In message CANX10hDf4y9KLMpLkgvLGhrr_-8yunW0xsijkNpGU_w6_7+8XA@mail.gmail.com, "Dr. David Kirkby" writes:

To rule out "because I say so" claims.

In the BIPM-adjucnt sphere, there is no such thing as "a measurement", there
are only "a measurement with uncertainty".

The objective evidence can either be that you have implemented the
definition of unit in question, (Cesium, Hall-effect etc.) in which
case the uncertainty is the uncertainty of your implementation.

Alternatively, you can establish a chain of measurements back to
such an implementation, in which case your uncertainty is the "proper
combination" of the uncertainties of each step in the chain, where
"proper combination" takes due account of the fact that time
elapses between each measurement in the chain.

In other word: Calibration Certificates.

--
Poul-Henning Kamp      | UNIX since Zilog Zeus 3.20
phk@FreeBSD.ORG        | TCP/IP since RFC 956
FreeBSD committer      | BSD since 4.3-tahoe
Never attribute to malice what can adequately be explained by incompetence.

Funny thing how things work out time-wise: I had a discussion yesterday
on the very topic during re-audit for ISO 9001.

In basic terms, verification in metrology is a very slimmed down
calibration: For a calibration, you essentially check every range of
your instrument at usually five or more spots within that range in
order to determine accurracy of your instrument in each range.
For a verification, you do this only at the spot where you intend to
measure. So if you were to measure a nominal 7.2V source, you'd compare
the reading of your meter with your, say, known good 7.5V reference
instead of doing a full calibration of the meter. It doesn't tell you
anything about, say, the offset error of your meter, or how big the
deviation is at the 1000V range, it just tells you if your meter meets
requirements of the one measurement you intend to do.

So, in order to determine whether or not your chinese voltage reference
meets its specs, you'd check your meter against, say, the
well-characterized LTZ1000A you happen to have in your lab.

Strictly speaking, you still have to do it as carefully as you would do
a real calibration, taking all known effects into account, but it's
still much less time-consuming than a full calibration as you check only
one single point instead of all possible ranges with five points each.

Does this help answer your questions or did I just bring up more
questions than answers?

best regards,
Florian

Am Wed, 3 Jun 2020 09:58:59 +0100
schrieb "Dr. David Kirkby" drkirkby@kirkbymicrowave.co.uk:

Along those very lines, HP/Agilent/Keysight sold both calibration kits and verification kits for VNAs. We did a full calibration using the mechanical calibration kit at the start of a measuring session where you needed the absolute best accuracy your system was capable of and used the verification kit as a sanity check during the daily operation.

Steve
WB0DBS

On Sat, 6 Jun 2020 at 12:18, Florian Teply usenet@teply.info wrote:

Yes, it is.

In basic terms, verification in metrology is a very slimmed down

Yes. I wish VIM was a bit more explicit about it. A single sentence just
does not do it justice.

best regards,

Dave

But in calibration you have to make some type of corrections and on verification you dont you just compare to be within the manufacturer ranges. Now for instance Tektronix has verification manuals to check their equipment, before they called them calibration manuals and included adjustment to meet the specifications of the manufacturer. Now the user cant adjust easily because most are via software and is not commercially available.

From: volt-nuts volt-nuts-bounces@lists.febo.com on behalf of Dr. David Kirkby drkirkby@kirkbymicrowave.co.uk
Sent: Saturday, June 6, 2020 9:19 AM
To: Discussion of precise voltage measurement volt-nuts@lists.febo.com
Subject: Re: [volt-nuts] What if “verification” in metrology?

On Sat, 6 Jun 2020 at 12:18, Florian Teply usenet@teply.info wrote:

Yes, it is.

In basic terms, verification in metrology is a very slimmed down

Yes. I wish VIM was a bit more explicit about it. A single sentence just
does not do it justice.

best regards,

Dave

volt-nuts mailing list -- volt-nuts@lists.febo.com
To unsubscribe, go to http://lists.febo.com/mailman/listinfo/volt-nuts_lists.febo.com
and follow the instructions there.

Calibration does not involve adjustment. In fact adjustment is ill advised as it invalidates all previous calibration history and breaks the trends. Calibration defined as only comparison to known reference with known uncertainty, no adjustment. Adjustments are service procedure which can be performed between two calibrations to get "before'' and 'after' conditions. Labs that do adjustment without customer consent have good chance to make things worse , if we talking about precision instruments. Equivalent in metrology would be trying to adjust national 1kg standards to IPK every time when those are sent in for comparison. That would add lot of unwanted uncertainty. In metrology drift itself is not much of a problem, problem is to have it drift in predictable , ideally linear matter, so corrections can be applied by math at the time of use.

On June 6, 2020 11:58:55 AM EDT, Miguel Yepes theamberco@hotmail.com wrote:

BR,
Illya Tsemenko

Am Sat, 6 Jun 2020 15:58:55 +0000
schrieb Miguel Yepes theamberco@hotmail.com:

That's a very popular misconception here: Calibration by itself does
not include adjustment. You can have a calibrated instrument which is
one order of magnitude off, or an instrument which fails manufacturer
specifications for measurement accurracy, and still the calibration is
valid. Also you can have an instrument which meets specification but
is not calibrated. As Ilya mentioned, calibration is essentially just a
quantified comparison to a known good reference, giving you an estimate
of how much the instrument reading is off under well-defined conditions.

Of course performing a calibration requires you to take all known
systematic effects into account, and it would be wise to do the same
for verification, but there's no need at all to touch the instrument.
If you happen to know through calibration that your volt meter has an
offset of 1.8236V in all ranges, there'sno need to change what the
instrument displays. You just take this offset into account whenever
you use the value you read. The same goes for all the other types of
corrections: if you know the instrument has a temperature dependency of
-50mV/K, and it has been calibrated at 23°C, but you're doing the
measurements in summer at 28°C, you better add 250mV to your reading to
cancel that temperature drift. And you do exactly the same for a
verification. In the end it doesn't matter if these influences are
taken into account before the instrument displays its reading or before
you write down the measurement result in the report.

Personally, I'd rather include these considerations in measurement
reports than adjusting the instrument for two reasons: a) it
demonstrates that you're aware of these effects and can compensate for
them, and b) as Ilya mentioned it doesn't change the instrument. Even
though nowadays I consider argument b) only half valid: It's true that
there's a certain risk of changing the instruments behaviour by turning
trim pots and the like, thereby invalidating prior knowledge about the
instruments behaviour. But purely mathematical corrections do not
change the behaviour of the instrument. There's still the issue of
comparing readings taken at different points in time: the same readings
will have different meanings before and after adjustment, while
properly calculated corrections should lead to the same result even at
different points in time.

I consider it good measurement practice to not only identify the
sources of error and their respecive contributions, but also the
uncertainties associated with these, as it clearly shows where
improvements have the most effect on your quality of measurement.

And if you implement all these corrections outside your instrument,
there's no need to buy a calibration manual as you take the instrument
as is.

HTH,
Florian

That's true, also took me some time and many discussions with
colleagues to wrap my head around the concept. Verification can be a
pretty powerful tool to save cost: at work we use it to significantly
extend calibration periods of our test equipment. And it helps keeping
measurement uncertainty in check.

Florian

Am Sat, 6 Jun 2020 07:59:42 -0500
schrieb Steve - Home steve-krull@cox.net:

Yes, and no. Unfortunately the term "calibration" is used differently
in the realms of metrology and Network Analyzers. What's typically
referred to as calibration when network analyzers are concerned is
aimed at canceling the effect of the environment (test leads
and their frequency response, time delay/phase shift), not actual
calibration of the instrument. You don't care to know how big the error
introduced by that actually is, you just want to have the instrument
substract it from the real measurement values. As you mention, the
verification kits serve as sanity checks if the "calibration" actually
was okay. But there's a number of parameters which none of the
calibration and verification kits I've seen until now addresses:
frequency accurracy is not checked, absolute power accurracy is not
checked.

Florian
DH7FET