If you’ve designed electronic systems for a while, or if you’ve repaired vintage electronic equipment, no doubt you’ve run into the fascinating term “unobtainium.” Here’s Wikipedia’s definition of the term:
“Unobtainium originally referred to materials that do not exist at all, but can also be used to describe real materials that are unavailable due to extreme rarity or cost.”
Wikipedia lists a few examples, including parts for old cars made before the 1960s, parts for reel-to-reel tape recorders, and rare vacuum tubes that haven’t been manufactured for many decades. Many ICs made in the 1960s and 1970s now fall into the unobtainium category. For example, I recently purchased the last two 9915 dual RTL NOR gates available on the Web for an earlier EEJournal article. (See “The First ICs on the Moon – The Apollo Guidance Computer, Part 2.”)
The Vintage HP Computer group frequently runs into unobtainium challenges. This band of hardy, knowledgeable, resourceful, and enthusiastic individuals has dedicated itself to the preservation and continued use of old HP computing machines made as far back as the 1960s, when HP first got into the computing business.
One particular Vintage HP Computer story involves the resuscitation of HP 98035A real-time clock (RTC) cards, which were designed for the HP 98×5 series of desktop calculator/computers back in the late 1970s. The 98035A, introduced in 1977, plugs into one of the real I/O slots of these computers and provides millisecond-accurate, time-of-day information to the computer as well as precisely timed interrupts. Many HP 98×5 desktop computers were incorporated into automated test and measurement systems, so time-of-day information was an important data collection component. (This card was designed a few feet from my own little engineering cubicle and workbench at HP’s Desktop Computer Division in Loveland, Colorado.)
The 98035A is based on two important components that are now made of unobtainium: the HP-manufactured Nanoprocessor, was an ALU-free microprocessor designed in the early 1970s for HP’s Loveland Instrument Division, and a Texas Instruments (TI) AC5954N digital watch chip, which kept time while the computer was switched off and no longer powering the real-time clock card. When the computer powers up, the 98035A boots the Nanoprocessor, which loads the time of day by reading the multiplexed 7-segment information from the TI watch chip.
A 2.5V NiCd battery internal to the 98035A is supposed to keep the watch chip running while the computer is powered down. These original NiCd batteries are now half a century old and, invariably, they leak electrolyte all over the 98035A internals, which causes all sorts of mischief including corroded circuit board traces and failed ICs. If either the Nanoprocessor or the TI watch chip fail, the I/O card is nearly unfixable because these two ICs are not available now except as reclaimed parts stripped from other equipment. (You can read a lot more about HP’s Nanoprocessor in this blog written by Ken Shirriff, a supreme wizard of IC reverse engineering who reverse engineered the Nanoprocessor from its original masks a few years ago.)
Four years ago, my friend Marc Verdeil, a member of the Vintage HP Computer group, managed to collect six 98035A clock cards. All of them were in non-working condition. Verdeil has a YouTube channel called “CuriousMarc” and, as part of a multi-video series on repairing these HP 98035A cards, he ordered some replacement TI clock chips from the only place he could find them: a vendor in China. Verdeil made at least nine YouTube videos about his repair adventures with the 98035A, and they’re listed in the reference section at the end of this article.
TI originally made this watch chip for its LED Black Watch, which revolutionized digital watches back in 1976 when it was introduced. The TI-501 Black Watch, made completely out of plastic, sold for $19.95 back in 1976. Other digital watches of that era cost hundreds of dollars. I bought a TI-501 as my first digital watch, and I wish I still had it. The heart of the TI-501 was an I2L (integrated injection logic) chip that was never offered as a packaged device to TI customers. However, HP was a very good TI customer back in those days and wanted the TL-501 chip to use in the 98035A, so TI accommodated HP by packaging the chip in a 24-pin plastic DIP and marked it as AC5954N. As far as I know, no other system vendor ever used this chip, and HP never used it in another product. Given the custom nature of these real-time clock chips, it’s surprising that Verdeil found any at all. However, he did find these chips in China and, incredibly, they worked. However, the clock chips did not turn out to be the problem with Verdeil’s 98035A.
More recently, Paul Berger, another member of the Vintage HP Computer group, was also trying to revive his 98035A clock card. He traced his problem to a failed output on the TI watch chip. Unable to source a chip from China as Verdeil did, Berger considered three alternatives to replacing the failed IC:
“The first would be to use a microcontroller to emulate the TI clock chip. If I was successful at this, nothing else would need to change in the 98035A. The second option was to rewrite the Nanoprocessor code that reads and sets the RTC to use a different RTC. The third option was to replace the Nanoprocessor with a more modern microcontroller and RTC chip.”
All those alternatives seemed like too much work. Then, Berger found an old article in the October 1978 issue of Kilobaud written by Emerson Brooks and titled “Let Your Computer Wear a Watch,” which described a project to directly interface a TI-501 Black Watch to an SWTPC 6800 microcomputer. Berger wrote:
“After a delay due to sudden retirement and needing to move homes again, I finally bought a plain brown TI series 500 watch from eBay. The first thing I did was clean it up a bit inside and install batteries to make sure it worked, which it did. So, will it work with a 98035A? Short answer: yes!”
(Note: If meatball surgery turns your stomach, you’d best stop reading at this point.)
Berger then wrote a detailed description of his Black Watch transplant into the 98035A:
“The guts of the watch is a molded piece of plastic that the chip appears to be encapsulated in. In the middle there is a gap that has a battery separator set into it. When this separator is removed you can see the connections to the back of the display. At the back of the display all of the digit and segment connections are available.
“It would seem that the author left the display in the watch, however I chose to remove the display. I made the connections to all of these pins using wire-wrap wire. I also made a connection to the two pins on the lower left of the diagram. If you look at the back of the display, there are ground (-) pins. I am not sure if these connections are important, but I did ground those two pins.
Interface connections to the front side of a TI Black Watch module (with the LED module removed), ready to transplant into an HP 98035A clock card. Image credit: Paul Berger
Interface connections to the back side of a TI Black Watch module, ready to transplant into an HP 98035A clock card. Image credit: Paul Berger
“The other connections you will need to make are the set and display button connections (make note of which is which by comparing to buttons on the watch case), and the power connections. I just soldered wires to the battery terminals. You now have all of the connections you need except the one labeled C15 on the schematic in the 98035A Installation and service manual.
“There are 6 pins protruding from the epoxy frame near the negative battery terminal that are not connected to anything, and I wondered if one of them might serve the same function as C15. Sure enough, one of them stopped the count when seconds are displayed, just like C15.
“So now I had all the connections, and it was time to fit it into my 98035A. I already had a socket in the position of the clock chip from earlier PD and I initially connected the wires from the watch to a 24-pin component carrier, so my first attempt was to plug the component carrier into the socket. But this proved to be too thick, so I removed the socket and tried putting the component carrier directly into the board. However, the carrier is wider that a normal IC and it was too cramped, so I ended up soldering the wires directly to the board.
“I am not using an internal rechargeable battery, so I located the watch module where the battery would normally be and insulated both sides with tape, just in case. Since the watch has a crystal and trimmer cap already on it, I am not using the onboard crystal and trimmer cap.”
So, there you have it, another ingenious way to overcome an unobtainium challenge. These days, all of the 98035A’s circuitry can be emulated by a small FPGA or a microcontroller from the usual vendors (Altera, AMD, Lattice, Microchip, NXP, STMicro, TI, etc.) or a $5 Raspberry Pi Pico, but with the addition of a lot of design work. In some ways, I much prefer Paul Berger’s more “Mad Max” approach to solving the unobtainium problem for the 98035A clock card.
How about you? Have you had adventures with unobtainium? Let me know about them in the comments below.
References
Ken Shirriff, “Inside the HP Nanoprocessor: a high-speed processor that can’t even add”
Marc Verdeil (CuriousMarc), “Chinese-sourced vintage chips, real or fake?”
Marc Verdeil (CuriousMarc), “HP 98035 Clock Module Repair-A-Thon – Part 1: 12 bad boards, how many can we save?”
Marc Verdeil (CuriousMarc), “HP 98035 Clock Module Repair-A-Thon – Part 2: HP’s Nanoprocessor.”
Marc Verdeil (CuriousMarc), “HP 98035 Clock Module Repair-A-Thon – Part 3: magic smoke escapes.”
Marc Verdeil (CuriousMarc), “HP 98035 Clock Module Repair-A-Thon – Part 4: getting the ghosts out of the interface boards.”
Marc Verdeil (CuriousMarc), “HP 98035 Clock Module Repair-A-Thon – Part 5: 100% success or bust!”
Marc Verdeil (CuriousMarc), “HP 9825 calculator with HP 98035 clock module demo (and an Apollo switch!)”
Marc Verdeil (CuriousMarc), “LED watch chip from Texas Instruments (HP 98035 Repair-a-thon Part 0)”
Marc Verdeil (CuriousMarc), “Bubble LED clock hack with 1970’s tech”
Emerson Brooks, “Let Your Computer Wear a Watch,” Kilobaud, October 1978, pages 52-60
