FIR Digital Filter Design
FIR Fractional Delay Filter Design
Example 2: Rational ApproximationSearch Spectral Audio Signal Processing
Would you like to be notified by email when Julius Orion Smith III publishes a new entry into his blog?
Instead of raising
to the non-integer power
,
suppose we express
as
and compute the frequency response
as
? Perhaps working with integer powers and
integer root extractions will be better behaved.
Generalizing Fig.B.28 of the previous section
(page
), the wrapped phase response of
, for
any integer
, consists of N negatively sloped linear segments
(like those in Fig.B.28), with one segment always centered
about
, and with each segment traversing the vertical range
to
from left to right. The slope of each segment is
radians per radian.
Now, when we take the
th root of
to get
, we see
that the phase of
is divided by
. Thus, the phase
response of
will still have
linear segments, but now
traversing the vertical range
to
. In our present
example, we take the square root of
to get
to obtain the phase response shown in Fig.B.25b on
page
. This phase response is what is computed by the
matlab expression angle(sqrt(exp(-j*9*omega*T))). As can be
seen in the plot, there are now two bandpass regions in which
the sign is inverted by the phase jumps of
radians. This is even
worse than the result of Example 1.
