Re: differences between colorimeter & spectrophotometer
Re: differences between colorimeter & spectrophotometer
- Subject: Re: differences between colorimeter & spectrophotometer
- From: Don Hutcheson <email@hidden>
- Date: Fri, 05 Oct 2001 01:41:14 -0400
The myth that colorimeters out-perform spectrophotometers on CRTs was
perpetrated several years ago by X-Rite in an attempt to defend their DTP-92
colorimeter against the superior (but more expensive) GretagMacbeth
SpectroLino. It is of course bunkum. The claim that three or four
mass-produced "broad-band" RGB filters which vary from batch to batch and
instrument to instrument can somehow simulate CIEXYZ more effectively than a
properly-calibrated spectrophotometer is a classic example of
pseudo-science. Don't believe it.
Rather than relying on inaccurate, mass-produced colored filters that can at
best only approximate CIEXYZ, a true spectrophotometer 'constructs' very
precise XYZ values from even more precise data, multiplying 16, 32 or 64
overlapping "narrow-band" spectral samples by separate D-50 coefficients for
each sample. This not only allows a spectro to deliver much truer CIEXYZ
than any colorimeter, but also permits accurate XYZ values for any white
point - whereas 3-channel colorimeters only work correctly at one white
point (usually D50.)
But we really shouldn't be measuring with XYZ filters in the first place, as
that's NOT how the eye sees. The ideal "synthetic eyeball" would use Roh
(Long), Gamma (Medium), Beta (Short) filters which represent our actual RGB
sensitivity curves, then transform them into XYZ using a simple LMS-CIE
matrix. This is quite impossible in a colorimeter as it involves negative
values (you can't have filters with negative absorbtions) but it's quite
easy with a spectro and the right set of coefficients.
The claim that a colorimeter's broad-band filters somehow "smooth out" a CRT
phosphor's spikes and troughs better than a spectro is another bit of
poppycock. A spectro produces an XYZ curve set with precisely the same
"smoothing" power as a colorimeter, only more consistently device to device.
In fact the presence of sharp spikes in a monitor spectrum requires a more
accurate synthesis of XYZ than most mass-produced colorimeters can achieve.
I teach real-world color management six times a year at GATF and part of the
course involves profiling 14 monitors in various stages of decay. We use
four SpectroLinos and the rest are profiled with the DTP-92 Monitor
Optimizer. Guess what? When you stand at the back of the room and look at
the same gray test target displayed on all 14 monitors, four look virtually
identical (guess which ones) and the rest make up a rainbow of different
pastel colors. I've done the same test with one colorimeter in turn on a
dozen monitors and the results still vary, indicating that the colorimeter
performs differently from monitor to monitor. The Spyder is considerably
better than the DTP-92, however, but then it has many more sensors. If I use
the SpectrLino, SpectrCam or Eye-One (all true spectrophotometers) I get
more consistent, accurate results than any colorimeter.
There are enough real problems in color management without deliberate
misinformation.
Buying advice? The Eye-One is my clear hardware favorite, but I prefer
ProfileCity's ICC Display over the Eye-One software. This will set you back
a total of about $700. If you can't afford the Eye-One, get the Spyder and
ICC Display for less than $300 total. I've seen some pretty impressive
results with this combo, even on the Apple Cinema Display, which is much the
best monitor available for color-critical work.
If anything better comes out I'll let you know.
*************************************
Don Hutcheson
Hutcheson Consulting
(Color Management Solutions)
Phone: (908) 689 7403
Mobile: (908) 500 0341
email@hidden
*************************************