[Radio Wrangler]

Calibration is an interesting minefield.

A general certificate of calibration simply means that an instrument was checked on a specified date and was found to comply with its manufacturer's specifications.

Thereafter, several things go wrong:

1) The couriers bounce the thing down several flight of stairs on its way home from the cal lab. Maybe this affects the accuracy, maybe it doesn't. How lucky do you feel?

2) Once home, most users see the cal certificate and then treat the instrument as gospel, ignoring or unaware of the allowable amounts of error specified by the manufacturer.

3) Most users ignore or are unaware of the specified uncertainties of the cal lab they used.

4) Calibration certificates, like MOT certificates relate only to the condition at the time of the test. What if the couriers chucked it out of one van/plane too many? Everything you adjust using it from then on will be wrong. You may only find out at the next calibration time when you get notified that it failed. Calibration is BACKWARD LOOKING. You have to decide if you need to get things back from the field to correct them. Unfortunately, most people see a cal certificate as a prophecy that the instrument can be trusted for the next year. What it does do is justify what you did with it in the PREVIOUS year. Here is an application just ready for someone with a time machine, but they may not be interested in things less lucrative than cleaning up on the football pools, the stock market etc.

Once you start digging into calibration and accuracy it comes as a surprise how nasty it gets.

Individual correction tables can be useful, but they transfer the demand for accuracy into a need for repeatability and stability.

I was involved in the HP/Agilent noise sources used for measuring noise figure. Each noise source has an individual sheet of Excess Noise Ratio and impedance figures for frequencies at 1GHz intervals across its range. Each of these figures has an associated value for uncertainty. They are all different because calibration at different frequencies may trace back to different thermal standard hardware, and different labs may trace back to different hardware sets as well. As time goes by, national standards labs revise the mean and uncertainty figures for their thermal standards hardware and uncertainty figures change from year to year. So, if you measure noise figure, the noise source you need as a standard doesn't have one global accuracy figure, each frequency point in its cal chart has a different uncertainty and the effect on noise fig results is very significant. Ian White and myself gave a presentation on errors in noise fig measurements at the EMC 2012 conference (Churchill College, Cambridge) and it caused a little bit of a stir.

Maybe it's like what they say about sausages and laws; you really don't want to see how they are made?

Avo probably had a Weston cell and a potentiometer/galvanometer as their root standard. Production instruments would have seen a transfer standard instrument checked against it. Perhaps they had several cells to compare against each other as a safety check.

David