# [time-nuts] HP 5370B

Mike S mikes at flatsurface.com
Sat May 10 13:36:24 EDT 2008

```At 10:43 AM 5/10/2008, Didier Juges wrote...
(corrected for top-posting)
>At 10:18 AM 5/10/2008, Mike S wrote...
> > At 09:22 AM 5/10/2008, Didier Juges wrote...
> > >I have had to argue too many times that a piece of equipment
> > with a 2dB
> > >p-p requirement on flatness was just fine when it measured 2.01dB
> on
> > >the HP network analyzer. I would not have gotten in that argument
> if
> > >the data had been 1.99dB. Go figure.
> >
> > I figure you're both wrong. If the measuring instrument is no
> > better than 1 dB absolute, you can't expect to do it - a flat
> > line measurement could represent an actual + or - 1 db, the
> > limit of the spec.

>I knew would get comments on that. I simply did not provide enough
>information. I thought most everyone who has done it would know what I
>was
>refering to.
>
>Acuracy of these instruments is not one number, it is a 3 dimensional
>chart,
>and time is the 4th dimension.

Exactly what (real world, not something made up to match the claim like
-.02 to -1.00) accuracy spec allows a reading of 2.01 to meet a spec of
max 2? Please tell us what you're referring to. I don't claim to be an
expert at math or statistics, but I just don't see it.

I understand that there is the _possibility_ that the DUT is in spec
due to measurement uncertainty, but to say it's "just fine" (i.e.
guaranteed to  be within spec) can't be claimed. Aren't you measuring
so you can guarantee the spec is met? If you say 2.01 is as acceptable
as 1.99, then what about 2.02? 2.03?...999.00?

You seem to be trying to make the point that a measurement difference
of .02 is meaningless if the accuracy is some amount greater than that.
That's a losing argument, since there has to be _some_ defined point
beyond which the spec can't be guaranteed. For your example, it might
be impossible except for a perfect measurement, as in the example I
gave. In any case, the measurement would always have to be less than
the specification. How much less would depend on the accuracy. Maybe
it's 1.50, maybe it's 1.49, but it's a pass/fail test, either the
specification is guaranteed by the measurement, or it isn't (excepting
statistical specs).

```