curiousmarc
patreon
The HP 9825 is alive again!
🕑 Added 2021-05-15 23:16:37 +0000 UTCYay! Our HP 9825 is back! At least in its "B" guise. We finally got the RAM working on the A24 board, and it booted right up.
The Logisim simulation of the refresh noodle logic led me to suspect the Intel 3242 helper chip rather than the refresh logic itself.
It's the TTL hub chip that sits in-between the refresh noodle logic and the RAM.
Off the board it goes:
It is one of the early medium scale integration chips from Intel, released in 1976. It's an address multiplexer that does the row/column multiplexing of the 14 bit addresses into the two row and column 7 bit chunks that the MK 4116 DRAM chip wants. It also contains the refresh counter that is incremented at each refresh cycle by the noodle logic.
I had a brief outburst of Ben Eater-ness and made myself a vintage bench tester for the chip. The chip tested entirely bad, always giving 77 octal output no matter what. Ken took it for an autopsy.
So all the address accesses, including the refreshes, went to a single location in memory! Which is why the memory tested good as seen from the processor: it was always testing the same location, without knowing it of course. And when it tried to boot up, all writes to RAM overwrote each other at that single location, hence the bad subroutine return addresses.
A new old chip procured from eBay tested good. Here you can see it in the row mode, and it faithfully transfers the "54" row input from the dial wheels at the left to the LED output. I could see it switch to the column (dial on the right), and I could increment the refresh counter (push button on the lower right).
I was still unsure if this was the only fault in our memory. But after I put the new Intel 3242 in, the computer booted right back up. So that was the only faulty chip left in the RAM system.
The final tally of the damage on the A24 board consists of three regular TTL chips, one Intel MSI TTL chip, and one ROM.
I suspect the failure rate is going to be relatively similar on the A25 32k RAM extension board, which we still need to repair to make it back to a true 9825T. Right now it's back to a 9825B. That level of damage seems manageable. And we actually had zero damage on the processor board, so there is evidence of at least one better case.
That leaves us with the keyboard, which we know is bad, and the tape drive board, which status is unknown. But unless the HP proprietary chip on the keyboard is zapped, we should have a good shot at bringing the 9825 back to 100% functional, and learning how it works while doing it.
Marc
Comments
"a brief outburst of Ben Eater-ness" LOL
Dan Swinehart
That chip needed one more optimization: Always writing 0. Then you could replace memory with a block of wood.
