Re: Colorsync-users Digest, Vol 5, Issue 340
Re: Colorsync-users Digest, Vol 5, Issue 340
- Subject: Re: Colorsync-users Digest, Vol 5, Issue 340
- From: email@hidden
- Date: Tue, 30 Sep 2008 01:09:35 EDT
In a message dated 9/29/08 12:04:08 PM,
email@hidden writes:
> Spectrometers aren't "tuned" for any particular
> device, they are calibrated to measure the spectrum as accurately as
> possible,
> and the XYZ values flow from the spectral values.
>
As I said, at a given (low) price, the above is not generally true nowadays.
If you have a nearly ideal (very expensive) spectro (constant bandpass,
constant wavelength interval, ideal bandshape, perfect SNR) then it is indeed as you
say, except that wavelength and radiometric calibration tuning is usually
required. This is the case in the very high-end units. However, manufacturers
learned even as far back as the 1970s when filter spectros started to come out
that achieving ideal behavior was a difficult and expensive proposition, and to
a certain extent, unnecesary. What they learned has been mapped to modern
units of many designs: grating, filter, leds, etc. Its basically this:
mathematically optimize the instrument transform (nXm) in the spectro so that it produces
the optimal spectral estimates that minimize spectral errors for training
data, or frequently more usefully, optimal spectral estimates that minimize
colorimetric errors for training data. If you have enough spectral bands, then you
can do a very good job at this for all kinds of "scenes", monitors and
reflection data. And of course, changing modes clues device what transform to use. If
you have any doubt, the patent and technical literature in the instrument
field is replete with examples on how to do these kinds of mathematical mappings.
Its the camera profiling example on steroids...
<<Yes spectrometers have smaller bands,
so for a given amount of light each band has a lower S/N, but there
as precisely as many more bands as they are narrow, canceling out the
overall effect.>>
Unfortunately, not always does the sum of the parts equal the whole because
of non-ideal design particularly: SNR at a given sample, error propogation
(errors square by the gains squared in the transforms when trying to "re-assemble"
the colorimetry that the colorimeter measures more directly with smaller
sigmas and gains in the transform) and many other reasons that probably are not of
interest to this list!
Karl put it very simply and basically has it right in Andrew's reference.
Eric Walowit
Koln
**************
Looking for simple solutions to your real-life financial
challenges? Check out WalletPop for the latest news and information, tips and
calculators.
(http://www.walletpop.com/?NCID=emlcntuswall00000001)
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Colorsync-users mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden