Timo -
Calibrate v. Linearize

Timo states on a web page that he considers "calibrated" and "linear" to be synonyms:

The words "linear" and "calibrated" are or should be synonyms when the best possible accuracy is the target.

Timo thinks that "calibration" makes a system operate linearly:

To calibrate a system or subsystem, you bring it into conformance, within a certain tolerance, with a predefined characteristic. When you "tune" a piano, you could be said to be calibrating it. But you do not calibrate so that the frequency of each key is linearly related to its neighbor! You calibrate so that the keys have the correct frequency - that is, so that each key has a frequency related to its neighbor by the correct ratio.

Timo fails to understand that you can calibrate a nonlinear system.

If Timo were to calibrate a piano, I suppose that Timo would prefer the notes of a piano to have frequencies related by an absolute frequency increment.

Timo's "Uncalibrated" means "conventional"

When Timo says "uncalibrated," he means "not linearized." To Timo, nonlinear is bad. A system calibrated nonlinearly - like a piano, or even a Barco Calibrator monitor - is to Timo, "uncalibrated."

When you read Timo's documents, for his word "un-calibrated," you must substitute "conventional."

Timo wishes to linearize CRTs. He wishes to make them operate in an unconventional manner. I claim that for images intended for viewing, this is a very bad idea.

Timo says, "Calibrate your system."

Timo wants you to "calibrate" your system, using his definition of "calibrate." Using my clarification above, he wants to linearize your system.

The first problem with this is that if you have 8 bits per component, this will deliver poor image quality. I recommend against linear 8-bit coding, for reasons summarized on the page about Weber and contrast ratios.

If you were to linearize your system, in addition to degrading the performance of your system, you will operate against the conventions of image scanning, exchange, and display, which are based upon perceptual principles. You will tune your piano to linearly-spaced frequencies. You will find that you cannot play any of the classical scores, because they are written for conventional tuning, tuning that sounds good. (You will also find that anything you play won't sound very good, because the keys have been assigned frequencies without the benefit of knowing about perception - without the benefit of knowing how things sound.)

I am not arguing for complete subjectivity here. We design a piano for optimum perceptual performance. But we build it using mechanics and physics, and we tune it using instruments. The point is that the abstract model of a piano is based upon perceptual principles, and we bury those principles among the engineering, the nuts and bolts, the physical pieces.

Main thread:

Timo thread:

See also: Gamma FAQ - Linear and nonlinear coding - Timo and linear coding

Charles Poynton
Copyright © 1998-03-12