Newer
Older
mwyman
committed
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
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
import numpy as np
from scipy.optimize import root_scalar
import xraylib
"""
pyDevice TO DO:
WHAT inputs change the focal size arrays? Energy, what else?
WHAT inputs change the search through the arrays? desired focal size, what else?
IOC init functions
-get lens stack parameters (# of lenses in each stack, radius, location, thickness, thickness error) -- from substitution file but put into PVs? Update with autosave?
-get source info
-energy from from ID IOC
-hor/vert sizes and divergence (also energy dependent)
-lens diameter table? What is it doing?
-desired focal size is changed --> what needs updating? --> nothing, just need to search focal size array again
-multiple flags: is focal size achievable? is it achievable at sample?
recalc function -- should probably be same as init function
-energy is updated --> what needs updating?
-what else could user/staff change? sample position?
"""
# Beamline input block
energy = 15000.0 # Energy in eV
energy_keV = energy/1000.0 # Energy in keV
wl = 1239.84 / (energy * 10**9)
d_StoL1 = 51.9 # Source-to-CRL1 distance, in m
d_StoL2 = 62.1 # Source-to-CRL2 distance, in m
d_Stof = 66.2 # Source-to-focus distance, in m
#slit1_H = 500.0e-6 # H slit size before CRL 1
#slit1_V = 300.0e-6 # V slit size before CRL 1
# CRL input block
d_min = 3.0e-5 # Minimum thickness at the apex in m
stack_d = 50.0e-3 # Stack thickness in m
L1_n = np.array([1, 1, 1, 1, 1, 1, 2, 4, 8, 16]) # CRL1 number of lenses in each stack
L1_R = np.array([2.0e-3, 1.0e-3, 5.0e-4, 3.0e-4, 2.0e-4, 1.0e-4, 1.0e-4, 1.0e-4, 1.0e-4, 1.0e-4]) # CRL1 lens radius in each stack
L1_mater= np.array(["Be", "Be", "Be", "Be", "Be", "Be", "Be", "Be", "Be", "Be"]) # CRL1 lens material in each stack
L1_loc = np.array([4.5, 3.5, 2.5, 1.5, 0.5, -0.5, -1.5, -2.5, -3.5, -4.5])*stack_d # CRL1 lens stack location relative to center stack, positive means upstream
L1_HE = np.array([1.0e-6, 1.0e-6, 1.0e-6, 1.0e-6, 1.0e-6, 1.0e-6, 1.4e-6, 2.0e-6, 2.8e-6, 4.0e-6]) # CRL1 lens RMS thickness error
# Source size input block
L_und = 4.7 # undulator length
sigmaH_e = 14.8e-6 # Sigma electron source size in H direction in m
sigmaV_e = 3.7e-6 # Sigma electron source size in V direction in m
sigmaHp_e = 2.8e-6 # Sigma electron divergence in H direction in rad
sigmaVp_e = 1.5e-6 # Sigma electron divergence in V direction in rad
sigmaH = (sigmaH_e**2 + wl*L_und/2/np.pi/np.pi)**0.5
sigmaV = (sigmaV_e**2 + wl*L_und/2/np.pi/np.pi)**0.5
sigmaHp = (sigmaHp_e**2 + wl/L_und/2)**0.5
sigmaVp = (sigmaVp_e**2 + wl/L_und/2)**0.5
# Lookup table where each entry is a tuple (column1, column2)
Lens_diameter_table = [
(50, 450.0),
(100, 632.0),
(200, 894.0),
(300, 1095.0),
(500, 1414.0),
(1000, 2000.0),
(1500, 2450.0),
]
# Convert the lookup table to a dictionary for faster lookup
Lens_diameter_dict = {int(col1): col2 for col1, col2 in Lens_diameter_table}
def lookup_diameter(lens_radius):
# Convert the input float to an integer
input_int = int(round(lens_radius*1.0e6))
return Lens_diameter_dict.get(input_int, (lens_radius*1.0e6)**0.5*63.222+ 0.73)/1.0e6
def index_to_binary_list(index, length):
"""
Converts an index number to its binary representation as a list of digits,
and pads the list with zeros in front to achieve the desired length.
Parameters:
index (int): The index number to be converted.
length (int): The desired length of the binary list.
Returns:
list: A list of digits representing the binary representation of the index.
"""
# Convert the index to a binary string and remove the '0b' prefix
binary_str = bin(index)[2:]
# Pad the binary string with zeros in front to achieve the desired length
#padded_binary_str = binary_str.zfill(length)
# Reverse the binary string
reversed_binary_str = binary_str[::-1]
# Convert the reversed binary string to a list of integers
binary_list = [int(digit) for digit in reversed_binary_str]
# Pad the list with zeros at the end to achieve the desired length
while len(binary_list) < length:
binary_list.append(0)
return binary_list
def binary_list_to_index(binary_list, length):
"""
Converts a list of binary digits in reverse order to its integer representation,
padding the list with zeros at the end to have a fixed number of elements.
Parameters:
binary_list (list): A list of digits representing the binary number in reverse order.
length (int): The fixed number of elements the list should have.
Returns:
int: The integer representation of the binary number.
"""
# Pad the list with zeros at the end to achieve the desired length
while len(binary_list) < length:
binary_list.append(0)
# Convert the binary list to an integer
index = 0
for i, digit in enumerate(binary_list):
index += digit * 2**i
return index
def materials_to_deltas(material_list, energy):
"""
Convert a list of material names to a list of delta values at a given energy.
Parameters:
material_list (list): A list of material names.
energy (float): The energy in keV.
Returns:
list: A list of delta values for the given materials at the given energy.
"""
# The list to store delta values
delta_list = []
# Iterate through each material in the input list
for material in material_list:
# Compute the delta value for the current material at the given energy
Z = xraylib.SymbolToAtomicNumber(material)
density = xraylib.ElementDensity(Z)
delta = 1.0-xraylib.Refractive_Index_Re(material, energy, density)
# Add the delta value to the delta list
delta_list.append(delta)
return delta_list
def materials_to_linear_attenuation(material_list, energy):
"""
Convert a list of material names to a list of linear attenuation coefficients at a given energy.
Parameters:
material_list (list): A list of material names.
energy (float): The energy in keV.
Returns:
list: A list of linear attenuation coefficient values (in m^-1) for the given materials at the given energy.
"""
# The list to store linear attenuation coefficient values
mu_list = []
# Iterate through each material in the input list
for material in material_list:
# Compute the delta value for the current material at the given energy
Z = xraylib.SymbolToAtomicNumber(material)
density = xraylib.ElementDensity(Z)
# Compute the mass attenuation coefficient in cm^2/g
#mass_attenuation = xraylib.CS_Photo(Z, energy)
mass_attenuation = xraylib.CS_Total(Z, energy)
# Convert mass attenuation coefficient to linear attenuation coefficient in m^-1
mu = mass_attenuation * density * 100.0
# Add the linear attenuation coefficient value to the list
mu_list.append(mu)
return mu_list
def absorptionaperture(x, n1mud, sigma, n1mur):
numerator = np.exp(-(x**2/(2*sigma**2))) * np.exp(-n1mur*(x**2) - n1mud)
denominator = np.exp(-n1mud)
return numerator / denominator - 0.5
def find_levels(array, levels, direction='forward'):
"""
Find the first indices at which the array crosses specified levels and the corresponding crossed values.
Parameters:
array (numpy.ndarray): An array of numbers.
levels (float or numpy.ndarray): A number or an array of levels to find crossings.
direction (str, optional): The searching direction. Defaults to 'forward'.
Can be either 'forward' or 'backward'.
Returns:
tuple: A tuple containing two arrays:
- An array of first indices at which the array crosses the specified levels.
- An array of first crossed values at the corresponding indices.
"""
# Convert a single level to a numpy array
if isinstance(levels, (int, float)):
levels = np.array([levels])
indices = []
values = []
# Compute the max and min of the array ignoring NaNs
max_val = np.nanmax(array)
min_val = np.nanmin(array)
for level in levels:
# If level is out of bounds
if level > max_val or level < min_val:
indices.append(-1)
values.append(np.nan)
continue
crossings = []
if direction == 'forward':
for i in range(1, len(array)):
if np.isnan(array[i - 1]) or np.isnan(array[i]):
continue
if (array[i - 1] < level <= array[i]) or (array[i - 1] > level >= array[i]):
crossings.append(i - 1)
break
elif direction == 'backward':
for i in range(len(array) - 2, -1, -1):
if np.isnan(array[i + 1]) or np.isnan(array[i]):
continue
if (array[i + 1] < level <= array[i]) or (array[i + 1] > level >= array[i]):
crossings.append(i)
break
else:
raise ValueError("Invalid direction. It should be either 'forward' or 'backward'.")
if len(crossings) > 0:
idx = crossings[0]
indices.append(idx)
values.append(array[idx])
else:
# In case no crossing is found within the range
indices.append(-1)
values.append(np.nan)
return np.array(indices), np.array(values)
def Single_CRL2D_control(fsize):
L1_D = np.zeros(L1_R.size) # CRL1 diameters for each stack
for i in range(L1_R.size):
L1_D[i] = lookup_diameter(L1_R[i])
L1_delta = materials_to_deltas(L1_mater, energy_keV) # delta values for CRL1 stacks
L1_mu = materials_to_linear_attenuation(L1_mater, energy_keV) # mu values for CRL1 stacks
L1_Feq = L1_R/(2*L1_n*L1_delta) + L1_loc # CRL1 equivalent f in m for each stack
L1_index_n = 2**L1_Feq.size # Total number of combinations for CRL1
L1_invF_list= np.zeros(L1_index_n) # List of equivalent 1/f in m^-1 for CRL1
for i in range(L1_index_n):
L1_invF_list[i] = np.sum(index_to_binary_list(i, L1_Feq.size)/L1_Feq)
# Sort the L1_invF list (to avoid zigzagging)
L1_invF_list_sort_indices = np.argsort(L1_invF_list)
L1_invF_list_sorted = L1_invF_list[L1_invF_list_sort_indices]
q1_list = 1/(L1_invF_list_sorted - 1/d_StoL1) # focal position of CRL1 for all configurations (sorted)
dq1_list = q1_list - (d_Stof - d_StoL1)
# Start generating focal size list as a function of CRL1 setting
sigma1H = (sigmaH**2 + (sigmaHp*d_StoL1)**2)**0.5 # sigma beam size before CRL1
sigma1V = (sigmaV**2 + (sigmaVp*d_StoL1)**2)**0.5 # sigma beam size before CRL1
L1_n1mud_list = np.zeros(L1_index_n) # List of n1*mu*d_min for all possible CRL1 configurations
L1_n1muR_list = np.zeros(L1_index_n) # List of n1*mu/R for all possible CRL1 configurations
aperL1H_list = np.zeros(L1_index_n) # absorption H aperture of CRL1 for all configurations
aperL1V_list = np.zeros(L1_index_n) # absorption V aperture of CRL1 for all configurations
diameter1_list = np.zeros(L1_index_n) # CRL1 diameter for all possible configurations
FWHM1H_list = np.zeros(L1_index_n) # H focal size at the focus of CRL1
FWHM1V_list = np.zeros(L1_index_n) # V focal size at the focus of CRL1
Strehl_list = np.zeros(L1_index_n) # Strehl ratio based on lens thickness error
for i in range(L1_index_n):
# absorption aperture is a function of CRL absorption/physical aperture, incident beam size, and physical slits
L1_n1mud_list[i] = np.sum(index_to_binary_list(L1_invF_list_sort_indices[i], L1_Feq.size)*np.array(L1_mu*L1_n*d_min))
L1_n1muR_list[i] = np.sum(index_to_binary_list(L1_invF_list_sort_indices[i], L1_Feq.size)*np.array(L1_mu*L1_n/L1_R))
solution = root_scalar(absorptionaperture, args=(L1_n1mud_list[i], sigma1H, L1_n1muR_list[i]), bracket=[0.0, 2*sigma1H], method='bisect')
aperL1H_list[i] = solution.root*2.0
solution = root_scalar(absorptionaperture, args=(L1_n1mud_list[i], sigma1V, L1_n1muR_list[i]), bracket=[0.0, 2*sigma1V], method='bisect')
aperL1V_list[i] = solution.root*2.0
mask = (np.array(index_to_binary_list(L1_invF_list_sort_indices[i], L1_Feq.size)) == 1)
if np.all(mask == False):
diameter1_list[i] = np.inf
else:
diameter1_list[i] = np.min(L1_D[mask])
aperL1H_list[i] = min(aperL1H_list[i], diameter1_list[i], slit1_H)
aperL1V_list[i] = min(aperL1V_list[i], diameter1_list[i], slit1_V)
phase_error_tmp = np.linalg.norm(index_to_binary_list(L1_invF_list_sort_indices[i], L1_Feq.size)*np.array(L1_HE*L1_delta)*2*np.pi/wl)
Strehl_list[i] = np.exp(-phase_error_tmp**2)
# FWHMbeam size at CRL1 focus
FWHM1H_list = ((0.88*wl*q1_list/aperL1H_list)**2 + (2.355*sigmaH*q1_list/d_StoL1)**2)**0.5
FWHM1V_list = ((0.88*wl*q1_list/aperL1V_list)**2 + (2.355*sigmaV*q1_list/d_StoL1)**2)**0.5
if flag_HE:
FWHM1H_list *= (Strehl_list)**(-0.5)
FWHM1V_list *= (Strehl_list)**(-0.5)
FWHM_list = (FWHM1H_list*FWHM1V_list)**0.5
indices, values = find_levels(FWHM_list, fsize, direction='backward')
index = indices[0]
if index == -1:
print(f"Cannot achieve the focal size {fsize*1.0e6:.2f} μm")
else:
print("======== Find size at focus ========================================")
print(f"Energy: {energy_keV} keV")
print(f"CRL1 configuration index in sorted list is {index}")
print(f"CRL1 configuration index is {L1_invF_list_sort_indices[index]} or {index_to_binary_list(L1_invF_list_sort_indices[index], L1_Feq.size)}")
print(f"CRL1 f is {1/L1_invF_list_sorted[index]:.2f} m, focus at q1 = {q1_list[index]:.2f} m")
print(f"Focal size is {FWHM1H_list[index]*1.0e6:.2f} μm x {FWHM1V_list[index]*1.0e6:.2f} μm at the focal point ({dq1_list[index]*1e3:.1f} mm from sample)")
FWHM1H_atsample_list = (FWHM1H_list**2 + (aperL1H_list*dq1_list/q1_list)**2)**0.5
FWHM1V_atsample_list = (FWHM1V_list**2 + (aperL1V_list*dq1_list/q1_list)**2)**0.5
FWHM_atsample_list = (FWHM1H_atsample_list*FWHM1V_atsample_list)**0.5
indices, values = find_levels(FWHM_atsample_list, fsize, direction='forward')
index2 = indices[0]
if index2 == -1:
print(f"Cannot achieve the bame size {fsize*1.0e6:.2f} μm at sample")
else:
print("======== Find size at sample =======================================")
print(f"CRL1 configuration index in sorted list is {index2}")
print(f"CRL1 configuration index is {L1_invF_list_sort_indices[index2]} or {index_to_binary_list(L1_invF_list_sort_indices[index2], L1_Feq.size)}")
print(f"CRL1 f is {1/L1_invF_list_sorted[index2]:.2f} m, focus at q1 = {q1_list[index2]:.2f} m ({dq1_list[index2]*1e3:.1f} mm from sample)")
print(f"Beam size is {FWHM1H_atsample_list[index2]*1.0e6:.2f} μm x {FWHM1V_atsample_list[index2]*1.0e6:.2f} μm at the sample position)")
indices, values = find_levels(dq1_list, 0.0, direction='backward')
index3 = indices[0]
if index == -1:
print(f"Cannot find combination to focus close to sample")
else:
print("======== Find configuration focus close to the sample ==============")
print(f"CRL1 configuration index in sorted list is {index3}")
print(f"CRL1 configuration index is {L1_invF_list_sort_indices[index3]} or {index_to_binary_list(L1_invF_list_sort_indices[index3], L1_Feq.size)}")
print(f"CRL1 f is {1/L1_invF_list_sorted[index3]:.2f} m, focus at q1 = {q1_list[index3]:.2f} m ({dq1_list[index3]*1e3:.1f} mm from sample)")
print(f"Beam size is {FWHM1H_atsample_list[index3]*1.0e6:.2f} μm x {FWHM1V_atsample_list[index3]*1.0e6:.2f} μm at the sample position)")
return
if __name__ == "__main__":
flag_HE = True
fsize = 50.0e-6 # Desired focal size in m (area average of h and v size)
#Single_CRL2D_control(fsize) # Find the best configuration for a single transfocator system
'''
Update the following to accommodate XS code
'''
class singleTF():
def __init__(self):
# Initialize beamline layout variables
self.d_StoL = 51.9 # Source-to-CRL1 distance, in m
self.d_Stof = 66.2 # Source-to-focus distance, in m
# Initialize source variables
self.L_und = 4.7
self.energy = 15000.0 # Energy in eV
self.energy_keV = self.energy/1000.0 # Energy in keV
self.wl = 1239.84 / (self.energy * 10**9) #Wavelength in nm(?)
self.sigmaH_e = 14.8e-6 # Sigma electron source size in H direction in m
self.sigmaV_e = 3.7e-6 # Sigma electron source size in V direction in m
self.sigmaHp_e = 2.8e-6 # Sigma electron divergence in H direction in rad
self.sigmaVp_e = 1.5e-6 # Sigma electron divergence in V direction in rad
self.sigmaH = (self.sigmaH_e**2 + self.wl*self.L_und/2/np.pi/np.pi)**0.5
self.sigmaV = (self.sigmaV_e**2 + self.wl*self.L_und/2/np.pi/np.pi)**0.5
self.sigmaHp = (self.sigmaHp_e**2 + self.wl/self.L_und/2)**0.5
self.sigmaVp = (self.sigmaVp_e**2 + self.wl/self.L_und/2)**0.5
# Initialize lens variables
self.d_min = 3.0e-5 # Minimum thickness at the apex in m
self.stack_d = 50.0e-3 # Stack thickness in m
self.L1_n = np.array([1, 1, 1, 1, 1, 1, 2, 4, 8, 16]) # CRL1 number of lenses in each stack
self.L1_R = np.array([2.0e-3, 1.0e-3, 5.0e-4, 3.0e-4, 2.0e-4, 1.0e-4, 1.0e-4, 1.0e-4, 1.0e-4, 1.0e-4]) # CRL1 lens radius in each stack
self.L1_mater= np.array(["Be", "Be", "Be", "Be", "Be", "Be", "Be", "Be", "Be", "Be"]) # CRL1 lens material in each stack
self.L1_loc = np.array([4.5, 3.5, 2.5, 1.5, 0.5, -0.5, -1.5, -2.5, -3.5, -4.5])*stack_d # CRL1 lens stack location relative to center stack, positive means upstream
self.L1_HE = np.array([1.0e-6, 1.0e-6, 1.0e-6, 1.0e-6, 1.0e-6, 1.0e-6, 1.4e-6, 2.0e-6, 2.8e-6, 4.0e-6]) # CRL1 lens RMS thickness error
self.Lens_diameter_table = [
(50, 450.0),
(100, 632.0),
(200, 894.0),
(300, 1095.0),
(500, 1414.0),
(1000, 2000.0),
(1500, 2450.0),
]
# Convert the lookup table to a dictionary for faster lookup
self.Lens_diameter_dict = {int(col1): col2 for col1, col2 in Lens_diameter_table}
# Initialize pre-CRL slit size
self.slit1_H = 500.0e-6 # H slit size before CRL 1
self.slit1_V = 300.0e-6 # V slit size before CRL 1
#Calc lookup table
self.lookup_table=calc_lookup_table(self, ...)
self.energy = 0 # gets value from an ao (incoming beam energy)
self.focalSize = 0 # get value from an ao (desired focal length)
self.lenses = 0 # sets integer (2^12) whose binary representation indicates which lenses are in or out
def setSource(self):
self.L_und = 4.7
self.energy = 15000.0 # Energy in eV
self.energy_keV = self.energy/1000.0 # Energy in keV
self.wl = 1239.84 / (self.energy * 10**9) #Wavelength in nm(?)
self.sigmaH_e = 14.8e-6 # Sigma electron source size in H direction in m
self.sigmaV_e = 3.7e-6 # Sigma electron source size in V direction in m
self.sigmaHp_e = 2.8e-6 # Sigma electron divergence in H direction in rad
self.sigmaVp_e = 1.5e-6 # Sigma electron divergence in V direction in rad
self.sigmaH = (self.sigmaH_e**2 + self.wl*self.L_und/2/np.pi/np.pi)**0.5
self.sigmaV = (self.sigmaV_e**2 + self.wl*self.L_und/2/np.pi/np.pi)**0.5
self.sigmaHp = (self.sigmaHp_e**2 + self.wl/self.L_und/2)**0.5
self.sigmaVp = (self.sigmaVp_e**2 + self.wl/self.L_und/2)**0.5
#any update calcs to be called?
def setBeamline(self):
self.d_StoL = # Source-to-CRL1 distance, in m
self.d_Stof = # Source-to-focus distance, in m
def setLenses(self, propertyType, propertyVal, lensNum):
self.d_min = 3.0e-5 # Minimum thickness at the apex in m
self.stack_d = 50.0e-3 # Stack thickness in m
self.L1_n = np.array([1, 1, 1, 1, 1, 1, 2, 4, 8, 16]) # CRL1 number of lenses in each stack
self.L1_R = np.array([2.0e-3, 1.0e-3, 5.0e-4, 3.0e-4, 2.0e-4, 1.0e-4, 1.0e-4, 1.0e-4, 1.0e-4, 1.0e-4]) # CRL1 lens radius in each stack
self.L1_mater= np.array(["Be", "Be", "Be", "Be", "Be", "Be", "Be", "Be", "Be", "Be"]) # CRL1 lens material in each stack
self.L1_loc = np.array([4.5, 3.5, 2.5, 1.5, 0.5, -0.5, -1.5, -2.5, -3.5, -4.5])*stack_d # CRL1 lens stack location relative to center stack, positive means upstream
self.L1_HE = np.array([1.0e-6, 1.0e-6, 1.0e-6, 1.0e-6, 1.0e-6, 1.0e-6, 1.4e-6, 2.0e-6, 2.8e-6, 4.0e-6]) # CRL1 lens RMS thickness error
#any update calcs to be called?
def setLensCount(self, lensCount):
pass
def calc_lookup_table(self):
#------------------> Needs refactoring <--------------------------------
L1_D = np.zeros(L1_R.size) # CRL1 diameters for each stack
for i in range(L1_R.size):
L1_D[i] = lookup_diameter(L1_R[i])
L1_delta = materials_to_deltas(L1_mater, energy_keV) # delta values for CRL1 stacks
L1_mu = materials_to_linear_attenuation(L1_mater, energy_keV) # mu values for CRL1 stacks
L1_Feq = L1_R/(2*L1_n*L1_delta) + L1_loc # CRL1 equivalent f in m for each stack
L1_index_n = 2**L1_Feq.size # Total number of combinations for CRL1
L1_invF_list= np.zeros(L1_index_n) # List of equivalent 1/f in m^-1 for CRL1
for i in range(L1_index_n):
L1_invF_list[i] = np.sum(index_to_binary_list(i, L1_Feq.size)/L1_Feq)
# Sort the L1_invF list (to avoid zigzagging)
L1_invF_list_sort_indices = np.argsort(L1_invF_list)
L1_invF_list_sorted = L1_invF_list[L1_invF_list_sort_indices]
q1_list = 1/(L1_invF_list_sorted - 1/d_StoL1) # focal position of CRL1 for all configurations (sorted)
dq1_list = q1_list - (d_Stof - d_StoL1)
# Start generating focal size list as a function of CRL1 setting
sigma1H = (sigmaH**2 + (sigmaHp*d_StoL1)**2)**0.5 # sigma beam size before CRL1
sigma1V = (sigmaV**2 + (sigmaVp*d_StoL1)**2)**0.5 # sigma beam size before CRL1
L1_n1mud_list = np.zeros(L1_index_n) # List of n1*mu*d_min for all possible CRL1 configurations
L1_n1muR_list = np.zeros(L1_index_n) # List of n1*mu/R for all possible CRL1 configurations
aperL1H_list = np.zeros(L1_index_n) # absorption H aperture of CRL1 for all configurations
aperL1V_list = np.zeros(L1_index_n) # absorption V aperture of CRL1 for all configurations
diameter1_list = np.zeros(L1_index_n) # CRL1 diameter for all possible configurations
FWHM1H_list = np.zeros(L1_index_n) # H focal size at the focus of CRL1
FWHM1V_list = np.zeros(L1_index_n) # V focal size at the focus of CRL1
Strehl_list = np.zeros(L1_index_n) # Strehl ratio based on lens thickness error
for i in range(L1_index_n):
# absorption aperture is a function of CRL absorption/physical aperture, incident beam size, and physical slits
L1_n1mud_list[i] = np.sum(index_to_binary_list(L1_invF_list_sort_indices[i], L1_Feq.size)*np.array(L1_mu*L1_n*d_min))
L1_n1muR_list[i] = np.sum(index_to_binary_list(L1_invF_list_sort_indices[i], L1_Feq.size)*np.array(L1_mu*L1_n/L1_R))
solution = root_scalar(absorptionaperture, args=(L1_n1mud_list[i], sigma1H, L1_n1muR_list[i]), bracket=[0.0, 2*sigma1H], method='bisect')
aperL1H_list[i] = solution.root*2.0
solution = root_scalar(absorptionaperture, args=(L1_n1mud_list[i], sigma1V, L1_n1muR_list[i]), bracket=[0.0, 2*sigma1V], method='bisect')
aperL1V_list[i] = solution.root*2.0
mask = (np.array(index_to_binary_list(L1_invF_list_sort_indices[i], L1_Feq.size)) == 1)
if np.all(mask == False):
diameter1_list[i] = np.inf
else:
diameter1_list[i] = np.min(L1_D[mask])
aperL1H_list[i] = min(aperL1H_list[i], diameter1_list[i], slit1_H)
aperL1V_list[i] = min(aperL1V_list[i], diameter1_list[i], slit1_V)
phase_error_tmp = np.linalg.norm(index_to_binary_list(L1_invF_list_sort_indices[i], L1_Feq.size)*np.array(L1_HE*L1_delta)*2*np.pi/wl)
Strehl_list[i] = np.exp(-phase_error_tmp**2)
# FWHMbeam size at CRL1 focus
FWHM1H_list = ((0.88*wl*q1_list/aperL1H_list)**2 + (2.355*sigmaH*q1_list/d_StoL1)**2)**0.5
FWHM1V_list = ((0.88*wl*q1_list/aperL1V_list)**2 + (2.355*sigmaV*q1_list/d_StoL1)**2)**0.5
if flag_HE:
FWHM1H_list *= (Strehl_list)**(-0.5)
FWHM1V_list *= (Strehl_list)**(-0.5)
FWHM_list = (FWHM1H_list*FWHM1V_list)**0.5
#------------------> End refactoring <--------------------------------
def find_config(self):
# match desired fsize to lookup table
pass
def updateSlitSize(self, size, slit):
if slit = 'hor':
self.slit1_H = float(size) # H slit size before CRL 1
elif slit == 'vert':
self.slit1_V = float(size) # V slit size before CRL 1
else
# Need error handling
break
self.calc_lookup_table()
def updateE(self, energy):
# Energy variable sent from IOC as a string
self.energy = float(energy)
self.calc_lookup_table()
def updateFsize(self, focalSize):
# focalPoint variable sent from IOC as a string
self.focalSize = float(focalSize)
self.find_config()
def calc_lenses(self):
self.lenses = (self.energy * self.focalPoint) % 4096
pydev.iointr('new_lens_config', self.lenses)