Written by MIke Wood
Tuesday, 01 February 2011 01:00
White is made of other colors—and so is your set, and costumes, and…
White is made of other colors—and so is your set, and costumes, and…
I wrote an article in Stage Directions back in September of last year suggesting some of the questions you should be asking about LEDs and white light, and why colors may look different when using them. In that article I made the following statement: “We always tend to test lights exclusively on a white wall. That’s a bad habit we need to get out of.” The Stage Directions editor, Jacob Coakley, spotted this and asked me to elaborate further on what I meant. Why shouldn’t we test lights on a white wall? That’s the way we’ve always done it, why should we stop now?
Figure 1 – HID lamp spectrum
The reasons for my comment are closely related to the subject of that prior article, but perhaps bear digging into a little deeper. My concerns stem from the same core problem: Just because a light looks correctly “white” on a white wall doesn’t mean that it will look correct on a colored surface. This problem is not new. As we discussed last time, our eyes have evolved to work best with a continuous light spectrum, such as you would get from the sun or an incandescent source such as a flame—however, man-made light sources tend to exhibit discontinuous spectra to some extent or other. The September article gave examples of the discontinuous spectra of LED sources, but many other lamps exhibit the same behavior. For example, Figure 1 shows the typical spectrum of an HID lamp that might be used in a moving light.
As you can see, the light within the visible part of the spectrum from 400 nm to 700 nm contains a number of large peaks, in particular two in the blue and two more in the green and green/yellow. (The spikes below 400 nm are in the UV area and not visible). Fluorescent tubes have similar spectra, perhaps with more peaks. In every case the lamp manufacturer adjusts the mix of peaks and troughs in the light’s output so that it looks white to the human eye on a white surface. In reality though, when we use these light sources in a theatre, we rarely light a white surface other than the cyc. Costumes and scenery are very rarely pure white. If we shine that same HID lamp on a cyan-colored surface or fabric that primarily reflects light around 500 nm then, as there is very little light energy of that wavelength, those colors will appear much darker than they should. Conversely, a green pigment reflecting light at 550 nm—which just happens to align with the large spike—will appear much brighter than it should. Overall the color balance may be very different than you expect.
Figure 2 – Four whites on fabric
Visual
Learning
A few pictures will help make the point
much more easily. First take a look at Figure 2. This shows the same
piece of striped fabric (actually a swimming costume) when
illuminated with different types of white light. (In order to try and
get as clean a comparison as possible the camera was left locked to
incandescent color balance and only exposure times were adjusted
between photographs. The brightness and contrast were slightly
adjusted in Photoshop to try and make all the light levels comparable
but no colors were touched.)
The fabric was lit by four different sources; incandescent, cold white fluorescent, a mix of amber and cyan LEDs balanced to a white, and finally a mix of red, green and blue LEDs also balanced to white. The final three all show a blue tint because the camera is white balanced to the incandescent. But allowing for this, I hope you’ll agree, in all cases the background whites are relatively similar. Which one is correctly rendered? You might initially be tempted to say the incandescent but, because of its lack of blue wavelengths, it does a poor job of rendering the different pink and magenta stripes near the bottom of the photo, which the fluorescent picks out well. On the other hand, the fluorescent does a poor job of the yellow and overemphasizes the blue. Interestingly, the amber + cyan mix is visually very pleasing—until you realize that everything that was yellow, red or pink now shows as different shades of amber, and the green has gone completely! The RGB does a good job on the whole but we lose the warmth in the reds and everything is a little hyper-real and cartoony.
Figure 3 – Four whites on test chart
Figure 3 shows a standard Macbeth color test chart under similar illumination conditions. Again, the mix was adjusted for each light source so that the bottom left white squares were as visually similar as I could get them and then the exposure was adjusted to try and get comparable images. This time I deliberately tweaked them a bit more aggressively in Photoshop to make all the whites look identical. This emphasizes the differences and helps us see the problem more clearly. (Note – these images will look different in this magazine than they did in real life or on my computer monitor. They have been through at least four different color processes between reality and your eye – the camera RGB color sensor, Photoshop processing, magazine layout and finally the CMYK printing process.)
Once
again we have incandescent, fluorescent and RGB LEDs, but this time I
used an amber + blue mix for the fourth version. The differences are
very clear again. Most striking of course is the amber + blue where
the white square and grey scale look pretty good but just about
everything else looks wrong. Only amber itself (second row, far left)
and pale blue (third row, far right) escape somewhat unscathed, while
anything with any green or red content is completely changed. RGB
shows some unsubtlety with over-emphasis of hue and has a
particularly hard time with the peach (top row, second from left),
which is rendered as a pale pink. That peach is a key color for us in
theatre as it’s close to light skin tone and, as such, is a
touchstone for color rendering. Your audience may not know what color
the costumes are, but they have years of experience with skin tones
and know exactly what they are supposed to look like!
So
Why Not?
Now, let’s get back to the
original question. Why not test lights on a white wall? If what you
are interested in is brightness, then a white wall is fine. If you
are looking at gobo focus or sharpness, then a white wall is fine. If
the luminaire is going to be used for lighting a cyc cloth, then a
white wall is fine. However, if you need a luminaire to light people
and costumes in specific colors and tones then a white wall is
inadequate. It’s always been this way to some extent, but we
could ignore the problem with incandescent sources—and we knew
HID based moving lights had a problem with color rendering anyway!
However, the rapid influx of LED light sources has brought this
concern to the front line and we should always, always look at
luminaires on a colored surface—if possible in the tones we are
planning to use. Get a fabric swatch from costume, have a painted
flat available—anything with some subtlety of tones that allow
you to make a value judgment about the color rendering. Your
designers will thank you.
Mike Wood provides design,
technical and intellectual property consulting services to the
entertainment technology industry. Send him colorful e-mails at 
mwood@stage-directions.com
