resulting in the distortions.

 

Does anyone have a take on that?

Edited January 19, 2010 by steinzeug

[luigi bertolotti]

Ok, also the RAW developer I have in this PC (ACR 3.7 - PS CS2) shows the effect...

Hello guest!
Please register or sign in to view the hidden content.

 

Hello guest!
Please register or sign in to view the hidden content.

 

I think is difficult it's the lens... the multiple reflections sensor - lens - sensor can result in "light spots" located in various point of the image (depending on angles - lenses) , NOT exactly in the light transition areas : we have seen some examples in the forum : typical case, which I remember has been posted some time ago, is street lamps at night... you could see some greenish spots on the road's surface.

I think of the de-mosaicizing algorithms...roughly, they took the digital values of 4 adjacent pixels (from the Bayer filter) to extrapolate the RGB value of a single pixel... with a so strong light transition along a straight line (but at an angle in respect to the pixels' matrix)... I think something strange can happen... I do not think the algorithms make a 2nd level "intelligent" analisys to verify if the 1st level computation has brought to "too regular" patterns (mazes and so) with the goal to "fix" them in case.

Edited January 19, 2010 by luigi bertolotti

[pgk]

My two pence worth.

 

I'd be as virtually certain that this is NOT a lens problem.

 

Given that different raw converters are producing varying results, all of which show poor translation of the raw file in extreme exposure differentiated hard edges (I suspect that the light area is fully clipped adjacent to the dark edge), I'd suggest that the raw file is being incorrectly 'coded' or 'described' in these areas, which to me would indicate a camera chip/software problem - I'd doubt the sensor, more likely within the internal electronics.

 

If I had this problem I would first contact Leica about it and then email them samples to illustrate the problem. Since no-one else has indicated it as being seen before it looks like something of a one off and Leica will no doubt be interested to find out the cause.

 

Hope you sort it out, and please, let us all know the solution.

[jaapv]

Another theory that I am hearing is that due to the lens being old (1974) and not having any antireflex coating on the back lens,

that the highlights are bouncing off the sensor and back onto the lens, and in effect exagerating the highlights resulting in the distortions.

 

Does anyone have a take on that?

Those mosaics realy do not look like a reflection. I see the idea of a failure of the A/D converter as most likely of all possibilities.

[pop]

From the raw file I form the tentative opinion that the problem occurs for a narrow strip of correctly exposed pixels between two areas where the pixels are either saturated or unexposed.

 

The M8 does not have multi-colored pixels. Rather, each sensor dot receives light which is filtered by one of three color filters. The color for each pixel is computed from the raw data by interpolating the values of the nearest sensor dots having the same color filter.

 

It seems quite possible that de-mosaiking algorhithm fails in the particular situation in the image (narrow strip between saturated and/or unexposed dots).

 

Now, if there was a way of "seeing" the raw image before de-mosaiking...

[steinzeug]

Advertisement (gone after registration)

This from Leica:

 

We must to inform you, that your CCD has some defect. This is not firmware / converter problem. Please contact your next Leica repair service.

 

Not sure what this means, but does sound like Leica think its an in-camera problem?

 

Mike

[Advertisement]

Hallo,

Look here: Leica M

[jaapv]

Oh yes- they think it is a sensor problem, so we all were looking in the wrong place. If I were you, I would send it directly to Solms.

[dNorm]

Wow, I would not have guessed based on my experience with my E-P1. Keep the lens, send the camera back for repair and then test again under the same conditions. I'd be curious to see if the problem got fixed with a sensor replacement. Cheers. -Norm

 

This from Leica:

 

We must to inform you, that your CCD has some defect. This is not firmware / converter problem. Please contact your next Leica repair service.

 

Not sure what this means, but does sound like Leica think its an in-camera problem?

 

Mike

[RichC]

Sorry to hear that the camera needs a repair - but there is a silver lining: as well as fixing the fault, the camera will get a thorough service, and the work will be guaranteed for a year (that was my experience when I dropped my camera and it had to go back to Solms).

[steinzeug]

Sorry to hear that the camera needs a repair - but there is a silver lining: as well as fixing the fault, the camera will get a thorough service, and the work will be guaranteed for a year (that was my experience when I dropped my camera and it had to go back to Solms).

 

Thanks to everyone for input. Will update this thread when the problem is solved.

 

Mike

[steinzeug]

An update:

 

I was sent a brand new Summarit 35mm to test the camera with, just to be 100% sure that the distortion wasn´t caused by the lens.

 

Unfortuantely for me the same artifacts can clearly be seen with both lenses (although the newer lens is somewhat sharper and has the encoding).

 

I am still surprised that no one else has come across this before, judging from Leicas email they seem to recognize the problem straight away.

 

Does anyone know how long repairs take, it´s not going to be fun parting with the camera, as i only just got it.

 

Here are the 100% crops form the DNG files. The sharper images are from the new Summarit. The artifacts are just as prevalent in the original DNG files and on the camera LCD.

Hello guest!
Please register or sign in to view the hidden content.

Hello guest!
Please register or sign in to view the hidden content.

Hello guest!
Please register or sign in to view the hidden content.

Hello guest!
Please register or sign in to view the hidden content.

[jaapv]

Most reports say that Solms is pretty good these days, NJ seems to have a backlog.

[Simon M]

Disappointingly I've just found I have the same issue.

 

Back to Solms...

[Guest WPalank]

Simon M,

 

Careful, there is a very similar issue that occurs when you are using "pre-Intel" Macintosh computers and Lightroom 2.5. You aren't using that combo by any chance are you?

[Advertisement]

Hallo,

Look here: Leica M

[steinzeug]

Disappointingly I've just found I have the same issue.

 

Back to Solms...

 

Hello Simon,

 

yes it might be the raw converters/software issues, however should you see the artifacts on the LCD screen before exporting then you will unfortunatley have the same in-camera problem that i have. My camera is now on its way to Leica, and will be interessting to see what they say.

 

Mike

[Simon M]

Hi William, Hi Mike,

 

I'm pretty sure it is not my computer - the problem is visible on the LCD before exporting also.

 

Mike, could you keep us updated on your results.

 

I'm working away at present so won't be able to return the camera to Solms until March.

 

Thanks,

 

Simon

[steinzeug]

Hi William, Hi Mike,

 

I'm pretty sure it is not my computer - the problem is visible on the LCD before exporting also.

 

Mike, could you keep us updated on your results.

 

I'm working away at present so won't be able to return the camera to Solms until March.

 

Thanks,

 

Simon

 

 

Will do!

[steinzeug]

 

UPDATE: well finaly I recieved the verdict from Solms today.

It is down to a faulty sensor and the CCD needs to be changed.

Cost approx 700 EUR plus taxes.

 

Unfortunately no details on how or why the problem occurred.

[AlanG]

 

UPDATE: well finaly I recieved the verdict from Solms today.

It is down to a faulty sensor and the CCD needs to be changed.

Cost approx 700 EUR plus taxes.

 

Unfortunately no details on how or why the problem occurred.

 

Do you reside in the EU? Does 700EUR plus taxes mean + VAT? If so it would end up being around 840 Euros? I thought repair prices were quoted inclusive of VAT.

 

It is hard for me to understand how a defective sensor could resolve everything about an image accurately except for an extremely bright to dark edge transition. That seems to me like a firmware/software issue unless something about the sensor design has been changed.