"While becoming a U.S. citizen, Kurt Gödel confided in his friend Albert Einstein that he had found an inconsistency in the U.S. Constitution that would allow the U.S. to become a dictatorship, causing Einstein to worry that Gödel's unpredictability would lead to his application being denied."
We are in sync! I also fell in love with this project after seeing it on Hackaday. At first I was just impressed, but the more I dug in (pcb, vhdl) the more I couldnt stop obsessing over it :) Its super well documented, well structured and easy to follow. True hello world of building a 386/486 chipset. My HaD comment from 3 weeks ago:
HaD blog entry doesnt do justice to this AMAZING project. Author implemented:
Intel 386/486 CPU bus handling
ISA bus handling
reused vintage 486 CPU
reused vintage 8259 PIT (timer)
reused vintage 8254 PIC (interrupts)
maniek86 build a legit vintage PC motherboard the way companies did back in mid eighties designing own Chipsets, all on his own in a span of few months. The only missing component is old school DRAM memory controller, skipping it is no brainer as driving DRAMs is almost an art form (as much digital as analog) and learning how to create one could take another year with most time spend chasing quirks and compatibility woes.
Want to hear something wild – this was maniek86s first 4 layer board ever :o Talk about jumping into deep water.
From reading maniek86 blog it all started when he got scammed buying Chinese no name ISA/PCI Post Code analyzer card that didnt really support ISA side https://maniek86.xyz/pl/blog.php?p=31 :
"It turned out that ISA part of the card was a scam – it could only measure voltages and show CLK, RDY, and reset signals. I was disappointed. I had to repair the motherboard without the help of POST codes. Eventually, I managed to fix it, but the card didn’t meet my expectations. That’s when I came up with the idea of building my own card instead of buying another one."
Indeed, by rights that shouldn't work. But it does and he threw in an ISA bus just for the heck of it and that works too. And all of this at a very respectable clock speed. Mad props.
Anything with latches above 10 MHz is pretty hard to get to run stable when using wirewrap. The Cray-1 was at its time absolute state-of-the-art. Replicating something like a 486 motherboard, which already had a whole raft of timing tricks to make sure that it all stayed synchronized is really not that easy.
I've spent more than one evening baffled by stuff that should have been trivial at clock speeds a lot lower than that. A 33 MHz clock has a cycle time of 1/33000000 = 33 nano seconds. But the rise time of that clock is much, much shorter, on the order of a few ns. At that sort of slew rate anything becomes an antenna. The backplane of the Cray-1 was set up as transmission lines with two spiral wound wires for each signal, cut exactly to size to make sure the signal arrived at the right moment, and without the bulk of the signal leaking away.
On this circuit board things are - let's put it friendly - a bit less organized than that. So by rights this really shouldn't have worked, the fact that it does absolutely amazes and inspires me.
https://en.wikipedia.org/wiki/Kurt_G%C3%B6del#Princeton,_Ein...