Calculating drift limits beyond the end of calibration validity

[nuno]

Is there a mathematically sound way of calculating limits on calibration drift of a (5.5 or 6.5 digits, LM399H based) DMM beyond calibration validity? Assuming it was not dropped on the floor, etc.
Let's say, if specs says +-100uV max error 1 year after calibration, is there a limit we can trust at 2 or 3 years since last calibration (trusted as much as the 100uV is trusted 1 year after calibration)? For my application it's not important that the DMM measures in spec, but more that I know a limit on its current max error. I can't find anything about this (maybe because nothing can be extrapolated?).
Thanks!

[CatalinaWOW]

Calibration intervals are a statistical estimate that all of the instruments made to that spec will remain within interval at the stated time.  It is intentionally conservative, and in most cases for older instruments is excessively conservative as they are well aged and not drifting as much as they did in their youth.  If you have calibration history on your instrument you can make an estimate specific to your instrument.  As with any extrapolation it has to taken with a grain of salt.  It is meaningful as long as nothing has changed.  That is what you are trying to capture with your "hasn't been dropped" statements, but there are state changes that can occur without physical abuse.  A capacitor leaking is one of the most common examples, but there are many others.  The good news is that many are not subtle.  You will know when they happen.

[nuno]

I have no history, it's a "new" equipment, less than 2.5 years, barely used.

[IanB]

There is no substitute for data.

I'd say if you really want confidence, you would have to have it measured by a calibration lab at regular intervals (annually?) and keep a record of the results. By plotting the results on a graph with uncertainty bands you will have some information to extrapolate from. Extrapolating from a single point (the as-new calibration) will give diverging uncertainty beyond a year since there will be no trend to base any predictions on.

[Andreas]

Hello,

unfortunately no longer on web:
Keithley had a statement that they usually do not adjust the DMMs during calibration if the deviation is within 70% of the limits.

So my conclusion would be:
If during the first year the instrument does not drift by more than 70% of the limits,
it will not drift by more than 30% of the limits during the 2nd + following years.

But without any calibration data a prediction is not possible.

with best regards

Andreas

[ap]

Having calibrated and adjusted many high precision DMMs (6+ digits), unfortunatelly the drift is not homogenous, i.e. you can e.g. have all except one within spec, and that one has drifted way beyond. If a meter was out of spec, that is what I saw most. Difting really depends on internal circuitry and so. Owning a precision DMM, and using it in a reasonable manner, also means calibration from time to time. Otherwise, what sense does it make to have one? Look for a cal lab that can do it at a reasonable cost. 150 euro with VAT or so including adjustment and report.

[nuno]

I see, thank you guys. The instrument is not mine; I also told the owner before purchase that it was just "too much" for his use, but hey, the money wasn't mine either.