Chirp Z transform question:
If N is the number of input data samples and M is the number of output
samples is there a limit, due to errors for M > N? If so, can the
chirp z transform be segmented or applied again with different start
and stop frequencies?

Hi: The advantages the CZT (chirp z-transform) offer are that it can
be used to evaluate the z-transform of an N-point sequence at an
arbitrarily number of points (this is your M) in the complex plane;
the name "chirp" comes from the shape of the "locus" of the set of
points you choose.  These advantages however comes with its own
"baggage:" the CZT imposes a higher computational burden.
 
As far I know, no restriction need be placed on M (the # of output
samples) except for this computational complexity that you might incur
in computing the CZT.  The linear convolution that appears in the CZT
is typically performed using FFTs whose length is completely
determined by both M and N; the higher the value of M, the more time
you would need to do the computations.

kprema