> I am  measuring the time period of an accelerating spindle which
> can reach anwhere from 600rpm to 25000rpm in a matter of
> seconds

Even the fastest speed is still quite a long time for the PIC @ 4MHz

1,000,000 / (25,000 / 60) = 2400us / rev or 2400 instruction cycles

and resolution is being reduced by the /8 pre-scaler

> Table will have about 1000 lines (not app 2000 as my original post!)

Are 1000 really necessary ? Why not 256 ? (bearing in mind the p/s)

> The idea being TMR1 will give a more accurate value without rolling
> over

> Using TMRO does not give me accurate enough steps at low speed

How are you detecting the revs ? If using INTF for example and given
the length of time available to process INTF, it would be quite possible
to use TMR0 and count any rollovers in RAM (which obviously becomes
your high byte to use with PCLATH). As both timers increment on the
internal clock, in this application TMR0 is as "accurate" as TMR1

--
http://www.piclist.com hint: The list server can filter out subtopics
(like ads or off topics) for you. See http://www.piclist.com/#topics