Got the parallel port interface working, and I could make the computer control the voltmeter.
My dad then came over and we did a lot of research on different TTL and CMOS families. Our two basic design ideas are:
1) One big shift chain. Power on/off after shifting all 2048 pixels. But that's a shitload of wiring... even with high-current 16 bit shifters (serial in/out, parallel 20mA out), that's a lot of work. And the 16 bits ones are way more pricey than the 8 bit ones, so that's even more work.
2) Shift a row in, light it, shift the next row, light it. But since we're only lighting rows for 1/40 of the time we have to use 40x the voltage. With 2.4V lights we're talking 120v everywhere. (still low power overall) But this will probably drive the price up, but only along the edges. Haven't looked for those parts.
Anyway, we ended tonight with a C program (using libparapin) turning a single 2.4V 20mA white xmas light on and off in different duty cycles (full, 1/2, 1/4) and increasing the voltage as a proof-of-concept. Indeed, brightness is the same as long as you increase the voltage as you decrease the work duty.
Current problem is the library or parallel port just isn't fast enough. There's noticable flashing.... need to get the frequency up coming from the computer. The library also works in kernel space, so that's one avenue. Need to do some research first, and read the library source... see if I can cut some bloat.
Oh, and the cashier girl at Radio Shack was totally flirting with me... very amusing.
I'm off to get drinks with people.