[argyllcms] Re: Calibrate a HCFR hardware

  • From: Gerhard Fuernkranz <nospam456@xxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Mon, 14 Jan 2008 00:42:21 +0100

Frédéric Mantegazza wrote:
What do you mean by "spectral responses are far-off colorimetric"? My 
english is limited :o/

With "colorimetric" sensors I mean mean tri-chromatic sensors, whose
spectral sensitivities fulfill the Luther-Ives condition, which means that the spectral sensitivities are a linear transformation of the three CIE color matching functions. Only sensors with this property won't show observer metamerism with respect to the human vision, and only sensors with this property are able to measure light sources with arbitrary spectra without need for a dedicated calibration to the spectral characteristics of the light source being measured (at least in theory, assuming perfect sensors).

So ideally, from this point of view, a colorimeter should have "colorimetric" sensors. It is however very difficult to realize filters with the necessary spectral transmission characteristics. Only very expensive filter colorimeters approximate the
CIE color matching functions well enough to give a good accuracy when measuring light sources with arbitrary spectra. Low-cost monitor colorimeters need to be cheap, so they clearly need to make trade-offs and can't afford such a good approximation. And the shapes of the spectral sensitivities of the TAOS TCS230 sensor can not even be considered an approximation of the CIE color matching functions (or a linear transformation thereof), but their shapes are pretty different. That's what I mean with "its spectral responses are far-off colorimetric". This was to be expected of course, since the TCS230 was not designed as a sensor for colorimetric purposes, but it is just an RGB sensor.

If the sensor is not colorimetric, not everything is lost, though. If the set of spectra we intend to capture happen to have a rank of <= 3 (i.e. if they originate from a tri-chromatic additive output device such that they can be represented with sufficient accuracy as a linear combination of three primary spectra), then it is usually still possible to establish a linear mapping between the output from the sensor and the corresponding CIE XYZ numbers. This means, although an instrument with non-colorimetric sensors can no longer be considered an accurate general purpose colorimeter, it can still be calibrated for the
spectral characteristics of a particular display device, if this display device is strictly additive. The more the instrument's sensors deviate from the CIE color matching functions (or a linear transformation thereof), the larger will be the measurement error, when we measure spectra deviating from the spectral characteristics for which the instrument was calibrated, or when the display device is not strictly additive.

As supplementary reading material I suggest for instance this survey, mainly the section starting at the top of the right column of page 917.

There is no IR filter in front of the sensors. I'll suggest that to the HCFR team. Do you know where such filters can be found? I  could make a test...

From the TAOS
FAQ:
"How can I eliminate infrared (IR) and view only the visible light spectrum?
Place a Hoya CM500 or Schott BG18 (or BG39) filter on top of the device to remove the IR. You can purchase a Hoya filter from Newport Glass (www.newportglass.com)."
Consulting Google I found for instance this source for Schott BG 18 filters, but 10 GBP does not really sound cheap for a small piece of glass...

If you can arrange to make the measurements in an IR-free ambient, and if the display does not output any IR radiation, then an IR blocking filter won't change anything, of course. It's just a general issue that basically an IR blocking filter should be used if only visible light should be captured and if the sensors are IR sensitive (for instance, digicams have usually IR filters too, for the same reason). After attaching an IR filter, you would need to recalibrate the instrument, of course.

Regards,
Gerhard

Other related posts: