[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.
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!