Keithley 196 6.5 digit resolution test results

[Enginerding]

I recapped a Keithley 196 (note 1) and did some tests compared to an EDC 501J DC voltage reference (spec'ed to 100nV resolution + error).

I think this is about as good as this meter can get in terms of resolution on the 300mV range (lowest). The test lead setup is solid, but in no way perfect (note 2).  The results were recorded via GPIB (note 3).   They're relatively short tests, but these were about getting a baseline on resolution and noise, as opposed to drift and stability.

0mV output test, 10 minute:



1uV output test, 10 minute:



I thought about swapping the linear regulators on the analog rails for something higher spec'ed for noise and regulation.

Notes:
=============

1: This K196 was in beautiful shape, no cal stickers, no button wear; no scorching on the LM399.   If it was used at all, it had a good life.  I did test the ESR of all the electrolytic caps in-circuit (DE-5000).  After 30+ years the 85C caps on the analog rails (+/- 15v, +5v) were badly degraded but functional.  Nichicon 85C type, out-circuit tested with DFs 20%+, Tan Angles low 80s, and ESR still acceptable, but undesirable.  I did not test leakage.  The electrolytics on the digital rails tested used but OK, in and out-circuit, but changed them all anyway for modern Nichicon 105C offerings.  (these weren't particularly low ESR or anything in today's sense, but for the early 80s they would have been some high-speed caps).

2:   The test lead is 1 meter UTP of solid core (believed 22awg, copper coated with tin or silver), from binding posts on EDC 501j (sense wires not used; shorted), run to a screw terminal on a 4mm insulated boot banana jack (likely nickel over brass).

Big thing I've found is to get the point where dissimilar metals contact insulated from air currents, so that it can reach a better equilibrium; this will greatly reduce thermals.

3:  Big thanks to @WaveyDipole and his contributors, their GPIB to Arduino project works amazingly well, and cost me $6 (just make sure your Nano has a real FTDI chip).  I log the serial output via putty, and then paste it into LibreOffice Calc (or Excel).
https://www.eevblog.com/forum/projects/ar488-arduino-based-gpib-adapter/
https://github.com/Twilight-Logic/AR488/graphs/contributors

[bitseeker]

Looks like a success. If you get a chance, post some pics, too.