Articles

Chromaticity Calculator

Image Quality Issues

Understanding Color Measurement

Writings \ Articles >

Colorimeters

Page 2 of 4

Even though the eye mainly responds to Red, Green and Blue, there are actually four peaks to the eye's sensitivity to light in the visible spectrum, since the response for Red has a double-peak.  Here is a composite graph that displays the complete eye's response scaled correctly.  It is presented in false-color for ease of understanding.

The response pictured here is extremely important in colorimetry, and is called the "standard observer".  It represents the sensitivity of the eye to various wavelengths of light for an average human.  There are actually two different standard observer responses that are slightly different depending on the width of the field-of-view under consideration (2 degree or 10 degree field-of-view).  Unless specified otherwise, most of the time the 2 degree standard observer is being considered.

From this response graph, we can begin to have some idea for possible hardware instrumentation that may model this response.  If you only remember one thing from this article so far, let it be this:  A hardware instrument is accurate to the extent that it precisely models the 'standard observer' response.  It's that simple.

Match those curves closely, and you have a very accurate instrument.  The less closely the curves are matched, the less accurate the instrument will be.

Now, let's discuss possible solutions to the problem of modeling the standard observer using measuring instruments.

One logical solution would be a colorimeter with three photodetectors, each of them tuned to the corresponding three large peaks of the eye.  In general, a colorimeter with only three narrow pass filters would produce a rather low quality instrument.  Another logical solution would be a device with (at least) four filters, tuned to all four of the peaks within the visible range (including the smaller of the Red peaks).

That should be sufficient for a colorimeter to model color similar to the way the eye does, correct?  Indeed it can, and it will.  However, this technique is very tricky to accomplish successfully.  Remember the rule that the colorimeter needs to precisely model the graph at the top of this page in order to be accurate.  

There are two techniques available to make a tristimulus colorimeter (literally: returning data representing the three stimuli of the human eye).  The two techniques could be called:

• absolute accuracy
• controlled condition accuracy

Absolute accuracy would involve creating a colorimeter that precisely models all of the human responses above, so the instrument works for all displays.  In the past, building this type of device had required so much bandpass filtering that it transmitted very little light and was not useful in many practical circumstances.  However, there have been significant advances in improving tristimulus colorimeters in the past decade, and today colorimeters with absolute accuracy and very good transmission rates are available.

Controlled condition accuracy in a colorimeter is also possible with three filtered photodetectors (Red, Green and Blue) and a mathematical correction that allows the device to be highly accurate in specific circumstances.  If the response of each of the three or four photodetectors in the colorimeter is known, and the response of the light source is known, the eye can be very well approximated with such a device.

Let's take as an example, a three photodetector colorimeter containing a filter-set that passes colors that approximate the peaks of the human eye.  Sticking to the graphical explanation of the concept, let's say that the colorimeter has the following response.

Not exactly the "standard observer" response that is pictured at the top of this page, but it can be made to work.  How?  If we know what parts of the spectrum the light source will be displaying, these photodetectors can be mathematically 'tuned' to appear to be reading just like the eye for a specific display.  The discussion of the tuning is beyond the scope of this article (and difficult to explain without mathematical support), but it does present some interesting possibilities:

• A single-purpose photodetector colorimeter could be designed with a specific response for specific applications.  A popular and inexpensive way of measuring CRT monitors in the 1980s and 1990s used this technique.  As long as the CRT monitors contained the same phosphors as those used to calibrate the sensor, the colorimeter could do a very good job of measuring the monitor accurately.
• Although the colorimeter could be 'tuned' to do a good job with known primaries (like CRT monitors), it would be quite inaccurate with an output that does not match the output of the device used to set up the sensor, and using a colorimeter with the wrong type of display could cause large color errors to be produced in the measurements.  It is possible to perform a mathematical transformation of the data to allow the colorimeter to appear to be returning correct information for other light sources as well.  However, the mathematical transformation would be unique for each display device or light source, meaning that it would take an infinite number of such 'profiles' to be able to characterize any given display.

So, although three and four photodetector "controlled condition" colorimeters can be an inexpensive attempt at characterizing the human eye for specific single-use situations, they are not without some non-trivial compromises when used with other light sources.

We will explore why this is the case with the next example.

Next