The colors that a sensor "sees" (D70, 20D, e.g.)

[xpatUSA]

I've just read an interesting paper which shows the colors that a couple of camera sensor can detect. We are accustomed to CEI XYZ diagrams and, separately, sensor spectral response curves. It is also possible to find software that can display the colours in your image on a XYZ diagram, but this is the first time I've seen all three data types shown together in one document:



It is seen that sensors capture wavelengths way outside any known wide gamut. Fortunately for us, the camera manufacturers know all about this and thoughtful filter out all the extra stuff both optically and in software - some would say to excess ;-)

I've put the source article up on my site (warning - geeky stuff):

http://kronometric.org/phot/gamut/Ca...ysisGamuts.pdf
.

[davidedric]

Well, I've read the abstract - I'll save the rest for supper time Thanks for posting.

[revi]

One remark, before we get another Canon/Nikon discussion:
the two sensors also differ in technology: one is a CMOS sensor, the other a CCD.
That difference might be a lot more important than the brand on the box...

The method used for the camera measurements is described in more detail here:
http://art-si.org/PDFs/Metric/EPMurphyThesis05.pdf, page 41-44.
According to that text, the data is recorded as RAW data, so, if a sensor sees
infrared, that ends up in the RAW data. The manufacturers can not have done
any extra filtering beyond what appears in the measurements (in the optics/
hardware of the camera, what happens in their software is another question).

So if the CMOS/Canon sensor sees more IR, then that's what ends up in the RAW
file. Otoh, sensitivity at 700 nm is only about 0.05, and falling, so total area
under the curve is probably negligable, unless you have a spot in your scene that's
very intense on just that wavelength.

[xpatUSA]

One remark, before we get another Canon/Nikon discussion:
the two sensors also differ in technology: one is a CMOS sensor, the other a CCD.
That difference might be a lot more important than the brand on the box...

The method used for the camera measurements is described in more detail here:
http://art-si.org/PDFs/Metric/EPMurphyThesis05.pdf, page 41-44.
According to that text, the data is recorded as RAW data, so, if a sensor sees
infrared, that ends up in the RAW data. The manufacturers can not have done
any extra filtering beyond what appears in the measurements (in the optics/
hardware of the camera, what happens in their software is another question).

So if the CMOS/Canon sensor sees more IR, then that's what ends up in the RAW
file. Otoh, sensitivity at 700 nm is only about 0.05, and falling, so total area
under the curve is probably negligable, unless you have a spot in your scene that's
very intense on just that wavelength.

Thanks for highlighting that, Remco. Although it was clear to me that the example sensors were purposely chosen as CMOS and CCD, I really should have pointed out that the paper was making a comparison. Thanks for the link to the method used!