Xref: utzoo sci.electronics:3985 comp.graphics:3294
Path: utzoo!utgpu!attcan!uunet!ubvax!ardent!sleat
From: sleat@ardent.UUCP (Michael Sleator)
Newsgroups: sci.electronics,comp.graphics
Subject: Re: Perceptual Color (was: Re: Looking for Blue LEDs)
Keywords: black/white -> color
Message-ID: <619@ardent.UUCP>
Date: 5 Oct 88 04:30:03 GMT
References: <1138@nmtsun.nmt.edu> <862@ritcv.UUCP> <255@rna.UUCP> <4422@lynx.UUCP> <871@ritcv.UUCP> <870@dlhpedg.co.uk> <6101@watcgl.waterloo.edu> <790@ncar.ucar.edu> <2903@utah-gr.UUCP> <801@ncar.ucar.edu>
Reply-To: sleat@ardent.UUCP (Michael Sleator)
Organization: Ardent Computer
Lines: 71

In article <801@ncar.ucar.edu> cook@stout.UCAR.EDU (Forrest Cook) writes:
>Consider the possible advantage of this approach:
>ALL DIGITAL (I.E. Pulse Width/Frequency) color modulation.
>No messy DACs or analog amps needed.

I've lost the trail leading up to this, but the discussion was about
apparent color due to flashing white light.  If that's what the above
is suggesting, then I think there's a problem.  I believe that the
phenomenon of perceived color from flashing white light does not give
anything like the range of color that you can get with mixed primaries.
Also, I think it's very dependent on the particular observer.  In sum,
I don't think it's really a viable general purpose technique.

There is another interpretation, though, which leads me to the following
question:

	Suppose you have three LED's, red, green, and blue; the light
	from all three being combined by some diffuser.  If you control
	the current through each one (in an analog fashion), you can control
	the brightness of each, and hence the resulting perceived color.

	Now, for a single LED, you can also control the apparent brightness
	by varying the duty cycle (the pulse width modulation referred to
	above).  However, it's not obvious to me that varying the apparent
	brightness of each LED in this manner would result in the same
	control of color as would the analog method.

	Let's say that the outputs of the three LEDs are normalized such
	that when all three are on, you see white.  If you want to make
	a color that's comprised of, say, 33% of full red, 66% full green,
	and 100% full blue, you would in effect be producing three flashes
	of equal length: white, followed by cyan, followed by blue.  (Note
	that this sequence assumes the drives are synchronized and all
	turn on at the same time.)  So the question is, what will this
	look like?  What happens if the drives are phased differently,
	or asynchronous?

Here's what I *think* will happen:  If you do this at a high enough
frequency (a couple of MHz oughtta be safe :-) ), it will probably work
pretty well.  But as you decrease the frequency, things will start to get
weird somewhere, possibly well above the nominal flicker fusion frequency
for a constant-spectrum source.

Either control technique will give you reasonably precise control over the
energy output of the LED (integrated over some reasonable time period).
However, almost nothing in nature is linear.  The human eye is decidedly not.
I have the impression that a pulsed source (e.g., multiplexed LED displays)
appears brighter, at the same average power, than a DC source.  This
suggests that at the very least a different linearity correction function
might have to be applied to the raw color values for the PWM vs. the analog
technique.

Much of this is just speculation (bad pun) on my part.  Does anyone out
there have any other thoughts?


>While we're on the subject, would someone out there in LED manufacturing land
>please hurry up and make a simple color pixel consisting of RGB LEDs in one
>diffuse package?  I could find quite a few uses for such a device! :-)
>An imbedded controller would be nice, but I can wait for that.

I could use a few hundred thousand of such devices (or even a few thousand,
actually).  I can't think what I'd do with just one. :-) :-)  The interesting
question is just what would you have the imbedded controller do?  (I can
see it now:  PixelPlanes with a trio of LED's per processor!!!  Quick!
Someone call Henry Fuchs!  (you could make some pretty amazing billboards))


Michael Sleator
Ardent Computer
...!{decwrl | hplabs | ubvax | uunet}!ardent!sleat