I wonder how many of us have had to suffer through using
toggle quoted messageShow quoted text
I was given one for Christmas when I was about 14, with
the express understanding that my dad could use it if he
needed it.... He did, and burned out the front end by
probing around in a horizontal section of a TV set.
And thus began my great adventure fixing the all mighty
I had to scrape away the arc'd over section of micarta on
the vertical input attenuator switch, fill it with epoxy,
replace the burned out precision resistor with selected
carbon composition resistors, and replace a few capacitors
Then I replaced all of the bumble bee capacitors, as they
were all leaky, both physically and electrically. And then,
all of the Sprague 20uf-20uf, 450V FP style capacitors...
and most of the 1 and 2W carbon composition resistors that
fed them... and the screen of the forest of 6AG7 metal tubes...
And, then, one of the 5642 diodes in the HV can popped, making
the beam dim and very large... So, I learned to puddle in
an oil filled assembly...
I still have it at my mom's house. I wonder if it still
I too designed a ttl computer, but that was when I learned
that I am truly greedy. I couldn't be satisfied with TTL
speeds, so I had to use STTL. A modest instruction set wasn't
good enough, so I had to give it all 16 possible logical/arithmetic
functions of two inputs... and a full 74S181 ALU. An accumulator
and an address register wasn't enough, so I designed in a 16 word
register file. I couldn't even consider making it byte wide, as
that would take multiple fetches per instruction, so it was at
first 16 bit, and then became 24 bit...
I accumulated parts, spending all of my lunch money... until
Morrow's Microstuff announced its $130 front panel/CPU replacement
board for the S100 IMSAI 8080 computer... and my own design died.
I never had an assembler for the 8080, I assembled everything by
hand directly into located machine language.
I guess I have always been a CISC over RISC guy.
Harvey White wrote:
I (very early on) tried to build a Tektronix calculator out of TTL, jumps, program
flow, and then an FPU.
Got the jumps and program flow working, you could do an indirect jump through a
location in hardware (and yes, this was very early....)
I built an 8 trace scope switch to be able to look at digital traces on my 513D (I
got lucky getting the 513D, P11 phosphor).
Simple enough, sync a counter to the sweep trigger output, take the counter, run it
to a multiplexer and weight the multiplexer output as '1' in a very crude D/A.
Take the appropriate counter outputs and weight them 2, 4 and 8.
Instant 8 traces on a 1 channel scope, for 5 volt TTL, of course.
Then along came the 6502, for 30 dollars, one each (while the 8080 was still 100
dollars or so). I looked at my 130 chip wirewrapped design (still not done), and
felt a bit bad.......
I bought two (and literally, drove to the MOS technology factory to pick them up with
the original books). Still have them, I think.
There were some bugs in the microcode, some things not implemented, and I seem to
remember being bitten by that once I got the "newer" processors.
Naturally, the whole thing was in assembly language. Program the thing at work
(someone tweaked the DG macro assembler), program the UVPROMS, take it home and
single step (hardware) through the program.
Everything was wirewrap then. Just everything.
On 1/20/2020 9:39 AM, Chuck Harris wrote:
Many of us early adopters to microprocessors got bit by
the discovery that not all of the 8080's made were truly
Everyone knows that Intel developed the 8080, but some might
not remember that the Japanese company NEC, almost immediately
introduced a cheaper, unauthorized, second source to the 8080...
only theirs was partly done from Intel's design, and partly
done on their own.
Some of the subtle differences were in the settings of flags
after some arithmetic operations, and in the numbers of clock
cycles used by some common instructions.
I discovered the flag problem, when I found some code that
should have branched, only it didn't... NEC fixed a mistake in
the Intel microcode.
I discovered the clock cycle problem when I was writing a music
program, that played on a transistor radio... using timing loops.
I meticulously counted cycles, ala the Intel data book, but
several notes were coming out sharp.
It turns out that Intel used 5 cycles for one instruction, and
NEC found a way to do it in 4 cycles...
And, I had stacks of these instructions, inside of loops that
timed the note pitch.
Dave Seiter wrote:
Grin! Things were different back then. In high school we had dial up line to a
HP2000 at the Stanford business school using DECWriters, and one could wander
through the directories, browsing students projects. It was usually very
responsive, but sometimes it would really bog down, and we'd have to use one of
the in-class micros (PET/APPLEII/TRS80) to work on our projects. But they were
all more reliable than our PDP8/E; we had to use that for Fortran, and it was
always breaking down. I think I was one of the last people to learn how to use a