HP 8082A
PULSE GENERATOR
HP 8082A Pulse Generator
The HP 8082A is from 1974, but you would not know it from its specs. It offers pulses down to 2 ns with 1 ns rise time up to 250 MHz repetition rate. More over, both the leading and trailing edge rise times are independently adjustable. You also get complementary outputs, with optional polarity inversion and adjustable DC offset. It also supports ECL levels.
The very unique leading and trailing edge capabilities, together with the speed, make them ideal when working with core memory, and that is indeed how we made the core memory video.
And this is all achieved by superb analog design. The secret of the exceptional performance comes from HP proprietary ECL ICs, developed specifically for this instrument. Which is great when they work, and terrible when they don't, as in my instrument, because you cannot find replacements. It also seems that the HP 5081-3011 that failed in my instrument is prone to failure and is bad in mnay other instruments. But I made a replacement circuit using modern surface mount components to replace the failed gate, and in the end it worked very well, yiedling the original performance.
The HP 8082A was developed in Boeblingen, Germany. The HP Memory site and the HP Journals give a nice account of its development history:
HP 8082A Repair Videos
My HP 8082A had several faults in it, which were all repaired in this playlist of 3 videos. The tougher repair was designing the replacement for the failed HP proprietary ECL IC.
HP 8082A Documentation
I collected the relevant documentation below, including the excellent service manual and the HP Journal article. In addition to the scan of the manual available on the web, I added a high resolution pictures of the 6 pages that show the schematics of the proprietary ICs which are unreadable in the scan, from my the original copy of the manual I have in my collection.
HP 8082A ECL Inverter Gate Replacement
In order to repair my HP 8082A, I had to make a replacement for the failed gate of the U2 IC on board A3, part number HP 5081-3011. This is an HP proprietary ECL IC, but fortunately there was enough information in the original service documentation to show us the layout of the original IC and make a reproduction possible. Mattehw D'Asaro led the way with his resurrection of an earlier version of this chip for the previous generation of this pulser (the HP 8007):
I inspired myself from D'Asaro's work, but went with a slightly different solution, choosing to only reproduce the gate that had failed in a separate, RF connected PCB, which is inserted in the instrument circuit by simply moving some RF cables around.
Here is a rendering of the inverter gate replacement circuit.
The PCB is available for purchase from our sponsor PCBWay:
Here is the BOM for the board components.
You also need a 2 inch SMC to SMC cable. I found mine on eBay from vendor wifi_expert and they were both inexpensive and well made. Here is the link:
A viewer mentioned that it might be better to use BFR92 5GHz transistors as the BFS17 are being phased out. I have not tried to model nor use them, but there is a good chance they'll work just as well or better. Let me now if you find out.
Picture of the finished PCB.
It implements an inverter gate inspired by the original IC schematics. The complete KiCAD project is available for download:
Our ECL inverter gate installed on board A3. It connects to ground and -5V borrowed from a nearby filtering capacitor, and to the original ND pulse output on A3.
Board A3 back into the pulser. The ND and DEL coax cables now go to our new daughter board instead of the original A3 board.
The build of the board and repair installation are covered in the Part 3 of the video series.
HP 8082 ECL Inverter Gate (HP 5081-3011 replacement) Project Files
Epilogue: Repaired HP 8082A
After I finished the video, I tried the circuit with the faster transistors and it performed even better, giving shorter pulses below the required 2 ns. But the non-delayed pulse was still about 1 ns longer than the delayed pulse, itself now a bit too short. Turns out, there is a factory selected resistor (A42 on board A4) that needs to be adjusted to balance the pulses. After I did my own "factory" resistor selection, then adjusted the minimum width back to 2 ns using the R240 pot, the instrument exhibited ideal characteristics at the fastest pulse speeds.
HP 8082 pulses in double pulse mode after the "CuriousMarc factory" calibration. Bill and Dave would have been proud.
Resistor R42 on board A4 needs to be selected to match the non-delayed pulse width that of the delayed pulse. My factory resistor was 5.11k, middle of the range. I had to add a 3.3k resistor in parallel. Here is a view when I was trying it out.