Page 1 of 1

Not easy, but working

Posted: 18 Nov 2019, 10:04
by Colin
I have 50 years experience building electronics. I design electronics for a living and back in the 70s designed processor boards for mini-computers.
However even using a professional equipment, this kit was a challenge. It took about three hours to build and another hour to debug.

Once finished it was obviously doing something, but the LEDs were not blinking. (BTW I would argue with which end of the LED is marked +)
In the end I looked around the board for bad logic levels and found one. D19 was faulty. I replaced it and it now works.

Re: Not easy, but working

Posted: 18 Nov 2019, 16:26
by marcelk
Do you mean U19? Interestingly in one of my own experimental builds I had shorted a signal pin there: pin 7, it's sitting between two grounded pins. Was it the same by any chance?

I find it impressive that you can find that in 1 hour. (But the diode mode check may have found it, not sure.)

For your consideration and if still possible, I wouldn't mind to have that part back so we can send it to our supplier. We'll reimburse the shipping costs of course. Please contact us by mail if you like.

Anyway, in any such case of needing of replacement, we can send parts as well of course.

And oh: congratulations and thanks for sharing the result :-)

Re: Not easy, but working

Posted: 18 Nov 2019, 20:35
by Colin
It was D19. I was just scoping around the board looking for bad logic levels. The top of that diode showed obvious signs of two outputs fighting each other. (Half rail levels.)

I did destroy the diode removing it. Cut in half, drop out the bits and then suck out the holes.

There were two spare diodes in the kit.

Looking at the design I did wonder if running at that speed a 1-bit ALU might have worked. In the 70s I did a redesign of a Data General CPU that previously used a single 4-bit ALU (74181), used four times to generate a 16-bit result.

Re: Not easy, but working

Posted: 18 Nov 2019, 21:13
by marcelk
I seriously looked at slicing this design in 4 bits, just to get the part count down a bit. But it complicated the logic around it so much that it cost almost as much as it saved. And it slowed things down. So I aborted that path and went back to the 8+2 setup we have now. It's easier to understand, looks better, and is fast.

Re: Not easy, but working

Posted: 19 Nov 2019, 10:10
by Colin
I'll give you easier to understand.
(I've just always fancied the idea of designing a screamingly fast 1-bit pipelined ALU.)

Anyway the Gigatron is great fun and very clever, especially building in the display. I'll be off to show it to friends in the pub tomorrow, both of whom worked on processor less computers in the 70s and later on bit slice designs.