NHSC SPIRE Data Processing Webinars 7 th March

NHSC SPIRE Data Processing Webinars 7 th March 2012 SPIRE Photometer Map Making C. Kevin Xu (NHSC/IPAC) page 1 PACS

NHSC SPIRE Data Processing Webinars 7 th March 2012 Contents • SPIRE mapmakers: • Naïve Mapper (pipeline default) • MADMapper (more complicated but not better, in HIPE) • Baseline Removers: • Median baseline remover (pipeline default) • Polynomial baseline remover (available in HIPE) • Destriper (minimizing stripes iteratively, available in HIPE) page 2 PACS

NHSC SPIRE Data Processing Webinars 7 th March 2012 Scan Map Pipeline Flow Chart A “reversed sequence” relative to the chain of data acquisition Options: • Default: scan median baseline remover • Advanced : polynomial baseline remover • Most advanced: an iterative destriper Baseline Remover + Mapper • Default: naïve mapper • Alternative: madmapper page 3 PACS

NHSC SPIRE Data Processing Webinars 7 th March 2012 SPIRE Naïve Mapper Flux of a sky pixel is the simple average of all signal samplings (by all detectors) in the pixel, with no weighting for either the measurement error or the position of a sampling: Sky pixel f 4 f 1 f 5 f 3 f 2 signal samplings New version (being developed): with the option of inverse-noise weighted mean. page 4 PACS

NHSC SPIRE Data Processing Webinars 7 th March 2012 MADMapper • A mapper originally developed for the CMB measurements (MAD: Microwave Anisotropy Dataset). • The algorithm is FFT based. • Strength: Minimize stripes due to uncorrelated detector drifts (“ 1/f noise”). Example (taken from an earlier talk of PACS team): Naïve map of a PACS observation MADmap of the same observation • Weakness: ringing effect (in the time domain) at any places the timelines have sharp changes (bright sources, edges of scans, etc. ) page 5 PACS

NHSC SPIRE Data Processing Webinars 7 th March 2012 SPIRE Photometer : Comparison between MADMapper & Naïve. Mapper Diff. map: MADMapper: FFT based, minimize uncorrelated 1/f noise • map. Plw=mad. Scan. Mapper(level 1, array="PLW") Naïve. Mapper: simple average • map. Plw=naive. Scan. Mapper(level 1, array="PLW") page 6 • MADMap does not improve over the simple naïve map: weak uncorr. 1/f in SPIRE!! • MADmap has shadows around the central bright source, due to ringing!! 6 PACS

NHSC SPIRE Data Processing Webinars 7 th March 2012 Baseline Remover: Why Is It Needed? example: PSW map without baseline removal you see nothing but stripes • stripes: offsets of individual detector channels on the order of ~ 0. 1 Jy/beam. (mainly caused by errors of Temperature Drift Correction). page 7 7 PACS

NHSC SPIRE Data Processing Webinars 7 th March 2012 Baseline Removers in HIPE • 3 Baseline Removers: • Median baseline remover (pipeline default). • Polynomial baseline remover. • Destriper. page 8 8 PACS

Median Baseline Remover NHSC SPIRE Data Processing Webinars 7 th March 2012 Module name: baseline. Removal. Median Signal (Jy/beam) 1. Subtract the median scan-by-scan and detector-by-detector (pipeline default). 2. Subtract the median of each timeline detector-by-detector (not very useful). ncorre Le line (u e m i t 1 vel ctor a dete cted) of Med subtr, scan-by-scan Time (second) • Strength: simple (you know what you are doing). • Weakness: • residual drifts (as shown in the illustration); • over subtraction when there is extended emission. page 9 9 PACS

NHSC SPIRE Data Processing Webinars 7 th March 2012 Polynomial Baseline Remover Module name: baseline. Removal. Polynomial Signal (Jy/beam) 1. Subtract a polynomial baseline scan-by-scan and detector-by-detector. 2. Subtract a polynomial baseline of each timeline detector-by-detector e t e d a ted) of c e r r o c line (un e m i t 1 Level Poly-order =1 (linear) baseline subtr. , whole obs Med subtr, scan-by-scan Time (second) • Strength: more flexible than median removal (often means less residual drifts). • Weakness (compare to median removal): • more severe over-subtraction (in particular when the polynomial order is high). page 10 10 PACS

NHSC SPIRE Data Processing Webinars 7 th March 2012 Polynomial Baseline Remover: Example • Remove “burp” stripes with scan-by-scan polynomial baseline removal: Median baseline removal 5 th order polynomial baseline removal ~4. 5 deg scan length 20”/sec scan speed (parallel mode) OBSID = 1342189003 page 11 9 th order polynomial baseline removal Caveat: this won’t work if the background is structured. 11 PACS

NHSC SPIRE Data Processing Webinars 7 th March 2012 ROI: Region Of Interest ROI (available for both median & polynomial baseline removers): Excluding bright regions from baseline estimate to minimize the over-subtraction. An example for median remover diff. map Median baseline removal without ROI mask. page 12 Data inside this ROI are masked & excluded from baseline removal. 12 PACS

Destriper: How It Works NHSC SPIRE Data Processing Webinars 7 th March 2012 • Module name: destriper • Minimize map stripes by adjusting the baseline subtraction (median or polynomial) iteratively using the naïve mapper: Signal of j-th detector at i-th sampling Naive Mapper: Sky pixels 1 2 Pix 1 Detector 1 Signal (Jy/beam) Baseline of j-th detector at i-th sampling Pix 1 Pix 2 Pix 1 Detector 2 Baselines (poly-order=1) Pix 2 Residual = (signal-baseline) – fpix (this is to be minimized through the iterations. ) New baseline adjustments (δB, additive). time (second) page 13 13 PACS

Destriper: How It Works NHSC SPIRE Data Processing Webinars 7 th March 2012 • Module name: destriper • Minimize map stripes by adjusting the baseline subtraction (median or polynomial) iteratively using the naïve mapper: Signal of j-th detector at i-th sampling Naive Mapper: Sky pixels 1 2 Pix 1 Detector 1 Signal (Jy/beam) Baseline of j-th detector at i-th sampling Pix 1 +δB B ld = o : B new e n i l e s a New. Pixb 2 Pix 1 Detector 2 Pix 2 And a new iteration starts (the iteration stops when δB’s converge to zero or, occasionally, the residuals converge to 0). Baselines (poly-order=1) Residual = (signal-baseline) – fpix (this is to be minimized through the iterations. ) New baseline adjustments (δB, additive). time (second) page 14 14 PACS

NHSC SPIRE Data Processing Webinars 7 th March 2012 Destriper: Continued • Strength: • make better maps in general (no significant over-subtraction); • the only SPIRE baseline remover in HIPE for structured emission regions; • can do the 2 nd order deglitching (HIPE 9. 1). • Weakness: • longer computation time, higher demand on RAM. Median Baseline Remover: page 15 Destriper: 15 PACS

Destriper + Extended Gain Correction • Applying Destriper + Extended Gain Correction can further improve the mapmaking! NHSC SPIRE Data Processing Webinars 7 th March 2012 • Why? The timelines are calibrated using the point source calibration, while a map measures the extended emission. Errors due to the difference between point source calibration and extended source calibration affect the timelines of different detectors differently. The extended gain correction removes these (systematic) errors. Naïve Mapper + Destriper (amplifying low level signals). page 16 Detector 1 Detector 2 Naïve Mapper + Destriper + Extended Gain Correction 16 PACS

NHSC SPIRE Data Processing Webinars 7 th March 2012 Other Two HIPE Scripts Related to SPIRE Mapmaking • Map Merging: Make maps using timelines of more than 1 observations (particularly useful for parallel mode observations with cross-scans). • Solar System Object Motion Correction: Correct the proper motion of solar system objects (causing blurred images) before the mapmaking. page 17 17 PACS

NHSC SPIRE Data Processing Webinars 7 th March 2012 Summary • We do not recommend to use MADMapper for SPIRE photometer maps. • SPIRE maps, produced by standard pipeline (with median baseline remover and the naïve mapper), can be improved using more advanced polynomial baseline remover or destriper. • In general, high quality SPIRE maps can be produced using the destriper together with the naïve mapper. However, this takes longer than the standard processing, and requires more RAM (> 8 GB recommended). • In many cases, the polynomial baseline remover (with or without ROI masking) can do a significantly better job than the default median baseline remover. However, be aware of removing real signals. page 18 18 PACS