New banding issue comment

[jrc]

I also posted this message in the thread entitled "M8 1.09 & filter problems summary" which includes a discussion of a new banding problem seen in some cameras. But that thread is so long and convoluted I thought I should also start a new thread here.

This new banding issue looks exactly like the banding that some MF backs are struggling with. As I understand it, it has to do with separate readouts of the sensor going in different directions off the chip (for speed's sake, more than one readout is done, but simultaneously.) As I understand it, the hardware that does the readout must be adjusted to be precisely the same to make the border areas (between the two or more separate readout areas) look the same. If the readout hardware is not adjusted properly, to make the interpretation of the very subtle voltages coming off the sensor, you get a band. The reason that the band is in precisely the same place is that the readout areas of each chip is precisely the same -- but it's not the chip that's the problem, is the interpretive hardware. There are some indications in the MF forums that this may be adjustable in firmware. Whether this is so, I have no ability to judge.

Here's a discussion of the phenomenon, with pictures, on the Luminous Landscape:

http://luminous-landscape.com/forum/...howtopic=13103

JC

[Riley]

John
can you check that link, i dont think its right

[harmsr]

John,

I did not see this vertical band on the 1.06 camera, but only on the 1.09 camera. As I understand it, Leica changed the hardware which reads the sensor information to eliminated the previous horizontal bands coming off of light sources.

My fear is that it has created this new issue. Relative to whether it needs another hardware change or can be fixed in firmware, is unknown to me.

It does need to be eliminated.

Best,

Ray

[Riley]

i dont know how this sits for logic but...
if the firmware change created the problem
then new firmware will undo it
i believe that makes sense

given that, everyone ought to take a breath i think

[marknorton]

To recap, the pixel content is read out of the sensor by shifting the rows down into the next row below. Rows which "drop off the bottom" are copied into a read out register which has two outputs, one at each end. Samples on one side of the image go to one read out amplifier, samples on the other side of the image to the other.

If the outputs are kept separate for analog to digital conversion, this is where you would see the MF back problem. Due to fractional differences between the two analog to digital converters (ADCs) which convert the pixel content to a digital value, the same brightness level would lead to slightly different digital values on each channel. The two ADCs would have to match at black and full intensity and at every point in between. If not, then take a picture of a uniform surface and you wil see the "join" between the left and right hand sections.

My guess is Leica actually use a single ADC and switch between the two channels so that the outputs from each side of the sensor are measured by the same ADC. That's where the green blob came from - feeding the ADC successively very bright and very dark samples from opposite sides of the sensor caused it to mis-convert the dark signal, leading to a dark mirror image which was green because the Bayer pattern is mostly green.

What's odd about this new banding is that it is on the right hand side of the image without any corresponding artefact in the mirror image. My guess is that this is a firmware problem, an incorrect calculation somewhere. Needs to be fixed and the people who can reproduce the problem need to rattle Leica's cage until they ask you to stop.

[marknorton]

Originally Posted by Riley

given that, everyone ought to take a breath i think

Riley, you may well be right, and I hope you are, but what this whole saga shows up is that Leica are not doing enough testing. They should be testing this camera to death, both with structured testing to systematically exercise and limit check as many code paths in the firmware as possible and unstructured testing where they hand the camera to people who will tell it the way it is and say, "there you are, go break it".

It's worth noting that all the problems in the new camera are in the electronics and firmware. Both are new areas to Leica (I discount the M7's electronics which are a walk in the park by comparison). Those elements of the M8 which were derived from the M7/R9 - shutter, viewfinder, mechanicals are all proving fine.

[sagarwal]

Originally Posted by Riley

can you check that link, i dont think its right

Here's the complete link.

[Riley]

Originally Posted by marknorton

Riley, you may well be right, and I hope you are, but what this whole saga shows up is that Leica are not doing enough testing. They should be testing this camera to death, both with structured testing to systematically exercise and limit check as many code paths in the firmware as possible and unstructured testing where they hand the camera to people who will tell it the way it is and say, "there you are, go break it".

It's worth noting that all the problems in the new camera are in the electronics and firmware. Both are new areas to Leica (I discount the M7's electronics which are a walk in the park by comparison). Those elements of the M8 which were derived from the M7/R9 - shutter, viewfinder, mechanicals are all proving fine.

there is much truth in what you say Mark
my assessment of Leica staff for the moment is
they are chasing spot fuel fires with a bucket brigade
if these were any other people but germans i would be real concerned
but i have faith that they will get it together, for their core skill as a people is organisational preparedness
that said, Leica's life is on the line here, so there are two immediate needs that fairly intersect with one interest
get the cameras working and the customers by extension will be happy
if the customers are happy, then the worth of this camera will be revealed in their art

[booggerg]

Riley, the new german automobiles are horrendous for electronic bugs. Maybe we shouldn't be glad that Leica is a german company.

[carstenw]

Well, they are German after all. They will figure it out and fix it.

[MarkM6]

Originally Posted by booggerg

Riley, the new german automobiles are horrendous for electronic bugs. Maybe we shouldn't be glad that Leica is a german company.

Get with the program! The last German car infamous for the "electronic bugs" is the Audi in the early 80's...

When it comes to Leica M8, the excuses are among the lamest. M8 does produce pretty good images but let's not ignore it's problems are a norm.

[stnami]

Yea just like my brother inlaw's VW V10 deisel 4x4 tourag spends a lot of time off road............................ in a repair shop waiting for computer parts.........things will only get worse as in car computer applications takes over maintanance and warranty issues in future cars. Big bucks for Europeon car manufacturers, others will follow...

[DaveSee]

Originally Posted by harmsr

John,

I did not see this vertical band on the 1.06 camera, but only on the 1.09 camera. As I understand it, Leica changed the hardware which reads the sensor information to eliminated the previous horizontal bands coming off of light sources.

My fear is that it has created this new issue. Relative to whether it needs another hardware change or can be fixed in firmware, is unknown to me.

It does need to be eliminated.

Best,

Ray

Hi Ray, et al.

Sorry to say, but this attached is one of the better worse shots and the "new" band is present here too... M8, first batch processor board, f/w 1.06. ISO640, 3300K, 1/16...
so, this "processor band" is in the first batch too.

Learning to see things in pictures everyday

JPEG from DNG, rotated 90

[rvaubel]

Originally Posted by DaveSee

Hi Ray, et al.

Sorry to say, but this attached is one of the better worse shots and the "new" band is present here too... M8, first batch processor board, f/w 1.06. ISO640, 3300K, 1/16...
so, this "processor band" is in the first batch too.

Learning to see things in pictures everyday

JPEG from DNG, rotated 90

Well in a funny way it is a relief to see it was there all along and they didn't create a new problem. To give Leica a break, nobody on this forum noticed it enough on the original batch to make a big deal of it like we are now. That's not to excuse Leica, but they are under a microscope now.

Rex

[marknorton]

Since Dave posted his 1.06 image, I've gone looking for the band through hundreds of images and I just do not see it. It's a real mystery what's causing it.

[harmsr]

A few more shots in the house tonight, do NOT show the band at ISO 160 with a dark background when overexposed or correctly exposed with a light source in the frame. This also includes pumping EV up in C1.

The same shots do show the vertical band very faintly at 320 when a light source is in the frame and the background is dark, even when correctly exposed. It does become more apparent when you bump up EV in C1.

These were with the B+W 486 filter.

Going back and looking at 1.06 shots, I could not see any of the banding. However, that should be clarified that almost all of my shooting was at 160 & 320 in daylight due to the previous banding and blob issues.

Ray

[rainer viertlböck]

Originally Posted by jrc

I also posted this message in the thread entitled "M8 1.09 & filter problems
This new banding issue looks exactly like the banding that some MF backs are struggling with. JC

no thats definitife wrong, you can stop thinkng in this direction. the mf "banding" you refer to is no banding, it is caused by the sensor zones and here appear in some sensors exactly in the middle the"stitch" between the sensore, which means one half of the image looks a little bit darker than the other, the effect appear visible exactly in the middle of the image where the sensor zones are stitching. this dals sensors have 6 zones, the center stitch usually becomes most visible.

[Peter Branch]

As I posted earlier I only got the vertical stripe at 2500 ISO but that was with the original, unmodified hardware and firmware. So it is not just with the upgraded cameras. Also others have now demonstrated that it occurs at lower ISO settings.

To me the worry is that that the vertical stripe, (Some posts seem to be talking about the horizontal streaking/banding), appears to be constant in position and width on all the samples I’ve seen wherever in the world the pictures were taken and is quite independent of the subject matter.

I've looked again at my early test images and interestingly the vertical stripe is very evident on those taken with the 24mm and 35mm lenses but is not visible at all on those taken with the 50mm, 90mm or 135mm. Probably does not help much to know this because the camera has now been upgraded so I will have to do a lot more tests.

BTW I'm not a "Test" junky the pictures which revealed this problem were taken to check that the, expensive, 6-bit coding, was actually doing something. I know only too well from decades in the photographic industry that testing cameras and lenses is a long, complicated and difficult task if reliable and reproducible conclusions are to be drawn. As a retired person without access to the necessary high tech equipment I would not dream of attempting to do anything other than spot obvious faults.