1.. _theory: 2 3******************* 4Theory of Operation 5******************* 6 7The Radio Interferometer Measurement Equation 8============================================= 9 10OSKAR uses the Radio Interferometer Measurement Equation of 11Hamaker, Bregman & Sault (1996) to produce simulated visibility data 12(see also Smirnov, 2011 for an excellent overview). 13This equation describes the effect due to all visible sources 14(indexed by :math:`s`) on the complex, polarised visibility 15:math:`\mathbf{V}_{p,q}` for a baseline between stations :math:`p` 16and :math:`q`. 17 18The main terms in the Measurement Equation are Jones matrices: these are 19two-by-two complex quantities that effectively modify the original signal 20from each radio source, to give the signal that would be produced from each 21station due to that source. Each source has an additive effect on the 22measured visibilities, so the observed visibility on baseline :math:`p,q` is 23given by the sum over all visible sources of the product of each set of 24Jones matrices with the source coherency matrix :math:`\mathbf{B}`. 25The Measurement Equation currently implemented in OSKAR takes the form 26 27.. math:: 28 29 \mathbf{V}_{p,q} = \mathbf{G}_{p} \left( \sum_s 30 \mathbf{K}_{p,s} \mathbf{Z}_{p,s} \mathbf{E}_{p,s} 31 \left< \mathbf{B}_{s} \right> 32 \mathbf{E}_{q,s}^H \mathbf{Z}_{q,s}^H \mathbf{K}_{q,s}^H 33 \right) \mathbf{G}_{q}^H 34 35where the superscript-H denotes Hermitian transpose. 36 37The Jones matrices currently included in OSKAR are: 38 39- Optional station beam matrix :math:`\mathbf{E}`, including the effects of 40 parallactic angle rotation, the individual element response patterns, 41 and the shape of the beam on the sky as a function of time, frequency 42 and polarisation. 43 44- Optional ionospheric phase screen :math:`\mathbf{Z}`, as a function of time. 45 46- Interferometer phase :math:`\mathbf{K}`, as a function of time and frequency. 47 48- Optional direction-independent complex station gains :math:`\mathbf{G}`, 49 as a function of time, frequency and polarisation. 50 51For the following sections, in which the Jones terms are described, it is 52helpful to first introduce the coordinate systems used. 53 54Coordinate Systems 55------------------ 56 57Sources are specified in the equatorial system and have positions Right 58Ascension :math:`\alpha` and Declination :math:`\delta`. 59This spherical system has Cartesian :math:`(x',y',z')` axes, where 60the :math:`x'` axis points towards :math:`\alpha=0`, the :math:`z'` axis points 61towards the North Celestial Pole (NCP) and the :math:`y'` axis is perpendicular 62to both to make a right-handed system. The angle :math:`\alpha` increases 63from :math:`x'` towards :math:`y'`, and the angle :math:`\delta` increases 64from the :math:`x'\,y'`-plane towards :math:`z'`. 65The equatorial system is shown in dashed black in the figure below. 66 67Antennas are specified in the local horizontal coordinate system. 68This spherical system has Cartesian :math:`(x,y,z)` axes, where the x-axis 69points to the observer's geographic East, the y-axis points to 70geographic North, and the z-axis points to the observer's zenith. 71The local horizon is therefore in the :math:`xy`-plane. The angle :math:`\phi` 72is the co-azimuth, and increases from :math:`x` towards :math:`y`, and the 73angle :math:`\theta` is the polar angle, zenith distance, or co-elevation, 74which increases from z towards the :math:`xy`-plane. The horizontal system is 75shown in red in the figure below. 76 77.. _theory_coord_fig: 78 79.. figure:: coordsys_small.png 80 :width: 13cm 81 :align: center 82 :alt: Coordinate systems 83 84 85Source Brightness Matrix (B) 86---------------------------- 87The source brightness (or coherency) matrix represents the intrinsic, 88unmodified radiation from an astronomical source. It is constructed using 89the source Stokes parameters :math:`(I,Q,U,V)`, which completely describe the 90polarisation state of the radiation. Using the standard polarisation 91convention adopted by the International Astronomical Union for radio 92astronomy (IAU, 1974; see the figure below), the 93polarisation axes are defined on the tangent plane to the sphere 94in the equatorial system. The polarisation angle is measured due 95east (counter-clockwise) from the direction to the North Celestial Pole, 96so that 100% Stokes +Q corresponds to North-to-South polarisation, 100% 97Stokes +U corresponds to North-East-to-South-West polarisation, 98and 100% Stokes +V corresponds to right-handed circular polarisation. 99 100.. _theory_fig_pol_axes: 101 102.. figure:: sketch_pol.png 103 :width: 10cm 104 :align: center 105 :alt: Polarisation axes 106 107Using this convention, Hamaker & Bregman (1996) show that 108 109.. math:: 110 111 \left< \mathbf{B} \right> = 112 \left[ 113 \begin{array}{cc} 114 I + Q & U + i V \\ 115 U - i V & I - Q 116 \end{array} 117 \right] 118 119 120Station Beam (E) 121---------------- 122The station beam response is a function of various factors, 123including the parallactic angle, the response of the individual antenna 124elements, and the shape of the beam on the sky - which itself is determined 125by the projected shape of the station in the direction of the beam, 126and any errors introduced by the beamformer. 127 128Parallactic Angle Rotation 129^^^^^^^^^^^^^^^^^^^^^^^^^^ 130The emission from each source must first be expressed in the 131frame of the antenna, which is fixed to the ground. 132To do this, the equatorial Stokes parameters are transformed 133to the local horizontal system by rotating by the parallactic 134angle at the position of the source. 135 136The parallactic angle at a source position is defined as the angle between 137the direction of the North Celestial Pole and the local vertical on the sky 138(measured from north towards east), and depends on the observer's 139latitude :math:`\varphi` and the source hour angle :math:`H` and 140declination :math:`\delta`. The parallactic angle :math:`\psi_p` is 141 142.. math:: 143 144 \psi_p = \arctan\left( 145 \frac{\cos\varphi \sin H} 146 {\sin\varphi \cos\delta - \cos\varphi \sin\delta \cos H} 147 \right) 148 149Element Factors 150^^^^^^^^^^^^^^^ 151The station beam is a weighted sum of the response from each antenna :math:`a`, 152where the complex beamforming weights :math:`w_a` are generated to maximise 153the sensitivity of the array in a given direction as a function of time and 154frequency. The shape of the station beam is largely governed by the 155instantaneous projected spacing between individual antennas (the array factor). 156Assuming two dipoles labelled X and Y, which have their respective axes 157nominally along the x and y axes shown in the figure above, the beamforming 158is done independently for the two X and Y antenna polarisations (so there may 159be two sets of beamforming weights :math:`w_a^X` and :math:`w_a^Y`), and each 160antenna has a co-polar and cross-polar response pattern at each source 161position (so there are four values in total, 162:math:`g^X_X, g^X_Y, g^Y_X, g^Y_Y`). 163The co-polar responses are on the diagonal elements of the matrix, while the 164cross-polar responses are on the off-diagonal elements. 165 166The overall station beam response to a source in both polarisations can 167then be expressed in matrix form as a combination of these effects, 168with the sum taken over all antennas in the station :math:`a`. 169 170.. math:: 171 172 \mathbf{E} = 173 \left( 174 \sum_a \left[ \begin{array}{c} w^X_a \\ w^Y_a \end{array} \right] 175 \left[ 176 \begin{array}{cc} 177 g^X_X & g^X_Y \\ 178 g^Y_X & g^Y_Y 179 \end{array} 180 \right]_a 181 \right) 182 \left[ 183 \begin{array}{cc} 184 \cos\psi_p & -\sin\psi_p \\ 185 \sin\psi_p & \cos\psi_p 186 \end{array} 187 \right] 188 189If using an average embedded element pattern, the individual antenna responses 190are factored out, since all antennas are then the same. 191 192 193Ionospheric Screen (Z) 194---------------------- 195If specified, the ionospheric screen adds an additional phase factor 196for each source and station. It is modelled as a physical 197screen at a specified height above the ground, so that sources in different 198directions pierce the screen at different locations at each station: 199for any given direction, stations which are far apart will also show greater 200variation in their phases than stations which are closer together. 201 202Values in the screen are interpreted as changes in the total electron content 203(:math:`\Delta{\rm TEC}`) above the array, and converted to phase values as a 204function of frequency :math:`\nu` (in Hz) using: 205 206.. math:: 207 208 \mathbf{Z} = 209 \exp\left\{i 210 \left[-8.44797245 \times 10^9 \, \frac{\Delta{\rm TEC}}{\nu} \right] 211 \right\} 212 \left[ 213 \begin{array}{cc} 214 1 & 0 \\ 215 0 & 1 216 \end{array} 217 \right] 218 219 220Interferometer Phase (K) 221------------------------ 222The interferometer phase matrix depends only on the projected spacing 223between stations. This is polarisation-independent, so :math:`\mathbf{K}` is 224a scalar. The phase is (e.g. Thompson, Moran & Swenson, 2001): 225 226.. math:: 227 228 \mathbf{K} = 229 \exp\left\{-2\pi i \left[ul + vm + w(n - 1)\right]\right\} 230 \left[ 231 \begin{array}{cc} 232 1 & 0 \\ 233 0 & 1 234 \end{array} 235 \right] 236 237where :math:`(u,v,w)` are the station coordinates in the plane perpendicular 238to the phase centre, and :math:`(l,m,n)` are the direction cosines of the source 239relative to the phase centre. Using the normal conventions in radio 240astronomy, the :math:`u` and :math:`l` directions increase towards the East, 241the :math:`v` and :math:`m` directions increase towards the North, and 242the :math:`w` and :math:`n` directions increase towards the phase centre. 243 244 245Direction-independent Gains (G) 246------------------------------- 247If specified, direction-independent complex gains can be supplied as a 248function of time, frequency and polarisation, for each station in the array. 249(Since they are direction-independent, the same gain value is used for all 250sources.) The X and Y polarisations at each station are treated separately, 251so the complex gain matrix is diagonal: 252 253.. math:: 254 255 \mathbf{G} = 256 \left[ 257 \begin{array}{cc} 258 g^X & 0 \\ 259 0 & g^Y 260 \end{array} 261 \right] 262 263 264Visibilities to Stokes Parameters 265--------------------------------- 266Having obtained the simulated visibility correlation matrix 267 268.. math:: 269 270 \mathbf{V_{p,q}} = 271 \left[ 272 \begin{array}{cc} 273 XX & XY \\ 274 YX & YY 275 \end{array} 276 \right] 277 = 278 \left[ 279 \begin{array}{cc} 280 I + Q & U + i V \\ 281 U - i V & I - Q 282 \end{array} 283 \right] 284 285the Stokes parameters can then be recovered for the purposes of making 286images by rearranging the diagonal and off-diagonal elements: 287 288.. math:: 289 290 I &= \frac{1}{2}(XX+YY) \\ 291 Q &= \frac{1}{2}(XX-YY) \\ 292 U &= \frac{1}{2}(XY+YX) \\ 293 V &= -\frac{1}{2} i(XY-YX) 294 295Note, however, that this conversion does not involve polarisation 296calibration in any way: additional corrections for the parallactic angle and 297antenna response would need to be made in order to recover the true source 298polarisation in the equatorial frame. 299 300.. _theory_noise: 301 302Addition of Uncorrelated System Noise 303===================================== 304 305When performing interferometer simulations, OSKAR provides the option of 306adding uncorrelated Gaussian noise, :math:`\varepsilon`, to the 307simulated visibilities, :math:`\textbf{V}_0`. 308 309.. math:: \textbf{V} = \textbf{V}_0 + \varepsilon 310 311This is achieved by adding randomly generated values, drawn from a zero-mean 312Gaussian distribution, to the complex visibility amplitudes for each baseline, 313time integration, frequency channel and polarisation. Gaussian distributions 314are defined as a function of frequency, and can be given a different value for 315each station in the interferometer. Noise values are expressed as the RMS flux 316level of an unresolved, unpolarised source measured in a single polarisation of 317the receiver. 318 319.. As such, if one measures the noise statistics of visibilities expressed 320 in terms of Stokes parameters, the RMS obtained will be smaller by a factor 321 of :math:`\sqrt{2}` than visibilities expressed as linear polarisations 322 (XX,XY,YX, and YY). 323 324While OSKAR requires that the the noise is expressed as a RMS in Jy, 325one can easily convert to this value from a measure of the noise in terms of 326system sensitivity or system temperature and effective area using the standard 327formulae described by Thompson, Moran & Swenson and Wrobel & Walker. 328 329The noise power per unit bandwidth, received in one polarisation of an 330antenna from an unpolarised source of system equivalent flux density 331:math:`S_{\rm sys}`, is given by 332 333.. math:: k_{\rm B} T_{\rm sys} = \frac{S_{\rm sys} A_{\rm eff} \eta} {2} 334 335Here, :math:`T_{\rm sys}` is the system temperature, :math:`A_{\rm eff}` is 336the effective area of the antenna, :math:`\eta` is the system efficiency, 337and :math:`k_{\rm B}` is the Boltzmann constant. 338 339The RMS noise on a given baseline can then be expressed in terms of the 340system equivalent flux densities :math:`S_p` and :math:`S_q` of antennas 341(or stations) :math:`p` and :math:`q` that make up the baseline by 342 343.. math:: \sigma_{p,q} = \sqrt{\frac{S_p S_q}{2\, \Delta\nu\, \tau_{\rm acc}}} 344 345Here, :math:`\Delta\nu` is the bandwidth and :math:`\tau_{\rm acc}` is the 346correlator accumulation time. Note the 347term :math:`2\, \Delta\nu\, \tau_{\rm acc}` represents the number of 348independent samples of the signal for a band-limited signal sampled at 349the Nyquist rate. 350 351This equation can be re-expressed in terms of the individual system 352temperatures :math:`T_p` and :math:`T_q`, effective areas :math:`A_p` 353and :math:`A_q` and system efficiencies :math:`\eta_p` and :math:`\eta_q` of 354antennas (or stations) which make up the baseline as 355 356.. math:: 357 358 \sigma_{p,q} = k_{\rm B} 359 \sqrt{\frac{2 \, T_p T_q} 360 {A_p A_q \, \eta_p \eta_q \, \Delta\nu\, \tau_{\rm acc}}} 361 362Equally, given values of the RMS on individual baselines :math:`\sigma_p` 363and :math:`\sigma_q`, the baseline RMS is given by 364 365.. math:: \sigma_{p,q} = \sqrt{\sigma_p \sigma_q} 366 367Noise fluctuations in the real and imaginary parts of the complex correlator 368outputs are uncorrelated. The RMS uncertainty in the visibility, 369:math:`\varepsilon_{p,q}`, obtained from combining the real and imaginary 370outputs of the correlator will therefore be 371 372.. math:: 373 374 \varepsilon_{p,q} = 375 \sqrt{\left \langle \varepsilon\cdot\varepsilon\right \rangle} = 376 \sqrt{2}\sigma_{p,q}. 377 378Noise in the Synthesised Map 379---------------------------- 380For an array with :math:`n_b` antenna pairs which observes for a length of 381total observation time :math:`\tau_0`, the total number of independent data 382points in the :math:`(u,v)` plane for a single polarisation is 383 384.. math:: n_d = n_b \frac{\tau_0}{\tau_{\rm acc}} 385 386and therefore the noise in the image or map will decrease by a 387factor :math:`\sqrt{n_d}`. 388 389If we consider the special case where the system temperature, effective area, 390and system efficiency are the same for an array of :math:`n_a` antennas 391observing for total time :math:`\tau_0`, the following equation describes 392the total noise in the image plane of a single polarisation image. 393 394.. math:: 395 396 \sigma_{\rm im} = \frac{2 \, k_{\rm B} \, T_{\rm sys}} 397 {A_{\rm eff} \eta \sqrt{n_a (n_a - 1) \Delta\nu \tau_0}} 398 399This can be expressed in terms of the RMS noise on a given baseline as 400 401.. math:: 402 403 \sigma_{\rm im} = 404 \frac{\sigma_{p,q}} 405 {\sqrt{ \frac{n_a (n_a-1)}{2} \frac{\tau_0}{\tau_{\rm acc}} } } 406 = \frac{\sigma_{p,q}} 407 {\sqrt{n_d}} 408 409Note that for measurements comprised of combinations of single polarisation 410data (such as Stokes-I,Q,U,V) the RMS will be reduced by a further factor of 411:math:`\sqrt{2}`. 412 413References 414========== 415 416- Hamaker, J. P., Bregman, J. D. & Sault, R. J., 1996, A&AS, 117, 137 417- Hamaker, J. P., Bregman, J. D., 1996, A&AS, 117, 161 418- IAU, 1974, Transactions of the IAU Vol. 15B (1973) 166 419- Smirnov, O. M., 2011, A&A, 527, 106 420- Thompson, A. R., Moran, J. M., & Swenson, G.W., 2001, 421 *Interferometry and Synthesis in Radio Astronomy* 422- Wrobel, J.M., & Walker, R. C., 1999, 423 *Synthesis Imaging in Radio Astronomy II*, p. 171 424 425