Minutes of the Parallel mode data reduction telecon of 12 July 2011

On the call: Vera Könyves, Edo Ibar, Robbie Auld, Bruno Altieri,
Luca Conversi, Herve Ausel, Babar Ali, Elena Puga, Katrina
Exter, Veli-Matti Pelkonen, Julia Roman-Duval, Chad Engelbracht,
David Shupe, Helene Roussel, Martin Henneman, Matthew
Smith, Bruno Merin, others

*'Cross' artifacts on the map on compact sources, steep gradients, or on glitches,
likely caused by uncertainties in the astrometry.

- This artifact concerns the Maximum Likelihood based algorithms (ROMAGAL,

- faint sources are also affected by this pointing accuracy, but they don't

suffer from the cross-like artifact.

- Babar et al. uses/extended a method which was developed for the ROMAGAL
code used by the HIGAL Herschel Open Time Key Programme.
The method removes these point?source artifacts from the final maps by first
creating then subtracting an artifact map from
the final maps (document: FirstReport.pdf, written by Babar).
- At the moment it seems that photometry is different on sources, with, and without
the cross artifacts. Should be checked...
- Any user interested in trying this prototype code and understanding its limitations
in its current state, should contact the NHSC to ask for it.

Herve Aussel's comment: it is a kind of 'cosmetics', doesn't cease the origin of
the problem.

- Pierre Chanial, Herve Aussel et al. are working on improving the pointing accuracy
to remove these artifacts from the TAMASIS maps.

- It doesn't suffer from this artifact (yet)??? We suspect that on H-Atlas fields
there are no bright sources with sharp flux gradients.

- Helene Roussel excludes sources with steed gradients from the drift calculation...
- SCANAMORPHOS doesn't do back-projection either, where the errors could propagate
and finally create the cross-like artifact.

  • Still, we are within 10-20 % accuracy for the point source photometry, drawn from
different maps (photProject, scanamorphos, madMap...), true both for PACS and SPIRE wavelengths.
(document: Alvaro's presentation at the HIPE Forum 2011)

  • The maps are 'science dependent' (it concerns rather the PACS data)
  • Related issue: small-, and large-scale emission preserving:

- Different maps or different data processing and map making parameters should/could
be used for different purposes; for compact source studies and for extended emission studies.
It is not unique that on the same field of the Gould Belt survey we want to investigate both
small and larger scales...

- This is the qiestion of accuracy too, if we want to use differently prepared maps for different
scientific purposes...

- I miss some explanation for this issue:
- recommended in the blue band: point-source within an aperture of 12"???
- also there is PACS-MIPS comparison...

- Conclusion:
Preferred codes to preserve large-scales, large-structures: madMap???, sanepic, scanamorphos also?
Codes, doing well/better for compact sources: photProject, scanamorphos

- H-Atlas/sanepic people are making simulations for telling how much large-scale emission they miss.
100 micron IRAS data is used, scaled to 160 micron cirrus emission...
- This method can be used for madMap...

- Question for extended emission differences: isn't is just a sensitivity issue between madMap
and sanepic? madMap relies heavily on preprocessing, when it filters the timeline, unlike sanepic...

  • PACS correlated noise...

- The drift is thermal, it has time dependence, showing different nature from the beginning to the
end of the OD.

  • Strong/long glitches, caused by slow response of the bolometer, should be removed, as well as
bright artifacts after the calibration blocks.

- How to identify them in the timeline:

- How relevant they are in the observations:
H-Atlas: In a 18 hr observation => 10 big glitches are affecting the maps (+ more can be there
at lower amplitudes)
- The H-Atlas KP has a IDL routine in development to identify them and get rid of them.

  • Offsets between in-scan and cross-scan (more for PACS)

- No shift is expected for concatenated scans.

  • Offsets between PACS and SPIRE

- Few arcseconds (2-3") offsets can be 'normal' between PACS and SPIRE maps.

- Both relative inaccuracies can be treated by re-centering the scans or crossmatching PACS and
SPIRE maps.
- Crossmatching with existing (lots of) sources improves the absolute photometry.
- This effect can be more expressed with fast scanning, when the PSF is more elongated.

  • A very big offset has been reported from a KP between the SPIRE and PACS maps
from a parallel mode observation. The obsid was collected and it is being investigated
now at the HSC. Any observer having such problem should contact the HSC to investigate
its origin.

  • 2nd level deglitching in HIPE 8 is less memory consuming (Bruno Altieri)

- 10 hr observation of parallel-mode blue band with 32 GB was done!!!
- downside of this change, according to Herve Aussel: if we want to test different thresholds,
we need to redo the deglitching, since in this new HIPE 8 method the mapindex is not used???
- Herve A. developed a method, working in HIPE, to do the 2nd level deglitching by chunks.
One region of the map can be selected, or the whole map can be treated by chunks. It is less
memory consuming this way.

Next telecon:

The next telecon of this series will be annouced in September 2011.