HIFI instrument and calibration web page
Latest updates -- 12 April 2016
- HIPE 14.1 has been released !
- Check the important calibration updates and associated improvements expected in the HIPE 14.0.1 and upcoming 14.1 reprocessed data - this includes in particular updated sideband gain ratio in bands 1 to 4
- The HIFI products now come with a frequency-dependent intensity calibration uncertainty budget
- Channel flags for spurs are now populated in all HIFI observations (this was limited to spectral scan up to HIPE 13)
- Re-organisation of the HIFI quality flags for a better assessement of issues being relevant or not to the end products
- Users wishing to import their HIFI data into Gildas/CLASS can now do it directly with the Gildas oct15 release without the need of any previous conversion
Observing with HIFI
- HIFI calibration data
- A significant number of the data taken by HIFI in the framework of the Performance Validation (PV) and routine phase are now publicly available. There are particular caveats that apply to those data. They are described in this disclaimer note.
- See also
- HIFI fact sheet A 2-page guide to the HIFI instrument
- A&A paper describing the method used to measure stability performance with HIFI (added 4 Oct 2011)
- A&A paper about the optimisation of mapping modes for heterodyne instruments (added 18 March 2014)
Reducing HIFI data
Recommended User release
HIPE 14 and Java 8
: please consult the HIPE Known Issues
page regarding important information on issues with Java for different versions of HIPE.
Which data would most benefit from being reprocessed in HIPE 14.1 ?
If your data have been processed with the latest bulk reprocessing (13.0), HIPE 12.0 or HIPE 11.1
- Spectra mean and median statistical numbers
- The HIFI 14.0.1 products suffered from a minor bug leading to a wrong computation of the mean and median spectra values (i.e. a measure of the DSB continuum to first order). This has been repaied in 14.1
- Intensity calibration uncertainty
- HIPE 14 introduces a new product providing an estimated breakdown of the intensity calibration uncertainty. The uncertainties are given separately for each component, as well as in a quadratic sum fashion for all statistically independent elements. This sum applies to the calibration of data in the Ta* scale. Further systematic and random error apply for data converted into different intensity scales (e.g. Tmb or Jy). See further details in section 9 of the DRG.
- Pointing reconstruction
- In HIPE 14, those observations where the interlacing mode was activated (i.e. using more than just 9 guide stars in the star tracker) will benefit from an more accurate reconstructed astrometry and so positions may change in those cases. On top of that, the quality figure associated to the new gyro-propagated method introduced back in HIPE 13 (see also the following Pointing Information page) is now computed in a slightly different fashion. Since the application or not of the gyro-propagated pointing is, for HIFI, conditional upon a certain threshold on this quality (avoiding to apply the new pointing reconstruction to under-performing case), the number of case making us of one or the other approach will differ in HIPE 14 - we recall that in case the gyro-propagated method is discarded, the pointing is the same as that used back in HIPE 12.For the HIFI data such a correction is applied in a conditional fashion depending on a quality figure computed for each individual observation. The new pointing reconstruction will not apply to under-performing cases, and those latter will still use the pointing files used back in HIPE 12. Details about the new attitude reconstruction, and the way it is approached and may impact the HIFI data, can be found in this memo.
- Band 6 and 7 observations
- From HIPE 13 onward observations in bands 6 and 7 have been automatically corrected from the Electrical Standing Wave affecting those data. The correction is based on an optimised fit to the baseline artefact stored in the HIFI calibration files and applied by the pipeline. In HIPE 14, some of those stored solutions have been refined and so the resulting products will be improved. For instructions on how to benefit from this reprocessing and see the typical improvement expected in the data please refer to this section from the HIFI Data Reduction Guide (DRG).
- Spectral Scan observations
- The reprocessing of spectral scans in HIPE 13 made use of optimised mask tables for spurs and unruly baseline ranges, resulting in improved deconvolved solutions at the Level 2.5 (an example of such improvement between HIPE 12 (left - red) and HIPE 13 (right - black) is shown in this plot - see also section 5.4 of the DRG). In HIPE 14, a dozen of additional spectral scans (mostly from the calibration programme) have been added to this list.In order to benefit from this update you should reprocess your data from Level 0 with the new calibration tree.
- Spur warning flags
- In HIPE 14, a new channel flag has been introduced ("warning") that will be assigned to the data based on a knowledge base of spurious features built from the spectral scan flagging (see bullet above). These flags will be applied to all point and mapping observations. They should be taken as indicative as they not necessarily accurately match spurious features in the data where they got applied. For this reason, this new flag is not honoured by any of the standard interactive post-processing tasks.
If your data have been processed with a version earlier than 11.1
- Flags in OFF positions.
- In HIPE 13 the data used in the OFF positions were already processed up to an equivalent Level 2 calibration (both in intensity and frequency) in order to be directly comparable to the ON-target data. In HIPE 14, those OFF spectra will also hold spur and warning flags that will be propagated from the ON-targert ones. This feature is for example very interesting in order to perform a deconvolution of the OFF spectras in a spectral scan and be able to compare the OFF spectra to the level 2.5 deconvolved products of those observations.
If your data have been processed with a version earlier than 10.3
- Solar System Object ephemerides
- A bug (see here) was fixed in the calculation of SSO ephemeris positions (ra_centre/dec_centre) that led to offsets of up to ~10". This bug affected users that use cubes in a co-moving frame (including the standard Level 2.5 cubes), use the doOffset task, or make explicit use of ra_centre / dec_centre in their HIPE scripts. The offset is practically constant over a map and could, e.g., make the emission of a comet appear to be off-center when it really isn't. The fixed bug is in the Level 0 pipeline, which users cannot run easily. The issue was fixed in HIPE 11.1.
- DBS Observations If you have data processed with < HIPE 10
- A new pipeline step, mkDbsReference, calculates the differences in the chop positions in all DBS observations, and after applying the band-pass correction, stores them in a product in calibration->pipeline-out called ReferenceSpectra. This allows you to check for contamination in chop positions for all types of DBS observations.
- Improved pointing reconstruction for observations taken between OD 320 and 761 If you have data processed with < HIPE 9
- The pointing information attached to the data for observations taken between OD 320 and OD761 did not use the most accurate representation of the star tracker focal length. This was done for the bulk reprocessing with HIPE 9. As a consequence, some observations will experience a shift in astrometry, that can be as high as 8 arcsec. Details about the consequences for a particular obsid, and recipes to reconstruct the improved pointing yourself can be found at http://herschel.esac.esa.int/twiki/bin/view/Public/HowToUseImprovedPointingProducts as well as in the Level 0 section of the HIFI Data Reduction Guide.
- Backfilling of observational parameters If you have data processed with < HIPE 9
- From 9.1 onwards, most of the observational parameters that got optimised in HSpot will be propagated into the observation context, under a new product called HifiUplinkProduct (in the auxiliary product branch). Some of these parameters will be used in the pipeline to e.g. estimate the dimension of the cubes in the mapping and have them more representative of how the map was really obtained.
- Solar System Object maps If you have data processed with < HIPE 9
- Cubes for moving targets are now created in the comoving frames in HIPE 9. In order to benefit from that, you should re-pipeline from Level 0 up to Level 2.5. Note that the hifiPipeline task should be called with the option "Aux=True". Please check the Pipeline chapter of the HIFI Data Reduction Guide for more details about how to do that.
- See the What's New page for more details of improvements in HIPE 14.0
Documentation and Cookbooks
- The HIFI Launch Pad is intended to help you quickly off the ground with HIFI data reduction
- Cookbooks now exist for each HIFI observing mode
- The HIFI Data Reduction Guide and the HIFI AOT Observing Mode Release and Performance Notes contain information you need to know about HIFI data, regardless of the software you use for data reduction
- A series of useful scripts can be found following the HIPE menu Scripts -> HIFI Useful Scripts
- If you want to know more about the details of the HIFI pipeline, you should look at the HIFI Pipeline Specification Document
- If you are working in HIPE, the HIFI Data Reduction Guide will be your prime resource for all things HIFI but you should also look at the Herschel Data Analysis Guide for information about general tools, such as those for viewing and manipulating spectra and spectral cubes, and for obtaining data from the archive.
- The full set of online documentation for the current user release also contains detailed information about the HIFI pipeline, information about scripting in HIPE, and reference manuals for command listings. New users of HIPE are recommended to read through the Quick Start Guide and the HIPE Owner's Guide
- The full set of Herschel documentation for the developer track is also available. This documentation contains revisions and updates to the documentation associated with the current User Release but you should also be aware that it may describe functionality not available in the current User Release
- The General HCSS Public Twiki page provides general framework information and updates:
Typical Data Reduction Workflow
- Download data from the HSA. If you are downloading several observations it is better to use the tar ball.
- Inspect data quality. Both point spectra and spectral cubes can be viewed in the SpectrumExplorer. In addition to inspecting the quality of your Level 2 data, you should always check the Level 1 data for any bad scans. If data at Level 2 appears strange then looking at Level 1 data may provide insight to the problem, see the Data Primer chapter of the HIFI Data Reduction Guide for more information about levels of data
- Assess whether data should be re-pipelined.
- Look at the section above and also at the What's New in the latest User Release to decide if you should re-pipeline.
- Note that the ICC recommends that data processing should not be allowed to fall more than one version behind the HIPE version being used.
- Everything you need to re-pipeline data is available in the ObservationContext, and the instructions for using the pipeline, including tips on re-pipelining using new calibration data and customising the pipeline, can be found in the Pipeline chapter of the HIFI Data Reduction Guide
- Flag data. A user-friendly task has been developed in order to easily flag your data: flagTool. It can be used both as a GUI (similar concept as for the fitBaseline task) or in command line
- Remove standing waves. Standing waves are a common problem in HIFI data, particularly in bands 3, 4, 6, and 7. The HIFI Data Reduction Guide provides some information about typical standing waves in HIFI data and describes how to remove them in HIPE using fitHifiFringe and the modified band pass technique
- Correct baseline drift. Baselines can be flattened by subtracted, or division in the case of real continuum. This can be done in HIPE with the fitBaseline task
- Fold frequency switch data. Frequency switch data is not folded in the pipeline and can be done with the doFold task.
- Average together H and V polarisation.
- The noise estimates given by HSpot assume that the H and V polarisations are averaged together. This can be done in HIPE with the PolarPair task, which resamples the spectra to the same frequency scale (they are frequency calibrated with different comb measurements) and then averages them. Alternatively, you can use the Accumulate task, which allows you to specify the resampling width if you desire.
- Note that differences may be seen in H and V profiles, see the note on H and V positions. If you are particularly interested in the spatial structure of your source you may prefer not to average the H and V polarisations together.
- You can also average maps taken with the H and V polarisations:
- Spectral manipulation. The Spectral Toolbox in HIPE allows you to perform typical spectral arithmetic tasks on spectra and spectral cubes
- Fit lines. The SpectrumFitterGUI, which is also a part of the Spectral Toolbox, can be used to fit line profiles.
- Save modified data You can save data to a pool, or store your session in order to continue working in HIPE.
- You may wish to export data as FITS, ASCII or use the hiClass task to export data as CLASS readable FITs files.
- Saving as FITS
- Saving as ASCII
- HiClass task - Note, however, that GILDAS/Class can now directly read the HCSS-generated FITS files (as of release October 2015) and so it is no longer required to convert the HCSS FITS files into a dedicated format. More details about this new FITS reader can be found in this report.
- Line identification. IdentifyLines and exportLines are two new tasks that allow you to identify and export lines in your spectrum, and then run a comparison of known lines with a Linelist. To learn how to use the tasks identifyLines and exportLines, you will need the following three files, and we direct you to HIFI Data Reduction Guide for further details on how to run the tasks:
Dedicated data reduction tips for Spectral Scan data
- The data needs to be deconvolved to a single sideband solution. This can be done with the doDeconvolution task, which works with an ObservationContext.
- It is important to clean the data before deconvolving it: this means removing standing waves, correcting baselines, and flagging out any spurs or bad data that the pipeline missed. The calibration tree for HIPE 14 provides optimised mask tables for each spectral scan obsids so it is recommended that your make use of those, typically be reprocessing your data with HIPE 14 and this calibration file (see above).
- If you are planning to deconvolve your data with CLASS, please contact the Helpdesk, and cc to Claudia Comito (email:firstname.lastname@example.org) if you need further details.
Dedicated data reduction tips for Mapping observations
- Spectral cubes from all HIFI mapping observations are produced as part of the SPG pipeline (that is, the pipeline run for the HSA) and are now found in the Level 2.5 product. However, baseline and standing wave corrections are not done automatically in the pipeline because of the risk of harming the scientific content of the data. This can now be customised in the Level 2.5 step and be done prior to the cube creation.
- It is strongly recommended that you inspect your Level 2 HTPs for baseline drift and residual standing waves, and decide if some, or all, datasets need to be cleaned up before re-running the gridding task.
- Regridding in HIPE > 9 will automatically use the comoving frame for SSOs. Considering the section above, mapping observations of SSOs will benefit from reprocessing in HIPE > 11.1.
- You can then grid your cleaned data into a spectral cube using the doGridding task, you can also use this task to customise the cube creation to best suit your science goals.
Dedicated data reduction tips for Strong continuum observations
- Strong continuum sources
- An optional step in the pipeline can be used to remove standing waves arising in the loads, this is particularly effective for strong continuum sources. A report describing the technique can be found here. Instructions for using this modified passband technique are available in the Standing Wave Removal chapter of the HIFI Data Reduction Guide:
HIFI performance and calibration
- General: In-orbit performance of Herschel-HIFI, Roelfsema et al. 2012
- Intensity calibration: Intensity Calibration Framework document
- Sideband ratio
- Sideband ratio technical note: This note describes the general calibration problem associated with the sideband gain imbalance existing in double-sideband heterodyne systems such as HIFI, and provides some recipes and calibration tables applicable to the particular case of the HIFI mixers.
- http://arxiv.org/abs/1404.2806: "The effect of sideband ratio on line intensity for Herschel/HIFI", Experimental Astronomy.
- After an extensive investigation, the HIFI ICC has introduced updated sideband gain ratio in the calibration tree. A final documentation is still in preparation but the overview of the sideband gain ratio frequency dependence in each band is illustrated in this plot.
- Beam calibration (see also HIFI beam sub-section thereafter)
- Spectral calibration
- This FTP repository of system temperature plots across the IF for each HIFI subband (updated 24 Jan 2011) inform about the variation in system temperature (sensitivity) across the HIFI intermediate frequency band at various frequencies.
Calibration error budget
Line Intensity Calibration Accuracy
The following table provides the percentage flux error associated with each component of the error budget. These numbers are updates from the Roelfsema et al. 2012 paper and reflect what was implemented in the calibration files of HIPE 14. The detailed frequency dependence of this uncertainty budget is shown in this plot
*Note that the "sideband gain ratio" convention used in the HIFI pipeline does not correspond to the ratio between the respective sideband gains (equal to 1 for a balanced system), but to the gain in one sideband normalised by the combined gain of the two sidebands. Consequently, perfectly-balanced sidebands will feature a sideband ratio of 0.5
| Error source
|| Bands 1/2
|| Bands 3/4
|| Band 5
|| Bands 6/7
| Sideband ratio*
| Hot load coupling
| Cold load coupling
| Hot load temperature
| Cold load temperature
| Opt. standing waves: loads
| Opt. standing waves: diplexer
| Planetary model error
| Beam/aperture efficiency
| Flux loss due to pointing**
(see also section 5.3.1 of the Observers' Manual
, and the Intensity Calibration Framework document
**For a point source (worst case scenario) - this flux loss estimates takes into account an APE of 1.2", a residual SIAM bias of order 1.5" and the fact that the H/V respective apertures are slightly offset from the middle synthetic aperture used to point the telescope (offset ranging between 3" in band 1 and 0.5" in band 7)
Frequency Calibration Accuracy
- HRS accuracy: Master Oscillator accuracy (1 part in 10^8)
- HRS resolution: LO signal frequency profile, as determined by the LO Source Unit mixers; band-dependent
- WBS accuracy: COMB algorithm and interpolation interval
- WBS resolution: optical alignment & thermal expansion effects within spectrometer
HIFI beam coupling efficiencies
A revision of the HIFI beam analysis has been performed, based on a multi-epoch set of measurements on Mars. As a result, a more accurate description of the HIFI beams has been obtained, leading to some change in the beam coupling efficiencies, typically a drop. The details of the analysis and the revised coupling efficiencies are given in this release note
. For the record, the first release of the HIFI beam efficiencies was described in this document
(17 Nov 2010). Please note that these updates will only be featured in the HIFI calibration tree from HIPE 13 onwards
HIFI beam maps and Encircled Energy Fraction
In addition to this, a detailed description of the HIFI beam models is provided as 2-D beam maps, 1-D azimuthal beam averages, and Encircled Energy Fractions (EEF). Each are provided at two frequencies per mixer band and polarisation, and can be extrapolated (within the LO frequencies of a given band) as a function of wavelength. The release notes
provides further details about the delivered files and how to use them. In particular, some useful functions are provided in this Jython script
The beam material is concatenated as a single FITS file per mixer, frequency, and polarisation. Each file has three extensions tables: one for the 2-D beam model maps (2705x2705 pixels), one for the EEF, and one for the 1-D azimuthal beam averages. Note that 2-D beam model of smaller size (201x201 pixels) will also be made available in the calibration tree (not before HIPE 13), and used in the generation of Level 2.5 cubes by the pipeline. The following page
provides links to each individual file (27 Mb each). Plots of the 1-D azimuthal beam averages, and EEF for each band can be found in the Appendix B
of the release note.
Access to the calibration tree:
- Information about what calibration is found in the HIFI ObservationContext is found in the Data Primer of the HIFI Data Reduction Guide.
- Information about how to find what calibration version was used on your data, how to get the latest calibration, and reprocess observations with it is found in the Pipeline chapter of the HIFI Data reduction Guide.
- The latest calibration tree pool can also be retrieved directly here (1 April 2016).
- Updates to the HIFI calibration data are generally concurrent with the release of each major version of the HCSS-HIFI software. However, it is possible to have updates to the calibration data in between major releases of the software as the software and data are independent of each other, in the table below the HIPE or the OD from which the calibration updates apply are given. The calibration versions available since HIPE 5 are listed in the table below:
| Calibration version number
|| Release date
|| HIPE version
| IA_CAL_USER_ or HIFI_CAL_
|| Beam efficiency parameters introduced, updates to spur table
|| Beam efficiency parameters updated
|| Spur table update
|| Smoothing widths of OFF positions updated
|| Sideband ratios in band 2a, addition of strong spur at 1108 GHz in band 5a to spur table
|| Prevention of unnecessary Quality Flags by correction and addition of units in calibration products and correction of thresholds, addition of Quality Flags when LO multiplier currents are out of limits, improved saturated pixel flagging, addition to spur table, uplink product
|| Update of a priori table of IF saturations, used in order to flag bad LO data to be discarded in the deconvolution
|| Sideband ratios in bands 5a and 5b, update of a number of engineering threshold for more accurate quality flagging related to hardware housekeeping, introduction of a list of a priori known corrupted data-frames for dedicated flagging in Level 0 data
|| Introduction of a new HifiUplink product to back-fill most the observational parameters as of HSpot optimisation back into the Uplink product, update of the list of corrupted data-frames for flagging at Level 0, update of a number of engineering threshold for more accurate quality flagging related to hardware housekeeping
|| Update of the list of corrupted data-frames for flagging at Level 0, Quality flags created for SEUs (Singe Event Upsets leading to on-board software corruption)
|| Update of the list of corrupted data-frames for flagging at Level 0, updates to quality flag meta-data and uplink information for mapping modes
|| Update of the list of corrupted data-frames for flagging at Level 0, updates to spur table to account for band 3B purification
|| Update of the list of corrupted data-frames for flagging at Level 0
|| Update of the list of corrupted data-frames for flagging at Level 0, ingestion of the Electrical Standing Wave correction reference table
|| Update of the list of corrupted data-frames for flagging at Level 0, Absolute Pointing Error (APE) added to allow a quality flag to be raised in future releases in case of large pointing offsets, updates to allow IF-dependent corrections to sideband ratio, update to spur tables to store entry of spurs found by the ICC or a Key Programme, addition of an instrument uncertainty table (currently only placeholder)
|| Update of the beam parameters (coupling efficiencies, HPBW, 2-D beam maps - see also the beam section above), addition of a new table of expected spectral spurs to be masked by the pipeline in Spectral Scans, addition of a database of Electrical Standing Wave correction solution for observations in bands 6 and 7, addition of a new Uncertainty table that will be propagated into an error budget in HIPE 14 (currently only placeholder), introduction of IF-dependent sideband gain ratio (currently only true for the lower end of band 2a), update of the list of corrupted data-frames for flagging at Level 0
|| Major update of all sideband ratio tables. Introduction of uncertainty budget tables. Introduction of spur warning tables for all point and mapping observations. Update of some Electrical Standing Wave correction solutions. Added some spur flag tables for a handful of spectral scans, as well as some remaining corrupted data-frame entries for flagging at level 0. Updated smoothing width for load chop processing in band 6a, allowing to mitigate the affect of strong artefacts present at tunings between 1457 and 1459 GHz.
|| Minor update to table of BBIDs in order to allow processing of a handful of calibration observations.
Outstanding calibration issues
- Standing waves:
- there are several standing waves that can affect the HIFI data at various processing levels. Their nature and impact on the HIFI calibration are described in the Standing Wave technical note and, more briefly, in the Standing Wave removal chapter of the HIFI Data Reduction Guide
- There are various techniques currently offered to clean these baseline distortions:
- Remove sine waves, or combinations of sine waves. Check the bullet about usage of the fitHifiFringe task in the data reduction section above.
- Electrical Standing Wave (aka Matching technique): to correct for the IF standing wave present in bands 6 and 7
- A new task hebCorrection has been implemented in HIPE 12.0 and is being used by default by the pipeline in HIPE 13 (see also the Reducing HIFI data section above). We warn that the task does not guarantee an accurate correction in all circumstances .
- The task is described in the following document. The mathematical background can be found in this proceeding. Unlike the script made available in previous version of HIPE, this task is applicable to any HIFI observing mode.
- The task needs to work from Level 1 data, then reprocess the corrected data up to Level 2.5. Two example codes are provided here:
- A thorough description of the technique, and the underlying causes, is in Ronan Higgins' thesis: Advanced optical calibration of the Herschel HIFI heterodyne spectrometer
- in case of strong source continuum some standing waves can be enhanced, as described in the Alternative Calibration Scheme report. Such standing waves can be significantly reduced in amplitude using an alternative pipeline algorithm known as the Modified Passband Technique , which is described in the Standing Wave Removal chapter of the HIFI Data Reduction Guide. Alternatively, you can run the Level1PipelineAlgo_hc_filtered_v1.py script.
- Baseline removal: imperfect ON-OFF calibration scheme can result in residual baseline distortion (not necessarily standing waves). Check the bullet about usage of the fitBaseline task in the data reduction section above.
- Spectral purity: there are places over the HIFI frequency range where the Local Oscillator does not offer a single frequency tone, so that spurious spectral line may end up in the data, and the calibration of the targeted line gets affected. A dedicated cookbook on the matter is in preparation. In the meantime, we refer to section 5.4.6 of the Observer's Manual, and to the release notes provided above.
- Spurs: Spectral Spurs are narrow and broad spectral features due to instabilities in the Local Oscillator. Up to HIPE 12, spurs have been assigned by the pipeline in an automatic fashion. In HIPE 13, this is no longer the case and the masks are assigned based on a priori tables provided by the ICC in the calibration tree. As of now, all spectral scans benefit from such tables, but only a handful of single point and spectral mapping observations. This limitation has little impact on the top level data though since no pipeline module makes use of these masks in those modes. A list of the spurs registered up to HIPE 12 can be found in section 5.4.6 of the Observer's Manual. This table will be updated soon based on the new statistics derived from the HIPE 13 mask tables.
Interest Groups and Scripts
- HIPI: a library of HIFI Plug-In's has been prepared by the NHSC - check the HIPI website for more information. From HIPE 13 onwards, most of those utilities have been ported to Useful scripts (Menu Scripts > HIFI Useful Scripts)
- The following interest groups relate to processing of observations taken with HIFI. The links provided allow subscription to these interest groups.
- subscribe to PACS, SPIRE and HIFI spectral maps interest group
- subscribe to HIFI point source and spectral scan interest group
- User scripts: Users are welcome to submit scripts they believe could be of general interest to the community to the Herschel Helpdesk.
- 11 April 2016 -