source of graininess in IR photos

[Lisa Nikodym]

I have a Nikon D70 that was converted to full-time IR use.  I've noticed that the images appear much grainier/blotchier/noisier than they used to be in color from the same camera before conversion.  I have two theories as to why, but don't know which (if either) is correct:

(1)  Because IR effectively uses only the red channel (the brightness of the B & G channels are very much less), you only have 1/3 of the pixels as before.  What you're seeing is an image with not enough useful pixels.  

(2)  Because I have to set the exposure compensation in the camera to about -2 in order to not blow out the red channel, I'm effectively underexposing overall compared with before, so the image is noisier.

I *think* reason #1 makes more sense, but I don't think the graininess looks quite of the right sort for that (but I may be wrong).  I'm thinking that reason #2 doesn't quite make sense, but I'm not sure.  Or is there some other reason?

If the real reason is #1, then I can make a case for getting a converted IR camera with more pixels (a D200, for example).

Attached is an IR image taken with the D70, and a full-resolution crop of a sky area showing the graininess.

Lisa

(P.S.  I'll be out of town for a few days starting tomorrow, but will check in here again when I get back Wednesday.)

[kal]

I have a Nikon D70 that was converted to full-time IR use.  I've noticed that the images appear much grainier/blotchier/noisier than they used to be in color from the same camera before conversion.  I have two theories as to why, but don't know which (if either) is correct:

(1)  Because IR effectively uses only the red channel (the brightness of the B & G channels are very much less), you only have 1/3 of the pixels as before.  What you're seeing is an image with not enough useful pixels. 

(2)  Because I have to set the exposure compensation in the camera to about -2 in order to not blow out the red channel, I'm effectively underexposing overall compared with before, so the image is noisier.
[a href=\"index.php?act=findpost&pid=138195\"][{POST_SNAPBACK}][/a]

Aren't they the same thing? You are underexposing green and blue, so you have only 1/4 of real image data and 3/4 of pure noise.

[Lisa Nikodym]

Not quite the same thing.  The first problem is just not enough "real" pixels, sort of like taking a 2 megapixel camera image and blowing it up to poster size.  The second problem is purely a signel-to-noise issue, where you have enough pixels, but they aren't very good pixels.  Replacing it with a camera with more pixels would fix the first problem, but not the second.

Don't we have anyone else here with IR experience and/or a desire to speculate about camera physics?

Lisa

(P.S.  OK, off to the airport now.  Back Wednesday.)

[kal]

Not quite the same thing.  The first problem is just not enough "real" pixels, sort of like taking a 2 megapixel camera image and blowing it up to poster size.  The second problem is purely a signel-to-noise issue, where you have enough pixels, but they aren't very good pixels.  Replacing it with a camera with more pixels would fix the first problem, but not the second.

Don't we have anyone else here with IR experience and/or a desire to speculate about camera physics?

Lisa

(P.S.  OK, off to the airport now.  Back Wednesday.)
[a href=\"index.php?act=findpost&pid=138428\"][{POST_SNAPBACK}][/a]

But they are almost the same if you look at signal power per unit area and noise power per unit area, and in the final print what really counts is how much "real" detail you have at a given size, and how much noise. Does this make any sense?

[disclaimer: I know absolutely nothing about digital IR. Everything is based only on my (hopefully quite strong) digital signal processing background]