Mike,
I'm sure that there are other, better ways to do this, but I've implemented a really crude version of what you describe that may be of some use.
What I did was take the output from a PID routine and converted it to a % of the next second that it should turn the relay on. I had 24 possible levels (so pretty coarse proportional control, and kluge-ily set by the number of channels in my system).
The routine runs through, checks a channel, computes the proportional output for the next second and writes that to a row of a 2-D square array... if it's 50%, then it writes 8 elements T and 8 F. It then takes the column of that array (which represents 1 point in time for all channels) and writes that to the relays. THen it cycles to the next channel. About every second, it would cycle through all channels. Thus, a channel might get turned on at step one and then turned off at step 2 (representing 1/24 duty cyle) or turned on at step 1 and off at step 12 (12/24 duty cycle). I made it run fairly slowly to reduce the cycles on the mechanical relays.
It was slow, made the mechanical relays click like mad, and held temp really well.
I've attached the code which might show you something or might just be confusing. Unfortunately, I was using Measurement Computing boards when I wrote this, so you will get errors when you try to open it. If you ignore them all, the missing vi's are pretty self explanatory. Another thing this code is doing that may be confusing is using a multiplexer to read 8 thermometers... that's the DIG OUT and READ TEMP vis... this version only controlled 8 chambers using the parallel port for output.
Hope this helps for something. Anyone who knows of better ways to do this, chime in.
cheers,
mike
