Occasionally people buy clothes with colours that match under electric light in the shop, only to find that they don't match outside, under daylight. This is because the colour-rendering properties of daylight and artificial light are slightly different. The difference is quite subtle, and perhaps not surprising either, as daylight and most artificial lights are visibly different in colour.
It occurred to me that, given what I know about colour theory, I should be able to produce two lights that appear identical but which render colours radically differently. Below I describe a prototype exhibit that achieves this aim. Along the way, it vividly demonstrates a fundamental truth about colour vision: that we are all colour-blind.
What does the user do?
The photograph on the right shows the apparatus. Not shown is a viewing tube that stands on top of the box and helps to exclude ambient light. The user looks down the tube into the box, through a hole in the wooden lid. They see the inside of the box divided into two halves by a central partition. Both sides of the box are lit by apparently identical yellow lights. The user can slide their hand (or any other thin object) into the box through the fabric light-trap at the front, so that they can see it as they look down the tube. One one side of the box, their hand looks a dull, lifeless grey, but on the other side of the box, it looks a vivid pink. A colour photograph will appear almost completely lacking in colour on one side of the box, with red objects very dark, but will appear with rich colours on the other side. The colour changes are dramatic, yet the lights on each side of the box appear identical!
This picture shows a red pen in the box. Notice how, although the light appears to be very nearly the same on the two sides of the partition, the pen appears a very different colour on the two sides.
I have shown this demonstration to many visitors at Glasgow Science Centre. Some gasped when they saw their hand change colour, others physically jumped, and children would try it and then run off to fetch their parents to show them.
(Similar exhibits do exist, but lack the drama of this one. For example Techniquest have an exhibit where identical displays of objects are lit by two different white lights. However the white lights do not appear identical, and the differences in the apparent colours of objects under the two lights are subtle.)
What does the user get out of it?
Firstly, they get a big surprise, and their ideas of how colour works are challenged. Secondly, they learn at first-hand that they are colour-blind. When we say that someone is colour-blind, we mean that there are colours or lights that appear identical to that person, but which we (the colour 'normal') can see are different. When a colour 'normal' person looks down the tube, they see two lights that appear identical. But they soon discover that the lights are not identical, because they render the colours of objects very differently. This shows directly that the colour 'normal' person is failing to distinguish two physically different lights. They are colour-blind, just to a lesser degree than people who are traditionally called colour-blind.
More about colour blindness later.
What's inside the box?
The box is lit by light-emitting diodes (LEDs). On one side of the box there is a single kind of LED that gives a yellow light. The other side of the box is lit by a combination of red and emerald-green LEDs. The intensity of the two kinds of LED is adjustable. With the correct settings, the combination of these two LEDs gives a light that is identical in colour to that from the yellow LEDs.
The photograph on the right shows the box with the lid open. The main part of the box is divided into two chambers by the central partition. At the far end you can see the two large yellow LEDs that light the right-hand chamber, and at the near end is the cluster of red and green LEDs that light the left-hand chamber. Mixing of the lights is achieved by shining them onto the matt white underside of the box lid - objects in the chamber are light by the reflected light. The electronics that run the LEDs are in the rear compartment, and the controls are on a panel on the back of the box.
Why does the exhibit work?
LEDs produce narrow bands of wavelengths, so there is very little, if any, overlap between the spectrum of the yellow LED and the spectrums of the red and green LEDs. The two sides of the box are therefore lit by lights that, although they are identical in their effect upon the human visual system, have very little physically in common.
The difference becomes apparent when the lights shine on a surface that reflects some wavelengths of lights better than others. For example, the yellow light produces very little red (long-wavelength) light, so red objects appear dark under this light. However, the red+green light does contain red light, and so red objects appear vivid and bright under this light.
More about colour blindness
Rather rarely, a man (not a woman) will find that objects look identical in each side of the box. If you remove the lid of the box and show him the red and green LEDs, he will say that they are identical in colour. Such a man is truly red-green colour-blind. Instead of the three different kinds of light-sensitive cells (cones) that most of us have in our eyes, he has only two.
Rather more frequently (about 1 in 20), you will find men who look down the tube and immediately say that the two sides of the box are not lit identically, and that the difference is very substantial. These people are colour anomalous. They have three kinds of cone in their eyes, but one of these cone types is sensitive to a slightly different region of the spectrum than its counterpart in colour 'normal' people. The colour anomalous person needs a different balance of red and green lights to match the yellow light. It's easy to adjust the red and green lights to achieve this match for the colour-anomalous person. When this is done, the colour 'normal' person sees one side of the box as being very clearly green - nothing like the other, yellow, side of the box. This experiment gives us an idea of how different the visual world of the colour-anomalous person is.
The experience of the colour-anomalous person also shows us that there is no 'true colour' view of the world. Their way of seeing colour is not better or worse than the colour-normal person's - it is just different. Over the animal kingdom there is a great variety of types of colour vision, none of which can claim to the one, true, correct way to see the world.
Many people who describe themselves as colour blind are in fact colour anomalous.
Technical issues
I had to overcome two technical hurdles to make this exhibit work well. The first was physiological, and the second was to do with controlling the LEDs.
1. I've got the colour-vision blues
The problem
In the first version of this exhibit I tried to create white lights on each side of the box. To understand how I did this, take a look at the diagram at the right. It is called the 1931 CIE Chromaticity Diagram, and it is the colour scientist's map of all possible colours. The actual colours in this illustration are very approximate. The diagram is adapted from one at Peter Kaiser's page at York University, Canada. Pure spectral colours plot around the curved edge of the diagram (the numbers are wavelengths in nanometres). The colours along the lower straight edge are purples made by mixing deep blues and deep reds. All other colours plot in the interior. LEDs produce narrow bands of wavelengths; they are approximately spectral and therefore their colours plot close to the curved edge of the diagram.
The diagram enables you to predict the results of mixing coloured lights. Suppose that you have two coloured lights and that you plot points on the diagram representing the colours of these lights. The straight line joining the two points shows you all of the colours that you could obtain by mixing these lights. For example, take the bottom-left (deep blue) and top-right (greenish-yellow) black blobs. By varying how bright each light is, you could mix these lights to produce any colour from deep blue, through paler blues, white, pale greenish-yellows to deep greenish-yellow.
Similarly, with the colours represented by the blobs at top-left (blue-green) and bottom right (deep red) you could mix any colour along the line joining those blobs. Again, this range includes white.
It was by mixing two pairs of colours like these that I produced white on both sides of the box. I then encountered three problems:
- I couldn't get the illumination to appear even - the colour seemed to vary from place to place
- If I adjusted the colours to look as identical as I could, they would no longer look identical if I viewed the box from a distance
- Not everybody (especially older people) agreed that my settings produced a good match between the two sides of the box
The explanation
I eventually traced these problems to a quirk of physiology. In the centre of the retina of our eyes is a region of yellow pigment - the macular pigment - that filters out some of the incoming blue light. Thus the colour vision of our central vision is different to the colour vision of our peripheral vision - we are less sensitive to blue light in the centre. This was the cause of the apparently uneven illumination that I saw - it was my own eyes that were non-uniform, not the illumination.
The macular pigment also explains the effect of viewing distance. Close up, nearly all of the inside of the box is imaged in peripheral vision; far away, it is all imaged in central vision. As the colour vision of these two regions is different, lights that match close up don't match when seen at a distance.
The density of the macular pigment also varies hugely between people - some of us are more sensitive to blue light than others. Thus two light mixtures that match for me would not necessarily match for anyone else. In addition, the lens of our eyes gets yellower as we get older, further reducing our sensitivity to blue light. Thus two light mixtures that match for a young person would not necessarily match for an old person.
The solution
The solution was suggested to me by people from the 2002 European Conference on Visual Perception, who held an evening event at Glasgow Science Centre. It was to avoid using blue light in any of my mixtures.
If you look at the CIE diagram, and remember the colour-mixing rule I described earlier, you'll see that to mix a white light using spectral colours (the ones around the curved edge of the diagram, which LEDs nearly are) you can't avoid using a light that is blue or greenish-blue. Therefore I had to abandon the idea of mixing a white light, and compromise by mixing a yellow light instead. I could do this using only three kinds of LEDs, because the colour of a yellow LED lies on the straight line joining the colours of deep red and emerald green LEDs, as shown in the diagram on the right. (Note that the central blob represents a yellow LED - the colours on the diagram are not completely accurate.) This compromise not only eliminated all of the problems, but also brought two advantages.
Firstly, the visual effect is more dramatic. In the original (white light) version, objects appeared coloured whichever side of the box they were on - the colours were just somewhat different. In the new, yellow-light version, one side of the box (the one with the single yellow LED) doesn't render colours at all, and gives a powerful contrast to the vivid colours seen on the other side of the box.
Secondly, adjusting the balance of the lights to achieve a good visual match is very much easier with only three lights than with four.
2. Controlling LED brightnesses
The other technical issue was about how to control the brightnesses of the LEDs. In the first version, I used the simple approach of putting a variable resistor (potentiometer) in series with the LEDs. This is not at all satisfactory. Adjusting the current through an LED to get the LED to go from zero brightness to full brightness needs a very specific range of resistance, which differs for different types of LED and which is most unlikely to match the range provided by any standard type of potentiometer. The result is that you end up using only a small fraction of the range of the pot, making sensitive control very difficult
In the second version I used pulse-width modulation (PWM) to control the LEDs. The power to the LEDs was switched back and forth from zero to full power about 200 times a second. By adjusting the relative lengths of the on-periods and the off-periods, the effective brightness of the LED can be varied - the eye smooths out the flicker. I programmed a microcontroller chip (a PIC chip) to provide the PWM signals to each group of LEDs. For each LED group, I used a potentiometer to provide a control voltage to tell the PIC chip how bright to make that group.
The PWM system was very successful and well worth the extra effort involved in setting it up. I was able to use the full range of my control potentiometers, with the added bonus that LED brightness was perfectly linearly related to the setting of the potentiometer.