On Sat, 18 Oct 1997 15:17:59 -0700, you wrote:

>I would like to design my projects/products so the ubiquitous ON indicator
>LED is serving the secondary purpose of sending debug info serially, to a
>phototransistor or some such which I have placed over it.  This would be
>useful for development, but also for periodic sample testing of production
>devices, don't you think?
>
>I've found that spacing out the bytes, and sending at reasonably high data
>rates results in a visibly steady LED glow (comm feature transparent).  My
>problem is that I can't find phototransistors fast enough to switch fully
>on and off with so little illumination (a few mA through a standard
>brightness LED).
>
>Yet it seems to me that this SHOULD work.  How about you smart folks who
>have done this kind of thing a million times giving me the benefit of your
>wisdom?
>
>Bruce Cannon
Always use a red LED - phototransistors have poor response to anything
else. Use a low-value pullup on the p/trans (10K or less). Consider
using a protocol other than RS232, as unpredictable coupling can alter
pulse width, and rise/fall times will usually not be the same. 
Consider using a high-frequency carrier (20-40KHz) and send bursts of
it with different lengths for '1' and '0' bits - this is what IR
remote control chips do. This makes it easier to reject ambient light
and noise - you can use an IR pre-amp chip to demodulate it.
For short-distances (<10mm), you shouldn't need to go to these lengths
- a 10K pullup with the right photodiode gives a good 5V signal which
should manage a few KBits/sec. 
A 2-way link can also be done in this way for lid-on calibration - the
LED sends a timing reference pulse, and the test jig wiggles the
supply voltage (or some other accessible line)slightly within defined
timeslots, which is read by the on-Pic ADC
Most importantly, though  USE A RED LED!!!