[642b259] | 1 | /** |
---|
| 2 | This software was developed by the University of Tennessee as part of the |
---|
| 3 | Distributed Data Analysis of Neutron Scattering Experiments (DANSE) |
---|
| 4 | project funded by the US National Science Foundation. |
---|
| 5 | |
---|
| 6 | If you use DANSE applications to do scientific research that leads to |
---|
| 7 | publication, we ask that you acknowledge the use of the software with the |
---|
| 8 | following sentence: |
---|
| 9 | |
---|
| 10 | "This work benefited from DANSE software developed under NSF award DMR-0520547." |
---|
| 11 | |
---|
| 12 | copyright 2009, University of Tennessee |
---|
| 13 | */ |
---|
| 14 | #include "smearer.hh" |
---|
| 15 | #include <stdio.h> |
---|
| 16 | #include <math.h> |
---|
| 17 | using namespace std; |
---|
| 18 | |
---|
[b23b722] | 19 | #if defined(_MSC_VER) |
---|
[09e89b7] | 20 | extern "C" { |
---|
[b23b722] | 21 | #include "winFuncs.h" |
---|
[09e89b7] | 22 | } |
---|
[b23b722] | 23 | #endif |
---|
[642b259] | 24 | /** |
---|
| 25 | * Constructor for BaseSmearer |
---|
| 26 | * |
---|
| 27 | * @param qmin: minimum Q value |
---|
| 28 | * @param qmax: maximum Q value |
---|
| 29 | * @param nbins: number of Q bins |
---|
| 30 | */ |
---|
| 31 | BaseSmearer :: BaseSmearer(double qmin, double qmax, int nbins) { |
---|
| 32 | // Number of bins |
---|
| 33 | this->nbins = nbins; |
---|
| 34 | this->qmin = qmin; |
---|
| 35 | this->qmax = qmax; |
---|
| 36 | // Flag to keep track of whether we have a smearing matrix or |
---|
| 37 | // whether we need to compute one |
---|
| 38 | has_matrix = false; |
---|
| 39 | even_binning = true; |
---|
| 40 | }; |
---|
| 41 | |
---|
| 42 | /** |
---|
| 43 | * Constructor for BaseSmearer |
---|
| 44 | * |
---|
| 45 | * Used for uneven binning |
---|
| 46 | * @param q: array of Q values |
---|
| 47 | * @param nbins: number of Q bins |
---|
| 48 | */ |
---|
| 49 | BaseSmearer :: BaseSmearer(double* q, int nbins) { |
---|
| 50 | // Number of bins |
---|
| 51 | this->nbins = nbins; |
---|
| 52 | this->q_values = q; |
---|
| 53 | // Flag to keep track of whether we have a smearing matrix or |
---|
| 54 | // whether we need to compute one |
---|
| 55 | has_matrix = false; |
---|
| 56 | even_binning = false; |
---|
| 57 | }; |
---|
| 58 | |
---|
| 59 | /** |
---|
| 60 | * Constructor for SlitSmearer |
---|
| 61 | * |
---|
| 62 | * @param width: slit width in Q units |
---|
| 63 | * @param height: slit height in Q units |
---|
| 64 | * @param qmin: minimum Q value |
---|
| 65 | * @param qmax: maximum Q value |
---|
| 66 | * @param nbins: number of Q bins |
---|
| 67 | */ |
---|
| 68 | SlitSmearer :: SlitSmearer(double width, double height, double qmin, double qmax, int nbins) : |
---|
| 69 | BaseSmearer(qmin, qmax, nbins){ |
---|
| 70 | this->height = height; |
---|
| 71 | this->width = width; |
---|
| 72 | }; |
---|
| 73 | |
---|
| 74 | /** |
---|
| 75 | * Constructor for SlitSmearer |
---|
| 76 | * |
---|
| 77 | * @param width: slit width in Q units |
---|
| 78 | * @param height: slit height in Q units |
---|
| 79 | * @param q: array of Q values |
---|
| 80 | * @param nbins: number of Q bins |
---|
| 81 | */ |
---|
| 82 | SlitSmearer :: SlitSmearer(double width, double height, double* q, int nbins) : |
---|
| 83 | BaseSmearer(q, nbins){ |
---|
| 84 | this->height = height; |
---|
| 85 | this->width = width; |
---|
| 86 | }; |
---|
| 87 | |
---|
| 88 | /** |
---|
| 89 | * Constructor for QSmearer |
---|
| 90 | * |
---|
| 91 | * @param width: array slit widths for each Q point, in Q units |
---|
| 92 | * @param qmin: minimum Q value |
---|
| 93 | * @param qmax: maximum Q value |
---|
| 94 | * @param nbins: number of Q bins |
---|
| 95 | */ |
---|
| 96 | QSmearer :: QSmearer(double* width, double qmin, double qmax, int nbins) : |
---|
| 97 | BaseSmearer(qmin, qmax, nbins){ |
---|
| 98 | this->width = width; |
---|
| 99 | }; |
---|
| 100 | |
---|
| 101 | /** |
---|
| 102 | * Constructor for QSmearer |
---|
| 103 | * |
---|
| 104 | * @param width: array slit widths for each Q point, in Q units |
---|
| 105 | * @param q: array of Q values |
---|
| 106 | * @param nbins: number of Q bins |
---|
| 107 | */ |
---|
| 108 | QSmearer :: QSmearer(double* width, double* q, int nbins) : |
---|
| 109 | BaseSmearer(q, nbins){ |
---|
| 110 | this->width = width; |
---|
| 111 | }; |
---|
| 112 | |
---|
| 113 | /** |
---|
| 114 | * Compute the slit smearing matrix |
---|
| 115 | * |
---|
| 116 | * For even binning (q_min to q_max with nbins): |
---|
| 117 | * |
---|
| 118 | * step = (q_max-q_min)/(nbins-1) |
---|
| 119 | * first bin goes from q_min to q_min+step |
---|
| 120 | * last bin goes from q_max to q_max+step |
---|
| 121 | * |
---|
| 122 | * For binning according to q array: |
---|
| 123 | * |
---|
| 124 | * Each q point represents a bin going from half the distance between it |
---|
| 125 | * and the previous point to half the distance between it and the next point. |
---|
| 126 | * |
---|
| 127 | * Example: bin i goes from (q_values[i-1]+q_values[i])/2 to (q_values[i]+q_values[i+1])/2 |
---|
| 128 | * |
---|
| 129 | * The exceptions are the first and last bins, which are centered at the first and |
---|
| 130 | * last q-values, respectively. The width of the first and last bins is the distance between |
---|
| 131 | * their respective neighboring q-value. |
---|
| 132 | */ |
---|
| 133 | void SlitSmearer :: compute_matrix(){ |
---|
| 134 | |
---|
| 135 | weights = new vector<double>(nbins*nbins,0); |
---|
| 136 | |
---|
| 137 | // Check the length of the data |
---|
| 138 | if (nbins<2) return; |
---|
| 139 | int npts_h = height>0.0 ? npts : 1; |
---|
| 140 | int npts_w = width>0.0 ? npts : 1; |
---|
| 141 | |
---|
| 142 | // If both height and width are great than zero, |
---|
| 143 | // modify the number of points in each direction so |
---|
| 144 | // that the total number of points is still what |
---|
| 145 | // the user would expect (downgrade resolution) |
---|
| 146 | //if(npts_h>1 && npts_w>1){ |
---|
| 147 | // npts_h = (int)ceil(sqrt((double)npts)); |
---|
| 148 | // npts_w = npts_h; |
---|
| 149 | //} |
---|
| 150 | double shift_h, shift_w, hbin_size, wbin_size; |
---|
| 151 | // Make sure height and width are all positive (FWMH/2) |
---|
| 152 | // Assumption; height and width are all same for all q points |
---|
| 153 | if(npts_h == 1){ |
---|
| 154 | shift_h = 0.0; |
---|
| 155 | } else { |
---|
| 156 | shift_h = fabs(height); |
---|
| 157 | } |
---|
| 158 | if(npts_w == 1){ |
---|
| 159 | shift_w = 0.0; |
---|
| 160 | } else { |
---|
| 161 | shift_w = fabs(width); |
---|
| 162 | } |
---|
| 163 | // size of the h bin and w bin |
---|
| 164 | hbin_size = shift_h / nbins; |
---|
| 165 | wbin_size = shift_w / nbins; |
---|
| 166 | |
---|
| 167 | // Loop over all q-values |
---|
| 168 | for(int i=0; i<nbins; i++) { |
---|
| 169 | // Find Weights |
---|
| 170 | // Find q where the resolution smearing calculation of I(q) occurs |
---|
[2e2452b] | 171 | double q, q_min, q_max, q_0=0.0; |
---|
[642b259] | 172 | get_bin_range(i, &q, &q_min, &q_max); |
---|
| 173 | // Block q becomes <=0 |
---|
| 174 | if (q <= 0){ |
---|
| 175 | continue; |
---|
| 176 | } |
---|
| 177 | // Find q[0] value to normalize the weight later, |
---|
| 178 | // otherwise, we will have a precision problem. |
---|
| 179 | if (i == 0){ |
---|
| 180 | q_0 = q; |
---|
| 181 | } |
---|
| 182 | // Loop over all qj-values |
---|
| 183 | for(int j=0; j<nbins; j++) { |
---|
| 184 | double q_j, q_high, q_low; |
---|
| 185 | // Calculate bin size of q_j |
---|
| 186 | get_bin_range(j, &q_j, &q_low, &q_high); |
---|
| 187 | // Block q_j becomes <=0 |
---|
| 188 | if (q_j <= 0){ |
---|
| 189 | continue; |
---|
| 190 | } |
---|
| 191 | // Check q_low that can not be negative. |
---|
| 192 | if (q_low < 0.0){ |
---|
| 193 | q_low = 0.0; |
---|
| 194 | } |
---|
| 195 | // default parameter values |
---|
| 196 | (*weights)[i*nbins+j] = 0.0; |
---|
| 197 | // protect for negative q |
---|
| 198 | if (q <= 0.0 || q_j <= 0.0){ |
---|
| 199 | continue; |
---|
| 200 | } |
---|
| 201 | double shift_w = 0.0; |
---|
| 202 | // Condition: zero slit smear. |
---|
| 203 | if (npts_w == 1 && npts_h == 1){ |
---|
| 204 | if(q_j == q) { |
---|
| 205 | (*weights)[i*nbins+j] = 1.0; |
---|
| 206 | } |
---|
| 207 | } |
---|
| 208 | //Condition:Smear weight integration for width >0 when the height (=0) does not present. |
---|
| 209 | //Or height << width. |
---|
| 210 | else if((npts_w!=1 && npts_h == 1)|| (npts_w!=1 && npts_h != 1 && width/height > 100.0)){ |
---|
| 211 | shift_w = width; |
---|
| 212 | //del_w = width/((double)npts_w-1.0); |
---|
| 213 | double q_shifted_low = q - shift_w; |
---|
| 214 | // High limit of the resolution range |
---|
| 215 | double q_shifted_high = q + shift_w; |
---|
| 216 | // Go through all the q_js for weighting those points |
---|
| 217 | if(q_j >= q_shifted_low && q_j <= q_shifted_high) { |
---|
| 218 | // The weighting factor comes, |
---|
| 219 | // Give some weight (delq_bin) for the q_j within the resolution range |
---|
| 220 | // Weight should be same for all qs except |
---|
| 221 | // for the q bin size at j. |
---|
| 222 | // Note that the division by q_0 is only due to the precision problem |
---|
| 223 | // where q_high - q_low gets to very small. |
---|
| 224 | // Later, it will be normalized again. |
---|
| 225 | (*weights)[i*nbins+j] += (q_high - q_low)/q_0 ; |
---|
| 226 | } |
---|
| 227 | } |
---|
| 228 | else{ |
---|
| 229 | // Loop for width (;Height is analytical.) |
---|
| 230 | // Condition: height >>> width, otherwise, below is not accurate enough. |
---|
| 231 | // Smear weight numerical iteration for width >0 when the height (>0) presents. |
---|
| 232 | // When width = 0, the numerical iteration will be skipped. |
---|
| 233 | // The resolution calculation for the height is done by direct integration, |
---|
| 234 | // assuming the I(q'=sqrt(q_j^2-(q+shift_w)^2)) is constant within a q' bin, [q_high, q_low]. |
---|
| 235 | // In general, this weight numerical iteration for width >0 might be a rough approximation, |
---|
| 236 | // but it must be good enough when height >>> width. |
---|
| 237 | for(int k=(-npts_w + 1); k<npts_w; k++){ |
---|
| 238 | if(npts_w!=1){ |
---|
| 239 | shift_w = width/((double)npts_w-1.0)*(double)k; |
---|
| 240 | } |
---|
| 241 | // For each q-value, compute the weight of each other q-bin |
---|
| 242 | // in the I(q) array |
---|
| 243 | // Low limit of the resolution range |
---|
| 244 | double q_shift = q + shift_w; |
---|
| 245 | if (q_shift < 0.0){ |
---|
| 246 | q_shift = 0.0; |
---|
| 247 | } |
---|
| 248 | double q_shifted_low = q_shift; |
---|
| 249 | // High limit of the resolution range |
---|
| 250 | double q_shifted_high = sqrt(q_shift * q_shift + shift_h * shift_h); |
---|
| 251 | |
---|
| 252 | |
---|
| 253 | // Go through all the q_js for weighting those points |
---|
| 254 | if(q_j >= q_shifted_low && q_j <= q_shifted_high) { |
---|
| 255 | // The weighting factor comes, |
---|
| 256 | // Give some weight (delq_bin) for the q_j within the resolution range |
---|
| 257 | // Weight should be same for all qs except |
---|
| 258 | // for the q bin size at j. |
---|
| 259 | // Note that the division by q_0 is only due to the precision problem |
---|
| 260 | // where q_high - q_low gets to very small. |
---|
| 261 | // Later, it will be normalized again. |
---|
| 262 | |
---|
| 263 | // The fabs below are not necessary but in case: the weight should never be imaginary. |
---|
| 264 | // At the edge of each sub_width. weight += u(at q_high bin) - u(0), where u(0) = 0, |
---|
| 265 | // and weighted by (2.0* npts_w -1.0)once for each q. |
---|
[510e7ad] | 266 | //if (q == q_j) { |
---|
| 267 | if (q_low <= q_shift && q_high > q_shift) { |
---|
| 268 | //if (k==0) |
---|
| 269 | (*weights)[i*nbins+j] += (sqrt(fabs((q_high)*(q_high)-q_shift * q_shift)))/q_0;// * (2.0*double(npts_w)-1.0); |
---|
[642b259] | 270 | } |
---|
| 271 | // For the rest of sub_width. weight += u(at q_high bin) - u(at q_low bin) |
---|
[510e7ad] | 272 | else{// if (u > 0.0){ |
---|
[642b259] | 273 | (*weights)[i*nbins+j] += (sqrt(fabs((q_high)*(q_high)- q_shift * q_shift))-sqrt(fabs((q_low)*(q_low)- q_shift * q_shift)))/q_0 ; |
---|
| 274 | } |
---|
| 275 | } |
---|
| 276 | } |
---|
| 277 | } |
---|
| 278 | } |
---|
| 279 | } |
---|
| 280 | }; |
---|
| 281 | |
---|
| 282 | /** |
---|
| 283 | * Compute the point smearing matrix |
---|
| 284 | */ |
---|
| 285 | void QSmearer :: compute_matrix(){ |
---|
| 286 | weights = new vector<double>(nbins*nbins,0); |
---|
| 287 | |
---|
| 288 | // Loop over all q-values |
---|
| 289 | double q, q_min, q_max; |
---|
| 290 | double q_j, q_jmax, q_jmin; |
---|
| 291 | for(int i=0; i<nbins; i++) { |
---|
| 292 | get_bin_range(i, &q, &q_min, &q_max); |
---|
| 293 | |
---|
| 294 | for(int j=0; j<nbins; j++) { |
---|
| 295 | get_bin_range(j, &q_j, &q_jmin, &q_jmax); |
---|
| 296 | |
---|
| 297 | // Compute the fraction of the Gaussian contributing |
---|
| 298 | // to the q_j bin between q_jmin and q_jmax |
---|
[510e7ad] | 299 | long double value = erf( (q_jmax-q)/(sqrt(2.0)*width[i]) ); |
---|
[642b259] | 300 | value -= erf( (q_jmin-q)/(sqrt(2.0)*width[i]) ); |
---|
| 301 | (*weights)[i*nbins+j] += value; |
---|
| 302 | } |
---|
| 303 | } |
---|
| 304 | } |
---|
| 305 | |
---|
| 306 | /** |
---|
| 307 | * Computes the Q range of a given bin of the Q distribution. |
---|
| 308 | * The range is computed according the the data distribution that |
---|
| 309 | * was given to the object at initialization. |
---|
| 310 | * |
---|
| 311 | * @param i: number of the bin in the distribution |
---|
| 312 | * @param q: q-value of bin i |
---|
| 313 | * @param q_min: lower bound of the bin |
---|
| 314 | * @param q_max: higher bound of the bin |
---|
| 315 | * |
---|
| 316 | */ |
---|
| 317 | int BaseSmearer :: get_bin_range(int i, double* q, double* q_min, double* q_max) { |
---|
| 318 | if (even_binning) { |
---|
| 319 | double step = (qmax-qmin)/((double)nbins-1.0); |
---|
| 320 | *q = qmin + (double)i*step; |
---|
| 321 | *q_min = *q - 0.5*step; |
---|
| 322 | *q_max = *q + 0.5*step; |
---|
| 323 | return 1; |
---|
| 324 | } else if (i>=0 && i<nbins) { |
---|
| 325 | *q = q_values[i]; |
---|
| 326 | if (i==0) { |
---|
| 327 | double step = (q_values[1]-q_values[0])/2.0; |
---|
| 328 | *q_min = *q - step; |
---|
| 329 | *q_max = *q + step; |
---|
| 330 | } else if (i==nbins-1) { |
---|
| 331 | double step = (q_values[i]-q_values[i-1])/2.0; |
---|
| 332 | *q_min = *q - step; |
---|
| 333 | *q_max = *q + step; |
---|
| 334 | } else { |
---|
| 335 | *q_min = *q - (q_values[i]-q_values[i-1])/2.0; |
---|
| 336 | *q_max = *q + (q_values[i+1]-q_values[i])/2.0; |
---|
| 337 | } |
---|
| 338 | return 1; |
---|
| 339 | } |
---|
| 340 | return -1; |
---|
| 341 | } |
---|
| 342 | |
---|
| 343 | /** |
---|
| 344 | * Perform smearing by applying the smearing matrix to the input Q array |
---|
| 345 | */ |
---|
| 346 | void BaseSmearer :: smear(double *iq_in, double *iq_out, int first_bin, int last_bin){ |
---|
| 347 | |
---|
| 348 | // If we haven't computed the smearing matrix, do it now |
---|
| 349 | if(!has_matrix) { |
---|
| 350 | compute_matrix(); |
---|
| 351 | has_matrix = true; |
---|
| 352 | } |
---|
| 353 | |
---|
| 354 | // Loop over q-values and multiply apply matrix |
---|
| 355 | for(int q_i=first_bin; q_i<=last_bin; q_i++){ |
---|
| 356 | double sum = 0.0; |
---|
| 357 | double counts = 0.0; |
---|
| 358 | |
---|
| 359 | for(int i=first_bin; i<=last_bin; i++){ |
---|
| 360 | // Skip if weight is less than 1e-03(this value is much smaller than |
---|
| 361 | // the weight at the 3*sigma distance |
---|
| 362 | // Will speed up a little bit... |
---|
| 363 | if ((*weights)[q_i*nbins+i] < 1.0e-003){ |
---|
| 364 | continue; |
---|
| 365 | } |
---|
| 366 | sum += iq_in[i] * (*weights)[q_i*nbins+i]; |
---|
| 367 | counts += (*weights)[q_i*nbins+i]; |
---|
| 368 | } |
---|
| 369 | |
---|
| 370 | // Normalize counts |
---|
| 371 | iq_out[q_i] = (counts>0.0) ? sum/counts : 0.0; |
---|
| 372 | } |
---|
| 373 | } |
---|