[time-nuts] What is "accuracy"? (newbie timenut, hi folks!)

[Magnus Danielson]

Bill,

On 05/06/2016 06:22 AM, wb6bnq wrote:
> Hello Belinda,
>
> First off there is no such thing as accuracy, in and of itself.  I know
> many people on this list will call me on that, but accuracy requires a
> point of reference.  With regard to frequency that reference point has
> been defined by some World committee as a certain number of oscillations
> in a Cesium atom controlled within a specific set of conditions.

We do have an agreed point of reference, the SI second as realized in 
the forms of TAI and UTC we can get traceability of measurements all the 
way down to parts in 1E-16 which is where the currently best clocks are, 
and as a constellation it is even better. It's by far the best 
realization of any SI units, and because of that, more and more of the 
units becomes re-defined in terms of the SI second and the precision and 
accuracy of those will improve. At the same time redefinition of the SI 
second is being discussed, as the best optical clocks is now pushing 
into the low parts of 1E-18 and we soon expect 1E-19 numbers to be reported.

> The 100 ppm statement is talking about a change in frequency due to
> temperature,  The typical 100 ppm statement is saying for every change
> of 1 degree (usually "C") the oscillator (or other components such as
> resistors, capacitors, etc.) will shift in value by a worst case of 100
> ppm (parts per million).  This has nothing to do with accuracy except
> that it would not be considered accurate relative to a reference point.
> What it does address is specifically the stability, but is not the only
> condition affecting stability.

Yes and no. 100 ppm oscillators have pretty bad temperature coefficient 
(tempco) because they don't need to be better. There is also variation 
in their manufacturing that is not adjusted for, so that's why they are 
cheap.

> With respect to accuracy and stability, they are not related.  That is
> to say you could have extreme stability (say parts in 10 to the minus
> 21) and it could be way off from the recognized standard reference.  In
> the other direction you could have something that is adjusted to be
> precisely in agreement with the reference standard but will only hold
> that value for a very brief period of time.  The first case is a very
> good (and quite expensive) oscillator and the second example is a poor
> (and not expensive) oscillator.
>
> With regard to precision, the best example would be shooting at a target
> and how tight the grouping is maintained.  The tighter the grouping the
> better the precision.  You could have a tightly well defined small group
> of holes from the bullets but they could be anywhere on the paper
> target.  The only time you have accuracy (with respect to the shooting)
> is if all the bullets were directly in the bulls eye (center of
> target).  And if you can consisterntly repeat hitting just the bulls eye
> then you have stability.

This is one of the pictures often shown. The more entertaining is a play 
with Willhem Tell. :)

Cheers,
Magnus