Spectrum baseline removing: running fit

[EttoreCarretti]

To implement spectrum baseline estimate (and removing) by a polynomial running fit.

It is similar to a running median (or running mean), but, for any channel, a polynomial fit is performed over a narrow range of channels centred on the channel itself. The baseline estimate is then the fit evaluated in the position of the channel. Both range width and polynomial degree should be set by the user. The polynomial degree default value could be 2 (quadratic fit) which I usually found very effective in other applications.

I implemented this method for other applications and found it very effective with complex baselines, which otherwise would require high degree polynomials if fit over the whole range with poor performances.

[Malte Marquarding]

I have added this in changesets [1570,1571,1574].

It is available via

scantable.smooth(kernel='poly', order=2)

The order keyword argument is new.

[Malte Marquarding]

To use it for baselining simply do:

s = scantable('xyz.rpf')
polyfitted = s.smooth('poly', order=3, insitu=False)
baselined = s-polyfitted