Banding, some more tests

[fuchs]

Advertisement (gone after registration)

As you can see, the problem is beginning to show up at ISO640, and is very evident at 1250 and 2500, precisely the ISO's typically used indoors at night without a flash. If you look hard at the tiff file, you can actually see faint hints of banding even at ISO320!

 

Clay,

looking carefully, one can see "banding" (sensor streaking) also on your first picture, at ISO 160.

Is it noticeable also on the full DNG file, or does jpeg compression make it somewhat worse?

 

Best,

 

Ed

[clayh]

You're right, if you look very carefully, you can see it on the .tif file at all ISO's. It is not objectionable (and easily repairable in photoshop) until 640. But you are right, it is there at all ISO's. I do think the jpeg compression makes it a little more evident, but it is there in all its glory on the raw files and the tiff files. The Canon 5d does not even flinch at this sort of commonly encountered pictorical situation. Let's hope the fix is easy.

 

Clay,

looking carefully, one can see "banding" (sensor streaking) also on your first picture, at ISO 160.

Is it noticeable also on the full DNG file, or does jpeg compression make it somewhat worse?

 

Best,

 

Ed

[aj37]

As an interesting aside, Mark Norton has done some test today that would indicate that the RD1 has an problem also, but in the opposite direction. Well I've shot over 10,000 frames with my RD1 and never noticed it. Of course I'm going to confirm marks tests but I wouldn't be surprised if he's right.

 

The R-D 1 definitely can exhibit some color "bleeding" under the same sorts of extreme conditions that provoke M8 "bleeding." The R-D 1's bleeding seems to be less compact and well-defined than the types of examples we've seen posted from the M8.

Here's one of my R-D 1 pictures, originally posted in another forum, in which the effect is visible as the horizontal purple streak on the left side:

 

 

This is my first R-D 1 image in thousands of shots on which I noticed this phenomenon, although I suspect it might lurk in others if I were to look for it. I wonder if it might be present to some extent in photos made with other types of digital cameras as well, although not in a form as pronounced as we are seeing in the worst-case M8 examples...?

[clayh]

Looks like a manifestation of the same phenomenon. The thing that is giving me heartburn is that I don't even have to try very hard to get the M8 to fail in this way. That EV range I quoted is in no way extreme or something that is rarely encountered when out shooting photos. And the fact that other cameras seem to have no problem with it.....

 

For example. Here is a shot from my three year old 3.2 MP Canon S230. No issues that I can discern, other than a not-so-great lens. I think I paid $200 for it.

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

[rvaubel]

The R-D 1 definitely can exhibit some color "bleeding" under the same sorts of extreme conditions that provoke M8 "bleeding." The R-D 1's bleeding seems to be less compact and well-defined than the types of examples we've seen posted from the M8.

Here's one of my R-D 1 pictures, originally posted in another forum, in which the effect is visible as the horizontal purple streak on the left side:

 

 

This is my first R-D 1 image in thousands of shots on which I noticed this phenomenon, although I suspect it might lurk in others if I were to look for it. I wonder if it might be present to some extent in photos made with other types of digital cameras as well, although not in a form as pronounced as we are seeing in the worst-case M8 examples...?

 

Your Right!! Holy Moly, I've been living with the totally unacceptable banding disease for 10,000 frames and more than a year and didn't even know it! They whole year has been wasted! I better start editing my files to get rid of those nasty bands!

 

Seriously, it is a little reassuring to see that I have been living with this problem without noticable bad effects. Of course now that I am aware.....

 

Hey, what about you other camera owners? Maybe your files aren't as clean as you imagine. Maybe there's a little band blood in your background that you don't know about

 

Rex

feeling a little better

[jrc]

Okay, I surrender.

 

I can't seem to make my M8 band.

 

I shot twinkle lights against dark, medium and lighter backgrounds, in focus, out of focus, over and under exposed.

 

 

I don't think the light is bright enough and the adjacent area is not dark enough. To get the banding, you need an unfrosted light, and a black surround -- essentially, a light that is adjacent to a big dark area, like a night sky. I forced mine to band last night, and tonight, went out to see how much I could hold without banding, and it was quite a bit. Most distinct: super-bright halogen lights, where the sensor is looking right at the halogen element (not sideways) with a totally black surround. In the pictures taken at the photo show, if the phtoorapher hadn't included the ceiling in the shots, there wouldn't have been banding.

 

I'm also beginning to wonder if wave length might not have some effect (after arguing last night that it probably didn't.) Tonight I shot the outside of a Border's Books with some of those orange sodium vapor lights in the parking lot. The sodium lights are quite bright, but I got no banding against a dark sky.

 

JC

[fuchs]

Advertisement (gone after registration)

but it is there in all its glory on the raw files and the tiff files. The Canon 5d does not even flinch at this sort of commonly encountered pictorical situation. Let's hope the fix is easy.

 

I'm trying to make my 20d streak, even at the H-ISO 3200 setting, with no luck. Color noise banding IS an issue at this high iso setting, though. Let's hope Leica finds the fix soon.

 

Best,

Ed

[rvaubel]

Looks like a manifestation of the same phenomenon. The thing that is giving me heartburn is that I don't even have to try very hard to get the M8 to fail in this way. That EV range I quoted is in no way extreme or something that is rarely encountered when out shooting photos. And the fact that other cameras seem to have no problem with it.....

 

For example. Here is a shot from my three year old 3.2 MP Canon S230. No issues that I can discern, other than a not-so-great lens. I think I paid $200 for it.

 

They highlights really have to be blown. I have been studying published RAW files that have banding and I have dixcovered fro banding to occurr the highlite must be blown to the extent that reducing the "exposure" in photoshops RAW plug in by 4 F stops still will not "rescue" the highlight. In most causes a "blown" highlight can be 90% rescued and the unrescued part is the only thing to cause banding. Thats why you see a band that apparently is not as wide as the blown highlight!

Its weird but I can tell by looking at the band which part of the highlight can not be rescued before a do the manupulation in PS!

If you can rescue all the highlights in a photo, there will be no banding. Practically speaking this doesn't solve anything because bare lightbulbs should be (and look) blown. However, I now know whats going on.

 

Try it, you'll like it

 

Rex

[sean_reid]

Is this magenta color cast maybe associated with the operating system. For example are the people not having the magenta cast problem using a different version of the Mac OS or maybe Windows machines?

 

I don't know. I'm on XP and haven't seen a magenta cast. I'm not sure what Pascal's on and he's asleep now.

 

Cheers,

 

Sean

[clayh]

Well, there is still a problem. I understand what is going on. The problem is that this camera cannot handle it. Other cameras can. I measured this light with a calibrated spot meter. EV12. Background EV4. I run across these sort of situations all the time. I cannot reduce the exposure as you suggest without getting into other issues such as increasing the noise in already noisy shadow areas. And making sure that there are no lights in the framing of my photos is not really a solution either. This has got to be addressed.

 

 

They highlights really have to be blown. I have been studying published RAW files that have banding and I have dixcovered fro banding to occurr the highlite must be blown to the extent that reducing the "exposure" in photoshops RAW plug in by 4 F stops still will not "rescue" the highlight. In most causes a "blown" highlight can be 90% rescued and the unrescued part is the only thing to cause banding. Thats why you see a band that apparently is not as wide as the blown highlight!

Its weird but I can tell by looking at the band which part of the highlight can not be rescued before a do the manupulation in PS!

If you can rescue all the highlights in a photo, there will be no banding. Practically speaking this doesn't solve anything because bare lightbulbs should be (and look) blown. However, I now know whats going on.

 

Try it, you'll like it

 

Rex

[clayh]

FWIW, I notice the magenta cast mostly when I am shooting under incandescent lighting. Outdoors, it all looks pretty good. Also, I am using a calibrated Apple Cinema Display. So it ain't hardware. I do think the capture one profiles need a lot of work at the lower color temperatures. I am getting some lighting that I can't balance even after shooting through an ExpoDisk and using that for the white point setting.

 

I don't know. I'm on XP and haven't seen a magenta cast. I'm not sure what Pascal's on and he's asleep now.

 

Cheers,

 

Sean

[Jamie Roberts]

FWIW, I notice the magenta cast mostly when I am shooting under incandescent lighting. Outdoors, it all looks pretty good. Also, I am using a calibrated Apple Cinema Display. So it ain't hardware. I do think the capture one profiles need a lot of work at the lower color temperatures. I am getting some lighting that I can't balance even after shooting through an ExpoDisk and using that for the white point setting.

 

If the Canons are anything to go by, it takes C1 a few revs to get the profile right.

[sean_reid]

Your Right!! Holy Moly, I've been living with the totally unacceptable banding disease for 10,000 frames and more than a year and didn't even know it! They whole year has been wasted! I better start editing my files to get rid of those nasty bands!...

 

Rex

feeling a little better

 

That's interesting about the R-D1 and light streaks. I've been working professionally with two R-D1 bodies for two years (including weddings) and haven't ever noticed this nor had a client notice it. Very interesting....I wonder if it's because OOF bright lights in the frame tend to show flare, reflect off other surfaces, etc...in other words, lots of stray light tends to be bouncing around in those bright light areas....just thinking aloud...

 

Cheers,

 

Sean

[jrgeoffrion]

Is this magenta color cast maybe associated with the operating system. For example are the people not having the magenta cast problem using a different version of the Mac OS or maybe Windows machines?

 

I've looked at two of Pascal's DNG and the color cast is actually in the raw file itself. He posted the MacBeth chart previously somewhere (can't remember the exact thread). When the grays are white balanced, the rest shift to magenta under the halogen lights Pascal used. This did not happen with his DMR.

[robsteve]

I don't know. I'm on XP and haven't seen a magenta cast. I'm not sure what Pascal's on and he's asleep now.

 

Cheers,

 

Sean

 

I hadn't notice any casts in Sean's pictures either, even the indoor, incandescent lit still lifes.

 

Pascal is on a Mac, and I think a lot of the others complaining about the cast are on Macs too. If Pascal wants, he could post a DNG and some of us can try it and see what we get. I don't have Capture One though, just Lightroom or Adobe Capture Raw.

 

I suspect it is either an OS problem or maybe an old profile being used by Capture One. If there is a way to get the profile file version that might help in tracking down the problem.

[robsteve]

I've looked at two of Pascal's DNG and the color cast is actually in the raw file itself. He posted the MacBeth chart previously somewhere (can't remember the exact thread). When the grays are white balanced, the rest shift to magenta under the halogen lights Pascal used. This did not happen with his DMR.

 

 

Did you use a Mac version or Windows version of Capture One?

[mwilliamsphotography]

JRC, so the overhead light I shot directly and posted here wouldn't cause the banding to happen?

 

I didn't post the twinkle lights photo you quoted.

 

Here is that shot. They were against a window at night.

 

Nothing.

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

[jrgeoffrion]

Did you use a Mac version or Windows version of Capture One?

 

Win.

[jrgeoffrion]

For those trying to "make it happen", my guess is that if you shoot street lights in front of your home, it will happen starting at 640 ISO up to 2500. Make sure the lights are in the corners and much brighter than the scene et voila.

[Advertisement]

Hallo,

Look here: Interesting Leica gear

[aj37]

That's interesting about the R-D1 and light streaks. I've been working professionally with two R-D1 bodies for two years (including weddings) and haven't ever noticed this nor had a client notice it. Very interesting....I wonder if it's because OOF bright lights in the frame tend to show flare, reflect off other surfaces, etc...in other words, lots of stray light tends to be bouncing around in those bright light areas....just thinking aloud...

 

If it were flare, stray light, etc., presumably it would appear in various directions depending on conditions. There would be no reason for it to bounce in one direction more than another. The distinctive thing about the type of bleeding we're seeing in the example pictures is that it is always along the long axis of the frame and always exactly matches the height of the overexposed area. This is also true of the M8 bleeding of which we've seen examples posted.

 

Early in this thread, a couple of people posted to explain that in a CCD sensor, the charge is read off each photosite by being passed along down the row to the adjacent photosites, until it reaches a takeoff point where the charge is amplified and digitized. It's then supposed to be erased so that the next site's charge can be read.

 

They theorized that what's happening is that because the very overexposed area generates such a strong charge, the charge isn't getting erased fully and instead continues to "contaminate" subsequent reads from other photosites along the same row. This theory would fully explain the properties being observed in this bleeding: it's always along the long axis (because that's the direction the charge is passed) and it's always the same height as the hot spot (because the charges are passed horizontally.) It may also explain why people aren't seeing the same type of result from cameras that use other sensor technologies than CCD.

 

The picture in this thread that shows the soccer field with two bright lights is especially illuminating (no pun intended!) In this picture you can see separate bleeds from the two lights, not quite meeting in the middle. The poster noted that the M8 reads its sensor in two halves (presumably so it can read faster) and again, this would explain why the bleeds in the two halves of the image are separate.

 

As to why the M8's bleeds seem fairly well-defined, while those from the R-D 1 (and possibly other CCD cameras) are relatively vague and faint, there could be any number of reasons. I'm sure that the design of digital-camera firmware involves various compromises: For example (I am guessing here) the designer could read data off the sensor at a slower rate, which would allow more time for excess charge to deplete but which also would force the photographer to wait longer between shots. Different programmers might give different weight to the importance of these two competing factors, leading to different results.

 

Leica (I am really guessing here!) might consciously have chosen to opt for a faster-operating camera, accepting the risk that very overexposed areas might bleed somewhat and assuming that the effect would be rare enough to be acceptable. What we are seeing now may show that the effect isn't all that rare, at least for certain kinds of photography, and Leica may find that a certain amount of re-thinking is in order.