Polyphase Filter

12: Polyphase Filters

Heavy Lowpass ltering
Maximum Decimation
Frequency
Polyphase decomposition
Downsampled Polyphase
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
DSP and Digital Filters (2013-3816) Polyphase Filters: 12 1 / 10
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
Filter Specication:
Sample Rate: 20 kHz
Passband edge: 100 Hz (
1
= 0.03)
Stopband edge: 300 Hz (
2
= 0.09)
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
Filter Specication:
Sample Rate: 20 kHz
1
= 0.03)
2
= 0.09)
Passband ripple: 0.05 dB ( = 0.006)
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
Filter Specication:
Sample Rate: 20 kHz
1
= 0.03)
2
= 0.09)
Stopband Gain: 80 dB ( = 0.0001)
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
Filter Specication:
Sample Rate: 20 kHz
1
= 0.03)
2
= 0.09)
This is an extreme lter because the cutoff frequency is only 1% of the
Nyquist frequency.
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
Filter Specication:
Sample Rate: 20 kHz
1
= 0.03)
2
= 0.09)
Nyquist frequency.
Symmetric FIR Filter:
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
Filter Specication:
Sample Rate: 20 kHz
1
= 0.03)
2
= 0.09)
Nyquist frequency.
Design with Remez-exchange algorithm
Order = 360
0 1 2 3
-80
-60
-40
-20
0
M=360
(rad/s)
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
Filter Specication:
Sample Rate: 20 kHz
1
= 0.03)
2
= 0.09)
Nyquist frequency.
Design with Remez-exchange algorithm
Order = 360
0 1 2 3
-80
-60
-40
-20
0
M=360
(rad/s)
0 0.05 0.1
-80
-60
-40
-20
0
1

2
(rad/s)
Maximum Decimation Frequency
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
If a lter passband occupies only a small fraction
of [0, ], we can downsample then upsample
without losing information.
0 1 2 3
-60
-40
-20
0
1

2

Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
0 1 2 3
-60
-40
-20
0
1

2

Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
0 1 2 3
-60
-40
-20
0
1

2
Downsample: aliased components at offsets of

2
K
are almost zero because of H(z)
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
0 1 2 3
-60
-40
-20
0
1

2

2
K
Upsample: Images spaced at
2
K
can be
removed using another low pass lter
0 1 2 3
-60
-40
-20
0
= 2 /4

Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
0 1 2 3
-60
-40
-20
0
1

2

2
K
2
K
can be
0 1 2 3
-60
-40
-20
0
= 2 /4
To avoid aliasing in the passband, we need

2
K

2

1
K
2
1
+
2
0 1 2 3
-60
-40
-20
0
= 2 /7

Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
0 1 2 3
-60
-40
-20
0
1

2

2
K
2
K
can be
0 1 2 3
-60
-40
-20
0
= 2 /4

2
K

2

1
K
2
1
+
2
Normally place the centre of the transition band
at the intermediate Nyquist frequency.
0 1 2 3
-60
-40
-20
0
= 2 /7

Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
0 1 2 3
-60
-40
-20
0
1

2

2
K
2
K
can be
0 1 2 3
-60
-40
-20
0
= 2 /4

2
K

2

1
K
2
1
+
2
Normally place the centre of the transition band
at the intermediate Nyquist frequency.
0 1 2 3
-60
-40
-20
0
= 2 /7
We must add a lowpass lter to remove the images:

Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
For our lter: original Nyquist frequency = 10 kHz and transition band
centre is at 200 Hz so we can use K = 50.
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
For convenience, zero-pad h[n] to order M
= RK 1.
M = 360
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
= RK 1.
M = 360 R = 8
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
= RK 1.
M = 360 R = 8 M
= 399
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
= RK 1.
M = 360 R = 8 M
= 399
H(z) =
m=0
h[m]z
m
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
= RK 1.
M = 360 R = 8 M
= 399
H(z) =
m=0
h[m]z
m
=
K1
m=0
h[m]z
m
+
2K1
m=K
h[m]z
m
+
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
= RK 1.
M = 360 R = 8 M
= 399
H(z) =
m=0
h[m]z
m
=
K1
m=0
h[m]z
m
+
2K1
m=K
h[m]z
m
+
=
K1
m=0
R1
r=0
h[m+Kr]z
mKr
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
= RK 1.
M = 360 R = 8 M
= 399
H(z) =
m=0
h[m]z
m
=
K1
m=0
h[m]z
m
+
2K1
m=K
h[m]z
m
+
=
K1
m=0
R1
r=0
h[m+Kr]z
mKr
=
K1
m=0
z
m

R1
r=0
h
m
[r]z
Kr
where h
m
[r] = h[m+Kr]
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
= RK 1.
M = 360 R = 8 M
= 399
H(z) =
m=0
h[m]z
m
=
K1
m=0
h[m]z
m
+
2K1
m=K
h[m]z
m
+
=
K1
m=0
R1
r=0
h[m+Kr]z
mKr
=
K1
m=0
z
m

R1
r=0
h
m
[r]z
Kr
where h
m
[r] = h[m+Kr]
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
= RK 1.
M = 360 R = 8 M
= 399
H(z) =
m=0
h[m]z
m
=
K1
m=0
h[m]z
m
+
2K1
m=K
h[m]z
m
+
=
K1
m=0
R1
r=0
h[m+Kr]z
mKr
=
K1
m=0
z
m

R1
r=0
h
m
[r]z
Kr
where h
m
[r] = h[m+Kr]
Example:
h
0
[r] =
h[0] h[50] h[350]
h
1
[r] =
h[1] h[51] h[351]
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
= RK 1.
M = 360 R = 8 M
= 399
H(z) =
m=0
h[m]z
m
=
K1
m=0
h[m]z
m
+
2K1
m=K
h[m]z
m
+
=
K1
m=0
R1
r=0
h[m+Kr]z
mKr
=
K1
m=0
z
m

R1
r=0
h
m
[r]z
Kr
where h
m
[r] = h[m+Kr]
Example:
h
0
[r] =
h[0] h[50] h[350]
h
1
[r] =
h[1] h[51] h[351]
This is a polyphase implementation of the lter H(z)

Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
= RK 1.
M = 360 R = 8 M
= 399
H(z) =
m=0
h[m]z
m
=
K1
m=0
h[m]z
m
+
2K1
m=K
h[m]z
m
+
=
K1
m=0
R1
r=0
h[m+Kr]z
mKr
=
K1
m=0
z
m

R1
r=0
h
m
[r]z
Kr
where h
m
[r] = h[m+Kr]
Example:
h
0
[r] =
h[0] h[50] h[350]
h
1
[r] =
h[1] h[51] h[351]
This is a polyphase implementation of the lter H(z)

Split H(z) into K lters each of order R 1
Downsampled Polyphase Filter
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
H(z) is low pass so we downsample
its output by K without aliasing.
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
The number of multiplications per input
sample is M + 1 = 361.
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
Note that H
0
(z) to H
10
(z) are of order
7 while the rest are of order 6.
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
Note that H
0
(z) to H
10
(z) are of order
Using the Noble identities, we can move
the resampling back through the adders
and lters. H
m
(z
K
) turns into H
m
(z)
at a lower sample rate.
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
Note that H
0
(z) to H
10
(z) are of order
and lters. H
m
(z
K
) turns into H
m
(z)
We still perform 361 multiplications but
now only once for every K input
samples.
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
Note that H
0
(z) to H
10
(z) are of order
and lters. H
m
(z
K
) turns into H
m
(z)
We still perform 361 multiplications but
now only once for every K input
samples.
Multiplications per input sample = 7.2 (down by a factor of 50 ) but v[n]
has the wrong sample rate ().
Polyphase Upsampler
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
To restore sample rate: upsample and
then lowpass lter to remove images
Polyphase Upsampler
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
We can use the same lowpass lter,
H(z), in polyphase form:
K1
m=0
z
m

R1
r=0
h
m
[r]z
Kr
Polyphase Upsampler
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
K1
m=0
z
m

R1
r=0
h
m
[r]z
Kr
Polyphase Upsampler
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
K1
m=0
z
m

R1
r=0
h
m
[r]z
Kr
This time we put the delay z
m
after
the lters.
Polyphase Upsampler
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
K1
m=0
z
m

R1
r=0
h
m
[r]z
Kr
m
after
the lters.
Multiplications per output sample = 361
Polyphase Upsampler
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
K1
m=0
z
m

R1
r=0
h
m
[r]z
Kr
m
after
the lters.
the resampling forwards through the
lters. H
m
(z
K
) turns into H
m
(z) at a
lower sample rate.
Polyphase Upsampler
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
K1
m=0
z
m

R1
r=0
h
m
[r]z
Kr
m
after
the lters.
the resampling forwards through the
lters. H
m
(z
K
) turns into H
m
(z) at a
lower sample rate.
Multiplications per output sample = 7.2
(down by a factor of 50 ).
Complete Filter
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
The overall system implements:
Complete Filter
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
Need an extra gain of K to compensate for the downsampling energy loss.
Complete Filter
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
Filtering at downsampled rate requires 14.4 multiplications per input
sample (7.2 for each lter). Reduced by
K
2
from the original 361.
Complete Filter
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
K
2
H(e
j
) reaches 10 dB at the
downsampler Nyquist frequency of

K
.
0 0.05 0.1
-80
-60
-40
-20
0
1

2
/50
(rad/s)
Complete Filter
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
K
2
H(e
j

K
.
Spectral components >

K
will be aliased
down in frequency in V (e
j
).
0 0.05 0.1
-80
-60
-40
-20
0
1

2
/50
(rad/s)
Complete Filter
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
K
2
H(e
j

K
.

K
will be aliased
j
).
For V (e
j
), passband gain (blue curve)
follows the same curve as X(e
j
).
0 0.05 0.1
-80
-60
-40
-20
0
1

2
/50
(rad/s)
0 1 2 3
-80
-60
-40
-20
0
1
(downsampled)
Complete Filter
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
K
2
H(e
j

K
.

K
will be aliased
j
).
For V (e
j
j
).
Noise arises from K aliased spectral
intervals.
0 0.05 0.1
-80
-60
-40
-20
0
1

2
/50
(rad/s)
0 1 2 3
-80
-60
-40
-20
0
1
(downsampled)
Complete Filter
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
K
2
H(e
j

K
.

K
will be aliased
j
).
For V (e
j
j
).
Noise arises from K aliased spectral
intervals.
Unit white noise in X(e
j
) gives passband
noise oor at 69 dB (red curve) even
though stop band ripple is below 83 dB
(due to K 1 aliased stopband copies).
0 0.05 0.1
-80
-60
-40
-20
0
1

2
/50
(rad/s)
0 1 2 3
-80
-60
-40
-20
0
1
(downsampled)
Upsampler Implementation
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
We can represent the upsampler
compactly using a commutator.
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
H
0
(z) comprises a sequence of
7 delays, 7 adders and 8 gains.
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
H
0
We can share the delays between
all 50 lters.
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
H
0
all 50 lters.
We can also share the gains and
adders between all 50 lters and
use commutators to switch the
coefcients.
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
H
0
all 50 lters.
We can also share the gains and
adders between all 50 lters and
use commutators to switch the
coefcients.
We now need 7 delays, 7 adders and 8 gains for the entire lter.
Downsampler Implementation
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
We can again use a commutator.
The outputs from all 50 lters are
added together to form v[i].
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
We use the transposed form of
H
m
(z) because this will allow us
to share components.
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
H
m
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
H
m
We can sum the outputs of the
gain elements using an
accumulator which sums blocks
of K samples.
w[i] =
K1
r=0
u[Ki r]
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
H
m
of K samples.
Now we can share all the
components and use
commutators to switch the gain
coefcients.
w[i] =
K1
r=0
u[Ki r]
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
H
m
of K samples.
Now we can share all the
components and use
commutators to switch the gain
coefcients.
We need 7 delays, 7 adders, 8
gains and 8 accumulators in total.
w[i] =
K1
r=0
u[Ki r]
Summary
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
Filtering should be performed at the lowest possible sample rate
reduce lter computation by K
actual saving is only
K
2
because you need a second lter
downsampled Nyquist frequency max (
passband
) +

2
Summary
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
K
2
passband
) +

2
Polyphase decomposition: split H(z) as

K1
m=0
z
m
H
m
(z
K
)
each H
m
(z
K
) can operate on subsampled data
combine the ltering and down/up sampling
Summary
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
K
2
passband
) +

2

K1
m=0
z
m
H
m
(z
K
)
each H
m
(z
K
Noise oor is higher because it arises from K spectral intervals that
are aliased together by the downsampling.
Summary
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
K
2
passband
) +

2

K1
m=0
z
m
H
m
(z
K
)
each H
m
(z
K
Share components between the K lters
multiplier gain coefcients switch at the original sampling rate
need a new component: accumulator/downsampler (K : )
Summary
Maximum Decimation
Frequency
Filter
Polyphase Upsampler
Complete Filter
Upsampler
Implementation
Downsampler
Implementation
Summary
K
2
passband
) +

2

K1
m=0
z
m
H
m
(z
K
)
each H
m
(z
K
Share components between the K lters
multiplier gain coefcients switch at the original sampling rate
need a new component: accumulator/downsampler (K : )
For further details see Harris 5.

Polyphase Filter

Hochgeladen von

Dokumentinformationen

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Polyphase Filter

Hochgeladen von

Copyright:

Verfügbare Formate

12: Polyphase Filters

12: Polyphase Filters

Maximum Decimation Frequency

Maximum Decimation Frequency

Downsample: aliased components at offsets of

Downsample: aliased components at offsets of

Maximum Decimation Frequency

Downsample: aliased components at offsets of

To avoid aliasing in the passband, we need

Maximum Decimation Frequency

Downsample: aliased components at offsets of

To avoid aliasing in the passband, we need

Maximum Decimation Frequency

Downsample: aliased components at offsets of

To avoid aliasing in the passband, we need

We must add a lowpass lter to remove the images:

h[0] h[50] h[350]

h[1] h[51] h[351]

h[0] h[50] h[350]

h[1] h[51] h[351]

This is a polyphase implementation of the lter H(z)

h[0] h[50] h[350]

h[1] h[51] h[351]

This is a polyphase implementation of the lter H(z)

Das könnte Ihnen auch gefallen