Using a C-BASS like design, one can make a system that measures I in a gain-fluctuation nulling fashion, at a cost of a little bit of SNR. Chris says:
Polarized foregrounds: Richard Shaw has done quite a bit of work making
semi-realistic simulations of polarized foregrounds. Figure 8 of
http://arxiv.org/abs/1401.2095 is a nice illustration. We wondered why
Faraday rotation of galactic synch results in non-smooth spectral features
despite the fact that the rotation angle goes smoothly as lambda^2. It is
because the total integrated emission comes from a range of distances with
different Faraday depths in the galaxy. The rotation angle is then a
complicated function of frequency. I think the lambda^2 scaling is a red
herring. The Figure in Shaw et al shows filtered Stokes Q contamination
being sub-percent relative to the 21-cm signal. This is sort of a worst
case for an instrument that is sensitive to only one polarization. With
dual pol, we can make stokes I maps that, modulo instrumental
non-idealities that mix P->I, are insensitive to polarization. This makes
the contamination go down. So maybe we're okay without having to worry too
much. I'd be curious to see the simulated polarized synch spectrum with
some very naive filtering rather than this sophisticated m-mode analysis,
e.g. a 4th order poly removed from each column of the spectrum in the top
middle panel of Fig 8. But from Fig 8 and Fig 9, it does look like beam
frequency dependence is by far the dominant source of spectral
contamination post-filtering.
Note that CRIME does make some guess at polarization. We should look into that.
Using a C-BASS like design, one can make a system that measures I in a gain-fluctuation nulling fashion, at a cost of a little bit of SNR. Chris says:
Polarized foregrounds: Richard Shaw has done quite a bit of work making semi-realistic simulations of polarized foregrounds. Figure 8 of http://arxiv.org/abs/1401.2095 is a nice illustration. We wondered why Faraday rotation of galactic synch results in non-smooth spectral features despite the fact that the rotation angle goes smoothly as lambda^2. It is because the total integrated emission comes from a range of distances with different Faraday depths in the galaxy. The rotation angle is then a complicated function of frequency. I think the lambda^2 scaling is a red herring. The Figure in Shaw et al shows filtered Stokes Q contamination being sub-percent relative to the 21-cm signal. This is sort of a worst case for an instrument that is sensitive to only one polarization. With dual pol, we can make stokes I maps that, modulo instrumental non-idealities that mix P->I, are insensitive to polarization. This makes the contamination go down. So maybe we're okay without having to worry too much. I'd be curious to see the simulated polarized synch spectrum with some very naive filtering rather than this sophisticated m-mode analysis, e.g. a 4th order poly removed from each column of the spectrum in the top middle panel of Fig 8. But from Fig 8 and Fig 9, it does look like beam frequency dependence is by far the dominant source of spectral contamination post-filtering.
Note that CRIME does make some guess at polarization. We should look into that.