Banding, some more tests

[pascal_meheut]

I just went out my home tonight and shot some pictures to have more informations about banding. Raw mode, C1, AWB in C1. I've used a tripod, a 28mm/2 ASPH at 2.8 on the M8 and the 28mm/2.8 at 2.8 on the DMR.

First it appears at any shutter speed. Here is an example at 640 ISO for 1/125s to 1/8s.

[pascal_meheut]

Of course, the banding is more important when the ISO are high, exposure being the same.
Here are the examples from 320 ISO to 2500 ISO.

[pascal_meheut]

Third, the DMR is immune. Here are examples at 400 ISO and 800 ISO.

[pascal_meheut]

So I think we need to report this to Leica if not already done. The problem is quite rare, appeared only in 1 over 100 (or 1 over 500 depending on how you count) of the pictures I shot with the M8 but it is real.

[jflachmann]

We are also dicussing this issue in the German forum. I have just posted a picture taken a few hours ago. M8 at ISO 2500 and 1/30 - f4, with the new Elmarit-M 2.8/ 28 mm ASPH.

http://www.leica-camera-user.com/dig...html#post84457

Jens

[pascal_meheut]

Yes, someone pointed me to such threads but my German is quite poor. Do you know if someone reported it to Leica, which seems the logical next step to me now?

[eronald]

Originally Posted by pascal_meheut

Yes, someone pointed me to such threads but my German is quite poor. Do you know if someone reported it to Leica, which seems the logical next step to me now?

Leica will never fix this for us, they have to get a new chip made, and would have to replace every single sensor out there.

They will fix it in the next generation of Kodak chips, when the present series has been sold out. In a few months, we can all sell our prototype M8 cameras and get fully working ones.

People at Canon once told me "Please do not buy any camera we make for the first six months after it comes out".

Edmund

[pascal_meheut]

If so, why is the DMR banding free? How did Canon fixed the same problem on the Eos1D ?

[jflachmann]

Originally Posted by pascal_meheut

Yes, someone pointed me to such threads but my German is quite poor. Do you know if someone reported it to Leica, which seems the logical next step to me now?

I didn't until now. I am still looking for some more owners with that banding problem, just to get an idea if it's the same with every M8 or not. Very interesting for me is to hear, that Ronald Schmidt (one of the mods here in the forum), who has taken several shot this summer with an early version of the M8, has not noticed banding in his files (some were published in the last issue of the "Leica World News").

Jens

[pascal_meheut]

I've sent a mail to Leica.

[eronald]

Originally Posted by pascal_meheut

If so, why is the DMR banding free? How did Canon fixed the same problem on the Eos1D ?

It's a sensor issue. In fact I'm surprised that Kodak messed up this design, by now it's unusual. Canon replaced my sensor for me in the 1Ds, as a CPS member, and that gave me a camera that was really usable at high ISO.

Leica have an absolute winner in their hands with the M8. They should address these minor annoyances quickly in after-sales; having problems is normal, fixing them gets you a reputation for good client service.

Edmund

[gogopix]

Pascal, very nice series. You point out exactly when we should be careful, rather than why we should be sending back for sensor replacement.

Sure it is a sensor issue. And the DMR had blooming and the M8 has a bit of banding.

It is like this; you try to put 5 liters in a 3 liter bucket-can you complain to the bucket manufacturer that the water spilled on the floor?

this is happening because you are throwing too many photons into the sensor, and it is overflowing, except not to sides,but stright along a line at readout.

If this happened below 255,255,255 maybe you would have a point, but I find this one more case of damning the camera for not overcoming the laws of physics (those electrons you produced by throwing too much light in the cells have to go somewhere - in other cases it would produce HEAT, really bad.

How about flare in the 15mm
how about fuzziness of most canon optics
how about 1.4 lenses not as sharp wide open as at 2.8

the list can go on


And to think, I used to be considered a perfectionist! :-)

Victor

[dritz]

Forgive me, but banding at 255,255,255 still is unacceptable. We work with frast prime lenses because we are working in low light, and that low light often has strong pointed light sources in it. Shoot at a music concert, a outdoor cafe at night, etc.

What you're describing is a behavior that eliminates a substantial number of my historical and anticipated shooting sessions -- especially at higher ISO settings like 1200 and 1600.

Is this irrational thinking?

Dean
Seattle, WA

[innerimager]

Here's significant magenta banding in the upper left shadows and over skin, shot at Iso 640, using a 28/2.8, 1/30 at Photo Plus today. Not a terribly extreme condition, really a problem that would need addressing if this is how all the M8 sensors operate....Peter

[blado]

This banding thing is kind of a pain! That is a nasty example. I was wondering if this could be caused by a particular kind of Mercury-vapor or Halide light source or something. All the examples I've seen of banding have been of bizarre arena, halogen or street lighting. Or is this a nonsensical idea? Also, maybe a filter could prevent it? I guess that's not the point, though...

[gogopix]

The light wavelegth may matter, what is more isthat these sources are much stronger than normal lighting (tungsten, fluorescent etc)

The 255,255,255 is the max that is registered. My suspician is that it indicated a simple fact, blown highlights
what people dont realize is that the source can be 10,000 time brighter!
In many of the situations, aperture priority with large zone auto exposure is going to WAY overexpose, not just a little. If you used spot mode i would bet EV would be verymuch lower. The actual scene could be 24 stops not just 13 or 14 (that is, just alittle overblown)

In addition here and it the bad example on the german site the sources are even lined up! that is, multiple overexposure in the same read out line.

I think Pascals test shows that this is a reasonable interpretation, since with a drop in amplification, it goes away

I think there is a specific oversaturation that any sensor can take (it is in the Kodak spec) but these images are way beyond that
there are plentof example i see with normal bright spots that have no banding.

In addition, the lack of anti alias filter exaerbatesthe problem since these bright spots are not spread at all, they go right for that micro lens!

:-)

Big time

[blado]

I was wondering about a filter on the front of the lens, like a UV or ultra-something-else.

[jrc]

Don't think it's the wavelength. I went out and shot my garage lights -- ordinary 100-watt incandescent bulbs -- to see if I could force banding, and I could. Not only did I get the typical horizontal bands, I also got a vertical band that neither began nor ended in a bright light. I also got an odd green artifact that looked like the inside of a lens design -- so I think the sensor may have been picking up some internal reflections as well.

I'm edging toward the conclusion that you tend to get the banding only in a few circumstances -- you need a light source that is way, way, way over-exposed in a shot that is otherwise quite a bit darker. I'm also beginning to think that the light sources have to be fairly large -- I haven't been able to force any bands with very small but very-bright lights. (My neighbor's lights, which are bright, but relatively pin-pricks compared to the size of the overall frame.) I'm also beginning to suspect that lens flare is involved in some way that I don't understand.

It'd be nice if Leica could fix it with firmware, or that we could get a software fix, just to know that it's out there, but frankly, I'm not going to have any problem shooting around it. One a companion thread, there are some quite nice posts from nighttime London, which were shot at f1.4 and 320. No banding.

JC

[eronald]

Victor,
I have the problem on my M8, Pascal on his, other people. It's a design fault.
Do the current CCD and no anti-alias filter backs have this issue ? No.
Does the DMR have it ? No.
Does the 1DsII have it ? No.
So I cannot expect this this on a 2006 Leica either.

Edmund

Originally Posted by gogopix

The light wavelegth may matter, what is more isthat these sources are much stronger than normal lighting (tungsten, fluorescent etc)

The 255,255,255 is the max that is registered. My suspician is that it indicated a simple fact, blown highlights
what people dont realize is that the source can be 10,000 time brighter!
In many of the situations, aperture priority with large zone auto exposure is going to WAY overexpose, not just a little. If you used spot mode i would bet EV would be verymuch lower. The actual scene could be 24 stops not just 13 or 14 (that is, just alittle overblown)

In addition here and it the bad example on the german site the sources are even lined up! that is, multiple overexposure in the same read out line.

I think Pascals test shows that this is a reasonable interpretation, since with a drop in amplification, it goes away

I think there is a specific oversaturation that any sensor can take (it is in the Kodak spec) but these images are way beyond that
there are plentof example i see with normal bright spots that have no banding.

In addition, the lack of anti alias filter exaerbatesthe problem since these bright spots are not spread at all, they go right for that micro lens!

:-)

Big time

[innerimager]

Originally Posted by gogopix

The light wavelegth may matter, what is more isthat these sources are much stronger than normal lighting (tungsten, fluorescent etc)

The 255,255,255 is the max that is registered. My suspician is that it indicated a simple fact, blown highlights
what people dont realize is that the source can be 10,000 time brighter!
In many of the situations, aperture priority with large zone auto exposure is going to WAY overexpose, not just a little. If you used spot mode i would bet EV would be verymuch lower. The actual scene could be 24 stops not just 13 or 14 (that is, just alittle overblown)

In addition here and it the bad example on the german site the sources are even lined up! that is, multiple overexposure in the same read out line.

I think Pascals test shows that this is a reasonable interpretation, since with a drop in amplification, it goes away

I think there is a specific oversaturation that any sensor can take (it is in the Kodak spec) but these images are way beyond that
there are plentof example i see with normal bright spots that have no banding.

In addition, the lack of anti alias filter exaerbatesthe problem since these bright spots are not spread at all, they go right for that micro lens!

:-)

Big time

but- certainly the area of banding is not itself blown out, in fact it is largely in shadow tones. I don't think overexposure is the issue here. I do wonder about wavelength, how the micro lenses stack up those wavelengths, and also wonder about the sensor being too sensitive in the magenta spectrum. Looking at that color space Michael Riechman showed on his review, the M8 has a large area of red sensitivity compared to other sensors. I wonder if it's related to this issue.....Peter