[sdiy] HP 54602B selling on eBay.. DC vertical offsets on some channels or what ??

Florian Teply usenet at teply.info
Wed Apr 7 23:50:26 CEST 2021


Am Wed, 7 Apr 2021 10:08:02 +0200
schrieb Roman Sowa <modular at go2.pl>:

> If you don't mind 15% error on a scope then it's fine, I get your
> point. It's easy to check a multimeter just by comparing with another
> one to see if it's way off. And everybody has at least few of them.
> Checking if a scope is more or less within specs is not that easy 
> especially if old ebay scope is the only one in the lab.
> 
> Calibration means it was fully checked and nothing's broken.
> Taking that to synth terms - once I have serviced some big polysynth, 
> and it was quite OK when finished, but because not so perfect as I 
> wanted, I went through meticulous 2-days long full calibration 
> procedure. And only then I could find that there were still 2 or 3 
> faults, including a need to replace vintage chip being out of specs
> so much that it was impossible to adjust with trimmers.
> 
And here we enter "that depends" regime of things: I do agree that
usually the customer expects that equipment which has been calibrated
underwent rigorous testing and is performing to specs. But this
expectation is not always met in reality. In the end, it all depends on
what has been agreed on. Unfortunately, quite often in the field of
calibration strongly believe that it's so obvious that they don't
even consider actually discussing their expectations but rather assume
that there is only one possible way of doing things.
One example: At work we had for years some of our microwave network
analyzer calibrated only up to 18 GHz because the calibration lab
couldn't go higher. That's pretty pointless when you're regularly use
that equipment between 30 and 67 GHz and base pass/fail decisions on the
measurement result. But the guy who was taking care of calibration
before me didn't know (or didn't care). Luckily I noticed before one of
our customers did...

First and foremost, calibration is just a comparison with a known
value. Calibration in strict metrological sense does not involve any
adjustment of the equipment being calibrated. 
Think along the lines of a voltmeter which consistently gives readings,
say, 13 volts high. Or the abovementioned example of a scope with 15%
error. Calibration only gives you the knowledge, that indeed on day X
(the time of calibration) the readings were off by whatever the actual
amount of error is. But, assuming the error is constant and known, you
can correct your readings for it to get a much better idea of the true
value of what has been measured. One example here would be one of our
precision resistance thermometers at work. It reads okay at room
temperature, but the readings are off once we deviate from room
temperature, and have been that way for at least thirty years. There are
no trimpots or similar inside the instrument which could possibly be
adjusted. But we have sufficient data to reliably predict the error and
correct the readings for it. Still during last calibration the
calibration technician refused to sign the calibration certificate at
first because the equipment was not meeting specs and he couldn't
adjust it. It turned out that at the cal lab where he worked before
handled things differently...
Especially in precision metrology, people quite often ask explicitly
that adjustments are NOT to be done.

> All I'm saying, there's not much that can happen in multimeter to
> make it lie and not just go dead. But in a huge device like vintage
> scope with crazy high voltages, heat, airflow, early technologies
> stretched to their limits, and who knows what else, there's so many
> things that may drift in time.
> 
Umm, well, that depends on what you consider "lying" in that regard.
In recent years at work, I have seen quite a number of multimeters
which seemed to be working okay but nonetheless were not meeting specs
in various and non-obvious ways:
a) Readings were off by more than 10% in just one particular range
  a.1) across the whole range. Could be user error (using too small a
  range for a measurement and therefore altering part of the
  measurement circuitry, mostly due to thermal overload)
  a.2) for values above a certain threshold. This was a nasty one to
  find: The multimeter in question had a hybrid resistor network as
  switchable voltage divider, which was supposed to be hermetic.
  Through a combination of factors this one developed a sort of
  conductive filament with a spark gap which shorted part of the
  resistor network when specific conditions were met: high humidity,
  voltages above a certain threshold and of course the right range
  selected. Even calibration didn't show this.
b) readings are off occasionally, but often okay. I had fractures in
such a hybrid voltage divider network, which caused open circuits in
parts of the network or not, depending on how the equipment was bumped
when placed on the desk, causing readings to be either 11% high or good.
c) all readings consistently of by a significant but small percentage.
That's quite often a reference drifting over time.
d) all readings off by a fixed amount, regardless of range. Could be
parasitic thermocouples due to corrosion and thermal imbalance in the
equipment...

And there's probably more potential issues even simple multimeters can
have which are not obvious, and some of it even need a bit of luck in
order to be found during calibration, even if it is performed to
specification.

HTH,
Florian


More information about the Synth-diy mailing list