[time-nuts] Short-Term Stability
phk at phk.freebsd.dk
Thu Mar 16 11:36:56 EST 2006
In message <006001c6490d$abf4fdc0$0201a8c0 at n4fs>, "Mike Feher" writes:
>There are several
>mechanisms that cause the spurious, including the phase and amplitude
>truncation due to the finite amount of bits,
A researcher at Bruel&Kjaer wrote an IEEE paper about how filtering
the rounding errors on integer FIR filters could reduce spurious
by a LOT of dB for typical filters, his canonical example is 20dB
on a 69 tap FIR:
The same technique applies to DDS, but is usually too expensive to
perform in real-time. For fixed frequencies however, it is possible
to do the math beforehand, and add a "rounding" bit stream which
will be added to the accumulator along with delta-omega.
I've lost the reference to this however.
Poul-Henning Kamp | UNIX since Zilog Zeus 3.20
phk at FreeBSD.ORG | TCP/IP since RFC 956
FreeBSD committer | BSD since 4.3-tahoe
Never attribute to malice what can adequately be explained by incompetence.
More information about the time-nuts