Phase fitter that can force phase solutions over frequency onto a TEC model. More...

#include <PhaseFitter.h>

Public Member Functions
	PhaseFitter ()

	PhaseFitter (size_t channelCount)

void	FitDataToTEC1Model ()

double	FitDataToTEC1Model (double &alpha)

double	FitDataToTEC1ModelWithInitialValues (double &alpha)

void	FitDataToTEC2Model ()

double	FitDataToTEC2Model (double &alpha, double &beta)

double	FitDataToTEC2ModelWithInitialValues (double &alpha, double &beta)

void	FitTEC1ModelParameters (double &alpha) const

void	FitTEC2ModelParameters (double &alpha, double &beta) const

double	FittingAccuracy () const

const std::vector< double > &	GetFrequencies () const

void	Initialize (const std::vector< double > &frequencies)

double *	PhaseData ()

const double *	PhaseData () const

void	SetFittingAccuracy (double newAccuracy)

size_t	Size () const

double	TEC1ModelCost (double alpha) const

double	TEC2ModelCost (double alpha, double beta) const

double *	WeightData ()

const double *	WeightData () const

Static Public Member Functions
static double	AlphaToTEC (double alpha)

static double	TEC1ModelFuncWrapped (double nu, double alpha)

static double	TEC2ModelFunc (double nu, double alpha, double beta)

static double	TEC2ModelFuncWrapped (double nu, double alpha, double beta)

Detailed Description

Phase fitter that can force phase solutions over frequency onto a TEC model.

To use:

Construct and set the frequencies (with FrequencyData() ) and if already possible the weights (WeightData()).
Perform a calibration iteration.
Set the phase values (PhaseData()) and, if not yet done, the weights (WeightData()) to the current phase solutions / weights of a single antenna.
Call FitDataToTEC1Model() or FitDataToTEC2Model().
Read the new phase values.
When fitting multiple polarizations, the above steps should be done twice for each diagonal value of the Jones matrix. Also repeat for all antennae.
Continue the iteration with the new phase values.

Methods with name containing TEC1 refer to a single-parameter TEC model (no delay), while methods with TEC2 refer to dual-parameter TEC model (TEC + delay).

Constructor & Destructor Documentation

◆ PhaseFitter() [1/2]

PhaseFitter::PhaseFitter ( )

inline

◆ PhaseFitter() [2/2]

PhaseFitter::PhaseFitter ( size_t channelCount )

inline

Construct a phase fitter for the given number of channels. Weights are initialized to unity. The fitting accuracy is initialized to 1e-6.

Parameters

channelCount number of channels.

Member Function Documentation

◆ AlphaToTEC()

static double PhaseFitter::AlphaToTEC ( double alpha )

inlinestatic

◆ FitDataToTEC1Model() [1/2]

void PhaseFitter::FitDataToTEC1Model ( )

inline

Like FitDataToTEC1Model(double&), but without returning the parameters.

◆ FitDataToTEC1Model() [2/2]

double PhaseFitter::FitDataToTEC1Model ( double & alpha )

Dummy comment for making the reference near FitDataToTEC1Model() work.

◆ FitDataToTEC1ModelWithInitialValues()

double PhaseFitter::FitDataToTEC1ModelWithInitialValues ( double & alpha )

inline

◆ FitDataToTEC2Model() [1/2]

void PhaseFitter::FitDataToTEC2Model ( )

inline

Like FitDataToTEC2Model(double&,double&), but without returning the parameters.

◆ FitDataToTEC2Model() [2/2]

double PhaseFitter::FitDataToTEC2Model	(	double &	alpha,
		double &	beta
	)

Fits the given phase values to a TEC model and returns the parameters. This function is robust even when phase wrapping occurs. After this call, the PhaseData() satisfy the TEC model. The TEC model has two parameters and are fitted as described in FitTEC2ModelParameters().

Parameters

alpha	Found value for the alpha parameter.
beta	Found value for the beta parameter.

Returns: Cost of the found solution.

◆ FitDataToTEC2ModelWithInitialValues()

double PhaseFitter::FitDataToTEC2ModelWithInitialValues	(	double &	alpha,
		double &	beta
	)

inline

Fits the given phase values to a TEC model using prior estimates of the model parameters. This method is similar to FitDataToTEC2Model(), except that it will use the provided initial values of alpha and beta to speed up the solution. When the provided initial values are not accurate, the fit might not be accurate.

Todo:: No fast method has been implemented – instead it will perform a full parameter search.

Parameters

alpha	Estimate of alpha parameter on input, found value on output.
beta	Estimate of beta parameter on input, found value on output.

◆ FitTEC1ModelParameters()

void PhaseFitter::FitTEC1ModelParameters ( double & alpha ) const

◆ FitTEC2ModelParameters()

void PhaseFitter::FitTEC2ModelParameters	(	double &	alpha,
		double &	beta
	)		const

Fit the data and get the best fitting parameters. The model used is f(nu) = alpha/nu + beta, with possible 2pi wrappings in the data. The phase data is not changed. The alpha parameter is linearly related to the TEC. The beta parameter is a constant phase offset, given in radians. The fitting algorithm uses a combination of brute force searching and binary-like searching (ternary search).

Parameters

alpha	Will be set to the fitted value for the alpha parameter (value on input is not used).
beta	Will be set to the fitted value for the beta parameter (value on input is not used).

◆ FittingAccuracy()

double PhaseFitter::FittingAccuracy ( ) const

inline

Get the fitting accuracy. The fitter will stop once this accuracy is approximately reached. The default value is 1e-6.

Returns: Fitting accuracy.

◆ GetFrequencies()

const std::vector<double>& PhaseFitter::GetFrequencies ( ) const

inline

Get the frequency values.

Returns: Vector of Size() doubles with the frequencies in Hz.

◆ Initialize()

void PhaseFitter::Initialize ( const std::vector< double > & frequencies )

inline

Change the number of channels to be fitted.
Set the frequency data.
Discard the phase and weight data.
Parameters

frequencies The new frequencies, such that frequencies[ch] is the frequency corresponding to the phase value PhaseData()[ch]. The number of channels becomes the length of this vector.

◆ PhaseData() [1/2]

double* PhaseFitter::PhaseData ( )

inline

Get a pointer to the array of phase values. This array should be filled with the phases of solutions before calling one of the fit methods. FitDataToTEC1Model() and ~TEC2~ sets this array to the fitted phases. Normally, these values should be set to std::arg(z) or atan2(z.imag(), z.real()), where z is a complex solution for one polarizations. All phases should correspond to the same polarizations, i.e., different polarizations (xx/yy/ll/rr, etc.) should be independently fitted.

Returns: Array of Size() doubles with the phases.

◆ PhaseData() [2/2]

const double* PhaseFitter::PhaseData ( ) const

inline

Get a constant pointer to the array of values.

Returns: Constant array of Size() doubles with the phases.

◆ SetFittingAccuracy()

void PhaseFitter::SetFittingAccuracy ( double newAccuracy )

inline

Change the fitting accuracy. See FittingAccuracy().

Parameters

newAccuracy New accuracy.

◆ Size()

size_t PhaseFitter::Size ( ) const

inline

Number of channels used for the fitter.

Returns: Number of channels.

◆ TEC1ModelCost()

double PhaseFitter::TEC1ModelCost ( double alpha ) const

Evaluate the cost function for given TEC model parameter. The higher the cost, the worser the data fit the given parameters.

Parameters

alpha TEC parameter

Returns: sum of | alpha / nu_i - theta_i |, and each sum term is phase unwrapped.

◆ TEC1ModelFuncWrapped()

static double PhaseFitter::TEC1ModelFuncWrapped	(	double	nu,
		double	alpha
	)

inlinestatic

Parameters

nu	Frequency in Hz
alpha	TEC parameter (in undefined units)

Returns: | alpha / nu_i + beta | % 2pi

◆ TEC2ModelCost()

double PhaseFitter::TEC2ModelCost	(	double	alpha,
		double	beta
	)		const

Evaluate the cost function for given TEC model parameters. The higher the cost, the worser the data fit the given parameters.

Parameters

alpha	TEC parameter
beta	Phase offset parameter

Returns: sum of | nu_i / alpha + beta - theta_i |, and each sum term is phase unwrapped.

◆ TEC2ModelFunc()

static double PhaseFitter::TEC2ModelFunc	(	double	nu,
		double	alpha,
		double	beta
	)

inlinestatic

Like TEC2ModelFunc(), but 2-pi wrapped.

Parameters

nu	Frequency in Hz
alpha	TEC parameter (in undefined units)
beta	Phase offset parameter (in radians)

Returns: | nu_i / alpha + beta | % 2pi

◆ TEC2ModelFuncWrapped()

static double PhaseFitter::TEC2ModelFuncWrapped	(	double	nu,
		double	alpha,
		double	beta
	)

inlinestatic

Like TEC2ModelFunc(), but 2-pi wrapped.

Parameters

nu	Frequency in Hz
alpha	TEC parameter (in undefined units)
beta	Phase offset parameter (in radians)

Returns: | nu_i / alpha + beta | % 2pi

◆ WeightData() [1/2]

double* PhaseFitter::WeightData ( )

inline

This array should be filled with the weights of the channel phase solutions. If the solver supports weights during solving, this value should be set to the sum of weights of all visibilities that are used for the solution of this channel. If the solver does not support weights, it should be set to the number of unflagged visibilities used by the solver to generate the corresponding phase. Another way of saying this, is that the weights should be set to the esimated inverse variance of the phase solutions.

The use of weights will make sure that noisy channels do not bias the result. Weighting is for example helpful to avoid that the few remaining samples in a badly RFI contaminated channels cause the fit to be inaccurate.

While the weights could be different for each antenna solution, generally the weight of a channel is constant over the antennas. The latter implies that the weights can be set once before the solution starts, and only the PhaseData() need to be changed within solution iterations.

The weights are initially set to one.

Returns: Array of Size() doubles with the weights.

◆ WeightData() [2/2]

const double* PhaseFitter::WeightData ( ) const

inline

Constant array of weights, as described above.

Returns: Constant array of Size() doubles with weights.

The documentation for this class was generated from the following file:

PhaseFitter.h

Public Member Functions

Static Public Member Functions

Detailed Description

Constructor & Destructor Documentation

◆ PhaseFitter() [1/2]

◆ PhaseFitter() [2/2]

Member Function Documentation

◆ AlphaToTEC()

◆ FitDataToTEC1Model() [1/2]

◆ FitDataToTEC1Model() [2/2]

◆ FitDataToTEC1ModelWithInitialValues()

◆ FitDataToTEC2Model() [1/2]

◆ FitDataToTEC2Model() [2/2]

◆ FitDataToTEC2ModelWithInitialValues()

◆ FitTEC1ModelParameters()

◆ FitTEC2ModelParameters()

◆ FittingAccuracy()

◆ GetFrequencies()

◆ Initialize()

◆ PhaseData() [1/2]

◆ PhaseData() [2/2]

◆ SetFittingAccuracy()

◆ Size()

◆ TEC1ModelCost()

◆ TEC1ModelFuncWrapped()

◆ TEC2ModelCost()

◆ TEC2ModelFunc()

◆ TEC2ModelFuncWrapped()

◆ WeightData() [1/2]

◆ WeightData() [2/2]