The HP/Agilent 16700A/B and 16702A/B logic analyzer mainframes were originally released somewhere between 1998 and 1999, and though they have now been obsolete for many years, they can still be useful for hobbyists who need to debug digital circuits with large parallel buses (especially the homebrew CPU community, who build computers from scratch, often using nothing but 7400-series logic and parallel SRAM chips). These "boat anchors" were the last logic analyzers produced which do not use commodity PC hardware or run Windows; instead they run HP's unix variant (HP-UX) on a 150MHz PA-RISC processor. All the operating system and analysis software is still available, and can be run from a SCSI hard drive emulator like SCSI2SD or ZuluSCSI. And since it is just an X-windows application, you can run an X server on a modern machine with a high resolution monitor and then view the analysis software on that instead of being limited to the built-in 1024x768 LCD.
The actual triggering, sampling, and storage of waveform data is done on up to five modules. HP/Agilent produced a variety of these, but the ones most commonly found today are:
- 16710A, 16711A, 16712A: up to 128k sample depth, 104 channels (max. 208 with 2 modules), 100MHz state analysis, 500MHz timing analysis
- 16715A, 16716A, 16717A: up to 4M sample depth, 68 channels (max. 340 with 5 modules), 333MHz state analysis, 2GHz timing analysis (except 16715A)
There's only two different PCBs used for these and the lower model numbers are effectively just software limited to a lower memory depth and can be "upgraded" by adding or removing a resistor (apart from the 16715A which lacks all the timing zoom chips).
All the 167xxA modules had plastic sliders on the bottom to prevent damage to any components on that side if not installed carefully. Unfortunately, these rails were attached with a foam/rubber adhesive that over time either becomes corrosive, or absorbs water from the air which then becomes corrosive. This has been documented extensively on the EEVblog forums. It seems this corrosion begins to appear pretty reliably around 5 years after the board was made. As of 2025, any board which has not already had the rails removed is almost guaranteed to no longer be fully functional. Depending on the damage, it may still be possible to repair, but if any corrosion managed to work its way into a via connecting to one of the internal signal layers, it will likely be unfixable. I have a handful of 16717A boards like this, and multi-layer PCBs have become surprisingly affordable in recent years, so I decided to desolder all the components from one and sand down the PCB layer by layer to reverse engineer it. You can find the scans from this in the scans directory.
In the original directory you can find my reverse-engineered version of the 16717A that is as close to the original board design as I could make it based on my scans. Due to distortion, there may be some errors in component and trace placement, however it includes all test points from the original and should be useful when debugging failed 16717A boards. Unfortunately there are a number of signals that I can't confidently guess at their purpose, so I just named them things like CTRL1 etc.
I wasn't happy with a number of things about the original design, so I decided to create a modified version that I think should be more reliable, and if you're going to have these boards made, it's the one I would suggest. This version is pictured above, and the details can be found in the redux directory. As of this writing, ordering a set of 5 PCBs from JLCPCB will cost between $260 - $310 depending on solder mask color and shipping selections.