Quoting accuracy of color measurement instruments
Quoting accuracy of color measurement instruments
- Subject: Quoting accuracy of color measurement instruments
- From: tom lianza <email@hidden>
- Date: Tue, 30 Oct 2007 20:42:29 -0400
Hi to all,
The subject of quoting accuracy often comes up and people ask why we
don't quote accuracy numbers even more often. The simple answer to that
: we really don't know how to quote accuracy of a spectral device in a
meaningful way.
Roger wrote the following
What I'm particularely after is the "Accuracy" in x,y for a white point
(presumably 6500K, on a choice of Reference Monitor) and, if possible, as
that would be the cherry on the icing, "Accuracy" in x,y for R,G,B.
My research into other instrument's performance typically turned out that,
on a white point, Accuracy is often quoted in the order of ±0.002 while it
is often in the order of ±0.004 or (worse) ±0.006 for R,G,B.
Here is a quote from the NIST laboratory about how they calibrate a
standard instrument:
from the link:
http://ts.nist.gov/MeasurementServices/Calibrations/photometric.cfm
NIST will calibrate submitted colorimeters or spectroradiometers for
chromaticity coordinates (x,y) of several colors of a display. The
instrument under test will be calibrated against a NIST reference
spectroradiometer, using a standard CRT display or and LCD, or any other
display submitted by the customer. The calibration will be performed for
the primary colors (red, green, and blue) plus white of the display, and
a correction matrix for the Four-Color Method will be provided, which
can be used to correct the errors for any other colors of the display
measured with the test instrument. The expanded uncertainty (/k/=2) of
the calibration is typically 0.001 to 0.002 in (x,y) depending on the
color and display type.
Note two points: they mention an "expanded uncertainty" and they note
that it is dependent upon display type. The method, quoted above is
used to calibrate i1Displays relative to a population of reference i1Pro
Devices. The expanded uncertainty is larger than NIST's standard
because we are one step further from the source and we include a
population average of the i1Pro's in the mix. A spectral measurement
instrument is very difficult to specify because of the bandwidth of the
device. If we calibrated an i1Pro and a Minolta CS1000 to a tungsten
source, they would agree nearly exactly when looking at that source, but
they would have very large differences when looking at a complex spectra
such as CRT or CCFL LCD. These differences could be as large as .01 xy.;
huge. When an i1Pro is calibrated, it is subjected to a number of
calibrated sources with structure and without structure. Part of this
process of calibration includes a "deconvolution" step which is used to
minimize the effects of bandwidth, but there is no way that we can
determine the effective performance on every source, so we can't quote
it. My own internal studies with my i1Pro and the CS1000 in our lab
show numbers on the order +/- .002 on white on an LCD with CCFL (not
wide gamut), and differences on the order .003-.004 on the primaries
with an uncertainty in measurement setup of .0015 (k=2). Things start
to depart very rapidly when looking at wide gamut ccfl or led
backlights, particularly in the green, because of the shift of the
primary wavelength from 550 towards 520. But PLEASE, don't quote me.
Regards,
Tom
_______________________________________________
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