Rebuilding A Custom IC Saves HP Pulse Generator


Rebuilding an HP Custom IC[Matthew] got himself into a real pickle.  It all started when he was troubleshooting a broken Hewlett Packard 8007A pulse generator. While trying to desolder one of the integrated circuits, [Matthew] accidentally cracked it. Unfortunately, the chip was a custom HP Pulse shaper IC – not an easy part to source by any means. That broken chip began a 5 year mission: to explore strange new repair methods. To seek out new life for that HP 8007A. To boldy fix what no one had fixed before.

[Matthew’s] first repair attempt was to build a drop in replacement for the HP chip. He took a look at the block diagram, and realized the chip was just some simple logic gates. He built his version with a small PCB and Fairchild TinyLogic gates. Unfortunately, the TinyLogic series is fast CMOS, while HP’s original chip used Emitter-coupled Logic (ECL). Thanks to the wildly different voltage levels of the two logic families, this design had no chance of working.

Five years later, [Matthew] was going to school at MIT, and had access to a wire bonding machine. He rebuilt the package using some epoxy, and managed to re-run the various bond wires. While everything looked promising, this attempt was also a failure. After all that work, the chip was blown.

Block diagram of the HP 1820-0285

If you haven’t figured it out yet, [Matthew] is a persistent fellow. While setting up for wire bonding, he’d gotten a good look at that HP die. The HP chip was a relatively simple design, so simple that he was able to reverse engineer the entire schematic from the die images. Similar to his TinyLogic design, he built a drop in replacement on a two layer PCB. This time he used discrete transistors and resistors to replicate the ECL logic. By using both sides of the PCB, he was able to fit everything into a 16 pin DIP footprint. The result almost worked. The two layer board had some long traces. With low frequency transistors, the circuit would work – but not up to 105MHz. Switching to high frequency transistors caused the entire circuit to oscillate.

[Matthew] laid the board out one more time using power and ground planes. The simplified layout, coupled with BFS17 transistors worked. It wasn’t quite as good as the original HP design, but for the purposes of the pulse generator, it worked perfectly. He didn’t even have to recalibrate.

We love seeing old test equipment brought back to life. If you know any stories like this, drop us a tip!

38 thoughts on “Rebuilding A Custom IC Saves HP Pulse Generator

  1. This is what I love to see, maybe I’m getting nasty with age, but next comes the Cease and desist from HP’s lawyers. But I’d say that Bill and Dave, if they were still alive, would have offered Matthew a job.

    1. While I do enjoy the occasional AVR article, I would have to parrot that this is what HaD needs more of.

      Don’t forget, not all of us have full access to labs at MiT or ever hope to so people like me need the AVR (non-Arduino) or PIC article so we don’t feel left out ;)

  2. No BS, No nonsense, no compromise, old school hack! The site even has scope shots for comparison. You have thrown down the gauntlet (metaphorically speaking). I am truly amazed.

  3. Truth while you are right there may be copyright issues we are talking about something from the 50’s or 60’s even possible as early as the 40’s so a 70 year old technology is hardly protectable by copyrights.

    though hp could try

    heck much of the machine could be done in software now adays.

    1. Not entirely true, most Military Spec parts are still made and have to be made using Ceramic bodies with Gold legs to prevent corrosion in extreme environments. So judging age by the color of the legs is not accurate.

  4. If he had access to a wire bonding machine, surely he could’ve found someone with mosis (I think that’s the right name for it in the US) / Cadence access too, found a compatible process, and just tape out a copy of the chip – it is indeed rather simple, and all the detail required to duplicate it is visible in the image. It wouldn’t be nearly so easy with newer tech kits where the whole chip is full of fill cells to meet the metal density rules, so that you can’t actually see the circuit without deprocessing some metal layers off the top first.

    1. Could you elaborate on the phrase “tape out a copy of the chip” using mosis? Do you mean getting a chip fabricated somewhere, or is there a process that would allow us to build ecl parts a bit more elegantly? What would the process in between “working schematic” and “working part” be?

  5. And I just sold a bunch of those off for the scrap gold content. If only I had known…

    Just kidding! It amazes me the lengths that we as hackers will go to resurrect an old piece of equipment, be it from personal/sentimental attachment or just the mere satisfaction of meeting a challenge. I agree: we need more articles like these.

  6. I’ll add my kudos. I’m impressed. I also would like to see more of this type of post on HaD, but I can understand the scarcity. There just aren’t that many people who are doing this level of work, at least not while posting it publicly.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.