[time-nuts] R&S XSRM Rubidium Standard

jimlux jimlux at earthlink.net
Sun Sep 17 14:00:44 EDT 2017

On 9/17/17 9:42 AM, KA2WEU--- via time-nuts wrote:
> Simply call it " Make it to meet specification", N1UL
> In a message dated 9/17/2017 12:39:32 P.M. Eastern Daylight Time,
> magnus at rubidium.dyndns.org writes:
> Hi,
> The word "calibration" is overloaded with multiple  meanings, and
> incompatible too.
> "calibration" can thus imply different things.
> I  regularly see people use these terms inconsistently. That people get
> disappointed when they get the wrong thing is to be  expected.
> Cheers,
> Magnus

Indeed - in my business, we have to be careful about the word "test" vs, 
say, "characterization".  A "test" has a pass/fail criteria associated 
with it, while a "characterization" might just need recording what the 
value is.  "tests" have institutional requirements for witnessing, etc. 
that "characterizations" do not, for instance.

And this gets into the whole "knowledge" vs "control" - I may not care 
if a XO changes frequency 10ppm over temperature, as long as it's 
repeatable and I can know the actual frequency within 0.5 ppm.

As Magnus points out, "calibration" can mean many different things, and 
some of them are historically derived.  Back in the day, there may have 
been some sort of physical adjustment required to, for instance, set the 
scale factor of a display - but now, it's "calibrated by design", and 
the "calibration process" (as in "sending it to the cal lab" to get a 
"cal cert") is more about verification that it's not "broken".

I also used to (and still do) get bent out of shape when you'd send 
something like a power supply out for cal, and it would come back with 
some problem (like a non-functioning pilot light).  And the cal lab 
would say: "we checked the output voltage and it is within spec".  Yeah, 
but isn't "proper function of all features" covered - and the response 
would be "no, it is not, we don't see 'verify function of pilot light' 
on the cal procedure we have"

ANd then there's the "calibration/validation" phase of instruments in 
space - that's a completely different kind of thing, more akin to 
characterization.. The 'val' part is yes, verifying that the instrument 
is working as designed,but the 'cal' part is more about relating 
instrument measurements to some other reference, and from that, relating 
it to some physical property of interest.  And that can happen at many 

A spaceborne scatterometer to measure winds can be cal/val at
1) Does the instrument work, and are the instrumental effects accounted 
for - if it measure a particular backscatter cross section, is that what 
the backscatter cross section really is?
2) Does the "retrieval model function" that turns backscatter 
measurements into wind speed and direction work?

And particularly for #2, a lot of it is inferential - comparing one 
model against another, since there's not really "ground truth" available.

More information about the time-nuts mailing list