Automating full bench calibrations

[rhb]

Having acquired a very richly equipped test bench over the last two years, I am now attempting to address the problem of maintaining it.

My goal is to be able to cable up my bench to a relay system and run a comprehensive suite of tests to verify proper operation from DC to 3 GHz with source harmonics to 22 GHz once a year.

Control will be via an Arduino Mega2560 using the AR488 FW by WaveyDipole.  I'm not a big Arduino fan, but they are cheap and are able to drive 13 GPIB devices on the bus.  And most of the heavy lifting has already been done by WaveyDipole.

For my particular bench, a pair of ST6P Radiall SMA relays will handle my current RF gear and the DC and low frequency AC component will be handled by a 44421A relay card from a 3497A.  However, I am allowing for TEA syndrome in the overall design.

As a practical matter, I do not see a way of automating impedance measurements, so I am not including LCR meters and VNAs in the automation plan.

Because of the magnitude of the task I expect to write a test description language that feeds off a table of instrument specifications to actually generate the instrument commands.  So a 3478A file will test those and a 34401A file will test those across all ranges.  The goal is to define a test suite once and never have to touch it again.

Has anyone attempted anything similar?

[TiN]

Yup, I've automated calibrations of various DC/LF equipment I own, because paying to get everything calibrated with calibration lab would cost more than actually cost of running/supporting my own calibrator (tens of thousands).
And sad part - even if you interconnect everything with fancy relays and cables, you still must pay for external calibration lab for your sources/meters to get at reference standards in lab. There is no way around this.

To control insturments/collect/process data I use Raspberry Pi 3 or FPGA-SoC board, running linux and python app, called "calkit". I'm busy developing this app since August 2016 on free time.
Currently it supports full verification and calibration for 34401, 3457A, 3458A, K2001, K2002, K2182, K2400, K6517, R6581T, F8508A and Datron 1281, 4920, 4920M.
Calibration sources supported are Fluke 5700A, 5720A, 5720+5725A, 5730A and Datron 4808.

Example calkit report for 34401A.

Example calkit report of resistance standard calibration

Because software also stores each instrument errors/serial numbers and customer information for traceability, it is not open-source nor public.

[Kosmic]

Control will be via an Arduino Mega2560 using the AR488 FW by WaveyDipole.  I'm not a big Arduino fan, but they are cheap and are able to drive 13 GPIB devices on the bus.

Just a heads up, I'm not sure you will be able to drive 13 devices on a GPIB bus with that controller. I was using a simple pic based controller in the past and it was struggling with 3 devices. Was just not able to source the current.

I'm now using a Keysight 82357B, works a lot better.

[rhb]

And sad part - even if you interconnect everything with fancy relays and cables, you still must pay for external calibration lab for your sources/meters to get at reference standards in lab. There is no way around this.

That's what the Cal Club is for.

Because software also stores each instrument errors/serial numbers and customer information for traceability, it is not open-source nor public.

Would you mind writing out the mathematics for the various calibrations?  Scan or photo of hand written math will do fine.  No need to bother with typesetting equations.  Thanks.

[rhb]

There might be an issue with certain choices of address, but in general the ATMega2560 should work OK.  GPIB spec is 48 mA total per pin.

I can always use a 2nd unit for the RF deck if it gets flaky at more than 7-8 devices.

[rhb]

As an alternative to standard GPIB cables or plain flat ribbon, I'm considering using 3M 3782 24 conductor twisted pair ribbon cable to make up  GPIB cables for the test system using IDC connectors.  Standard is 20" between flat sections on the cable which should work out and keep the total bus length fairly short.  Does anyone have experience with such cabling for GPIB?

Aside from cost, the ribbon cable reduces the number of mechanical connectors and shortens the total length.