Sensor Problem (I hope)

Arkady · 12-03-2006, 10:30

Take a look at these and see what you think.
Images were captured as RAF/NEF as always and processed in CS2.

First image is opened using the camera default settings (as exposes, basically).
Second image is opened with the Auto function on and has increased exposure to highlight the shadow areas.

Image 001

Image 002

It's a bit smaller, but look at the noise in the shadow areas and in particular the horizontal banding. Anyone else come across this?
I've noticed that in dim light the excessive 'noise' is very noticable even at 100ISO.

Both cameras are going in to NIKON on monday for a good clean and I'll betaking these along in case I have a duff sensor. It seems more noticable on one body than the other.

CT · 12-03-2006, 10:52

Weird! Are you sure Rob that it wasn't a different metering pattern for each shot? As you say the first shot is exposed for the highlights and the second for the shadows.

I thoght at first the banding on the left was possibly just different coloured stonework, but I see you have the same issue in the stonework on the right hand side and at the same level, so it does look like banding. I don't see the banding right across the image - just those darker parts, which is also where you're getting the noise, so it does smack of being a sensor problem?

Arkady · 12-03-2006, 10:57

Both from the same RAW file - one prcessed 'as is' and the other to highlight this problem (just moved the 'exposure slider right across) - I took about three and can email the RAW files to anyone who can stand to wait three days for them to download.
I exposed for the outside, obviously. Other shots taken the same day on the other body are nowhere near this bad, but I didn't realise there was a problem at the time, otherwise I'd have swapped bodies and made a proper comparison.
Looking at other images from both, the ones taken on this body do seem more 'grainy' in retrospect.

heheheh... 19Mb RAW files... I love it...

CT · 12-03-2006, 11:01

I just increased the gamma on your first shot. Obviously, there's a limit to how far you can go, but it looks like the banding may actually be there in that first shot too.

Joe T · 12-03-2006, 11:07

Hmm, I see what you mean.

I assume these are your D2X bodies. I havent heard of any problems with banding or similar on these, so maybe it is a dodgy sensor as you say.

Hope you get it sorted.

Arkady · 12-03-2006, 11:19

Quote:

Originally Posted by CT

I just increased the gamma on your first shot. Obviously, there's a limit to how far you can go, but it looks like the banding may actually be there in that first shot too.

It's in the RAW image from which both the above images derive - the first shot was merely as a comparison to show how it looked at the time of capture and how the image was meant to appear as a finished item - I'm more concerned with the fact that it's there at all....

Gandhi · 12-03-2006, 23:57

Those are big fat bands! The only banding issues I've seen from a nikon were on the d200, but they were vertical and very thin. Much reference to them over on dpreview. I would think it's a sensor problem. Is it present on all your shots? just under certain lighting conditions? Have you/colleagues had similar problems on other bodies? At the end of the day, it aint gonna cost you to send it back as faulty, so I'd despatch it asap. Stuuf does slip under the QC radar occasionaly.

SammyC · 13-03-2006, 09:06

Is there an older shot you can compare it too for that body to see if it is a recent effect?

CCDs are prone to banding effects but not like that! Something strange is afoot.

Matt · 13-03-2006, 10:00

Rob, it could be down to the RAW converter.

I've just opened one of my shots in the PS CS RAW converter
Moved the exposure slider all the way to the right (this is equivilant to 4 stops compensation).
The shot is full of noise and banding.
Even at two stop EC the PS CS RAW converter shows a considerable amount of noise.

Open the same RAW file in PRSP, adjusted the EC to +3 stops and there are no signs of the noise or banding.

Arkady · 13-03-2006, 13:44

I've sent the RAW file to Matt for a look-see.
Both bodies are with NIKON now for a deep-clean (and a bit of a shouting match in the Service department over turn-around times, but I'll not go into that here) and if they have time they'll look at the noisy/bandy sensor issue.

shiato storm · 13-03-2006, 15:12

i see noise/slight vertical banding in the second shot on the left only. in the shadows. horizontal band...? there looks to be a bit of flare, or what looks like it just above a think vertical band. could this be an internal lens reflection that only appears when you postprocess/push the exposure up?
i believe the problem comes from trying to push the exposure too high in PP, yes the noise increases when you do that to raw files, but doing so will bring out artifacts that under normal conditions are not visible at all... I agree its not pretty but had you exposed for the shadows it would not have been there, i suspect. artificial exposure adjustment through software maybe the cause to this, it would be interesting to see the result with alternative methods...
its a bit like those complaining about the d200 banding, although an accepted problem some were exposing a few stops off the meter reading and then pushing it up in post-processing...surprise surprise the dynamic range of the sensor is reached and artifacts appear, so people make a fuss, when under 'normal' exposures banding is nowhere to be seen at all. I'm not questioning your ability as a photographer, thats not my intention and i'm hardly in a position to criticise, i'm just saying you have to know the limits of the equipment you're using and if you're asking it to do something a little beyond itself in a particular situation to maybe back of a touch and find a work around to bring the best results it can...

SammyC · 13-03-2006, 15:25

I see four very thick horizontal bands but it is so faint that I could be mistaken.

First is a red colour, then yellow, then red again and then yellow again.

shiato storm · 13-03-2006, 15:34

like a rainbow!
sounds like a lens imprefection/funny diffraction with spectral colours like that...

Arkady · 13-03-2006, 18:42

Those of you who use the RAW processor in CS2 will know that the 'AUTO' adjust feature is the default and that you have to 'Ctrl+U' to revert to the settings as shot by you - the lighter of the two images is from the 'auto' setting - the darker (as I intended it to be) is from the 'Ctrl+U' setting with no other adjustment.
+2 EV is well within the tolerances quoted by Nikon - this amount of noise is unacceptable - the banding is very apparent at larger sizes - trust me on this.
Had I exposed for the shadows, the outside area would be blown out - even the D2X doesn't have that amount of latitude.

If it were a lens problem I would expect it to be circular in nature, not manifested as horizontal banding - think about the physics guys - lenses are circular therefore flare is manifested as circular (or as near circular as the aperture leaves will allow) highlights.

I'm glad for all your input but think about the photographic basics before you post things that are obviously not right...

Arkady · 13-03-2006, 18:55

Hopefully this is larger - the site's been automatically resizing, so bear with me.

Mods; if it's oversize, please leave it for a day of so as people can't see in the smaller image at the top of the page.

Horizontal banding - four distinct bands - from the top magenta, yellow, magenta, yellow.

Slight bit of flare upper left caused by dirt on the lens more likely than a duff bit of glass (NIKKOR 17-55 f/2.8 DX @ £1500-ish - not a sawn-off milk bottle by any means)

Massive noise @ +2EV

CT · 13-03-2006, 19:22

Apart from the sensor or problem with RAW conversion, the only thing I can think of to which horizontal banding might be related is the vertically travelling FP shutter, but I'd expect more obvious and pronounced banding than just those colour casts?

shiato storm · 13-03-2006, 19:26

Quote:

Originally Posted by Arkady

If it were a lens problem I would expect it to be circular in nature, not manifested as horizontal banding - think about the physics guys - lenses are circular therefore flare is manifested as circular (or as near circular as the aperture leaves will allow) highlights.

true but light travels in a straight line...