RicksAstro
Senior Member
On others advice, I used dcraw to convert the raw files I took last night of extremely low light imaging. What I posted last night http://forums.dpreview.com/forums/read.asp?forum=1031&message=19887473 , with ridiculously large uncropped images) was using the Canon Utilities to convert the RAW to 16 bit TIF.
What I learned was the Canon utilities do manipulate the raw image extensively... a disappointment. The results from dcraw were SIGNIFICANTLY noisier, but a bit sharper. I'd rather have control of any manipulation, but I guess that's why there are other RAW converters for those more serious.
I had dcraw convert to raw B&W bayer pattern and used my imaging SW (MaximDL) to convert to color using absolutely no color adjustment at all (no black offsets, no color scaling) and used their high quality debayer algorithm. What I got was as as raw of images as you can get with no noise reduction, no sharpening and no color balancing...just what the sensor put out.
All shots are 2 minute exposures through the EF70-200L f/4 lens at f32 and ISO800. No noise reduction on either camera.
The first was under brighter room conditions (desk lamp and 2 monitors on). These both came out with the histogram trailing off in the high 20k's, with the XTi's black histogram edge being a little lower and white histogram edge being a little higher (the XT was more compressed towards the middle). In MaximDL, I set the whitepoint to 32768 and did a gamma stretch of 0.6 to 8 bits. That's it, no other processing whatsoever, and both images treated identically.
Here is a wide wiew crop of both images, with the XTI's scaled down to match the XT's:
Here is a full scale crop of there images, left at their native resolution:
The second series was taken in much lower lighting conditions (only 1 monitor on, and with a dark screen background) Same exact camera setup/exposure as the series above. The histograms showed a similar pattern...the XTi used a wider range of the histogram (lower black, higher white) . On closer inspection, the blue channel of the XTi was lower than the XT and the red channel was higher. This accounted for much of the wider histogram.
I processed them exactly the same as above except I used a whitepoint of 6900 for the 0.6 gamma stretch.
Here is a wide wiew crop of both images, with the XTI's scaled down to match the XT's:
And here is the full resolution crop:
In that last image, you can see the histogram differences, with the blacks grayer in the XT's image and the colors brighter in the XTi. If processed differently to optimize for the histogram for both, these would look more similar. But I didn't want to process these images any differently at all...they went through exactly the same process.
Do these apply to normal imaging? I think so in general. I can say pretty well that, under the circumstances I set up, the sensor on the XTi is not less sensitive overall than the XT. The blue seems less sensitive, but the red more. The green was a wash. This was at f/32 and microlenses have a way of shifting light gathering ability at different light cone angles, so the results could be a bit different with a fast lens.
But for now, I'm convinced that the XTi is not a step down from the XT. It seems similarly noisy (maybe a bit better) and is similarly sensitive (worse blue, better red), so the S/N should be similar to the XT...a significant acheivement with the extra Mpix...
Rick
What I learned was the Canon utilities do manipulate the raw image extensively... a disappointment. The results from dcraw were SIGNIFICANTLY noisier, but a bit sharper. I'd rather have control of any manipulation, but I guess that's why there are other RAW converters for those more serious.
I had dcraw convert to raw B&W bayer pattern and used my imaging SW (MaximDL) to convert to color using absolutely no color adjustment at all (no black offsets, no color scaling) and used their high quality debayer algorithm. What I got was as as raw of images as you can get with no noise reduction, no sharpening and no color balancing...just what the sensor put out.
All shots are 2 minute exposures through the EF70-200L f/4 lens at f32 and ISO800. No noise reduction on either camera.
The first was under brighter room conditions (desk lamp and 2 monitors on). These both came out with the histogram trailing off in the high 20k's, with the XTi's black histogram edge being a little lower and white histogram edge being a little higher (the XT was more compressed towards the middle). In MaximDL, I set the whitepoint to 32768 and did a gamma stretch of 0.6 to 8 bits. That's it, no other processing whatsoever, and both images treated identically.
Here is a wide wiew crop of both images, with the XTI's scaled down to match the XT's:
Here is a full scale crop of there images, left at their native resolution:
The second series was taken in much lower lighting conditions (only 1 monitor on, and with a dark screen background) Same exact camera setup/exposure as the series above. The histograms showed a similar pattern...the XTi used a wider range of the histogram (lower black, higher white) . On closer inspection, the blue channel of the XTi was lower than the XT and the red channel was higher. This accounted for much of the wider histogram.
I processed them exactly the same as above except I used a whitepoint of 6900 for the 0.6 gamma stretch.
Here is a wide wiew crop of both images, with the XTI's scaled down to match the XT's:
And here is the full resolution crop:
In that last image, you can see the histogram differences, with the blacks grayer in the XT's image and the colors brighter in the XTi. If processed differently to optimize for the histogram for both, these would look more similar. But I didn't want to process these images any differently at all...they went through exactly the same process.
Do these apply to normal imaging? I think so in general. I can say pretty well that, under the circumstances I set up, the sensor on the XTi is not less sensitive overall than the XT. The blue seems less sensitive, but the red more. The green was a wash. This was at f/32 and microlenses have a way of shifting light gathering ability at different light cone angles, so the results could be a bit different with a fast lens.
But for now, I'm convinced that the XTi is not a step down from the XT. It seems similarly noisy (maybe a bit better) and is similarly sensitive (worse blue, better red), so the S/N should be similar to the XT...a significant acheivement with the extra Mpix...
Rick
