| 1 | from __future__ import absolute_import
|
|---|
| 2 |
|
|---|
| 3 | import collections
|
|---|
| 4 | import math
|
|---|
| 5 | import numpy as np
|
|---|
| 6 | import pp
|
|---|
| 7 |
|
|---|
| 8 | import common.commonobjects as co
|
|---|
| 9 |
|
|---|
| 10 | def fftshift(data, shift):
|
|---|
| 11 | """Method to shift a 2d complex data array by applying the
|
|---|
| 12 | given phase shift to its Fourier transform.
|
|---|
| 13 | """
|
|---|
| 14 | # move centre of image to array origin
|
|---|
| 15 | temp = numpy.fft.fftshift(data)
|
|---|
| 16 | # 2d fft
|
|---|
| 17 | temp = numpy.fft.fft2(temp)
|
|---|
| 18 | # apply phase shift
|
|---|
| 19 | temp *= shift
|
|---|
| 20 | # transform and shift back
|
|---|
| 21 | temp = numpy.fft.ifft2(temp)
|
|---|
| 22 | temp = numpy.fft.fftshift(temp)
|
|---|
| 23 |
|
|---|
| 24 | return temp
|
|---|
| 25 |
|
|---|
| 26 |
|
|---|
| 27 | class DoubleFourier(object):
|
|---|
| 28 | """Class to compute interferograms.
|
|---|
| 29 | """
|
|---|
| 30 |
|
|---|
| 31 | def __init__(self, parameters, previous_results, job_server):
|
|---|
| 32 | self.parameters = parameters
|
|---|
| 33 | self.previous_results = previous_results
|
|---|
| 34 | self.job_server = job_server
|
|---|
| 35 |
|
|---|
| 36 | self.result = collections.OrderedDict()
|
|---|
| 37 |
|
|---|
| 38 | def run(self):
|
|---|
| 39 | print 'DoubleFourier.run'
|
|---|
| 40 |
|
|---|
| 41 | fts = self.previous_results['fts']
|
|---|
| 42 | fts_wn = fts['fts_wn']
|
|---|
| 43 | opd_max = fts['opd_max']
|
|---|
| 44 | fts_nsample = fts['ftsnsample']
|
|---|
| 45 | vdrive = fts['vdrive']
|
|---|
| 46 | delta_opd = fts['delta_opd']
|
|---|
| 47 |
|
|---|
| 48 | times = np.arange(int(fts_nsample), dtype=np.float)
|
|---|
| 49 | times *= (opd_max / vdrive) / float(fts_nsample-1)
|
|---|
| 50 |
|
|---|
| 51 | beamsgenerator = self.previous_results['beamsgenerator']
|
|---|
| 52 |
|
|---|
| 53 | uvmapgenerator = self.previous_results['uvmapgenerator']
|
|---|
| 54 | bxby = uvmapgenerator['bxby']
|
|---|
| 55 |
|
|---|
| 56 | skygenerator = self.previous_results['skygenerator']
|
|---|
| 57 | skymodel = skygenerator['sky model']
|
|---|
| 58 | spatial_axis = self.result['spatial axis'] = skygenerator['spatial axis']
|
|---|
| 59 | self.result['frequency axis'] = fts_wn
|
|---|
| 60 |
|
|---|
| 61 | # assuming nx is even then transform has 0 freq at origin and [nx/2] is
|
|---|
| 62 | # Nyquist frequency. Nyq freq = 0.5 * Nyquist sampling freq.
|
|---|
| 63 | # Assume further that the fft is shifted so that 0 freq is at nx/2
|
|---|
| 64 | nx = len(spatial_axis)
|
|---|
| 65 | spatial_freq_axis = np.arange(-nx/2, nx/2, dtype=np.float)
|
|---|
| 66 | sample_freq = (180.0 * 3600.0 / np.pi) / (spatial_axis[1] - spatial_axis[0])
|
|---|
| 67 | spatial_freq_axis *= (sample_freq / nx)
|
|---|
| 68 | self.result['spatial frequency axis'] = spatial_freq_axis
|
|---|
| 69 |
|
|---|
| 70 | self.result['baseline interferograms'] = collections.OrderedDict()
|
|---|
| 71 | # for baseline in bxby:
|
|---|
| 72 | for baseline in bxby[:3]:
|
|---|
| 73 | print baseline
|
|---|
| 74 | measurement = np.zeros(np.shape(times))
|
|---|
| 75 |
|
|---|
| 76 | # FTS path diff and possibly baseline itself vary with time
|
|---|
| 77 | for tindex,t in enumerate(times):
|
|---|
| 78 |
|
|---|
| 79 | # calculate the sky that the system is observing at this
|
|---|
| 80 | # moment, incorporating various errors in turn
|
|---|
| 81 | sky_now = skymodel
|
|---|
| 82 |
|
|---|
| 83 | # 1. baseline should be perp to centre of field.
|
|---|
| 84 | # If baseline is tilted then origin of sky map shifts.
|
|---|
| 85 | # (I think effect could be corrected by changing
|
|---|
| 86 | # FTS sample position to compensate.?)
|
|---|
| 87 | # for now assume 0 error but do full calculation for timing
|
|---|
| 88 | # purposes
|
|---|
| 89 |
|
|---|
| 90 | # perfect baseline is perpendicular to direction to 'centre'
|
|---|
| 91 | # on sky. Add errors in x,y,z - z towards sky 'centre'
|
|---|
| 92 | bz = 0.0
|
|---|
| 93 | # convert bz to angular error
|
|---|
| 94 | blength = math.sqrt(baseline[0]*baseline[0] +
|
|---|
| 95 | baseline[1]*baseline[1])
|
|---|
| 96 | bangle = (180.0 * 3600.0 / math.pi) * bz / blength
|
|---|
| 97 | bx_error = bangle * baseline[0] / blength
|
|---|
| 98 | by_error = bangle * baseline[1] / blength
|
|---|
| 99 |
|
|---|
| 100 | # calculate xpos, ypos in units of pixel - numpy arrays
|
|---|
| 101 | # [row,col]
|
|---|
| 102 | nx = len(spatial_axis)
|
|---|
| 103 | colpos = float(nx-1) * bx_error / \
|
|---|
| 104 | (spatial_axis[-1] - spatial_axis[0])
|
|---|
| 105 | rowpos = float(nx-1) * by_error / \
|
|---|
| 106 | (spatial_axis[-1] - spatial_axis[0])
|
|---|
| 107 |
|
|---|
| 108 | # calculate fourier phase shift to move point at [0,0] to
|
|---|
| 109 | # [rowpos, colpos]
|
|---|
| 110 | print 'shift', colpos, rowpos
|
|---|
| 111 | shiftx = np.zeros([nx], np.complex)
|
|---|
| 112 | shiftx[:nx/2] = np.arange(nx/2, dtype=np.complex)
|
|---|
| 113 | shiftx[nx/2:] = np.arange(-nx/2, 0, dtype=np.complex)
|
|---|
| 114 | shiftx = np.exp((-2.0j * np.pi * colpos * shiftx) / float(nx))
|
|---|
| 115 |
|
|---|
| 116 | shifty = np.zeros([nx], np.complex)
|
|---|
| 117 | shifty[:nx/2] = np.arange(nx/2, dtype=np.complex)
|
|---|
| 118 | shifty[nx/2:] = np.arange(-nx/2, 0, dtype=np.complex)
|
|---|
| 119 | shifty = np.exp((-2.0j * np.pi * rowpos * shifty) / float(nx))
|
|---|
| 120 |
|
|---|
| 121 | shift = np.ones([nx,nx], np.complex)
|
|---|
| 122 | for j in range(nx):
|
|---|
| 123 | shift[j,:] *= shiftx
|
|---|
| 124 | for i in range(nx):
|
|---|
| 125 | shift[:,i] *= shifty
|
|---|
| 126 |
|
|---|
| 127 | jobs = {}
|
|---|
| 128 | # go through freq planes and shift them
|
|---|
| 129 | for iwn,wn in enumerate(fts_wn):
|
|---|
| 130 | # submit jobs
|
|---|
| 131 | indata = (sky_now[:,:,iwn], shift,)
|
|---|
| 132 | jobs[wn] = self.job_server.submit(fftshift,
|
|---|
| 133 | indata, (), ('numpy',))
|
|---|
| 134 |
|
|---|
| 135 | for iwn,wn in enumerate(fts_wn):
|
|---|
| 136 | # collect and store results
|
|---|
| 137 | temp = jobs[wn]()
|
|---|
| 138 | sky_now[:,:,iwn] = temp
|
|---|
| 139 |
|
|---|
| 140 | if t == times[0]:
|
|---|
| 141 | self.result['sky at time 0'] = sky_now
|
|---|
| 142 |
|
|---|
| 143 | # 2. telescopes should be centred on centre of field
|
|---|
| 144 | # Telescopes collect flux from the 'sky' and pass
|
|---|
| 145 | # it to the FTS beam combiner. In doing this each
|
|---|
| 146 | # telescope multiplies the sky emission by its
|
|---|
| 147 | # amplitude beam response - always real but with
|
|---|
| 148 | # negative areas. Is this correct? Gives right
|
|---|
| 149 | # answer for 'no error' case.
|
|---|
| 150 |
|
|---|
| 151 | # multiply sky by amplitude beam 1 * amplitude beam 2
|
|---|
| 152 | amp_beam_1 = beamsgenerator['primary amplitude beam'].data
|
|---|
| 153 | amp_beam_2 = beamsgenerator['primary amplitude beam'].data
|
|---|
| 154 |
|
|---|
| 155 | for iwn,wn in enumerate(fts_wn):
|
|---|
| 156 | # calculate shifted beams here
|
|---|
| 157 | # for now assume no errors and just use beam
|
|---|
| 158 | # calculated earlier
|
|---|
| 159 | pass
|
|---|
| 160 |
|
|---|
| 161 | # multiply sky by amplitude beams of 2 antennas
|
|---|
| 162 | sky_now *= amp_beam_1 * amp_beam_2
|
|---|
| 163 |
|
|---|
| 164 | if t == times[0]:
|
|---|
| 165 | self.result['sky*beams at time 0'] = sky_now
|
|---|
| 166 |
|
|---|
| 167 | # 3. baseline error revisited
|
|---|
| 168 | # derive baseline at this time
|
|---|
| 169 | #
|
|---|
| 170 | # Perhaps baseline should be a continuous function in
|
|---|
| 171 | # time, which would allow baselines that intentionally
|
|---|
| 172 | # smoothly vary (as in rotating tethered assembly)
|
|---|
| 173 | # and errors to be handled by one description.
|
|---|
| 174 | #
|
|---|
| 175 | # what follows assumes zero error
|
|---|
| 176 |
|
|---|
| 177 | baseline_error = 0.0
|
|---|
| 178 | baseline_now = baseline + baseline_error
|
|---|
| 179 |
|
|---|
| 180 | fft_now = np.zeros(np.shape(sky_now), np.complex)
|
|---|
| 181 | spectrum = np.zeros(np.shape(fts_wn), np.complex)
|
|---|
| 182 | for iwn,wn in enumerate(fts_wn):
|
|---|
| 183 |
|
|---|
| 184 | # derive shift needed to place baseline at one of FFT coords
|
|---|
| 185 | # this depends on physical baseline and frequency
|
|---|
| 186 | baseline_now_lambdas = baseline_now * wn * 100.0
|
|---|
| 187 |
|
|---|
| 188 | # calculate baseline position in units of pixels of FFTed
|
|---|
| 189 | # sky - numpy arrays [row,col]
|
|---|
| 190 | colpos = float(nx-1) * \
|
|---|
| 191 | float(baseline_now_lambdas[0] - spatial_freq_axis[0]) / \
|
|---|
| 192 | (spatial_freq_axis[-1] - spatial_freq_axis[0])
|
|---|
| 193 | rowpos = float(nx-1) * \
|
|---|
| 194 | float(baseline_now_lambdas[1] - spatial_freq_axis[0]) / \
|
|---|
| 195 | (spatial_freq_axis[-1] - spatial_freq_axis[0])
|
|---|
| 196 | colpos = 0.0
|
|---|
| 197 | rowpos = 0.0
|
|---|
| 198 | print 'spatial colpos, rowpos', colpos, rowpos
|
|---|
| 199 |
|
|---|
| 200 | # calculate fourier phase shift to move point at [rowpos,colpos] to
|
|---|
| 201 | # [0,0]
|
|---|
| 202 | shiftx = np.zeros([nx], np.complex)
|
|---|
| 203 | shiftx[:nx/2] = np.arange(nx/2, dtype=np.complex)
|
|---|
| 204 | shiftx[nx/2:] = np.arange(-nx/2, 0, dtype=np.complex)
|
|---|
| 205 | shiftx = np.exp((-2.0j * np.pi * colpos * shiftx) / \
|
|---|
| 206 | float(nx))
|
|---|
| 207 |
|
|---|
| 208 | shifty = np.zeros([nx], np.complex)
|
|---|
| 209 | shifty[:nx/2] = np.arange(nx/2, dtype=np.complex)
|
|---|
| 210 | shifty[nx/2:] = np.arange(-nx/2, 0, dtype=np.complex)
|
|---|
| 211 | shifty = np.exp((-2.0j * np.pi * rowpos * shifty) / float(nx))
|
|---|
| 212 |
|
|---|
| 213 | shift = np.ones([nx,nx], np.complex)
|
|---|
| 214 | for j in range(nx):
|
|---|
| 215 | shift[j,:] *= shiftx
|
|---|
| 216 | for i in range(nx):
|
|---|
| 217 | shift[:,i] *= shifty
|
|---|
| 218 |
|
|---|
| 219 | # move centre of sky image to origin
|
|---|
| 220 | temp = np.fft.fftshift(sky_now[:,:,iwn])
|
|---|
| 221 | # apply phase shift
|
|---|
| 222 | temp *= shift
|
|---|
| 223 | # 2d fft
|
|---|
| 224 | temp = np.fft.fft2(temp)
|
|---|
| 225 | fft_now[:,:,iwn] = temp
|
|---|
| 226 |
|
|---|
| 227 | # set amp/phase at this frequency
|
|---|
| 228 | spectrum[iwn] = temp[0,0]
|
|---|
| 229 |
|
|---|
| 230 | if t == times[0]:
|
|---|
| 231 | self.result['skyfft at time 0'] = fft_now
|
|---|
| 232 |
|
|---|
| 233 | axis = co.Axis(data=fts_wn, title='wavenumber',
|
|---|
| 234 | units='cm-1')
|
|---|
| 235 | temp = co.Spectrum(data=spectrum, axis=axis,
|
|---|
| 236 | title='Detected spectrum', units='W sr-1 m-2 Hz-1')
|
|---|
| 237 | self.result['skyfft spectrum at time 0'] = temp
|
|---|
| 238 |
|
|---|
| 239 | # 3. FTS sampling should be accurate
|
|---|
| 240 | # derive lag due to FTS path difference
|
|---|
| 241 | # 0 error for now
|
|---|
| 242 |
|
|---|
| 243 | if t == times[0]:
|
|---|
| 244 | # test version
|
|---|
| 245 | # inverse fft of emission spectrum at this point
|
|---|
| 246 | temp = np.fft.ifft(spectrum)
|
|---|
| 247 | pos = np.fft.rfftfreq(len(spectrum))
|
|---|
| 248 |
|
|---|
| 249 | # move 0 frequency to centre of array
|
|---|
| 250 | temp = np.fft.fftshift(temp)
|
|---|
| 251 | pos = np.fft.fftshift(pos)
|
|---|
| 252 |
|
|---|
| 253 | axis = co.Axis(data=pos, title='path difference',
|
|---|
| 254 | units='cm')
|
|---|
| 255 | temp = co.Spectrum(data=temp,
|
|---|
| 256 | title='Detected interferogram', units='')
|
|---|
| 257 |
|
|---|
| 258 | self.result['test FTS at time 0'] = temp
|
|---|
| 259 |
|
|---|
| 260 | mirror_error = 0.0
|
|---|
| 261 | opd = 2.0 * (vdrive * t + mirror_error)
|
|---|
| 262 | print 'opd', opd, vdrive, t, mirror_error
|
|---|
| 263 | opd_ipos = opd / delta_opd
|
|---|
| 264 |
|
|---|
| 265 | # make spectrum symmetric
|
|---|
| 266 | nfreq = len(fts_wn)
|
|---|
| 267 | symmetric_spectrum = np.zeros([2*nfreq])
|
|---|
| 268 | symmetric_spectrum[:nfreq] = spectrum
|
|---|
| 269 | symmetric_spectrum[nfreq:] = spectrum[::-1]
|
|---|
| 270 |
|
|---|
| 271 | # calculate shift needed to move point at opd to 0
|
|---|
| 272 | shift = np.zeros([2*nfreq], dtype=np.complex)
|
|---|
| 273 | shift[:nfreq] = np.arange(nfreq, dtype=np.complex)
|
|---|
| 274 | shift[nfreq:] = np.arange(-nfreq, 0, dtype=np.complex)
|
|---|
| 275 | shift = np.exp((-2.0j * np.pi * opd_ipos * shift) / float(nfreq))
|
|---|
| 276 |
|
|---|
| 277 | # apply phase shift and fft
|
|---|
| 278 | symmetric_spectrum *= shift
|
|---|
| 279 | spectrum_fft = np.fft.fft(symmetric_spectrum)
|
|---|
| 280 | measurement[tindex] = spectrum_fft[0]
|
|---|
| 281 |
|
|---|
| 282 | self.result['baseline interferograms'][tuple(baseline)] = \
|
|---|
| 283 | measurement
|
|---|
| 284 |
|
|---|
| 285 | return self.result
|
|---|
| 286 |
|
|---|
| 287 | def matlab_transform(self):
|
|---|
| 288 | # readers should look at Izumi et al. 2006, Applied Optics, 45, 2576
|
|---|
| 289 | # for theoretical background. Names of variables in the code correspond
|
|---|
| 290 | # to that work.
|
|---|
| 291 |
|
|---|
| 292 | # For now, assume 2 light collectors giving one baseline at a time.
|
|---|
| 293 | interferograms = {}
|
|---|
| 294 |
|
|---|
| 295 | for baseline in self.baselines:
|
|---|
| 296 | interferogram = 0
|
|---|
| 297 |
|
|---|
| 298 | # baseline length (cm) and position angle
|
|---|
| 299 | bu = baseline[0]
|
|---|
| 300 | bv = baseline[1]
|
|---|
| 301 | mod_b = np.sqrt(pow(bu,2) + pow(bv,2))
|
|---|
| 302 | ang_b = np.arctan2(bv, bu)
|
|---|
| 303 |
|
|---|
| 304 | # loop over sky pixels covered by primary beam
|
|---|
| 305 | nx = self.sky_s.shape[1]
|
|---|
| 306 | ny = self.sky_s.shape[2]
|
|---|
| 307 | for j in range(ny):
|
|---|
| 308 | for i in range(nx):
|
|---|
| 309 |
|
|---|
| 310 | # inverse fft of emission spectrum at this point
|
|---|
| 311 | temp = np.fft.ifft(self.sky_s[:,j,i])
|
|---|
| 312 |
|
|---|
| 313 | # move 0 frequency to centre of array
|
|---|
| 314 | temp = np.fft.fftshift(temp)
|
|---|
| 315 |
|
|---|
| 316 | # length (radians) and position angle of theta vector
|
|---|
| 317 | mod_theta = np.sqrt(pow(self.sky_x[i],2) + pow(self.sky_y[j],2))
|
|---|
| 318 | ang_theta = np.arctan2(self.sky_y[j], self.sky_x[i])
|
|---|
| 319 |
|
|---|
| 320 | # calculate b.theta (the projection of b on theta)
|
|---|
| 321 | # and the corresponding delay in units of wavelength at
|
|---|
| 322 | # Nyquist frequency
|
|---|
| 323 | delay = mod_theta * mod_b * np.cos(ang_b - ang_theta) * self.freqs[-1]
|
|---|
| 324 |
|
|---|
| 325 | # sampling is done at twice Nyquist freq so shift transformed
|
|---|
| 326 | # spectrum by 2 * delay samples (approximated to nint)
|
|---|
| 327 | # NOTE factor of 2 discrepency with matlab version! I think
|
|---|
| 328 | # this is because there the variable 'Nyq' is the Nyquist
|
|---|
| 329 | # sampling rate, not the Nyquist frequency.
|
|---|
| 330 | temp = np.roll(temp, int(round(2.0 * delay)))
|
|---|
| 331 |
|
|---|
| 332 | # want only the real part of the result
|
|---|
| 333 | interferogram += np.real(temp)
|
|---|
| 334 |
|
|---|
| 335 | def __repr__(self):
|
|---|
| 336 | return 'DoubleFourier'
|
|---|
| 337 |
|
|---|
