-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathnlfir.py
196 lines (183 loc) · 6.41 KB
/
nlfir.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
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
import numpy as np
import time
try:
import numba as nb
have_numba=True
except:
have_numba=False
print("Numba unavailable - some things may be slower.")
def extract_vec(vec,jl,jr):
if jl==0:
return vec[:jr+1]
else:
v1=vec[-jl:]
v2=vec[:jr+1]
return np.hstack([v1,v2])
def insert_vec(vec,jl,jr,n):
out=np.zeros(n)
if jl==0:
out[:jr+1]=vec
return out
else:
out[-jl:]=vec[:jl]
out[:jr+1]=vec[jl:]
return out
#@nb.njit(parallel=False)
def vec2dense_nb(vec,jl,jr,k1=10000,k2=-10000,mat=None,istart=0):
"""make a dense version of the FIR matrix for one vector. Allocate a matrix
or optionally take in a pre-existing matrix. copies centered on k1:k2 with shifts
going from -jl to jr will go into mat."""
n=len(vec[k1:k2])
if mat is None:
mat=np.zeros((n,jl+jr+1))
istart=0
for i in nb.prange(0,jl+jr+1):
mat[:,istart+i]=vec[k1+i-jl:k2+i-jl]
return mat
@nb.njit(parallel=True)
def vec2dense_trans_nb(vec,jl,jr,k1,k2,mat,istart):
for i in nb.prange(-jl,jr+1):
mat[istart+i+jl,:]=vec[k1+i:k2+i]
def vec2dense_trans(vec,jl,jr,k1=10000,k2=-10000,mat=None,istart=0,use_nb=True):
"""make a dense version of the FIR matrix for one vector. Allocate a matrix
or optionally take in a pre-existing matrix. copies centered on k1:k2 with shifts
going from -jl to jr will go into mat."""
n=len(vec[k1:k2])
jtot=int(jl+jr+1)
if mat is None:
mat=np.zeros((jtot,n))
istart=0
#mat=np.zeros((jtot,n))
if have_numba&use_nb:
vec2dense_trans_nb(vec,jl,jr,k1,k2,mat,istart)
else:
for i in np.arange(jtot):
mat[i,:]=vec[k1+i+jl:k2+i+jl]
return mat
class AdjustFit:
def __init__(self,pred,targ,thresh,order=3,use_thresh=None):
self.order=order
self.thresh=thresh
if use_thresh is None:
use_thresh=thresh
self.use_thresh=use_thresh
mask=pred>thresh
self.coeffs=np.polyfit(pred[mask],targ[mask],order)
def apply_adjust(self,pred):
out=pred.copy()
mask=pred>self.use_thresh
tmp=pred[mask]
out[mask]=np.polyval(self.coeffs,pred[mask])
return out
class FIRmat:
def __init__(self,vecs,jl,jr,k1,k2):
if isinstance(vecs,list):
self.vecs=vecs
else:
self.vecs=[vecs]
if isinstance(jl,list):
self.jl=jl
else:
self.jl=[jl]
if isinstance(jr,list):
self.jr=jr
else:
self.jr=[jr]
self.k1=k1
self.k2=k2
if isinstance(vecs,list):
self.nvec=len(vecs)
else:
self.nvec=1
self.n=len(self.vecs[0][k1:k2])
self.fts=None
self.coeffs=None
self.adjust=None
def add_vec(self,vec,jl=None,jr=None):
self.vecs.append(vec)
if jl is None: #default to the previous jl/jr if none specified
jl=self.jl[-1]
if jr is None:
jr=self.jr[-1]
self.jl.append(jl)
self.jr.append(jr)
self.nvec=self.nvec+1
def nrow(self):
n=0
for i in range(self.nvec):
n=n+(self.jr[i]+self.jl[i]+1)
return n
def dense(self):
nr=self.nrow()
mat=np.empty([nr,self.n])
istart=0
for i in range(self.nvec):
vec2dense_trans(self.vecs[i],self.jl[i],self.jr[i],self.k1,self.k2,mat,istart)
istart=istart+(self.jr[i]+self.jl[i]+1)
return mat
def get_fts(self):
self.fts=[None]*self.nvec
for i in range(self.nvec):
self.fts[i]=np.fft.rfft(self.vecs[i][self.k1:self.k2])
def vecmult(self,vec):
vecft=np.conj(np.fft.rfft(vec[self.k1:self.k2]))
vals=[None]*self.nvec
for i in range(self.nvec):
tmp=np.fft.irfft(self.fts[i]*vecft)
vals[i]=extract_vec(tmp,self.jl[i],self.jr[i])
return np.hstack(vals)
def sqr(self):
tmparr = [[None for j in range(self.nvec)] for i in range(self.nvec)]
for i in range(self.nvec):
for j in range(self.nvec):
tmp=np.fft.irfft(self.fts[i]*np.conj(self.fts[j]))
block=np.empty([self.jl[i]+self.jr[i]+1,self.jl[j]+self.jr[j]+1])
for ii_tmp in range(self.jl[i]+self.jr[i]+1):
ii=ii_tmp-self.jl[i]
#this loop could be a *lot* more efficient.
#it doesn't add to runtime, though so I'll be lazy and leave it.
for jj_tmp in range(self.jl[j]+self.jr[j]+1):
jj=jj_tmp-self.jl[j]
block[ii_tmp,jj_tmp]=tmp[ii-jj]
tmparr[i][j]=block
tmparr[j][i]=block.T
tmp2=[None]*self.nvec
for i in range(self.nvec):
tmp2[i]=np.hstack(tmparr[i])
return np.vstack(tmp2)
def get_coeffs(self,s):
lhs=self.sqr()
rhs=self.vecmult(s)
self.coeffs=np.linalg.inv(lhs)@rhs
def set_adjust(self,targ,thresh,adjust=AdjustFit,order=None):
if order is None:
order=len(self.vecs)-1
self.adjust=None
pred=self.get_pred(adjust=False)
self.adjust=adjust(pred,targ,thresh,order)
def get_pred(self,coeffs=None,adjust=False):
if coeffs is None:
coeffs=self.coeffs
pred=0.0
icur=0
for i in range(self.nvec):
nn=self.jl[i]+self.jr[i]+1
tmp=insert_vec(coeffs[icur:icur+nn],self.jl[i],self.jr[i],self.n)
icur=icur+nn
tmpft=np.fft.rfft(tmp)
pred=pred+np.fft.irfft(self.fts[i]*np.conj(tmpft))
#print('=====pred len: ', len(pred))
if adjust:
if self.adjust is None:
print('adjusted fit requested, but adjust has not been set. skipping.')
else:
pred=self.adjust.apply_adjust(pred)
pout=np.zeros(len(self.vecs[0]))
pout[self.k1:self.k2]=pred
return pout
def read_dir(dirname,subdir='0_digitization'):
adc=np.loadtxt(dirname+'/'+subdir+'/digits_out_sequence_eT.txt')
pileup=np.loadtxt(dirname+'/'+subdir+'/hit_eT_bck_sequence.txt')
sig=np.loadtxt(dirname+'/'+subdir+'/hit_eT_sig_sequence.txt')
pulse=np.loadtxt(dirname+'/'+subdir+'/ideal_output_sequence.txt')
return adc,pileup,sig,pulse