mirror of
https://github.com/hb9fxq/gr-digitalhf
synced 2024-12-22 23:19:59 +00:00
342 lines
16 KiB
Python
342 lines
16 KiB
Python
## -*- python -*-
|
|
|
|
from __future__ import print_function
|
|
import numpy as np
|
|
import common
|
|
from digitalhf.digitalhf_swig import viterbi27
|
|
|
|
## ---- Walsh-8 codes -----------------------------------------------------------
|
|
WALSH8 = np.array([[0,0,0,0, 0,0,0,0], # 0 - 000
|
|
[0,1,0,1, 0,1,0,1], # 1 - 001
|
|
[0,0,1,1, 0,0,1,1], # 2 - 010
|
|
[0,1,1,0, 0,1,1,0], # 3 - 011
|
|
[0,0,0,0, 1,1,1,1], # 4 - 100
|
|
[0,1,0,1, 1,0,1,0], # 5 - 010
|
|
[0,0,1,1, 1,1,0,0], # 6 - 011
|
|
[0,1,1,0, 1,0,0,1]], # 7 - 111
|
|
dtype=np.uint8)
|
|
|
|
FROM_WALSH8 = -np.ones(256, dtype=np.int8)
|
|
for i in range(8):
|
|
FROM_WALSH8[np.packbits(WALSH8[i][:])[0]] = i
|
|
|
|
## ---- Walsh-4 codes -----------------------------------------------------------
|
|
WALSH4 = np.array([[0,0,0,0], # 0 - 00
|
|
[0,1,0,1], # 1 - 01
|
|
[0,1,1,0], # 3 - 11 modified gray coding!
|
|
[0,0,1,1]], # 2 - 10 modified gray coding!
|
|
dtype=np.uint8)
|
|
|
|
FROM_WALSH4 = -np.ones(256, dtype=np.int8)
|
|
for i in range(4):
|
|
FROM_WALSH4[np.packbits(WALSH4[i][:])[0]] = i
|
|
|
|
## ---- tri-bit codes -----------------------------------------------------------
|
|
TRIBIT = np.zeros((8,32), dtype=np.uint8)
|
|
for i in range(8):
|
|
TRIBIT[i][:] = np.concatenate([WALSH8[i][:] for j in range(4)])
|
|
|
|
## ---- tri-bit scramble sequence for preamble ----------------------------------
|
|
TRIBIT_SCRAMBLE = np.array(
|
|
[7,4,3,0,5,1,5,0,2,2,1,1,5,7,4,3,5,0,2,6,2,1,6,2,0,0,5,0,5,2,6,6],
|
|
dtype=np.uint8)
|
|
|
|
## ---- preamble symbols ---------------------------------------------------------
|
|
D1=D2=C1=C2=C3=0 ## not known
|
|
PRE_SYMBOLS = common.n_psk(2, np.concatenate(
|
|
[TRIBIT[i][:] for i in [0,1,3,0,1,3,1,2,0,D1,D2,C1,C2,C3,0]]))
|
|
PRE_SYMBOLS[9*32:14*32] = 0
|
|
|
|
## ---- preamble scramble symbols ------------------------------------------------
|
|
PRE_SCRAMBLE = common.n_psk(8, np.concatenate([TRIBIT_SCRAMBLE for _ in range(15)]))
|
|
|
|
## ---- data scrambler -----------------------------------------------------------
|
|
class ScrambleData(object):
|
|
"""data scrambling sequence generator"""
|
|
def __init__(self):
|
|
self.reset()
|
|
|
|
def reset(self):
|
|
self._state = 0xBAD
|
|
self._counter = 0
|
|
|
|
def next(self):
|
|
if self._counter == 160:
|
|
self.reset()
|
|
for _ in range(8):
|
|
self._advance()
|
|
self._counter += 1
|
|
return self._state&7
|
|
|
|
def _advance(self):
|
|
msb = self._state>>11
|
|
self._state = (self._state<<1)&4095
|
|
if msb:
|
|
self._state ^= 0x053
|
|
return self._state
|
|
|
|
## ---- constellatios -----------------------------------------------------------
|
|
BPSK=np.array(zip(np.exp(2j*np.pi*np.arange(2)/2), [0,1]), common.CONST_DTYPE)
|
|
QPSK=np.array(zip(np.exp(2j*np.pi*np.arange(4)/4), [0,1,3,2]), common.CONST_DTYPE)
|
|
PSK8=np.array(zip(np.exp(2j*np.pi*np.arange(8)/8), [0,1,3,2,6,7,5,4]), common.CONST_DTYPE)
|
|
|
|
## ---- constellation indices ---------------------------------------------------
|
|
MODE_BPSK=0
|
|
MODE_QPSK=1
|
|
MODE_8PSK=2
|
|
|
|
## ---- mode definitions --------------------------------------------------------
|
|
MODE = [[{} for _ in range(8)] for _ in range(8)]
|
|
MODE[7][6] = {'bit_rate':4800, 'ci':MODE_8PSK, 'interleaver':['N', 1, 1], 'unknown':32,'known':16, 'nsymb': 1, 'coding_rate': 'n/a', 'repeat': 1}
|
|
MODE[7][7] = {'bit_rate':2400, 'ci':MODE_8PSK, 'interleaver':['S', 40, 72], 'unknown':32,'known':16, 'nsymb': 1, 'coding_rate': '1/2', 'repeat': 1}
|
|
|
|
MODE[6][4] = {'bit_rate':2400, 'ci':MODE_8PSK, 'interleaver':['S', 40, 72], 'unknown':32,'known':16, 'nsymb': 1, 'coding_rate': '1/2', 'repeat': 1}
|
|
MODE[4][4] = {'bit_rate':2400, 'ci':MODE_8PSK, 'interleaver':['L', 40,576], 'unknown':32,'known':16, 'nsymb': 1, 'coding_rate': '1/2', 'repeat': 1}
|
|
|
|
MODE[6][5] = {'bit_rate':1200, 'ci':MODE_QPSK, 'interleaver':['S', 40, 36], 'unknown':20,'known':20, 'nsymb': 1, 'coding_rate': '1/2', 'repeat': 1}
|
|
MODE[4][5] = {'bit_rate':1200, 'ci':MODE_QPSK, 'interleaver':['L', 40,288], 'unknown':20,'known':20, 'nsymb': 1, 'coding_rate': '1/2', 'repeat': 1}
|
|
|
|
MODE[6][6] = {'bit_rate': 600, 'ci':MODE_BPSK, 'interleaver':['S', 40, 18], 'unknown':20,'known':20, 'nsymb': 1, 'coding_rate': '1/2', 'repeat': 1}
|
|
MODE[4][6] = {'bit_rate': 600, 'ci':MODE_BPSK, 'interleaver':['L', 40,144], 'unknown':20,'known':20, 'nsymb': 1, 'coding_rate': '1/2', 'repeat': 1}
|
|
|
|
MODE[6][7] = {'bit_rate': 300, 'ci':MODE_BPSK, 'interleaver':['S', 40, 18], 'unknown':20,'known':20, 'nsymb': 1, 'coding_rate': '1/4', 'repeat': 2}
|
|
MODE[4][7] = {'bit_rate': 300, 'ci':MODE_BPSK, 'interleaver':['L', 40,144], 'unknown':20,'known':20, 'nsymb': 1, 'coding_rate': '1/4', 'repeat': 2}
|
|
|
|
MODE[7][4] = {'bit_rate': 150, 'ci':MODE_BPSK, 'interleaver':['S', 40, 18], 'unknown':20,'known':20, 'nsymb': 1, 'coding_rate': '1/8', 'repeat': 4}
|
|
MODE[5][4] = {'bit_rate': 150, 'ci':MODE_BPSK, 'interleaver':['L', 40,144], 'unknown':20,'known':20, 'nsymb': 1, 'coding_rate': '1/8', 'repeat': 4}
|
|
## 75 bps othogonal WALSH modulation
|
|
MODE[7][5] = {'bit_rate': 75, 'ci':MODE_BPSK, 'interleaver':['S', 10, 9], 'unknown':160,'known': 0, 'nsymb':32, 'coding_rate': '1/2', 'repeat': 1}
|
|
MODE[5][5] = {'bit_rate': 75, 'ci':MODE_BPSK, 'interleaver':['L', 20, 36], 'unknown':160,'known': 0, 'nsymb':32, 'coding_rate': '1/2', 'repeat': 1}
|
|
|
|
## ---- deinterleaver -----------------------------------------------------------
|
|
|
|
class Deinterleaver(object):
|
|
"""deinterleave"""
|
|
def __init__(self, rows, cols):
|
|
self._a = np.zeros((rows, cols), dtype=np.float32)
|
|
self._i = 0
|
|
self._j = 0
|
|
self._di = 9 if rows==40 else 7
|
|
self._dj = -17 if rows==40 else -7
|
|
self._buffer = np.zeros(0, dtype=np.float32)
|
|
print('deinterleaver: ', rows, cols, self._di, self._dj)
|
|
|
|
def fetch(self, a):
|
|
pass
|
|
|
|
def load(self, a):
|
|
self._buffer = np.append(self._buffer, a)
|
|
print('interleaver load', self._a.shape, a.shape, self._buffer.shape)
|
|
if self._buffer.shape[0] < self._a.shape[0]:
|
|
return np.zeros(0, dtype=np.float32)
|
|
print('interleaver load buffer:', len(self._buffer),self._i,self._j)
|
|
i = np.arange(self._a.shape[0])
|
|
j = (self._j + self._dj*np.arange(self._a.shape[0])) % self._a.shape[1]
|
|
self._a[i,j] = self._buffer[0:self._a.shape[0]]
|
|
self._buffer = np.delete(self._buffer, i)
|
|
self._j += 1
|
|
print('interleaver load buffer:', len(self._buffer),self._i,self._j)
|
|
if self._j == self._a.shape[1]:
|
|
self._j = 0
|
|
print('==================== interleaver is full! ====================')
|
|
return np.concatenate([self._a[(self._di*i)%self._a.shape[0],j] for j in range(self._a.shape[1])])
|
|
else:
|
|
return np.zeros(0, dtype=np.float32)
|
|
|
|
## ---- physcal layer class -----------------------------------------------------
|
|
class PhysicalLayer(object):
|
|
"""Physical layer description for MIL-STD-188-110 Appendix A"""
|
|
|
|
def __init__(self, sps):
|
|
"""intialization"""
|
|
self._sps = sps
|
|
self._frame_counter = -1
|
|
self._constellations = [BPSK, QPSK, PSK8]
|
|
self._preamble = self.get_preamble()
|
|
self._pre_counter = -1
|
|
self._d1d2 = [-1,-1] ## D1,D2
|
|
self._mode = {}
|
|
self._scr_data = ScrambleData()
|
|
|
|
def get_constellations(self):
|
|
return self._constellations
|
|
|
|
def get_next_frame(self, symbols):
|
|
"""returns a tuple describing the frame:
|
|
[0] ... known+unknown symbols and scrambling
|
|
[1] ... modulation type after descrambling
|
|
[2] ... a boolean indicating if the processing should continue
|
|
[3] ... a boolean indicating if the soft decision for the unknown
|
|
symbols are saved"""
|
|
print('-------------------- get_frame --------------------',
|
|
self._pre_counter, self._frame_counter)
|
|
success = True
|
|
if self._frame_counter == -1: ## preamble mode
|
|
if len(symbols) == 0:
|
|
return [self._preamble,MODE_BPSK,success,False]
|
|
else:
|
|
success = self.decode_preamble(symbols)
|
|
if self._pre_counter != 0:
|
|
return [self._preamble,MODE_BPSK,success,False]
|
|
else:
|
|
self._frame_counter = 0
|
|
self._scr_data.reset()
|
|
return [self.get_next_data_frame(success),self._mode['ci'],success,success]
|
|
else: ## data mode
|
|
self._frame_counter += 1
|
|
##print('test:', symbols[self._mode['unknown']:], np.mean(np.real(symbols[self._mode['unknown']:])))
|
|
if self._mode['known'] == 0: ## orthogonal WALSH modulation
|
|
success = True
|
|
for i in range(5):
|
|
a = symbols[32*i:32*(i+1)]
|
|
success &= np.max(np.imag(np.mean(a.reshape(8,4),0))) < 0.25
|
|
elif self._frame_counter < self._num_frames_per_block-2:
|
|
success = np.mean(np.real(symbols[self._mode['unknown']:])) > 0.4 or np.max(np.imag(symbols[self._mode['unknown']:])) < 0.6
|
|
if not success:
|
|
print('aborting: ', symbols[self._mode['unknown']:])# np.mean(np.real(symbols[self._mode['unknown']:])),
|
|
#np.max(np.imag(symbols[self._mode['unknown']:])))
|
|
return [self.get_next_data_frame(success),self._mode['ci'],success,success]
|
|
|
|
def get_next_data_frame(self, success):
|
|
if self._frame_counter == self._num_frames_per_block:
|
|
self._frame_counter = 0
|
|
scramble_for_frame = common.n_psk(8, np.array([self._scr_data.next()
|
|
for _ in range(self._frame_len)]))
|
|
a = common.make_scr(scramble_for_frame, scramble_for_frame)
|
|
n_unknown = self._mode['unknown']
|
|
a['symb'][0:n_unknown] = 0
|
|
if self._mode['known'] != 0 and self._frame_counter >= self._num_frames_per_block-2:
|
|
idx_d1d2 = self._frame_counter - self._num_frames_per_block + 2;
|
|
a['symb'][n_unknown :n_unknown+ 8] *= common.n_psk(2, WALSH8[self._d1d2[idx_d1d2]][:])
|
|
a['symb'][n_unknown+8:n_unknown+16] *= common.n_psk(2, WALSH8[self._d1d2[idx_d1d2]][:])
|
|
if not success:
|
|
self._frame_counter = -1
|
|
self._pre_counter = -1
|
|
return a
|
|
|
|
def get_doppler(self, iq_samples):
|
|
"""quality check and doppler estimation for preamble"""
|
|
success,doppler = True,0
|
|
if len(iq_samples) != 0:
|
|
sps = self._sps
|
|
zp = np.array([z for z in PhysicalLayer.get_preamble()['symb']
|
|
for _ in range(sps)], dtype=np.complex64)
|
|
## find starting point
|
|
_,_zp = self.get_preamble_z()
|
|
cc = np.correlate(iq_samples, zp[0:3*32*sps])
|
|
imax = np.argmax(np.abs(cc[0:2*32*sps]))
|
|
print('imax=', imax, len(iq_samples), len(cc))
|
|
apks = np.abs(cc[(imax, imax+3*32*sps),])
|
|
tpks = np.abs(cc[imax+3*16*sps:imax+5*16*sps])
|
|
print('imax=', imax, 'apks=',apks,
|
|
np.mean(apks), np.mean(tpks))
|
|
success = np.mean(apks) > 5*np.mean(tpks) and apks[0]/apks[1] > 0.5 and apks[0]/apks[1] < 2.0
|
|
if success:
|
|
idx = np.arange(32*sps)
|
|
pks = [np.correlate(iq_samples[imax+i*32*sps+idx],
|
|
zp[ i*32*sps+idx])[0]
|
|
for i in range(9)]
|
|
doppler = common.freq_est(pks)/(32*sps)
|
|
print('success=', success, 'doppler=', doppler,
|
|
np.abs(np.array(pks)),
|
|
np.angle(np.array(pks)))
|
|
return success,doppler
|
|
|
|
def decode_preamble(self, symbols):
|
|
data = [FROM_WALSH8[np.packbits
|
|
(np.real
|
|
(np.sum
|
|
(symbols[i:i+32].reshape((4,8)),0))<0)[0]]
|
|
for i in range(0,15*32,32)]
|
|
print('data=',data)
|
|
self._pre_counter = sum([(x&3)*(1<<2*y) for (x,y) in zip(data[11:14][::-1], range(3))])
|
|
self._d1d2 = data[9:11]
|
|
self._mode = mode = MODE[data[9]][data[10]]
|
|
self._block_len = 11520 if mode['interleaver'][0] == 'L' else 1440
|
|
self._frame_len = mode['known'] + mode['unknown']
|
|
if mode['known'] == 0: ## orthogonal WALSH modulation
|
|
self._num_frames_per_block = mode['interleaver'][1]*mode['interleaver'][2]/2*32/160
|
|
else:
|
|
self._num_frames_per_block = self._block_len/self._frame_len
|
|
self._deinterleaver = Deinterleaver(mode['interleaver'][1], mode['interleaver'][2])
|
|
self._depuncturer = common.Depuncturer(repeat=mode['repeat'])
|
|
self._viterbi_decoder = viterbi27(0x6d, 0x4f)
|
|
print(self._d1d2, mode, self._frame_len)
|
|
return True
|
|
|
|
def set_mode(self, _):
|
|
pass
|
|
|
|
def decode_soft_dec(self, soft_dec):
|
|
print('decode_soft_dec', len(soft_dec), soft_dec.dtype)
|
|
if self._mode['known'] == 0: ## orthogonal WALSH modulation
|
|
n = len(soft_dec) // 32
|
|
soft_bits = np.zeros(2*n, dtype=np.float32)
|
|
for i in range(n):
|
|
w = np.sum(soft_dec[32*i:32*(i+1)].reshape(4,8),0)
|
|
b = FROM_WALSH4[np.packbits(w[0:4]>0)[0]]
|
|
print('WALSH', i, w, b)
|
|
abs_soft_dec = np.mean(np.abs(w))
|
|
soft_bits[2*i] = abs_soft_dec*(2*(b>>1)-1)
|
|
soft_bits[2*i+1] = abs_soft_dec*(2*(b &1)-1)
|
|
print('WALSH soft_bits=', soft_bits)
|
|
r = self._deinterleaver.load(soft_bits)
|
|
else:
|
|
r = self._deinterleaver.load(soft_dec)
|
|
print('decode_soft_dec r=', r.shape)
|
|
if r.shape[0] == 0:
|
|
return []
|
|
##print('deinterleaved bits: ', [x for x in 1*(r>0)])
|
|
rd = self._depuncturer.process(r)
|
|
self._viterbi_decoder.reset()
|
|
decoded_bits = self._viterbi_decoder.udpate(rd)
|
|
##print('bits=', decoded_bits)
|
|
print('quality={}%'.format(100.0*self._viterbi_decoder.quality()/(2*len(decoded_bits))))
|
|
return decoded_bits
|
|
|
|
@staticmethod
|
|
def get_preamble():
|
|
"""preamble symbols + scrambler"""
|
|
return common.make_scr(PRE_SCRAMBLE*PRE_SYMBOLS,
|
|
PRE_SCRAMBLE)
|
|
def get_preamble_z(self):
|
|
"""preamble symbols for preamble correlation"""
|
|
a = PhysicalLayer.get_preamble()
|
|
return 0,np.array([z for z in a['symb'][0:3*32]
|
|
for _ in range(self._sps)])
|
|
|
|
if __name__ == '__main__':
|
|
def gen_data_scramble():
|
|
def advance(s):
|
|
msb = s>>11
|
|
s = (s<<1)&((1<<12)-1)
|
|
if msb: s ^= 0x053
|
|
return s
|
|
a = np.zeros(160, dtype=np.uint8)
|
|
s = 0xBAD
|
|
for i in range(160):
|
|
for _ in range(8): s = advance(s)
|
|
a[i] = s&7;
|
|
return a
|
|
|
|
sps = 5;
|
|
p=PhysicalLayer(sps)
|
|
z1=np.array([x for x in PRE_SYMBOLS for _ in range(sps)])
|
|
z2=np.array([x for x in PRE_SCRAMBLE for _ in range(sps)])
|
|
z=z1*z2;
|
|
_,_z=p.get_preamble_z()
|
|
print(all(z[0:3*32*sps]==_z[0:3*32*sps]))
|
|
for i in range(3):
|
|
print(i, all(z[32*sps*i:32*sps*(i+1)] == z[32*sps*(3+i):32*sps*(3+i+1)]))
|
|
|
|
#print(np.sum(np.sum(z[0:32*5] * np.conj(z[32*5*3:32*5*4]))))
|
|
#print(WALSH8[1][:])
|
|
#print(sum(WALSH8[1][:]*(1<<np.array(range(7,-1,-1)))))
|
|
#print(FROM_WALSH8)
|
|
#print(gen_data_scramble())
|
|
|
|
s=ScrambleData()
|
|
#print([s.next() for _ in range(160)])
|
|
#print([s.next() for _ in range(160)])
|
|
#print(np.round(np.angle(PRE_SYMBOLS*PRE_SCRAMBLE)/np.pi*4))
|