1 | /* ======================================================================== *\
|
---|
2 | !
|
---|
3 | ! *
|
---|
4 | ! * This file is part of MARS, the MAGIC Analysis and Reconstruction
|
---|
5 | ! * Software. It is distributed to you in the hope that it can be a useful
|
---|
6 | ! * and timesaving tool in analyzing Data of imaging Cerenkov telescopes.
|
---|
7 | ! * It is distributed WITHOUT ANY WARRANTY.
|
---|
8 | ! *
|
---|
9 | ! * Permission to use, copy, modify and distribute this software and its
|
---|
10 | ! * documentation for any purpose is hereby granted without fee,
|
---|
11 | ! * provided that the above copyright notice appear in all copies and
|
---|
12 | ! * that both that copyright notice and this permission notice appear
|
---|
13 | ! * in supporting documentation. It is provided "as is" without express
|
---|
14 | ! * or implied warranty.
|
---|
15 | ! *
|
---|
16 | !
|
---|
17 | ! Author(s): Markus Gaug 05/2004 <mailto:markus@ifae.es>
|
---|
18 | !
|
---|
19 | ! Copyright: MAGIC Software Development, 2002-2004
|
---|
20 | !
|
---|
21 | !
|
---|
22 | \* ======================================================================== */
|
---|
23 |
|
---|
24 | //////////////////////////////////////////////////////////////////////////////
|
---|
25 | //
|
---|
26 | // MExtractTimeFastSpline
|
---|
27 | //
|
---|
28 | // Fast arrival Time extractor using a cubic spline algorithm of Numerical Recipes.
|
---|
29 | // It returns the position of the half maximum between absolute maximum
|
---|
30 | // and pedestal of the spline that interpolates the FADC slices.
|
---|
31 | //
|
---|
32 | // The precision of the half-maximum searches can be chosen by:
|
---|
33 | // SetPrecision().
|
---|
34 | //
|
---|
35 | // The precision of the maximum-finder is fixed to 0.025 FADC units.
|
---|
36 | //
|
---|
37 | //////////////////////////////////////////////////////////////////////////////
|
---|
38 | #include "MExtractTimeFastSpline.h"
|
---|
39 |
|
---|
40 | #include "MPedestalPix.h"
|
---|
41 |
|
---|
42 | #include "MLog.h"
|
---|
43 | #include "MLogManip.h"
|
---|
44 |
|
---|
45 |
|
---|
46 | ClassImp(MExtractTimeFastSpline);
|
---|
47 |
|
---|
48 | using namespace std;
|
---|
49 |
|
---|
50 | const Byte_t MExtractTimeFastSpline::fgHiGainFirst = 2;
|
---|
51 | const Byte_t MExtractTimeFastSpline::fgHiGainLast = 14;
|
---|
52 | const Byte_t MExtractTimeFastSpline::fgLoGainFirst = 3;
|
---|
53 | const Byte_t MExtractTimeFastSpline::fgLoGainLast = 14;
|
---|
54 | const Float_t MExtractTimeFastSpline::fgResolution = 0.003;
|
---|
55 | // --------------------------------------------------------------------------
|
---|
56 | //
|
---|
57 | // Default constructor.
|
---|
58 | //
|
---|
59 | // Calls:
|
---|
60 | // - SetRange(fgHiGainFirst, fgHiGainLast, fgLoGainFirst, fgLoGainLast)
|
---|
61 | //
|
---|
62 | // Initializes:
|
---|
63 | // - fResolution to fgResolution
|
---|
64 | //
|
---|
65 | MExtractTimeFastSpline::MExtractTimeFastSpline(const char *name, const char *title)
|
---|
66 | : fHiGainFirstDeriv(NULL), fLoGainFirstDeriv(NULL),
|
---|
67 | fHiGainSecondDeriv(NULL), fLoGainSecondDeriv(NULL)
|
---|
68 | {
|
---|
69 |
|
---|
70 | fName = name ? name : "MExtractTimeFastSpline";
|
---|
71 | fTitle = title ? title : "Calculate photons arrival time using a fast spline";
|
---|
72 |
|
---|
73 | SetResolution();
|
---|
74 | SetRange(fgHiGainFirst, fgHiGainLast, fgLoGainFirst, fgLoGainLast);
|
---|
75 |
|
---|
76 | }
|
---|
77 |
|
---|
78 | MExtractTimeFastSpline::~MExtractTimeFastSpline()
|
---|
79 | {
|
---|
80 |
|
---|
81 | if (fHiGainFirstDeriv)
|
---|
82 | delete fHiGainFirstDeriv;
|
---|
83 | if (fLoGainFirstDeriv)
|
---|
84 | delete fLoGainFirstDeriv;
|
---|
85 | if (fHiGainSecondDeriv)
|
---|
86 | delete fHiGainSecondDeriv;
|
---|
87 | if (fLoGainSecondDeriv)
|
---|
88 | delete fLoGainSecondDeriv;
|
---|
89 |
|
---|
90 | }
|
---|
91 |
|
---|
92 |
|
---|
93 | // --------------------------------------------------------------------------
|
---|
94 | //
|
---|
95 | // SetRange:
|
---|
96 | //
|
---|
97 | // Calls:
|
---|
98 | // - MExtractor::SetRange(hifirst,hilast,lofirst,lolast);
|
---|
99 | // - Deletes x, if not NULL
|
---|
100 | // - Creates x according to the range
|
---|
101 | //
|
---|
102 | void MExtractTimeFastSpline::SetRange(Byte_t hifirst, Byte_t hilast, Byte_t lofirst, Byte_t lolast)
|
---|
103 | {
|
---|
104 |
|
---|
105 | MExtractor::SetRange(hifirst,hilast,lofirst,lolast);
|
---|
106 |
|
---|
107 | if (fHiGainFirstDeriv)
|
---|
108 | delete fHiGainFirstDeriv;
|
---|
109 | if (fLoGainFirstDeriv)
|
---|
110 | delete fLoGainFirstDeriv;
|
---|
111 | if (fHiGainSecondDeriv)
|
---|
112 | delete fHiGainSecondDeriv;
|
---|
113 | if (fLoGainSecondDeriv)
|
---|
114 | delete fLoGainSecondDeriv;
|
---|
115 |
|
---|
116 | Int_t range = fHiGainLast - fHiGainFirst + 1;
|
---|
117 |
|
---|
118 | if (range < 2)
|
---|
119 | {
|
---|
120 | *fLog << warn << GetDescriptor()
|
---|
121 | << Form("%s%2i%s%2i%s",": Hi-Gain Extraction range [",(int)fHiGainFirst,","
|
---|
122 | ,fHiGainLast,"] too small, ") << endl;
|
---|
123 | *fLog << warn << GetDescriptor()
|
---|
124 | << " will move higher limit to obtain 4 slices " << endl;
|
---|
125 | SetRange(fHiGainFirst, fHiGainLast+4-range,fLoGainFirst,fLoGainLast);
|
---|
126 | range = fHiGainLast - fHiGainFirst + 1;
|
---|
127 | }
|
---|
128 |
|
---|
129 |
|
---|
130 | fHiGainFirstDeriv = new Float_t[range];
|
---|
131 | memset(fHiGainFirstDeriv,0,range*sizeof(Float_t));
|
---|
132 | fHiGainSecondDeriv = new Float_t[range];
|
---|
133 | memset(fHiGainSecondDeriv,0,range*sizeof(Float_t));
|
---|
134 |
|
---|
135 | range = fLoGainLast - fLoGainFirst + 1;
|
---|
136 |
|
---|
137 | if (range >= 2)
|
---|
138 | {
|
---|
139 |
|
---|
140 | fLoGainFirstDeriv = new Float_t[range];
|
---|
141 | memset(fLoGainFirstDeriv,0,range*sizeof(Float_t));
|
---|
142 | fLoGainSecondDeriv = new Float_t[range];
|
---|
143 | memset(fLoGainSecondDeriv,0,range*sizeof(Float_t));
|
---|
144 |
|
---|
145 | }
|
---|
146 | }
|
---|
147 |
|
---|
148 |
|
---|
149 | // --------------------------------------------------------------------------
|
---|
150 | //
|
---|
151 | // Calculates the arrival time for each pixel
|
---|
152 | //
|
---|
153 | void MExtractTimeFastSpline::FindTimeHiGain(Byte_t *first, Float_t &time, Float_t &dtime,
|
---|
154 | Byte_t &sat, const MPedestalPix &ped) const
|
---|
155 | {
|
---|
156 |
|
---|
157 | const Int_t range = fHiGainLast - fHiGainFirst + 1;
|
---|
158 | const Byte_t *end = first + range;
|
---|
159 | Byte_t *p = first;
|
---|
160 | Byte_t max = 0;
|
---|
161 | Byte_t maxpos = 0;
|
---|
162 |
|
---|
163 | //
|
---|
164 | // Check for saturation in all other slices
|
---|
165 | //
|
---|
166 | while (++p<end)
|
---|
167 | {
|
---|
168 | if (*p > max)
|
---|
169 | {
|
---|
170 | max = *p;
|
---|
171 | maxpos = p-first;
|
---|
172 | }
|
---|
173 |
|
---|
174 | if (*p >= fSaturationLimit)
|
---|
175 | {
|
---|
176 | sat++;
|
---|
177 | break;
|
---|
178 | }
|
---|
179 | }
|
---|
180 |
|
---|
181 | if (sat)
|
---|
182 | return;
|
---|
183 |
|
---|
184 | if (maxpos < 2)
|
---|
185 | return;
|
---|
186 |
|
---|
187 | Float_t pp;
|
---|
188 |
|
---|
189 | p = first;
|
---|
190 | fHiGainSecondDeriv[0] = 0.;
|
---|
191 | fHiGainFirstDeriv[0] = 0.;
|
---|
192 |
|
---|
193 | for (Int_t i=1;i<range-1;i++)
|
---|
194 | {
|
---|
195 | pp = fHiGainSecondDeriv[i-1] + 4.;
|
---|
196 | fHiGainSecondDeriv[i] = -1.0/pp;
|
---|
197 | fHiGainFirstDeriv [i] = *(p+2) - 2.* *(p+1) + *(p);
|
---|
198 | fHiGainFirstDeriv [i] = (6.0*fHiGainFirstDeriv[i]-fHiGainFirstDeriv[i-1])/pp;
|
---|
199 | p++;
|
---|
200 | }
|
---|
201 |
|
---|
202 | fHiGainSecondDeriv[range-1] = 0.;
|
---|
203 |
|
---|
204 | for (Int_t k=range-2;k>=0;k--)
|
---|
205 | fHiGainSecondDeriv[k] = fHiGainSecondDeriv[k]*fHiGainSecondDeriv[k+1] + fHiGainFirstDeriv[k];
|
---|
206 |
|
---|
207 | //
|
---|
208 | // Now find the maximum
|
---|
209 | //
|
---|
210 | Float_t step = 0.2; // start with step size of 1ns and loop again with the smaller one
|
---|
211 | Float_t lower = (Float_t)maxpos-1.;
|
---|
212 | Float_t upper = (Float_t)maxpos;
|
---|
213 | Float_t abmax = 0.;
|
---|
214 | Float_t x = lower;
|
---|
215 | Float_t y = 0.;
|
---|
216 | Float_t a = 1.;
|
---|
217 | Float_t b = 0.;
|
---|
218 | Int_t klo = maxpos-1;
|
---|
219 | Int_t khi = maxpos;
|
---|
220 | Float_t klocont = (Float_t)*(first+klo);
|
---|
221 | Float_t khicont = (Float_t)*(first+khi);
|
---|
222 | time = lower;
|
---|
223 |
|
---|
224 | //
|
---|
225 | // Search for the maximum, starting in interval maxpos-1. If no maximum is found, go to
|
---|
226 | // interval maxpos+1.
|
---|
227 | //
|
---|
228 | while (x<upper+0.1)
|
---|
229 | {
|
---|
230 |
|
---|
231 | x += step;
|
---|
232 | a -= step;
|
---|
233 | b += step;
|
---|
234 |
|
---|
235 | y = a*klocont
|
---|
236 | + b*khicont
|
---|
237 | + ( (a*a*a-a)*fHiGainSecondDeriv[klo]
|
---|
238 | +(b*b*b-b)*fHiGainSecondDeriv[khi] )/6.0;
|
---|
239 |
|
---|
240 | if (y > abmax)
|
---|
241 | {
|
---|
242 | abmax = y;
|
---|
243 | time = x;
|
---|
244 | }
|
---|
245 |
|
---|
246 | }
|
---|
247 |
|
---|
248 | if (time > upper-0.1)
|
---|
249 | {
|
---|
250 |
|
---|
251 | upper = (Float_t)maxpos+1.;
|
---|
252 | lower = (Float_t)maxpos;
|
---|
253 | x = lower;
|
---|
254 | a = 1.;
|
---|
255 | b = 0.;
|
---|
256 | khi = maxpos+1;
|
---|
257 | klo = maxpos;
|
---|
258 | klocont = (Float_t)*(first+klo);
|
---|
259 | khicont = (Float_t)*(first+khi);
|
---|
260 |
|
---|
261 | while (x<upper+0.1)
|
---|
262 | {
|
---|
263 |
|
---|
264 | x += step;
|
---|
265 | a -= step;
|
---|
266 | b += step;
|
---|
267 |
|
---|
268 | y = a* klocont
|
---|
269 | + b* khicont
|
---|
270 | + ( (a*a*a-a)*fHiGainSecondDeriv[klo]
|
---|
271 | +(b*b*b-b)*fHiGainSecondDeriv[khi] )/6.0;
|
---|
272 |
|
---|
273 | if (y > abmax)
|
---|
274 | {
|
---|
275 | abmax = y;
|
---|
276 | time = x;
|
---|
277 | }
|
---|
278 | }
|
---|
279 | }
|
---|
280 |
|
---|
281 | const Float_t up = time+step+0.015;
|
---|
282 |
|
---|
283 | x = time;
|
---|
284 | a = upper - x;
|
---|
285 | b = x - lower;
|
---|
286 |
|
---|
287 | step = 0.04; // step size of 83 ps
|
---|
288 |
|
---|
289 | while (x<up)
|
---|
290 | {
|
---|
291 |
|
---|
292 | x += step;
|
---|
293 | a -= step;
|
---|
294 | b += step;
|
---|
295 |
|
---|
296 | y = a* klocont
|
---|
297 | + b* khicont
|
---|
298 | + ( (a*a*a-a)*fHiGainSecondDeriv[klo]
|
---|
299 | +(b*b*b-b)*fHiGainSecondDeriv[khi] )/6.0;
|
---|
300 |
|
---|
301 | if (y > abmax)
|
---|
302 | {
|
---|
303 | abmax = y;
|
---|
304 | time = x;
|
---|
305 | }
|
---|
306 |
|
---|
307 | }
|
---|
308 |
|
---|
309 | Float_t lo = time-0.225;
|
---|
310 |
|
---|
311 | x = time;
|
---|
312 | a = upper - x;
|
---|
313 | b = x - lower;
|
---|
314 | khi--;
|
---|
315 | klo--;
|
---|
316 | klocont = (Float_t)*(first+klo);
|
---|
317 | khicont = (Float_t)*(first+khi);
|
---|
318 |
|
---|
319 | while (x>lo)
|
---|
320 | {
|
---|
321 |
|
---|
322 | x -= step;
|
---|
323 | a += step;
|
---|
324 | b -= step;
|
---|
325 |
|
---|
326 | y = a* klocont
|
---|
327 | + b* khicont
|
---|
328 | + ( (a*a*a-a)*fHiGainSecondDeriv[klo]
|
---|
329 | +(b*b*b-b)*fHiGainSecondDeriv[khi] )/6.0;
|
---|
330 |
|
---|
331 | if (y > abmax)
|
---|
332 | {
|
---|
333 | abmax = y;
|
---|
334 | time = x;
|
---|
335 | }
|
---|
336 |
|
---|
337 | }
|
---|
338 |
|
---|
339 | const Float_t pedes = ped.GetPedestal();
|
---|
340 | const Float_t halfmax = pedes + (abmax - pedes)/2.;
|
---|
341 |
|
---|
342 | //
|
---|
343 | // Now, loop from the maximum bin leftward down in order to find the position of the half maximum.
|
---|
344 | // First, find the right FADC slice:
|
---|
345 | //
|
---|
346 | klo = maxpos;
|
---|
347 | while (klo > maxpos-4)
|
---|
348 | {
|
---|
349 | if (*(first+klo) < (Byte_t)halfmax)
|
---|
350 | break;
|
---|
351 | klo--;
|
---|
352 | }
|
---|
353 |
|
---|
354 | //
|
---|
355 | // Loop from the beginning of the slice upwards to reach the halfmax:
|
---|
356 | // With means of bisection:
|
---|
357 | //
|
---|
358 | x = (Float_t)klo;
|
---|
359 | a = 1.;
|
---|
360 | b = 0.;
|
---|
361 | klocont = (Float_t)*(first+klo);
|
---|
362 | khicont = (Float_t)*(first+klo+1);
|
---|
363 | time = x;
|
---|
364 |
|
---|
365 | step = 0.5;
|
---|
366 | Bool_t back = kFALSE;
|
---|
367 |
|
---|
368 | while (step > fResolution)
|
---|
369 | {
|
---|
370 |
|
---|
371 | if (back)
|
---|
372 | {
|
---|
373 | x -= step;
|
---|
374 | a += step;
|
---|
375 | b -= step;
|
---|
376 | }
|
---|
377 | else
|
---|
378 | {
|
---|
379 | x += step;
|
---|
380 | a -= step;
|
---|
381 | b += step;
|
---|
382 | }
|
---|
383 |
|
---|
384 | y = a*klocont
|
---|
385 | + b*khicont
|
---|
386 | + ( (a*a*a-a)*fHiGainSecondDeriv[klo]
|
---|
387 | +(b*b*b-b)*fHiGainSecondDeriv[khi] )/6.0;
|
---|
388 |
|
---|
389 | if (y >= halfmax)
|
---|
390 | back = kTRUE;
|
---|
391 | else
|
---|
392 | back = kFALSE;
|
---|
393 |
|
---|
394 | step /= 2.;
|
---|
395 | }
|
---|
396 |
|
---|
397 | time = (Float_t)fHiGainFirst + x;
|
---|
398 | dtime = fResolution;
|
---|
399 | }
|
---|
400 |
|
---|
401 |
|
---|
402 | // --------------------------------------------------------------------------
|
---|
403 | //
|
---|
404 | // Calculates the arrival time for each pixel
|
---|
405 | //
|
---|
406 | void MExtractTimeFastSpline::FindTimeLoGain(Byte_t *first, Float_t &time, Float_t &dtime,
|
---|
407 | Byte_t &sat, const MPedestalPix &ped) const
|
---|
408 | {
|
---|
409 |
|
---|
410 | const Int_t range = fLoGainLast - fLoGainFirst + 1;
|
---|
411 | const Byte_t *end = first + range;
|
---|
412 | Byte_t *p = first;
|
---|
413 | Byte_t max = 0;
|
---|
414 | Byte_t maxpos = 0;
|
---|
415 |
|
---|
416 | //
|
---|
417 | // Check for saturation in all other slices
|
---|
418 | //
|
---|
419 | while (++p<end)
|
---|
420 | {
|
---|
421 | if (*p > max)
|
---|
422 | {
|
---|
423 | max = *p;
|
---|
424 | maxpos = p-first;
|
---|
425 | }
|
---|
426 |
|
---|
427 | if (*p >= fSaturationLimit)
|
---|
428 | {
|
---|
429 | sat++;
|
---|
430 | break;
|
---|
431 | }
|
---|
432 | }
|
---|
433 |
|
---|
434 | if (sat)
|
---|
435 | return;
|
---|
436 |
|
---|
437 | if (maxpos < 2)
|
---|
438 | return;
|
---|
439 |
|
---|
440 | Float_t pp;
|
---|
441 |
|
---|
442 | p = first;
|
---|
443 | fLoGainSecondDeriv[0] = 0.;
|
---|
444 | fLoGainFirstDeriv[0] = 0.;
|
---|
445 |
|
---|
446 | for (Int_t i=1;i<range-1;i++)
|
---|
447 | {
|
---|
448 | pp = fLoGainSecondDeriv[i-1] + 4.;
|
---|
449 | fLoGainSecondDeriv[i] = -1.0/pp;
|
---|
450 | fLoGainFirstDeriv [i] = *(p+2) - 2.* *(p+1) + *(p);
|
---|
451 | fLoGainFirstDeriv [i] = (6.0*fLoGainFirstDeriv[i]-fLoGainFirstDeriv[i-1])/pp;
|
---|
452 | p++;
|
---|
453 | }
|
---|
454 |
|
---|
455 | fLoGainSecondDeriv[range-1] = 0.;
|
---|
456 |
|
---|
457 | for (Int_t k=range-2;k>=0;k--)
|
---|
458 | fLoGainSecondDeriv[k] = fLoGainSecondDeriv[k]*fLoGainSecondDeriv[k+1] + fLoGainFirstDeriv[k];
|
---|
459 |
|
---|
460 | //
|
---|
461 | // Now find the maximum
|
---|
462 | //
|
---|
463 | Float_t step = 0.2; // start with step size of 1ns and loop again with the smaller one
|
---|
464 | Float_t lower = (Float_t)maxpos-1.;
|
---|
465 | Float_t upper = (Float_t)maxpos;
|
---|
466 | Float_t abmax = 0.;
|
---|
467 | Float_t x = lower;
|
---|
468 | Float_t y = 0.;
|
---|
469 | Float_t a = 1.;
|
---|
470 | Float_t b = 0.;
|
---|
471 | Int_t klo = maxpos-1;
|
---|
472 | Int_t khi = maxpos;
|
---|
473 | Float_t klocont = (Float_t)*(first+klo);
|
---|
474 | Float_t khicont = (Float_t)*(first+khi);
|
---|
475 | time = lower;
|
---|
476 |
|
---|
477 | //
|
---|
478 | // Search for the maximum, starting in interval maxpos-1. If no maximum is found, go to
|
---|
479 | // interval maxpos+1.
|
---|
480 | //
|
---|
481 | while (x<upper+0.1)
|
---|
482 | {
|
---|
483 |
|
---|
484 | x += step;
|
---|
485 | a -= step;
|
---|
486 | b += step;
|
---|
487 |
|
---|
488 | y = a*klocont
|
---|
489 | + b*khicont
|
---|
490 | + ( (a*a*a-a)*fLoGainSecondDeriv[klo]
|
---|
491 | +(b*b*b-b)*fLoGainSecondDeriv[khi] )/6.0;
|
---|
492 |
|
---|
493 | if (y > abmax)
|
---|
494 | {
|
---|
495 | abmax = y;
|
---|
496 | time = x;
|
---|
497 | }
|
---|
498 |
|
---|
499 | }
|
---|
500 |
|
---|
501 | if (time > upper-0.1)
|
---|
502 | {
|
---|
503 |
|
---|
504 | upper = (Float_t)maxpos+1.;
|
---|
505 | lower = (Float_t)maxpos;
|
---|
506 | x = lower;
|
---|
507 | a = 1.;
|
---|
508 | b = 0.;
|
---|
509 | khi = maxpos+1;
|
---|
510 | klo = maxpos;
|
---|
511 | klocont = (Float_t)*(first+klo);
|
---|
512 | khicont = (Float_t)*(first+khi);
|
---|
513 |
|
---|
514 | while (x<upper+0.1)
|
---|
515 | {
|
---|
516 |
|
---|
517 | x += step;
|
---|
518 | a -= step;
|
---|
519 | b += step;
|
---|
520 |
|
---|
521 | y = a* klocont
|
---|
522 | + b* khicont
|
---|
523 | + ( (a*a*a-a)*fLoGainSecondDeriv[klo]
|
---|
524 | +(b*b*b-b)*fLoGainSecondDeriv[khi] )/6.0;
|
---|
525 |
|
---|
526 | if (y > abmax)
|
---|
527 | {
|
---|
528 | abmax = y;
|
---|
529 | time = x;
|
---|
530 | }
|
---|
531 | }
|
---|
532 | }
|
---|
533 |
|
---|
534 | const Float_t up = time+step+0.015;
|
---|
535 |
|
---|
536 | x = time;
|
---|
537 | a = upper - x;
|
---|
538 | b = x - lower;
|
---|
539 |
|
---|
540 | step = 0.025; // step size of 165 ps
|
---|
541 |
|
---|
542 | while (x<up)
|
---|
543 | {
|
---|
544 |
|
---|
545 | x += step;
|
---|
546 | a -= step;
|
---|
547 | b += step;
|
---|
548 |
|
---|
549 | y = a* klocont
|
---|
550 | + b* khicont
|
---|
551 | + ( (a*a*a-a)*fLoGainSecondDeriv[klo]
|
---|
552 | +(b*b*b-b)*fLoGainSecondDeriv[khi] )/6.0;
|
---|
553 |
|
---|
554 | if (y > abmax)
|
---|
555 | {
|
---|
556 | abmax = y;
|
---|
557 | time = x;
|
---|
558 | }
|
---|
559 |
|
---|
560 | }
|
---|
561 |
|
---|
562 | Float_t lo = time-0.225;
|
---|
563 |
|
---|
564 | x = time;
|
---|
565 | a = upper - x;
|
---|
566 | b = x - lower;
|
---|
567 | khi--;
|
---|
568 | klo--;
|
---|
569 | klocont = (Float_t)*(first+klo);
|
---|
570 | khicont = (Float_t)*(first+khi);
|
---|
571 |
|
---|
572 | while (x>lo)
|
---|
573 | {
|
---|
574 |
|
---|
575 | x -= step;
|
---|
576 | a += step;
|
---|
577 | b -= step;
|
---|
578 |
|
---|
579 | y = a* klocont
|
---|
580 | + b* khicont
|
---|
581 | + ( (a*a*a-a)*fLoGainSecondDeriv[klo]
|
---|
582 | +(b*b*b-b)*fLoGainSecondDeriv[khi] )/6.0;
|
---|
583 |
|
---|
584 | if (y > abmax)
|
---|
585 | {
|
---|
586 | abmax = y;
|
---|
587 | time = x;
|
---|
588 | }
|
---|
589 |
|
---|
590 | }
|
---|
591 |
|
---|
592 | const Float_t pedes = ped.GetPedestal();
|
---|
593 | const Float_t halfmax = pedes + (abmax - pedes)/2.;
|
---|
594 |
|
---|
595 | //
|
---|
596 | // Now, loop from the maximum bin leftward down in order to find the position of the half maximum.
|
---|
597 | // First, find the right FADC slice:
|
---|
598 | //
|
---|
599 | klo = maxpos;
|
---|
600 | while (klo > maxpos-4)
|
---|
601 | {
|
---|
602 | if (*(first+klo) < (Byte_t)halfmax)
|
---|
603 | break;
|
---|
604 | klo--;
|
---|
605 | }
|
---|
606 |
|
---|
607 | //
|
---|
608 | // Loop from the beginning of the slice upwards to reach the halfmax:
|
---|
609 | // With means of bisection:
|
---|
610 | //
|
---|
611 | x = (Float_t)klo;
|
---|
612 | a = 1.;
|
---|
613 | b = 0.;
|
---|
614 | klocont = (Float_t)*(first+klo);
|
---|
615 | khicont = (Float_t)*(first+klo+1);
|
---|
616 | time = x;
|
---|
617 |
|
---|
618 | step = 0.5;
|
---|
619 | Bool_t back = kFALSE;
|
---|
620 |
|
---|
621 | while (step > fResolution)
|
---|
622 | {
|
---|
623 |
|
---|
624 | if (back)
|
---|
625 | {
|
---|
626 | x -= step;
|
---|
627 | a += step;
|
---|
628 | b -= step;
|
---|
629 | }
|
---|
630 | else
|
---|
631 | {
|
---|
632 | x += step;
|
---|
633 | a -= step;
|
---|
634 | b += step;
|
---|
635 | }
|
---|
636 |
|
---|
637 | y = a*klocont
|
---|
638 | + b*khicont
|
---|
639 | + ( (a*a*a-a)*fLoGainSecondDeriv[klo]
|
---|
640 | +(b*b*b-b)*fLoGainSecondDeriv[khi] )/6.0;
|
---|
641 |
|
---|
642 | if (y >= halfmax)
|
---|
643 | back = kTRUE;
|
---|
644 | else
|
---|
645 | back = kFALSE;
|
---|
646 |
|
---|
647 | step /= 2.;
|
---|
648 | }
|
---|
649 |
|
---|
650 | time = (Float_t)fLoGainFirst + x;
|
---|
651 | dtime = fResolution;
|
---|
652 | }
|
---|
653 |
|
---|
654 |
|
---|