The methods paper needs some justification of the number of PCA components used. This will require a map of some field with a range of number of PCA components. Plan: simulate a map of L111 (the most square field) with 0-20 PCA components x 21 iterations and a variety of source sizes and plot the recovered flux vs. number of PCA components. Ideally, do this with both deconvolution and not. Estimated processing time is ~24 hours. Also, a plot of flux vs. iteration number will be useful. Glitch filtering method has been modified: "Glitches are removed by drizzling each bolometer measurement into a given pixel using the mapping M[p], but retaining each pixel as an array of measurements. Then, measurements exceeding $3\times MAD$ (Median Average Deviation) are flagged out in the timestream. In cases where there were too few ($<3$) hits per pixel, the pixel was completely flagged out. This only occurred for pixels at scan edges." Data flagging: Partly covered by deglitching. Many scans were flagged by hand to remove overly noisy scans and those that were observed to confuse the iterative mapper. Hand flagging is more robust than automated and can remove features caused by the filter convolved with the glitch. Creation of astrophysical model: Not entirely sure what this section entails. Should have a subsection on deconvolution though. Jackknifing has not generally been done...
4.3 Relative Alignment and Mosaicing
Relative alignment was performed by finding the peak of the cross-correlation between images and a pointing master selected from the epoch with the best-constrained pointing model for that field. Each observation was initially mapped individually, then all observations of a given field were cross-correlated with a selected master image of that field. The cross-correlation peak was fit with a gaussian and the difference between the gaussian peak and the image center was used as the pixel offset. The offsets were recorded and written to the timestreams. Finally, all observations of a field were merged into a single timestream with pointing offsets applied to create the field mosaic.
Flagger images
James requested sample images from the flagger for the methods paper. Below are images + description: .. image:: http://3.bp.blogspot.com/_lsgW26mWZnU/SXHsZcqegtI/AAAAAAAAEs4/3PjUtU2IXtQ/s400/sample_waterfall_070911_o15_highFnoisepng.png This is a "noise-dominated" scan in the sense that the high/low pixels are set by noise, not signal. Despite the clear high frequency noise, this image actually maps out pretty well - I think the high-noise bolometers get downweighted and the high/low pixels probably get clipped by my hot pixel rejection procedure. .. image:: http://3.bp.blogspot.com/_lsgW26mWZnU/SXHsY7Rgq3I/AAAAAAAAEsY/0NT8IB26JUo/s400/flagger_marked_source_050708_o15.png An image of the galactic center with a source southeast of center identified. .. image:: http://1.bp.blogspot.com/_lsgW26mWZnU/SXHsYyHCEGI/AAAAAAAAEsg/3iwoklY0--0/s400/sample_waterfall_050708_o15_glitchandsources.png A scan from the GC image above. I forgot to mark the source, I should go back and do that. The glitch is obvious. .. image:: http://1.bp.blogspot.com/_lsgW26mWZnU/SXHsZHVZ4OI/AAAAAAAAEso/pOXwGFn7rGg/s400/sample_waterfall_050708_o15_glitchflagged.png I drew a box to flag out the region affected by the glitch. .. image:: http://4.bp.blogspot.com/_lsgW26mWZnU/SXHsZG_MmWI/AAAAAAAAEsw/RA6tqAWMnLM/s400/sample_waterfall_050708_o15_glitchgone.png This is what happens when I redraw after flagging out the glitch - the colorbar is rescaled and no more glitch. .. image:: http://4.bp.blogspot.com/_lsgW26mWZnU/SXIDILh3hfI/AAAAAAAAEtQ/cMuOhlFO_fA/s400/sample_waterfall_050708_o15_glitchandsources_marked.png Timestream with glitches and sources marked (one pixel in the map is hit by 3 different points in this scan). .. image:: http://4.bp.blogspot.com/_lsgW26mWZnU/SXIDH3uIjHI/AAAAAAAAEtI/vijTeEWf_nA/s400/sample_waterfall_050708_o15_glitchandsources_gray.png Grayscale version of above (ok, I lied about grayscale being impossible) with a different pixel marked. .. image:: http://1.bp.blogspot.com/_lsgW26mWZnU/SXIDHbYLbiI/AAAAAAAAEtA/nSh2rM6GhMc/s400/flagger_marked_source_footprint_050708_o15.png A zoom-in around the 'kidney bean' source with the Bolocam footprint overlaid and a pixel marked. Note that this pixel corresponds to the 3 points in the color waterfall above.
Paragraph for the methods paper
The raw data from Bolocam contains noise components from the atmosphere and instrument in addition to the astrophysical signal. To remove the atmospheric noise, an iterative approach was required.
- The median across all bolometers is subtracted
- A set number of principle components are subtracted. The principle components are the most correlated components between bolometers. In this process both the atmosphere above the telescope - which is assumed to be constant across the field of view - and any large-scale astrophysical structure are removed.
- The timestream data is mapped into the plane of the sky. Data points are mapped to the nearest pixel. 7.2" pixels are used so that sampling is better than Nyquist.
- The map is deconvolved using a maximum entropy deconvolution algorithm ( Based on paper by Hollis, Dorband, Yusef-Zadeh, Ap.J. Feb.1992, written by Frank Varosi at NASA/GSFC 1992)
- The deconvolved map is returned to a timestream and subtracted from the original to yield a noise-only timestream.
- Power spectral densities are calculated for each scan in the noise timestream, and weights are calculated from these. [At the moment, the weights are actually inverse-variance]
- The deconvolved map timestream is subtracted from the raw timestream, and then steps 1-6 are repeated on that timestream to recover flux that was oversubtracted in the first iteration.
Convergence takes ???? iterations.... ??? PCA components are subtracted [default 13]...
Page 1 / 1