Recall the standard interpolation procedure illustrated in
Figure 1.
Note that this procedure is computationally inefficient
because the lowpass filter operates on a sequence that is
mostly composed of zeros. Through the use of the Noble
identities, it is possible to rearrange the preceding block
diagram so that operations on zero-valued samples are avoided.
In order to apply the Noble identity for interpolation, we must
transform
Hz
H
z
into its upsampled polyphase components
H
p
zL
H
p
z
L
,
p=0…L−1
p
0
…
L
1
.
Hz=∑nnhnz−n=∑kk∑p=0L−1hkL+pz−(kL+p)
H
z
n
n
h
n
z
n
k
k
p
0
L
1
h
k
L
p
z
k
L
p
(1)
via
k≔⌊nL⌋
≔
k
n
L
,
p≔〈n〉L
≔
p
n
L
Hz=∑p=0L−1∑kk
h
p
kz−(kL)z−p
H
z
p
0
L1
k
k
h
p
k
z
k
L
z
p
(2)
via
h
p
k≔hkL+p
≔
h
p
k
h
k
L
p
Hz=∑p=0L−1
H
p
zLz−p
H
z
p
0
L1
H
p
z
L
z
p
(3)
Above,
⌊·⌋·
denotes the floor operator and
〈·〉M
·
M
the modulo-
MM operator. Note that the
pthpth
polyphase filter
h
p
k
h
p
k
is
constructed by downsampling the "master filter"
hn
h
n
at offset
pp. Using the
unsampled polyphase components, the
Figure 1 diagram can be redrawn as in
Figure 2.
Applying the Noble identity for interpolation to Figure 3 yields Figure 2. The ladder of upsamplers and delays on the
right below accomplishes a form of
parallel-to-serial conversion.