John Payson wrote :

>An alternative which will always give maximal accuracy, but which
requires
>a little more calculation effort, is to time as many pulses as will fit
in
>a second; then take the time from the first to the last and divide that
>into the number of event intervals received.  If the signal is, e.g.,
>100.7Hz you may receive 101 pulses, with 993,050us between the first and
>the last.  Dividing 0.993050 into 100 event intervals yields 100.700Hz
(if
>the time is accurate to within 10us).  This approach is often far more
acc-
>urate than pulse-counting or period-measurement alone.

NOTE: When using this scheme:

The beginning of the 1 second period (window) should be synchronous with
(begin at the same time as) the event (pulse) and only the subsequent
events are to be counted in the 1 second period.

 If the 1 second window begins more than one-half event period after the
event, then one too many events will fit into the 1 second window.

SO: the 1 second window must start less than one-half event period after
the event. The easiest way to ensure this is to start the window at the
event.

Roger
********************************
Certified Jowett Nut
(If you have to ask..you are better off not knowing about Jowetts)