Demixer

Demix bright sources, i.e. calibrate in the direction of the birght sources and subtract the bright sources from the data. Note that the calibration solution is assumed to be constant over the entire frequency range, so this will only work on relatively small bandwidth (e.g. a single sub-band).

inputs

type

type:

string

default:

demixer

doc:

Case-insensitive step type; must be ‘demixer’ (or ‘demix’) .

baseline

type:

string

default:

""

doc:

Baselines to demix. See Description of baseline selection parameters

blrange

type:

double[]

default

doc:

Baselines to demix. See Description of baseline selection parameters

corrtype

type:

string

default:

cross

doc:

Baselines to demix. Correlation type to match? Must be auto, cross, or an empty string .

defaultgain

type:

double

default:

1

doc:

The default and initial gain for the directional gains that are computed internally .

demixfreqresolution

type:

float

default:

0

doc:

Frequency resolution to use when demixing, in Hz (or append “MHz” or “kHz” to specify it in those units). If this is given, and is greater than zero, it overrides <step>.demixfreqstep. This will not exceed the bandwidth of a subband, even if a larger value is given .

demixfreqstep

type:

int

default:

0

doc:

Number of channels to average when demixing. It is truncated if exceeding the actual number of channels. It defaults to the averaging used for the subtract .

demixtimeresolution

type:

float

default:

0

doc:

Time resolution to use when demixing, in seconds. If this is given, and is greater than zero, it overrides <step>.demixtimestep .

demixtimestep

type:

int

default:

0

doc:

Number of time slots to average when demixing. It is truncated if exceeding the actual number of times. It defaults to the averaging used for the subtract .

freqstep

type:

int

default:

1

doc:

Number of channels to average when subtracting. It is truncated if exceeding the actual number of channels. Note that the data itself will also be averaged by this amount .

ignoretarget

type:

boolean

default:

False

doc:

false = project the target source away; true = ignore the target .

instrumentmodel

type:

string

default:

instrument

doc:

The name of the ParmDB to use. The ParmDB does not need to exist. If it does not exist it will be created .

lbfgs.historysize

type:

int

default:

10

doc:

If using the LBFGS solver, the history size, specified as a multiple of the parameter vector, to use to approximate the inverse Hessian .

lbfgs.robustdof

type:

double

default:

2.0

doc:

If using the LBFGS solver, the degrees of freedom (DOF) given to the noise model. The larger this becomes (more than 30), the noise model behaves as a Gaussian. The smaller values make the noise model more robust to outliers in the data .

maxiter

type:

int

default:

50

doc:

Maximum number of iterations used in the LM solve .

modelsources

type:

string[]

default:

[]

doc:

Names of sources with models to take into account when solving. the sources must exist as patches in the SourceDB. Note that the target should NOT be part of this parameter. If a model of the target has to be used, it has to be given in parameter targetsource.

ntimechunk

type:

int

default:

0

doc:

Number of demix time slots (after averaging) that are processed jointly in as much a parallel way as possible. If subtract uses different time averaging, it has to fit integrally .

othersources

type:

string[]

default:

[]

doc:

Names of sources of which the direction is taken into account when demixing by projecting the directions away. The direction needs to be specified if the source is unknown (which is usually the case). It can be done using parameters <step>.<sourcename>.phasecenter.

propagatesolutions

type:

boolean

default:

True

doc:

If set to true, solutions of a time slot are used as initial values for the next time slot. If set to false, the diagonal elements of the Jones matrix are initialized to one and the off-diagonal elements to zero .

skymodel

type:

string

default:

sky

doc:

Sourcedb (created with makesourcedb) with the sky model to calibrate on. When the path ends with .skymodel or .txt DP3 expects a skymodel file as used by makesourcedb. This makes it possible to directly use a skymodel file without using makesourcedb to convert the file.

step_name

type:

string

default:

demix

doc:

unique name for the step .

subtractsources

type:

string[]?

default:

[]

doc:

Names of the sources to subtract. If none are given, demixing comes down to averaging. The sources must exist as patches in the SourceDB .

targetsource

type:

string

default:

""

doc:

This setting can be used to specify the name of the source model of the target. If given, the target source model (its patch in the SourceDB) is taken into account when solving; in this case parameter othersources cannot be given. It cannot be given if ignoretarget=true. If not given, the target is projected away or ignored (depending on parameter ignoretarget).

timestep

type:

int

default:

1

doc:

Number of time slots to average when subtracting. It is truncated if exceeding the actual number of times. Note that the data itself will also be averaged by this amount .

uselbfgssolver

type:

boolean

default:

False

doc:

If true, use LBFGS solver instead of the LM solver (or LSQfit) .