M8 Banding Problem

The linked site demands that you register or you can't see the example pic.
 
Nikon experienced a similar problem when introducing the D200. It has been resolved but I do not know how.
 
I have D200 from first batch, It still exists on my camera but I don`t care much, you berarly will notice it, anyhow the one shown on example here is very hardcore, I have never had that...
 
I briefly read the italian article and indeed there are a couple of images that show banding.
The reviewer (Roberto Piero Ottavi) stated that the banding occured only with his Voigtlander 28 Ultron and it doesn't show when using any Leica lens (either asph or pre-asph) maybe because of "compatibility issue" between microlenses and light angle from non-Leica lenses.

But you can easily see the vertical (actually horizontal giving the portrait orientation) banding in the Noctilux image, so maybe we have a real issue here.
 
all sensor will show banding when pushed to its limit. I hope the M8 banding is not showing under normal circumstances.
 
Well, this could be an area that improves is they go back to 16 bit files (20 Mb) without the 16>8 non linear mapping. Large areas with similar tones will be improved, at the cost of speed
and any jpg conversion makes much worse
 
The Nikon's problem was caused by the way it scans the CCD to get the data off it. One reason the D200 is so fast is that it has four separate channels (data paths) to scan the CCD; if I remember the explanation correctly, some types of images could cause these scans to get slightly out of sync and cause a banded appearance in the images.

I've had a few images from my R-D 1 that exhibit what I might call "banding" or "smearing" -- generally they happen when there's one very overexposed area in an otherwise dark photo, such as a bright window in the back of a dark room. (I've got a perfect example image at home, but can't get at it here.)

This causes a light streak, the same height as the overexposed object, extending lengthwise along the frame. It fades out eventually.

I've read that this is also caused by the way the imager is scanned to get the data out; the very strong signal from the photosites in the overexposed area "leaks" onto the adjacent sites. Apparently it's somewhat related to the phenomenon of "blooming" that you get with a video camera.

I still can't see the sample images, but does this sound like a description? If so, the problem can't be solved by using higher bit depths; it happens at the analog stage of the signal being formed on the photosites.
 
I seem to remember a similar " issue " thread on FM when the 1Ds came out if memory serves. I never hear anything about it anymore nowadays... Welcome to the digital world, it has been said before.
 
jlw said:
I still can't see the sample images, but does this sound like a description? If so, the problem can't be solved by using higher bit depths; it happens at the analog stage of the signal being formed on the photosites.
Click to expand...

Yes, it's a horizontal band same height as the light source stretching across the image.

See page 19 in the Italian article: http://rpo.eranet.tv/files/Leicapassion 3-2006.pdf (slow download - it's 51 pages).
 
On the D200 it was mostly contained in one of the channels (ISTR blue)
I've seen the issue JLW reports both on a D70 and R-D1. So I wonder if its a data-shunting thing (D200) or a blooming thing. Blooming can't be resolved whereas the data-shunting thing can be - Nikon repaired D200s under warranty.
This might also be due to the absence of an AA filter in the image path...
It really doesn't surprise me with fast lenses, dark scenes, and a few blown highlights - a combination that will show up any shortcomings a sensor has...
 
I was looking at those photos and my thought is, I'm not sure you could hold anything, film or digital, with that much overexposure. The shot was taken at night, ISO 1250, f2, at 1/25. The apparent banding came from a row of very bright streetlights. So how many stops overexposure is that? Ten, maybe? More? And these are not point-source specular highlights, they are large streetlights. I suspect this might be beyond the ability of any sensor to hold without some kind of light spillover.
 
That street light shouldn't be too difficult of a subject for any modern imaging sensor to handle. My GRD handled these lights without banding... The wonderful star effect is the result of a dirty front element.

133699194_0ac1f6dab8.jpg
 
Last edited:
Ywenz,

I don't really know what I'm talking about here, but your shot is somewhat different than the ones that show banding in the M8 (I've seen two shots so far with banding.) In both, the banding came when the high light source was immediately adjacent to a dark color -- and only that way. When the high light source (in the same photos) grades into something else -- white curtains, then a wall, and so on, I don't see the banding. Your high light sources play out across a light-to-dark gradiant.

I also see a bit of a blooming effect in some M8 shots, similar to that lighter area in your shot, around the top of the low tan building in front of the Hancock; but that looks "okay'" to me -- like something I'd see with my eye.

Sean Reid has looked at a number of shots, and found the banding problem only in a couple (of many) that included bright lights at ISO 2500. At this point, I'd say it's a problem that resembles the moire thing -- it'll effect a few shots done in extreme conditions.

JC
 
Here's my banding example, made with an R-D 1 at a performance by a Sudanese choir. Note the purple streak running through the dark area on the left side, aligned vertically with the very bright window to its right.

I'm guessing that the signal is pulled off the sensor from left to right (remember that the image would be reversed in the camera) and if a really bright area oversaturates the imager in one area, the excess signal will be "smeared" horizontally across the adjacent areas of the image until it dissipates. The worst-case scenario would be a really bright area right at the edge that gets scanned first, with a deep black area all the rest of the way across the frame.

If my guess is right, I'd think it would be possible to create conditions where this might happen with almost any digital camera, but there could be interesting variations:

-- Would a CMOS sensor behave differently than a CCD?
-- Would a camera that pulls the data off the imager more slowly do better, by allowing more time for the excess signal to dissipate? Or would a fast-scanning camera do better, by getting past the "hot" area before the signal has time to cascade to the adjacent photosites?
-- Does a small sensor perform better than a large one, or vice-versa?

I know we've got some digital-sensor experts lurking about -- anyone care to provide an authoritative explanation?
 

Attachments

  • 06-10-02_013.jpg
    06-10-02_013.jpg
    94.2 KB · Views: 0
Last edited:
The DMR never did that in my experience. Not that it did not have several issues in the beginning.
 
You must log in or register to reply here.
Back
Top