forked from M-Labs/artiq
1
0
Fork 0
artiq/artiq/gateware/dsp/fir.py

173 lines
5.4 KiB
Python
Raw Normal View History

from math import floor
2016-12-08 02:14:23 +08:00
from operator import add
from functools import reduce
from collections import namedtuple
2016-12-08 02:14:23 +08:00
import numpy as np
2016-12-08 02:14:23 +08:00
from migen import *
2016-12-18 04:19:46 +08:00
def halfgen4(width, n, df=1e-3):
2016-12-08 02:14:23 +08:00
"""
http://recycle.lbl.gov/~ldoolitt/halfband
params:
2016-12-08 22:30:26 +08:00
* `up` is the passband/stopband width, as a fraction of
input sampling rate
2016-12-08 02:14:23 +08:00
* `n is the order of half-band filter to generate
returns:
* `a` is the full set of FIR coefficients, `4*n-1` long.
implement wisely.
"""
npt = n*40
2016-12-08 22:30:26 +08:00
wmax = 2*np.pi*width
2016-12-08 02:14:23 +08:00
wfit = (1 - np.linspace(0, 1, npt)[:, None]**2)*wmax
target = .5*np.ones_like(wfit)
basis = np.cos(wfit*np.arange(1, 2*n, 2))
weight = np.ones_like(wfit)
2016-12-18 04:19:46 +08:00
f0 = None
2016-12-08 02:14:23 +08:00
for i in range(40):
l = np.linalg.pinv(basis*weight)@(target*weight)
2016-12-18 04:19:46 +08:00
err = np.fabs(basis@l - .5)
f = np.max(err)/np.mean(err)
if f0 and (f0 - f)/(f0 + f) < df/2:
break
f0 = f
weight[err > (1 - df)*np.max(err)] *= 1 + 1.5/(i + 11)
2016-12-08 02:14:23 +08:00
a = np.c_[l, np.zeros_like(l)].ravel()[:-1]
a = np.r_[a[::-1], 1, a]/2
return a
_Widths = namedtuple("_Widths", "A B P")
2016-12-08 02:14:23 +08:00
_widths = {
"DSP48E1": _Widths(25, 18, 48),
}
2016-12-08 20:05:13 +08:00
class ParallelFIR(Module):
"""Full-rate parallelized finite impulse response filter.
Tries to use transposed form as much as possible.
:param coefficients: tap coefficients (normalized to 1.),
increasing delay.
2016-12-08 20:05:13 +08:00
:param parallelism: number of samples per cycle.
:param width: bit width of input and output.
:param arch: architecture (default: "DSP48E1").
2016-12-08 20:05:13 +08:00
"""
def __init__(self, coefficients, parallelism, width=16,
2017-06-29 17:24:56 +08:00
arch="DSP48E1", cull_delays=()):
2016-12-08 20:05:13 +08:00
self.width = width
self.parallelism = p = parallelism
n = len(coefficients)
# input and output: old to new, decreasing delay
2016-12-08 20:05:13 +08:00
self.i = [Signal((width, True)) for i in range(p)]
self.o = [Signal((width, True)) for i in range(p)]
2017-06-29 17:33:19 +08:00
self.latency = (n + 1)//2//p + 2
w = _widths[arch]
2016-12-08 20:05:13 +08:00
c_max = max(abs(c) for c in coefficients)
c_shift = bits_for(floor((1 << w.B - 2) / c_max))
self.coefficients = cs = [int(round(c*(1 << c_shift)))
for c in coefficients]
2017-06-13 02:07:23 +08:00
assert max(bits_for(c) for c in cs) <= w.B
2017-06-29 18:08:43 +08:00
max_out = sum(abs(c)*(1 << w.A - 1) for c in cs)
assert max_out <= (1 << w.P - 1) - 1, (bits_for(max_out), w)
2016-12-08 20:05:13 +08:00
###
# Delay line: increasing delay
2017-06-29 01:09:21 +08:00
x = [Signal((w.A, True), reset_less=True) for _ in range(n + p - 1)]
2017-06-13 02:07:23 +08:00
for xi, xj in zip(x, self.i[::-1]):
2017-06-29 01:55:15 +08:00
self.comb += xi.eq(xj)
for xi, xj in zip(x[len(self.i):], x):
self.sync += xi.eq(xj)
2016-12-08 20:05:13 +08:00
for delay in range(p):
2017-06-29 01:09:21 +08:00
o = Signal((w.P, True), reset_less=True)
2017-06-29 17:33:19 +08:00
self.sync += self.o[delay].eq(o >> c_shift)
2016-12-09 00:00:39 +08:00
# Make products
2017-06-29 17:24:56 +08:00
tap = delay
for i, c in enumerate(cs):
2016-12-08 20:05:13 +08:00
# simplify for halfband and symmetric filters
if not c or c in cs[:i]:
2016-12-08 20:05:13 +08:00
continue
js = [j + p - 1 for j, cj in enumerate(cs) if cj == c]
m = Signal.like(o)
o0, o = o, Signal.like(o)
q = Signal.like(x[0])
2017-06-29 17:24:56 +08:00
if tap + p <= js[0]:
self.sync += o0.eq(o + m)
2017-06-29 17:24:56 +08:00
tap += p
else:
self.comb += o0.eq(o + m)
2017-06-29 17:24:56 +08:00
assert min(js) - tap >= 0
2017-06-29 18:08:43 +08:00
js = [j for j in js
if (p - 1 - j - tap) % p not in cull_delays]
2017-06-29 17:24:56 +08:00
if not js:
continue
self.comb += q.eq(reduce(add, [x[j - tap] for j in js]))
self.sync += m.eq(c*q)
# symmetric rounding
2017-06-29 17:24:56 +08:00
if c_shift > 1 and delay not in cull_delays:
2017-06-29 01:55:15 +08:00
self.comb += o.eq((1 << c_shift - 1) - 1)
class FIR(ParallelFIR):
def __init__(self, *args, **kwargs):
super().__init__(self, *args, parallelism=1, **kwargs)
self.i = self.i[0]
self.o = self.o[0]
2016-12-08 22:30:26 +08:00
def halfgen4_cascade(rate, width, order=None):
"""Generate coefficients for cascaded half-band filters.
2017-06-29 18:08:43 +08:00
Coefficients are normalized to a gain of two per stage to compensate for
the zero stuffing.
2016-12-08 22:30:26 +08:00
:param rate: upsampling rate. power of two
:param width: passband/stopband width in units of input sampling rate.
:param order: highest order, defaults to :param:`rate`"""
if order is None:
order = rate
coeff = []
p = 1
while p < rate:
p *= 2
coeff.append(2*halfgen4(width*p/rate/2, order*p//rate))
2016-12-08 22:30:26 +08:00
return coeff
class ParallelHBFUpsampler(Module):
"""Parallel, power-of-two, half-band, cascading upsampler.
Coefficients should be normalized to overall gain of 2
(highest/center coefficient being 1)."""
def __init__(self, coefficients, width=16, **kwargs):
self.parallelism = 1 # accumulate
self.latency = 0 # accumulate
2016-12-08 22:30:26 +08:00
self.width = width
self.i = Signal((width, True))
###
i = [self.i]
for coeff in coefficients:
self.parallelism *= 2
2017-06-29 17:24:56 +08:00
hbf = ParallelFIR(coeff, self.parallelism, width=width,
cull_delays={0}, **kwargs)
2016-12-08 22:30:26 +08:00
self.submodules += hbf
self.comb += [a.eq(b) for a, b in zip(hbf.i[1::2], i)]
2016-12-08 22:30:26 +08:00
i = hbf.o
self.latency += hbf.latency
self.o = i