[argyllcms] Re: Display Calibration Hardware Capabilities

  • From: Graeme Gill <graeme@xxxxxxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Sat, 04 Feb 2012 14:46:39 +1100

Noses, That F. wrote:
> My understanding is that Adaptive mode only decreases the effect of
> random measurement errors by averaging more samples and increases the
> accuracy of the of dark measurements only.

It uses gain and integration time settings to optimise
the sensor value. This both decreases quantisation errors
and reduces noise in the readings.

> According to the data publicized in the review (the link in the first
> E-mail of this conversation), the mean and peak absolute error of the
> CM and i1Pro spectros are significantly (about four times) higher than
> what they got with the i1d3 colorimeters on various display types.

The practical significance of this may not be very great. For instance
there is a great deal of variation in absolute calibration for emissive
instruments (this seems to be a hard thing to pin down in the calibration
chain), but in practice this is typically of no consequence for display
profiling, because the profile is normalised to the white point.

Similarly, if I compare the i1d3 and CM that I have, then the
brightness normalised average error between them (measured
on 6 primaries + 6 grey step wedge) is peak 3.2 and average
1.7 CIEDE2K (3.2/1.7). If I normalise the white points I get
peak 3.5 and average 1.4 de CIEDE2K (3.5/1.4). I'm not at all
dismayed by this discrepancy given that this is two random
instances of consumer grade instruments that operate on quite
different principals.
(I get very similar numbers comparing a Rev A i1pro
 against the i1d3 and CM).

By contrast the numbers I get for (say) my Spyder 3 vs. i1d3 are
not quite as good ( brightness norm: 6.7/3.2, white norm: 6.5/2.8).

My i1d2 vs i1d3: brightness norm: 12.4/6.5, white norm: 7.0/2.8.

> This sounds like I probably get about the same or even higher absolute
> error if I create my own *.ccss files with my CM spectro.

Possibly. A spectrometer is a trade-off - there's more flexibility
(different observers) and better accuracy on an arbitrary spectra (unknown
displays, illuminants), but less absolute accuracy and stability than a 
carefully
built and calibrated colorimeter, of which there have been no consumer examples
that I'm aware of prior to the Discus & i1d3 (although the DTP92 & DTP94
seemed pretty good on certain displays).

Graeme Gill.

Other related posts:

  1. Home
  2. argyllcms
  3. Archive
  4. 02-2012