Thursday, September 23, 2021

CONTINUITY: HP 9825 Repair Part 12—Onward with the Keyboard, Display, and Printer Board

After replacement of all of its blown-out display chips, further diagnosis of the Keyboard/Display/Printer (KDP) module raises the disturbing possibility that its custom IC controller chip, which now might as well be made of a secret alloy of delirium and pandemonium, was destroyed by the power supply malfunction. Is it dead, or just compromised by further failures in the “glue” logic interfacing it to the rest of the board?

Posted at 10:41

Wednesday, September 1, 2021

CONTINUITY: HP 9825 Repair Part 11—Debugging the Keyboard, Display, and Printer Board

The repairs of earlier episodes got the Hewlett-Packard 9825 laboratory computer working, but only with a keyboard, display, and printer module borrowed from a working 9825. Now the focus moves on to repairing the module from the unit fried when its power supply blew up. This is a worrisome task, since this module's circuit board includes a custom Hewlett-Packard application-specific integrated circuit (ASIC) which, all these years later, might as well be made of gold-pressed unobtainium.

Posted at 15:15

Monday, June 14, 2021

CONTINUITY: HP 9825 Repair Part 9—Fixing the Memory Expansion Board

With the Hewlett-Packard 9825 laboratory computer booting in “classic” mode, it's time to move on to the memory expansion board, nicknamed “Skoal”, which is one of the supreme sledgehammers ever created to cope with an address space limit. The custom H-P microprocessor used in the computer had 16 address lines, limiting it to a 64 Kb address space, around half of which was occupied by the system firmware. But the customers demanded more! What to do? The designers were not permitted to change the original firmware, which had to remain identical for compatibility with existing software. That ruled out commonly-used schemes such as bank switching (as I used on the Marinchip 9900). So, the Skoal board ascended to a new summit of hackery by overlaying the original address space with additional RAM, then implementing nine rules to decide whether an instruction in the original ROM intended to access RAM or ROM. Eight of the rules handled the most common cases, but for those they didn't catch, there was a ROM with a bit for every instruction address in the original ROM that indicated whether it was accessing the original RAM. These settings were derived from analysis of the code with a logic analyser and the source listing. This is a level of kludge which deserved a Lifetime Achievement Golden Hammer Award, were I still giving them.

Posted at 13:11

Monday, May 17, 2021

CONTINUITY: HP 9825 Repair Part 8: It Boots Again!

After digging into the tangled logic of the RAM board, simulating it with Logisim Evolution, the problem is narrowed down to an Intel 3242 dynamic RAM controller chip. This was was a single chip solution which handled multiplexing of row and column addresses as well as the refresh address to an array of dynamic RAM chips, and automatically kept track of the refresh address. This eliminated a handful of MSI chips in a memory board design. I used one in my 1979 M9900 64K RAM board.

Replacing the chip and…it boots!

Posted at 12:23

Sunday, May 9, 2021

TRACKING WITH CLOSEUPS: HP 9825 Repair Part 7: How Dynamic RAM Works (or Doesn't)

In the next installment of the “everything is broken” saga, the focus moves on to the random access memory (RAM) of the Hewlett-Packard 9825 laboratory computer. The refresh circuitry appears not to be working, which causes the system to rapidly forget what's been stored. What's refresh? Perhaps a deep dive into the history and technology of dynamic RAM will refresh your memory!

The HP 9825 uses the 4116 dynamic RAM chip, which brought back…memories…as I designed a 64 Kb memory board in 1979 using this same iconic chip.

Posted at 11:41

Sunday, May 2, 2021

CONTINUITY: HP 9825 Repair Part 6: Can the HP 547A Probe Get Us Out of Trouble?

As analysis progresses of the damage done to a 1970s vintage Hewlett-Packard (H-P) 9825 laboratory computer by a single shorted transistor in its power supply, one of the worst nightmares in digital circuitry debugging manifests itself: a stuck bit in an on-board bus. The stuck bit could be due to any of the numerous components on the bus, all soldered in place. To deal with this problem, in the 1970s, H-P introduced a piece of test equipment, the HP 547A current tracer. The idea was that by sensing the magnetic field of current flowing to the fault in the circuit, it would be possible to localise the fault on a bus with many components attached. Can this H-P gadget diagnose the fault in this H-P computer?

Posted at 11:57

Wednesday, April 21, 2021

CONTINUITY: HP 9825 Repair Part 5: What Isn't Broken?

Or, “CPU board, we have a problem.”

Posted at 11:30

Monday, April 19, 2021

CONTINUITY: HP 9825 Repair Part 4: The Processor Reads and Executes ROM Code

Further probing with the logic analyser, correcting its interpretation of bus signals, board swapping with a working machine, and another mysterious release of magic smoke gets closer to the source of the problem(s).

Posted at 13:01

Friday, April 16, 2021

CONTINUITY: HP 9825 Repair Part 3: Logic Analyser and a 43 Year Old Patent to the Rescue

After going about as far as possible with an oscilloscope (although back in the day we went way deeper into the woods with just a ’scope), it's time to hook up a logic analyser and see what the CPU and memory are doing. Aiding in the process is U.S. Patent 4,075,679 [PDF], granted in February, 1978 and assigned to Hewlett-Packard, whose 606 pages contain, inter alia, a complete commented source code listing of the ROM and extensive logic, circuit, and timing diagrams. How deep was the damage to this vintage machine when its power supply went all berserker?

Posted at 13:42

Sunday, April 11, 2021

CONTINUITY: HP 9825 Repair Part 2: Is Our Rare 16-Bit Processor Fried?

In Part 1, the catastrophic failure of the power supply was analysed and repaired, but that didn't fix the computer, indicating damage elsewhere as a result. Now the investigation digs deeper into the circuitry, checking the clock generation and memory access signals from the processor. There's excellent background about Hewlett-Packard's pioneering 16-bit processor built from multiple chips on a hybrid substrate.

Posted at 13:59

Wednesday, April 7, 2021

CONTINUITY: Hewlett-Packard 9825T Repair Part 1: Power Supply

In January, 2021, Curious Marc's 1970s vintage Hewlett-Packard 9825 laboratory computer blew up when a single transistor in the power supply failed with a dead short from emitter to collector which placed 13 volts on the +5 power supply rail. This caused one integrated circuit on one of the boards to literally explode, with less apparent damage elsewhere the way to bet. The diagnosis and repair adventure begins with fixing the power supply.

Posted at 14:41