Controls the determination of multiple background signals

An
alternative to modelling the background in each bolometer using a simple common-mode (the COM
model) is to use principal component analysis (the PCA model). The PCA model calculates a new set
of basis vectors (components) from the bolometer time-series such that the components have
zero covariances. The amplitudes of these new correlated signals are then calculated, and
the largest amplitude components can be removed. Whereas the COM model assumes a
single background signal that is the same for all bolometers, the PCA model is capable
of detecting multiple background signals with different correlation patterns across the
array.

As with the common-mode, PCA can be used either as a distinct model within the iterative
map-making process (in which case `"`

PCA`"`

should be included in the string supplied for parameter
`"`

modelorder`"`

), or as a step in the cleaning performed prior to the start of the iterative map-making
loop (see parameter `"`

pcathresh`"`

).

The main parameter for the PCA model is the number of components to remove, specified by `"`

pca.pcathresh`"`

. If a positive value is supplied, a sigma-clipping algorithm is used to determine the
number of components to remove (see below). If a negative value is supplied, the nearest
integer (after removal of the minus sign) is found and used as the absolute number of
components to remove. Thus if pca.pcathresh is set to -10, the 10 strongest components are
removed.

If a positive value is supplied for pca.pcathresh, the RMS amplitude across all bolometers is calculated for each component – a single positive number related to the average strength of the component. An iterative sigma clipper is then used to flag components that are more than pcathresh$\ast $RMS away from the mean value. This approach is quite arbitrary, but a value of about 4 seems reasonable for some test data.

Be warned that this statistical black box will remove real correlated astronomical signals as well, thus slowing down the rate of convergence of iterative map-making algorithm. As with common-mode removal, the actual impact on science will need to be calibrated with simulations.

The PCA model can be masked in the same way as other models to reduce the influence of bright
sources on the model. A set of parameters `"`

pca.xxx`"`

can be used for this purpose, where `"`

xxx`"`

is any
of `"`

zero_accum`"`

, `"`

zero_circle`"`

, `"`

zero_freeze`"`

, `"`

zero_lowhits`"`

, `"`

zero_mask`"`

, `"`

zero_niter`"`

, `"`

zero_notlast`"`

, `"`

zero_snr`"`

, `"`

zero_snr_ffclean`"`

, `"`

zero_snr_hipass`"`

, `"`

zero_snr_lopass`"`

, `"`

zero_snr_fwhm`"`

, `"`

zero_snr_low`"`

, `"`

zero_snrlo`"`

or `"`

zero_union`"`

. See the documentation for the
equivalent `"`

com.xxx`"`

parameters for details. [4.0]

- Type:

MAKEMAP,
CALCQU

Copyright © 2012 University of British Columbia & the Science & Technology Facilities Council