On what was probably my third or fourth 805x app, back in '85 or so. Since the display didn't have to be updated unless a few particular operations were in use, it seemed like a reasonable approach. Few of the other tasks were particularly time-critical, and the whole task had previously been handled by a 6 MHz 8748 that bit-banged the serial I/O, which it also only used two or three times per day. Expanding its functionality seemed simple enough when going to an 805x. Oddly enough, I'd previously used Mode 0 to communicate with other 805x's in multi-MCU systems, and never previously used the async modes at all.

I don't know why you "would not even dream of driving such a display with a '51 that had to do other things as well, thus which frequency the "UART connected" unit runs at is irrelevant."

The 805x had 99.999% of its time available to do useful work. During the very few milliseconds per day that it had to interact with either the console or the CRT display, its main burden was getting in sync with the video circuit, which simply required a subroutine call. Because the oscillator was, in fact the timebase for the CRT adapter circuit as well, once in sync, it stayed there in sync. Getting there was just a question of waiting for a blanking pulse.

Apparently you just interconnect functional modules rather than designing the low-level hardware. I hope I'm wrong, of course, as that's a quick way to do things, but not very cost-effective in production quantity.

RE