Functions related to sensor modelling. More...

#include "sensor.h"
#include <cmath>
#include <list>
#include <stdexcept>
#include "arts.h"
#include "logic.h"
#include "matpackI.h"
#include "matpackII.h"
#include "messages.h"
#include "sorting.h"

Functions
void	antenna1d_matrix (Sparse &H, const Index &antenna_dim, ConstVectorView antenna_dza, const GriddedField4 &antenna_response, ConstVectorView za_grid, ConstVectorView f_grid, const Index n_pol, const Index do_norm)
	antenna1d_matrix More...

void	antenna2d_gridded_dlos (Sparse &H, const Index &antenna_dim, ConstMatrixView antenna_dlos, const GriddedField4 &antenna_response, ConstMatrixView mblock_dlos, ConstVectorView f_grid, const Index n_pol)
	antenna2d_interp_gridded_dlos More...

void	antenna2d_interp_response (Sparse &H, const Index &antenna_dim, ConstMatrixView antenna_dlos, const GriddedField4 &antenna_response, ConstMatrixView mblock_dlos, ConstVectorView f_grid, const Index n_pol)
	antenna2d_interp_response More...

void	gaussian_response_autogrid (Vector &x, Vector &y, const Numeric &x0, const Numeric &fwhm, const Numeric &xwidth_si, const Numeric &dx_si)
	gaussian_response_autogrid More...

void	gaussian_response (Vector &y, const Vector &x, const Numeric &x0, const Numeric &fwhm)
	gaussian_response More...

void	mixer_matrix (Sparse &H, Vector &f_mixer, const Numeric &lo, const GriddedField1 &filter, ConstVectorView f_grid, const Index &n_pol, const Index &n_sp, const Index &do_norm)
	mixer_matrix More...

void	mueller_rotation (Sparse &H, const Index &stokes_dim, const Numeric &rotangle)
	mueller_rotation More...

void	met_mm_polarisation_hmatrix (Sparse &H, const ArrayOfString &mm_pol, const Numeric dza, const Index stokes_dim, const String &iy_unit)
	Calculate polarisation H-matrix. More...

void	sensor_aux_vectors (Vector &sensor_response_f, ArrayOfIndex &sensor_response_pol, Matrix &sensor_response_dlos, ConstVectorView sensor_response_f_grid, const ArrayOfIndex &sensor_response_pol_grid, ConstMatrixView sensor_response_dlos_grid)
	sensor_aux_vectors More...

void	spectrometer_matrix (Sparse &H, ConstVectorView ch_f, const ArrayOfGriddedField1 &ch_response, ConstVectorView sensor_f, const Index &n_pol, const Index &n_sp, const Index &do_norm)
	spectrometer_matrix More...

void	stokes2pol (VectorView w, const Index &stokes_dim, const Index &ipol_1based, const Numeric nv)
	stokes2pol More...

bool	test_and_merge_two_channels (Vector &fmin, Vector &fmax, Index i, Index j)
	Test if two instrument channels overlap, and if so, merge them. More...

void	find_effective_channel_boundaries (Vector &fmin, Vector &fmax, const Vector &f_backend, const ArrayOfGriddedField1 &backend_channel_response, const Numeric &delta, const Verbosity &verbosity)
	Calculate channel boundaries from instrument response functions. More...

void	integration_func_by_vecmult (VectorView h, ConstVectorView f, ConstVectorView x_f_in, ConstVectorView x_g_in)
	integration_func_by_vecmult More...

void	integration_bin_by_vecmult (VectorView h, ConstVectorView x_g_in, const Numeric &limit1, const Numeric &limit2)
	integration_bin_by_vecmult More...

void	summation_by_vecmult (VectorView h, ConstVectorView f, ConstVectorView x_f, ConstVectorView x_g, const Numeric x1, const Numeric x2)
	summation_by_vecmult More...

Variables
const Numeric	PI

const Numeric	NAT_LOG_2

const Numeric	DEG2RAD

const Index	GFIELD1_F_GRID

const Index	GFIELD4_FIELD_NAMES

const Index	GFIELD4_F_GRID

const Index	GFIELD4_ZA_GRID

const Index	GFIELD4_AA_GRID

Detailed Description

Functions related to sensor modelling.

Author: Mattias Ekstr�m ekstr.nosp@m.om@r.nosp@m.ss.ch.nosp@m.alme.nosp@m.rs.se

Date: 2003-02-27

Functions to model sensor behaviour and integration calculated as vector multiplication.

Definition in file sensor.cc.

Function Documentation

◆ antenna1d_matrix()

void antenna1d_matrix	(	Sparse &	H,
		const Index &	antenna_dim,
		ConstVectorView	antenna_dza,
		const GriddedField4 &	antenna_response,
		ConstVectorView	za_grid,
		ConstVectorView	f_grid,
		const Index	n_pol,
		const Index	do_norm
	)

antenna1d_matrix

Core function for setting up the response matrix for 1D antenna cases.

Main task is to extract correct antenna pattern, including frequency interpolation. Actual weights are calculated in integration_func_by_vectmult.

Parameters

H	The antenna transfer matrix
antenna_dim	As the WSV with the same name
antenna_dza	The zenith angle column of antenna_dlos.
antenna_response	As the WSV with the same name
za_grid	Zenith angle grid for pencil beam calculations
f_grid	Frequency grid for monochromatic calculations
n_pol	Number of polarisation states
do_norm	Flag whether response should be normalised

Author: Mattias Ekstr�m / Patrick Eriksson

Date: 2003-05-27 / 2008-06-17

Definition at line 58 of file sensor.cc.

References GriddedField4::data, DEBUG_ONLY, GriddedField::get_numeric_grid(), GriddedField::get_string_grid(), GFIELD4_AA_GRID, GFIELD4_F_GRID, GFIELD4_FIELD_NAMES, GFIELD4_ZA_GRID, gridpos(), Sparse::insert_row(), integration_func_by_vecmult(), interp(), interpweights(), joker, Sparse::ncols(), Array< base >::nelem(), ConstVectorView::nelem(), Sparse::resize(), and ConstVectorView::sum().

Referenced by sensor_responseAntenna().

◆ antenna2d_gridded_dlos()

void antenna2d_gridded_dlos	(	Sparse &	H,
		const Index &	antenna_dim,
		ConstMatrixView	antenna_dlos,
		const GriddedField4 &	antenna_response,
		ConstMatrixView	mblock_dlos,
		ConstVectorView	f_grid,
		const Index	n_pol
	)

antenna2d_interp_gridded_dlos

The radiances are treated as a bi-linear function, but the antenna response is treated as step-wise constant function (in contrast to 1D). See also built-in doc.

Parameters

H	The antenna transfer matrix
antenna_dim	As the WSV with the same name
antenna_dza	The zenith angle column of antenna_dlos.
antenna_response	As the WSV with the same name
za_grid	Zenith angle grid for pencil beam calculations
f_grid	Frequency grid for monochromatic calculations
n_pol	Number of polarisation states
do_norm	Flag whether response should be normalised

Author: Patrick Eriksson

Date: 2020-09-01

Definition at line 199 of file sensor.cc.

References abs, i, ConstMatrixView::ncols(), ConstVectorView::nelem(), and ConstMatrixView::nrows().

Referenced by sensor_responseAntenna().

◆ antenna2d_interp_response()

void antenna2d_interp_response	(	Sparse &	H,
		const Index &	antenna_dim,
		ConstMatrixView	antenna_dlos,
		const GriddedField4 &	antenna_response,
		ConstMatrixView	mblock_dlos,
		ConstVectorView	f_grid,
		const Index	n_pol
	)

antenna2d_interp_response

For this option, each direction defined by mblock_dlos_grid is considered to represent the same size in terms of solid beam angle, and the antenna pattern is interpolated to these directions. See also built-in doc.

Parameters

H	The antenna transfer matrix
antenna_dim	As the WSV with the same name
antenna_dza	The zenith angle column of antenna_dlos.
antenna_response	As the WSV with the same name
za_grid	Zenith angle grid for pencil beam calculations
f_grid	Frequency grid for monochromatic calculations
n_pol	Number of polarisation states
do_norm	Flag whether response should be normalised

Author: Patrick Eriksson

Date: 2018-09-12

Definition at line 442 of file sensor.cc.

References GriddedField4::data, GriddedField::get_numeric_grid(), GriddedField::get_string_grid(), GFIELD4_AA_GRID, GFIELD4_F_GRID, GFIELD4_FIELD_NAMES, GFIELD4_ZA_GRID, gridpos(), Sparse::insert_row(), interp(), interpweights(), joker, Sparse::ncols(), ConstMatrixView::ncols(), Array< base >::nelem(), ConstVectorView::nelem(), ConstMatrixView::nrows(), Sparse::resize(), and ConstVectorView::sum().

Referenced by sensor_responseAntenna().

◆ find_effective_channel_boundaries()

void find_effective_channel_boundaries	(	Vector &	fmin,
		Vector &	fmax,
		const Vector &	f_backend,
		const ArrayOfGriddedField1 &	backend_channel_response,
		const Numeric &	delta,
		const Verbosity &	verbosity
	)

Calculate channel boundaries from instrument response functions.

This function finds out the unique channel boundaries from f_backend and backend_channel_response. This is not a trivial task, since channels may overlap, or may be sorted in a strange way. The function tries to take care of all that. If channels overlap, they are combined to one continuous frequency region. therefore the number of elements in the output vectors fmin and fmax can be lower than the number of elements in f_backend and backend_channel_response.

The function also does consistency checking on the two input variables.

The output vectors fmin and fmax will be monotonically increasing.

Author: Stefan Buehler

Parameters

[out]	fmin	Vector of lower boundaries of instrument channels.
[out]	fmax	Vector of upper boundaries of instrument channels.
[in]	f_backend	Nominal backend frequencies.
[in]	backend_channel_response	Channel response, relative to nominal frequencies.
[in]	delta	Extra margin on both sides of each band. Has a default value of 0.

delta;

Definition at line 1152 of file sensor.cc.

References CREATE_OUT2, get_sorted_indexes(), i, is_increasing(), last(), Array< base >::nelem(), ConstVectorView::nelem(), Vector::resize(), _CS_basic_sstream_base< _CS_cT, _CS_Tr, _CS_Al >::str(), and test_and_merge_two_channels().

Referenced by f_gridFromSensorAMSU(), f_gridFromSensorAMSUgeneric(), and f_gridFromSensorHIRS().

◆ gaussian_response()

void gaussian_response	(	Vector &	y,
		const Vector &	x,
		const Numeric &	x0,
		const Numeric &	fwhm
	)

gaussian_response

Returns a 1D gaussian response

y is the gaussian function on grid x, with max at x0 and width following fwhm.

Parameters

y	Calculated response.
x	Grid.
x0	The x-position of response centre/max.
fwhm	The full width at half-maximum of the response

Author: Patrick Eriksson

Date: 2009-09-20

Definition at line 629 of file sensor.cc.

References i, n, NAT_LOG_2, ConstVectorView::nelem(), PI, pow(), Vector::resize(), and sqrt().

Referenced by antenna_responseVaryingGaussian(), and gaussian_response_autogrid().

◆ gaussian_response_autogrid()

void gaussian_response_autogrid	(	Vector &	x,
		Vector &	y,
		const Numeric &	x0,
		const Numeric &	fwhm,
		const Numeric &	xwidth_si,
		const Numeric &	dx_si
	)

gaussian_response_autogrid

Returns a 1D gaussian response with a suitable grid

First a grid is generated. The grid is si*[-xwidth_si:dx:xwidth_si], where si is the "standard deviation" corresponding to the FWHM, and dx is biggest possible value < dx_si, to enusre an symmetric grid wth end points exactly at xwidth_si. That is, width and spacing of the grid is specified in terms of number of standard deviations. If xwidth_si is set to 2, the response will cover about 95% the complete response. For xwidth_si=3, about 99% is covered.

y is the gaussian function on grid x, with max at x0 and width following fwhm.

Parameters

x	Grid generated.
y	Calculated response.
x0	The x-position of response centre/max.
fwhm	The full width at half-maximum of the response
xwidth_si	The one-sided width of x. See above.
dx_si	The grid step size of x. See above.

Author: Patrick Eriksson

Date: 2009-09-20

Definition at line 606 of file sensor.cc.

References gaussian_response(), n, NAT_LOG_2, nlinspace(), and sqrt().

Referenced by antenna_responseGaussian(), antenna_responseVaryingGaussian(), and backend_channel_responseGaussian().

◆ integration_bin_by_vecmult()

void integration_bin_by_vecmult	(	VectorView	h,
		ConstVectorView	x_g_in,
		const Numeric &	limit1,
		const Numeric &	limit2
	)

integration_bin_by_vecmult

Calculates the (row) vector that multiplied with an unknown (column) vector, g, approximates the integral between limit1 and limit2, where limit1 >= limit2.

This can be seen as a special case of what is handled by integration_func_by_vecmult*, where the function f is a boxcar function. Or expressed differently, the function g is "binned" between limit1 and limit2.

The limits must be inside the range the x_g.

Parameters

h	The multiplication (row) vector.
x_g_in	The grid points of function g(x). Can be increasing or decreasing. Must cover a wider range than the boxcar function (in both ends).
limit1	The lower integration limit.
limit2	The upper integration limit.

Author: Patrick Eriksson

Date: 2017-06-02

Definition at line 1503 of file sensor.cc.

Referenced by yActive().

◆ integration_func_by_vecmult()

void integration_func_by_vecmult	(	VectorView	h,
		ConstVectorView	f,
		ConstVectorView	x_f_in,
		ConstVectorView	x_g_in
	)

integration_func_by_vecmult

Calculates the (row) vector that multiplied with an unknown (column) vector approximates the integral of the product between the functions represented by the two vectors: h*g = integral( f(x)*g(x) dx )

Basic principle follows Eriksson et al., Efficient forward modelling by matrix representation of sensor responses, Int. J. Remote Sensing, 27, 1793-1808, 2006. However, while in Eriksson et al. the product between f*g is assumed to vary linearly between the grid point, the expressions applied here are more advanced and are completly exact as long as f and g are piece-wise linear functions. The product f*g is then a quadratic funtion between the grid points.

Parameters

h	The multiplication (row) vector.
f	The values of function f(x).
x_f_in	The grid points of function f(x). Must be increasing.
x_g_in	The grid points of function g(x). Can be increasing or decreasing. Must cover a wider range than x_f (in both ends).

Author: Mattias Ekstr�m and Patrick Eriksson

Date: 2003-02-13 / 2008-06-12

Definition at line 1361 of file sensor.cc.

Referenced by antenna1d_matrix(), and spectrometer_matrix().

◆ met_mm_polarisation_hmatrix()

void met_mm_polarisation_hmatrix	(	Sparse &	H,
		const ArrayOfString &	pol,
		const Numeric	dza,
		const Index	stokes_dim,
		const String &	iy_unit
	)

Calculate polarisation H-matrix.

Takes into account instrument channel polarisation and zenith angle.

Parameters

[out]	H	Polarisation matrix
[in]	mm_pol	Instrument channel polarisations
[in]	dza	Zenith angle, from reference direction
[in]	stokes_dim	Workspace variable
[in]	iy_unit	Workspace variable

Definition at line 775 of file sensor.cc.

References abs, i, Sparse::insert_row(), mueller_rotation(), mult(), Array< base >::nelem(), stokes2pol(), _CS_basic_sstream_base< _CS_cT, _CS_Tr, _CS_Al >::str(), and w().

Referenced by sensor_responseMetMM().

◆ mixer_matrix()

void mixer_matrix	(	Sparse &	H,
		Vector &	f_mixer,
		const Numeric &	lo,
		const GriddedField1 &	filter,
		ConstVectorView	f_grid,
		const Index &	n_pol,
		const Index &	n_sp,
		const Index &	do_norm
	)

mixer_matrix

Sets up the sparse matrix that models the response from sideband filtering and the mixer.

The size of the transfer matrix is changed in the function as follows: nrows = f_mixer.nelem() ncols = f_grid.nelem()

The returned frequencies are given in IF, so both primary and mirror band is converted down.

Parameters

H	The mixer/sideband filter transfer matrix
f_mixer	The frequency grid of the mixer
lo	The local oscillator frequency
filter	The sideband filter data. See sideband_response for format and constraints.
f_grid	The original frequency grid of the spectrum
n_pol	The number of polarisations to consider
n_sp	The number of spectra (viewing directions)
do_norm	Flag whether rows should be normalised

Author: Mattias Ekstr�m / Patrick Eriksson

Date: 2003-05-27 / 2008-06-17

Definition at line 645 of file sensor.cc.

References GriddedField1::data, DEBUG_ONLY, GriddedField::get_numeric_grid(), GFIELD1_F_GRID, i, last(), ConstVectorView::nelem(), and Vector::resize().

Referenced by sensor_responseMixer().

◆ mueller_rotation()

void mueller_rotation	(	Sparse &	H,
		const Index &	stokes_dim,
		const Numeric &	rotangle
	)

mueller_rotation

Returns the Mueller matrix for a rotation of the coordinate system defining H and V directions.

The function follows Eq 9 in the sensor response article (Eriksson et al, Efficient forward modelling by matrix representation of sensor responses, IJRS, 2006).

The sparse matrix H is not sized by the function, in order to save time for repeated usage. Before first call of this function, size H as H.resize( stokes_dim, stokes_dim ); The H returned of this function can be used as input for later calls. That is, no need to repeat the resize command above.

Parameters

H	Mueller matrix
stokes_dim	Stokes dimensionality
rotangle	Rotation angle.

Author: Patrick Eriksson

Date: 2014-09-23

Definition at line 745 of file sensor.cc.

References DEG2RAD, Sparse::ncols(), Sparse::nrows(), and Sparse::rw().

Referenced by met_mm_polarisation_hmatrix(), and sensor_responseStokesRotation().

◆ sensor_aux_vectors()

void sensor_aux_vectors	(	Vector &	sensor_response_f,
		ArrayOfIndex &	sensor_response_pol,
		Matrix &	sensor_response_dlos,
		ConstVectorView	sensor_response_f_grid,
		const ArrayOfIndex &	sensor_response_pol_grid,
		ConstMatrixView	sensor_response_dlos_grid
	)

sensor_aux_vectors

Sets up the the auxiliary vectors for sensor_response.

The function assumes that all grids are common, and the full vectors are just the grids repeated.

Parameters

sensor_response_f	As the WSV with same name
sensor_response_pol	As the WSV with same name
sensor_response_za	As the WSV with same name
sensor_response_aa	As the WSV with same name
sensor_response_f_grid	As the WSV with same name
sensor_response_pol_grid	As the WSV with same name
sensor_response_dlos_grid	As the WSV with same name

Author: Patrick Eriksson

Date: 2008-06-09

Definition at line 938 of file sensor.cc.

References i, joker, n, ConstMatrixView::ncols(), Array< base >::nelem(), ConstVectorView::nelem(), ConstMatrixView::nrows(), Vector::resize(), and Matrix::resize().

Referenced by sensor_responseAntenna(), sensor_responseBackend(), sensor_responseBackendFrequencySwitching(), sensor_responseFrequencySwitching(), sensor_responseGenericAMSU(), sensor_responseInit(), sensor_responseMixer(), sensor_responseMixerBackendPrecalcWeights(), sensor_responsePolarisation(), and sensor_responseWMRF().

◆ spectrometer_matrix()

void spectrometer_matrix	(	Sparse &	H,
		ConstVectorView	ch_f,
		const ArrayOfGriddedField1 &	ch_response,
		ConstVectorView	sensor_f,
		const Index &	n_pol,
		const Index &	n_sp,
		const Index &	do_norm
	)

spectrometer_matrix

Constructs the sparse matrix that multiplied with the spectral values gives the spectra from the spectrometer.

The input to the function corresponds mainly to WSVs. See f_backend and backend_channel_response for how the backend response is specified.

Parameters

H	The response matrix.
ch_f	Corresponds directly to WSV f_backend.
ch_response	Corresponds directly to WSV backend_channel_response.
sensor_f	Corresponds directly to WSV sensor_response_f_grid.
n_pol	The number of polarisations.
n_sp	The number of spectra (viewing directions).
do_norm	Corresponds directly to WSV sensor_norm.

Author: Mattias Ekstr�m and Patrick Eriksson

Date: 2003-08-26 / 2008-06-10

Definition at line 975 of file sensor.cc.

References data, GFIELD1_F_GRID, Sparse::insert_row(), integration_func_by_vecmult(), Array< base >::nelem(), ConstVectorView::nelem(), Sparse::resize(), and ConstVectorView::sum().

Referenced by sensor_responseBackend().

◆ stokes2pol()

void stokes2pol	(	VectorView	w,
		const Index &	stokes_dim,
		const Index &	ipol_1based,
		const Numeric	nv = `1`
	)

stokes2pol

Sets up a vector to convert the Stokes vector to different polarsiations.

The measured value is the sum of the element product of the conversion vector and the Stokes vector. Schematically:

y[iout] = sum( w.*iy(iin,joker)

Vectors for I, Q, U and V are always normalised to have unit length (one value is 1, the remaining ones zero). The first element of remaining vectors is set to nv (and other values normalised accordingly), to allow that calibartion and other normalisation effects can be incorporated.

Parameters

s2p	Array of conversion vectors.
nv	Norm value for polarisations beside I, Q, U and V.

Author: Patrick Eriksson

Date: 2011-11-01 and 2018-03-16

Definition at line 1041 of file sensor.cc.

References Array< base >::nelem(), ConstVectorView::nelem(), and _CS_basic_sstream_base< _CS_cT, _CS_Tr, _CS_Al >::str().

Referenced by iy_transmitterMultiplePol(), iy_transmitterSinglePol(), met_mm_polarisation_hmatrix(), sensor_responsePolarisation(), and yActive().

◆ summation_by_vecmult()

void summation_by_vecmult	(	VectorView	h,
		ConstVectorView	f,
		ConstVectorView	x_f,
		ConstVectorView	x_g,
		const Numeric	x1,
		const Numeric	x2
	)

summation_by_vecmult

Calculates the (row) vector that multiplied with an unknown (column) vector approximates the sum of the product between the functions at two points.

E.g. h*g = f(x1)*g(x1) + f(x2)*g(x2)

The typical application is to set up the combined response matrix for mixer and sideband filter.

See Eriksson et al., Efficient forward modelling by matrix representation of sensor responses, Int. J. Remote Sensing, 27, 1793-1808, 2006, for details.

No normalisation of the response is made.

Parameters

h	The summation (row) vector.
f	Sideband response.
x_f	The grid points of function f(x).
x_g	The grid for spectral values (normally equal to f_grid)
x1	Point 1
x2	Point 2

Author: Mattias Ekstr�m / Patrick Eriksson

Date: 2003-05-26 / 2008-06-17

Definition at line 1607 of file sensor.cc.

◆ test_and_merge_two_channels()

bool test_and_merge_two_channels	(	Vector &	fmin,
		Vector &	fmax,
		Index	i,
		Index	j
	)

Test if two instrument channels overlap, and if so, merge them.

The channels boundaries are specified in two separate vectors, fmin and fmax. These vectors are both input and output. If merging has happened, they will each be one element shorter.

The positions of the channels to compare is given by the input parameters i and j. It is assumed that the minimum frequency of i is lower than or equal to that of j.

Furthermore, it is assumed that i itself is lower than j.

The range of the first channel (i) will have been extended to accomodate the second channel (j). The second channel will have been removed.

The function also handles the updating of index j: If the two channels do not overlap, j is increased by one.

Function returns true if merging has happened.