On Tuesday, March 4, 2014 6:00:36 PM UTC-6, Andrew Bingham wrote:
you're making a chip with several state machines programmed into it, where the S-100 address lines connected to the 4 EEPROM lines select which state machine is in operation and the counter goes through the steps to output the corresponding sequence to I2C to send the data.
Basically yes. This all comes from experiments I was doing in wanna-be homemade FPGAs some decades ago. At one point in history there were no FPGA boards under $5K and software licenses were all $75K etc etc. Not like the modern era of FPGA development at all.
But I really wanted to synthesize and download logic, and had access to an eprom burner and eproms and this newfangled language (at that time) called "perl" so I could write perl code to implement logic, then the perl code ran thru all possible address combinations (no big deal when all you have is 64K eproms...) then you stack several eproms in parallel to get more outputs, feed some outputs back via latches to get inputs and use latches as flipflops, etc. It was fun but of course entirely useless and shortly after, the FGPA mfgrs started sending out $50 dev boards that used freely downloadable software and that was the end.
I "synthesized" and burned eproms for a lot of discrete logic and adders and magcomps and ALUs and the like. I wanted to make a 1-bit serial ALU and never got around to finishing that. I also fooled around with interfacing like previous discussion.
It was/is possible to do some pretty weird stuff with eproms. I made a one digit voltmeter using A/D and 7 seg display. And what amounts to audio wavetable synthesis using discrete parts by burning free running counters with sine and triangle and whatever waves and stranger patterns. Rather than building a psuedo-random shift register to make white noise, I burned white noise pattern into an eprom simply because I could. I built a digital R/C servo driver which never really worked well, where you loaded the waveform you want from 8 bits parallel from a then new 68HC11 processor (ran into serious resolution vs repetition rate problems here which never really worked). Then I got the idea that the eprom could run any number of servos simultaneously in parallel for a walker and the processor would just output the position of gait for a really simple API including steering, which never worked anyway so it didn't matter. I took 8-bit recordings of people speaking and converted from wav file to binary and with 4 chips and an osc and a battery I made something like those speaking birthday cards although that never worked right either. Most screwing around seems to end with things only working about 90% of the way, in general.
Yes many weird things can be done by someone with access to an eprom burner who is bored. Some of which might still be useful today.