/
simcyc.py
157 lines (134 loc) · 5.34 KB
/
simcyc.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
"""
Library for simulating cyclic spectra
"""
import pycyc
reload(pycyc)
import numpy as np
from matplotlib import pyplot as plt
import os
import glob
def runLoopTest(tau=650.0,nlag=4096,bw=1.0,noise=0,tolfact=10, niter=5,maxfun=1000,maxinitharm=20):
filename = 'itersim1643_tau%.2fus_noise%.2f' % (tau,noise)
blah,fbase = os.path.split(filename)
plotdir = os.path.join(os.path.abspath(os.path.curdir),('%s_plots' % fbase))
if not os.path.exists(plotdir):
os.mkdir(plotdir)
ht0 = makeht(tau=tau,nlag=nlag,bw=bw)
np.save(os.path.join(plotdir,'ht0'), ht0)
CS = initSim(ht0,bw=bw,noise=noise)
CS.filename = filename
CS.initProfile(maxinitharm=maxinitharm)
np.savez(os.path.join(plotdir,'start'),cs=CS.cs,pp_ref=CS.pp_ref,
pp_start = CS.data.mean(2).squeeze())
for k in range(1,niter+1):
CS.pp_ref = CS.pp_int.copy()
CS.loop(make_plots=True,plotdir=plotdir,hf_prev=CS.hf_prev,maxfun=maxfun)
np.savez(os.path.join(plotdir,('iter%d' % k)),
pp_ref=CS.pp_ref,pp_int=CS.pp_int, hf = CS.hf_prev,
ht = pycyc.freq2time(CS.hf_prev), csm = CS.modelCS(hf=CS.hf_prev),
)
def runTest(tau,nlag,bw=1.0,noise=0.0,tolfact=10,initprof=False):
CS = initSim(makeht(tau=tau,nlag=nlag,bw=bw),bw=bw,noise=noise)
CS.filename = 'sim1643_tau%.2fus_noise%.2f' % (tau,noise)
if initprof:
CS.initProfile()
CS.loop(make_plots=True,tolfact=tolfact)
return CS
def simSuite(taus=[2.0,10.0,100.0],harms=[3,10],periods=[1.5e-3,4e-3,10e-3],snrs=[10,100],
maxfun = 100):
for tau in taus:
for harm in harms:
for period in periods:
for snr in snrs:
prof = makeProfile(harm)
ht = makeht(tau, nlag=2048)
CS = initSim(ht,prof,ref_freq=1/period,bw=1.0,noise=snr,source = ('tau_%.1f_nharm_%d_period_%.2f_snr_%.3f' % (tau,harm,period*1000.,snr)))
CS.initProfile()
CS.pp_meas = CS.pp_ref.copy()
CS.pp_ref = prof
CS.ph_ref = pycyc.phase2harm(CS.pp_ref)
CS.ph_ref = pycyc.normalize_profile(CS.ph_ref)
CS.ph_ref[0] = 0
CS.s0 = CS.ph_ref.copy()
CS.plotdir = '/home/gjones/workspace/pycyc/simsuite'
CS.pharm = harm
CS.tau = tau
CS.noise = snr
CS.loop(make_plots=False,maxfun=maxfun)
CS.noise = snr
CS.saveState(os.path.join(CS.plotdir,'cs_' + CS.source + '.pkl'))
pycyc.plotSimulation(CS)
def replotSims(simdir = '/home/gjones/workspace/pycyc/simsuite'):
sims = glob.glob(os.path.join(simdir, 'cs_*.pkl'))
for sim in sims:
cs = pycyc.loadCyclicSolver(sim)
pycyc.plotSimulation(cs)
def initSim(ht,prof,ref_freq,bw,rf = None,filename= None,source='fake',noise = None):
if rf is None:
rf = np.abs(bw)/2.0
if ht.ndim == 1:
ht = ht[None,:]
CS = pycyc.CyclicSolver()
CS.ht0 = ht
if filename:
CS.filename = filename
else:
CS.filename = 'sim%s_%.1fMHz_%.1fms' % (source,bw,1000.0/ref_freq)
CS.nchan = ht.shape[1]
CS.nlag = CS.nchan
CS.nphase = prof.shape[0]
CS.nbin = CS.nphase
CS.nharm = CS.nphase/2 + 1
CS.source = source
CS.nspec = ht.shape[0]
CS.dynamic_spectrum = np.zeros((CS.nspec,CS.nchan))
CS.optimized_filters = np.zeros((CS.nspec,CS.nchan),dtype='complex')
CS.intrinsic_profiles = np.zeros((CS.nspec,CS.nbin))
CS.nopt = 0
CS.nloop = 0
CS.nopt = 0
CS.ref_freq = ref_freq
CS.bw = bw
CS.ref_phase = 0
CS.rf = rf
CS.hf_prev = np.ones((CS.nchan,),dtype='complex')
CS.pp_int = np.zeros((CS.nphase)) #intrinsic profile
CS.pp_ref = prof.copy()
CS.ph_ref = pycyc.phase2harm(CS.pp_ref)
CS.ph_ref = pycyc.normalize_profile(CS.ph_ref)
CS.ph_ref[0] = 0
CS.s0 = CS.ph_ref.copy()
CS.data = np.empty((CS.nspec,1,CS.nchan,CS.nphase), dtype='complex')
for k in range(CS.nspec):
CS.data[k,0,:,:] = pycyc.cs2ps(CS.modelCS(ht[k,:]))
if noise is not None:
signal = np.abs(CS.data).sum()
CS.noise = noise
rn = (np.random.randn(CS.data.shape[0],CS.data.shape[1],CS.data.shape[2],CS.data.shape[3]).astype('complex')
+1j*np.random.randn(CS.data.shape[0],CS.data.shape[1],CS.data.shape[2],CS.data.shape[3]))
rnpow = np.abs(rn).sum()
fact = signal/rnpow
CS.data = CS.data + rn*(fact/noise)
return CS
def makeProfile(scale,nbins=1024):
"""
Construct a fake profile with exponentially decaying harmonics
scale : harmonic at which amplitude = exp(-1)
nbins : number of bins in the profile
"""
ph = np.exp(-np.arange(nbins/2 + 1)/(1.0*scale))
ph[0] = 0.0
pp = np.fft.irfft(ph)
return pp
def makeht(tau=1.0,nlag=1024,scale = 0,bw=1.0):
t = np.arange(nlag)/bw
ht = (np.random.randn(nlag)+1j*np.random.randn(nlag))*np.sqrt(np.exp(-t/tau))
ht = ht/np.abs(ht[0])
ht[0] += scale
ht[ht.shape[0]/2:] = 0
return ht
def makehtOne(tau=1.0,val = 0.1, nlag=1024,bw=1.0):
ht = np.zeros((nlag,),dtype='complex')
ht[0] = 1.0
ht[int(tau*bw)] = val
return ht