Difference between revisions of "Timelines"

From Planck Legacy Archive Wiki
Jump to: navigation, search
Line 60: Line 60:
  
 
=== LFI processing ===
 
=== LFI processing ===
As in the 2013 release, the processing of LFI data is divided into three levels. The main changes compared to the earlier release are related to the way in which we take into account the beam information in the pipeline processing, as well as an entire overhaul of the iterative algorithm used to calibrate the raw data. According to the LFI scheme, processing starts at Level 1, which retrieves all the necessary information from data packets and auxiliary data received from the Mission Operation Centre, and transforms the scientific packets and housekeeping data into a form manageable by Level 2. Level 2 uses scientific and housekeeping information
+
The processing of LFI data is very similar to the once applied to the previous release. The main changes compared to the earlier release are related to the way in which we take into account the beam information in the pipeline processing, as well as an entire overhaul of the iterative algorithm used to calibrate the raw data.  
to:
+
According to the LFI scheme, the input of the scientific TOI are the data in engineerig units [Volts]:
 +
 
 
– build the LFI reduced instrument model (RIMO), which contains the main characteristics of the instrument;
 
– build the LFI reduced instrument model (RIMO), which contains the main characteristics of the instrument;
 
– remove analogue-to-digital converter (ADC) non-linearities and 1 Hz spikes diode by diode;
 
– remove analogue-to-digital converter (ADC) non-linearities and 1 Hz spikes diode by diode;

Revision as of 21:56, 30 January 2015

General description[edit]

The timelines, or TOIs for Time-Ordered Information, are vectors of signal or of pointing or of some other quantity giving the value of that quantity as a function of time during the mission. The TOIs described here are sampled regularly at the (instrument dependent) detector sampling frequency and span the full science mission. They thus consist of a large amount of data. This section is concerned with the signal TOIs and their associated pointing TOIs. For selected detectors the DPCs provide a single timeline of cleaned and calibrated signal and 3 timelines of coordinates, corresponding to the two angular coordinates and one position angle.

Obviously these vectors are very long (~1.38E10 samples for HFI, from ~2.5E06 to ~5.5E06 for LFI) and thus need to be split into multiple files for export. Here the data are split by Operational Day (OD) as follows:

  • HFI: 12 files per OD, one for pointing and one for signal for each channel, covering ODs 91 -- 974, all sampled at Fsamp = 180.3737 Hz.
  • LFI: 6 files per OD, 3 each for signal and pointing, since the sampling frequency is channel dependent, and covering ODs 91 -- 1543, sampled at Fsamp = 32.5079, Fsamp = 46.5455 and Fsamp = 78.7692 Hz.

The signal timelines are encoded as Real*4, but the pointing vectors had to be encoded as Real*8 to maintain the required accuracy. The result is that the total volume of the full dataset is ~30 TB. All files also contain the OBT.

The signal timelines have been cleaned of all known instrumental systematic effects, they have been calibrated in flux, and the solar and earth motion dipole signals have been removed. But they have not been cleaned of the low-frequency, 1/f noise that needs to be removed via a destriping tool. The methods consists of removing offsets or baselines determined by minimizing the differences in the signals at points on the sky where they intersect. For HFI an offset per ring is determined; for LFI the the baseline is computed every 0.246, 0.988 and 1.000 second for the 30, 44 and 70 GHz respectively and maintain the same structure of the signal timelines. In case of HFI these offsets are determined using the full mission and all the valid detectors per channel, and they are then used for all the maps produced, i.e., for those maps using any fraction of the mission (year, survey) or any subset of the detectors (single detector, detector set), and it was shown that using offsets determined from a limited part of the mission and/or a subset of the detectors yields maps that are less consistent than otherwise.

In case of LFI these offsets are determined using the full mission and all the valid detectors per channel, those values has been used for the production of the full mission period maps. Note that baseline used for shorter period maps are determined on those data period to avoid noise cross-correlation effect and those are not delivered.

The offsets are delivered separately, as described below.

Maybe insert a figure for example?

Furthermore, for HFI and LFI, there are three sets of offsets produced: the primary set using the full rings, and the secondary ones using the first and second half rings, only. The difference between the primary and the secondary sets are fairly minor, but they are necessary to rebuild the maps as they were built by the HFI-DPC and LFI-DPC. Overall the offsets are useful for users wishing to build a map of a small area of the sky.

The HFI delivers its offsets in a ROI, or "Ring-Ordered Information" file. That file contains a table of Nrings = 26766 rows by Nbolometers in which each cell contains a 3-elements vector containing the full-ring, first half-ring, and second half-ring offsets. The indices of the signal TOI delimiting the rings are given in a second ROI file with the same Nrings rows containing global parameters.

The LFI delivers its offsets in a TOI format, the structure is EXACTLY the same used by the Science Timelines. The user can simply subtract one-by-one the offset timelines from the Science Timelines and then generate a map with the result. In case of the half-ring baseline, a vector has been added in the OBT extension; this vector contains 1 or 2 depending to which half ring should be applied.

HFI Indexing[edit]

Each FITS file contains keywords giving the first and last index of the section of the full TOI contained in the file. These indices are those of the internal DPC representation; they begin at 0 corresponding to the switch-on of the instrument, and run to some very large number (~25 billion for HFI) representing the full mission. Only the science part of the mission is exported, which is about 60% of the total for HFI. Are left out the early mission phase (cool-down and validation phases) and the warm phase of HFI, during which LFI continued to collect data. The HFI keywords also indicated which rings are included in each OD, but be aware that normally an OD does not begin at the same time (with the same index) as a ring.

Note that for historical reasons, the OD definitions of the two DPCs differ: for LFI they occur at the transition between pointing periods, whereas for HFI they do not. This has however no significant importance for the user, as this splitting is somewhat arbitrary anyway, and what counts is the full vector once it is rebuilt in the user's own work space.

HFI timelines at the DPC are indexed from 0 to ~25E9, that correspond to instrument switch-on to switch-off. Of these the indices ~1.4E9 to 151.5E9 correspond to the science mission and are exported and delivered. Each file contains a keyword giving the first and last index of the data in that file, and EndIndex(OD)+1 = BeginIndex (OD+1). The ROI Global file gives the Begin/End Indices of each Ring, or Pointing Period, and can be used to destripe the signal TOIs with the offsets provided in the Destriping-Offsets ROI file.

Production process[edit]

HFI processing[edit]

The input TOIs are in ADUs representing the voltage signal at the end of the electronics (see Detection chain for details). The processing applied to remove instrumental effects and to calibrate them is described in detail in the HFI TOI processing section, and in [Paper A08 link] we give a very brief summary here for convenience. That pipeline performs the following operations:

ADC correction
corrects for the uneven size of the ADC bins (see ADC correction).
demodulation
this is performed around a variable level which is determined from the valid input data (using a validity flag from a previous version of the processing), and the data are converted to engineering units (V) using known conversion coefficients.
despiking
the redundancy within a ring is used to determine where glitches occur. Once identified, the glitches are fitted with templates. A glitch flag is produced that identifies the strongest part of the glitches, and a timeline of glitch tails. produced from the template fits, is subtracted from the demodulated timeline. Finally, the flagged regions are replaced with data from an average over the pointing period in order to not leave holes in the timeline that would perturb the usage of Fourier methods on them
dark template removal
the two dark bolometers are demodulated and despiked as above; the resulting timelines are then smoothed and used as an indicator of the overall temperature variations of the bolometer plate. Where the variations are consistent with each other, they are combined and removed from the bolometer signal timelines using appropriate coupling coefficients. The few percent of the data where they are not consistent are flagged as not valid on the timelines.
conversion to absorbed power
the timeline is converted to watts of absorbed power using the bolometer function. This includes a non-linearity correction; removal of the 4K cooler lines: the electromagnetic interference of the 4K cooler with the bolometer readout wires induces some sharp lines in the signal power spectra at frequencies of the 4K cooler's fundamental and its multiples, folded by the signal modulations. Fourier coefficients of the relevant lines are determined on a per-ring basis, and then removed from the data. The quality of the removal depends on the bolometer.
deconvolution by the time transfer function
this is done to correct for the non-instantaneous time response of the bolometers. The function itself is modeled using up to 8 time constants which are adjusted primarily on the planet data and also from comparisons of the northward and southward scans of the Galactic Plane. It is then removed using Fourier techniques.
jump correction
corrects some pretty rare jumps in the signal baseline (there are on average ~ 0.3 jumps per bolometer per pointing period). The jumps are detected and characterised on smoothed TOIs, and corrected by adding a constant to part of the signal timeline. The origin of the jumps is not known.

The results of this processing are a timeline of signal (in absorbed watts) and a valid data flag timeline for each of the 50 valid bolometers which are processed, out of the 52 HFI bolometers; these timelines contain the full sky signal, i.e., including the solar and orbital dipoles, the Zodiacal light, and contributions from the Far-Side lobes. The dipoles are necessary for the flux calibration and are removed at the mapmaking stage. The remaining two bolometers (143-8 and 545-3) show semi-random jumps in the signal level, typically jumping over 2-5 different pseudo-baseline levels, a behavior known as Random Telegraphic Signal, so that these are commonly called the RTS bolometers. Finally, ring-level statistics of different types (mean, median, rms, kurtosis, etc.) are determined on a per-ring basis for all timelines, and a selection based on these statistics is used to discard anomalous rings, which are recorded in a ring-level flag for each valid bolometer timeline (see the Discarded rings section).

Throughout this processing, some planet signals are masked in the timeline in order to avoid ringing effects in the processing. It concerns only Jupiter at 100 and 143 GHz, Jupiter and Saturn at 217 GHz, Jupiter, Saturn and Mars at 353, 545 and 857 GHz (thanks to various improvements in the despiking and in the transfer functions it is no longer necessary to flag these three planets in the CMB channels as was done in Release 1). Since they move on the sky, the portion of the sky masked during one survey is observed during another one, and no hole is left in the final map. In parallel, the planet data are processed in a similar way, but with different parameters for the despiking step. These results are processed separately to determine the beam shapes and the focal plane geometry.

Next, the TOIs are calibrated in astrophysical units using the results of the calibration pipeline (see the Map-making and calibration section) and the best estimate of the zero-point offsets (a constant level for each bolometer). These values are given in the RIMO. Also, the solar and earth-motion dipole signals are computed and removed.

These TOIs are accompanied by several flags that are described below. The most important one is the Total flag, which identifies all the samples that were discarded by the DPC mapmaking, ad described in HFI TOI processing. This flag includes the portion beyond 72 min for the rings that are longer, which is not used because of the slight drift of the satellite pointing direction (spin axis) during these long acquisition periods.

At this point the TOIs still contain the low frequency (1/f) noise which should be removed before projection onto a map. That cleaning step is called "destriping" "baseline removal". The HFI-DPC does its destriping at ring-level, meaning that a constant is added to the signal of each ring in order to minimise the difference at the ring crossings, where the signal should be the same for all detectors (with the necessary precautions as described in Map-making and calibration section); should the user want to use the DPC's offsets, they are provided separately (see ROI files below). Maps produced from these TOIs, and after subtraction of the DPC's destriping offsets, are not identical to the maps delivered, the differences being of order XXXXX. The reasons for these differences are described in [###???? SECTION]

The pointing (see common Detector Pointing section for details) is determined from the AHF produced by MOC, which gives the direction and orientation of the LOS of a fiducial position in the focal plane at frequencies of 8Hz during stable pointing and 4 Hz during maneuvers. This is interpolated to the timestamps of data samples, corrected for the wobble and other time-dependent offsets determined from the observed positions of a large number of sources around the sky, and finally converted to the LOS of each detector using the quaternions in the IMO.

LFI processing[edit]

The processing of LFI data is very similar to the once applied to the previous release. The main changes compared to the earlier release are related to the way in which we take into account the beam information in the pipeline processing, as well as an entire overhaul of the iterative algorithm used to calibrate the raw data. According to the LFI scheme, the input of the scientific TOI are the data in engineerig units [Volts]:

– build the LFI reduced instrument model (RIMO), which contains the main characteristics of the instrument; – remove analogue-to-digital converter (ADC) non-linearities and 1 Hz spikes diode by diode; – compute and apply the gain modulation factor to minimize 1= f noise; – combine signals from the diodes; – compute the appropriate detector pointing for each sample, based on auxiliary data and beam information corrected by a model (PTCOR) built using solar distance and radiometer electronics box assembly (REBA) temperature information; – calibrate the scientific timelines to physical units (KCMB), fitting the total CMB dipole convolved with the 4� beam representation, without taking into account the signature due to Galactic straylight; – remove the solar and orbital dipole convolved with the 4�pi beam representation and the Galactic emission convolved with the beam sidelobes from the scientific calibrated timeline; LFI This section should be Reviewed


File Names[edit]

The file names are of the form:

{H,L}FI_TOI_{fff}-{SCI,PTG}_R2.nn_ODxxxx.fits

where

  • fff denotes the frequency
  • SCI or PTG denote signal or pointing TOIs
  • R2.nn is the version, and
  • ODxxxx indicates the OD.


At the present time, HFI is providing signal TOIs for all the SWB (non-polarised) bolometers and for the PSB (polarised) bolometers at 353GHz only. The reasons for this are detailed in [###REF###]. On the other hand, the pointing TOIs are available for all bolometers. LFI is providing all signal TOI.

FITS file structure[edit]

All FITS files begin with a primary extension containing no data and a minimal header, which is followed by one or more BINTABLE extensions with data, and with a description of the data in the header keywords. In what follows will usually ignore the primary extension, and count only the extensions containing data.

TOI files[edit]

The signal FITS files contain N ,'BINTABLE', data extensions, where N is the number of detectors in that frequency channel. The first extension contains the OBT and its flags, called the global flags since they apply to all detectors, which is followed by one extension for each detector, containing the signal and its local flag. The flags columns are written as byte in which each of the 8 bits (max) encodes one flag timeline. The meaning of each bit is given in the header comments and is subject to change. For the global flag they include:

  • Unstable pointing: 1= pointing is not stable (e.g., during repointing maneuvers)
  • Dark correlation: 1 = darks are uncorrelated and data are flagged
  • First/Second half ring: which samples are in which half (only covers the stable pointing part of the ring)
  • HCM: in HCM mode (unstable pointing)
  • and more

LFI This section should be Reviewed


And for the local flag they include

  • Total Flag: a combination of the various flags that is the one finally used in the map-making (all samples with Total Flag different from zero should not be used)
  • Data not valid: glitched samples
  • Despike Common: (for PSBs only) glitch on current or other of PSB pair
  • StrongSignal: on Galactic Plane
  • Strong Source: on point source
  • other


TOI file data structure
1. EXTNAME = 'OBT' : Data columns
Column Name Data Type Units Description
OBT Int*8 2-16 sec On-board time
FLAG Byte none the various bit-level flags
Keyword Data Type Value Description
OD Int OD covered (as in filename)
BEGIDX Int first sample index of given OD
ENDIDX Int last sample index of given OD
BEGRING Int first ring in given OD
ENDRING Int last ring in given OD
TIMEZERO String 1958-01-01z00:00 Origin of OBT
n. EXTNAME = DETNAME : Data columns
Column Name Data Type Units Description
SIGNAL Real*4 Kcmb or MJy/sr Value of signal
FLAG Byte none the various bit-level flags
Keyword Data Type Value Description
UNIT String Units of signal
DESTRIPE 1/0 whether timeline is destriped
OD Int OD covered (as in filename)
BEGIDX Int first sample index of given OD
ENDIDX Int last sample index of given OD
BEGRING Int first ring in given OD
ENDRING Int last ring in given OD


The pointing files have a similar structure, except that DETNAME extensions contain 3 columns of Real*8 with the phi, theta, and psi Galactic spherical coordinates of each sample in radians, as shown in the table below. There is no local flag for the coordinates.

Pointing TOI file DETNAME extension structure
n. EXTNAME = DETNAME : Data columns
Column Name Data Type Units Description
PHI Real*8 radian longitude
THETA Real*8 radian colatitude
PSI Real*8 radian roll angle


ROI files[edit]

The files provided by HFI are

  • HFI_ROI_GlobalParams_RelNum_full.fits
  • HFI_ROI_DestripeOffsets_RelNum_full.fits

which are described below.

Global parameters
this ROI file contains a single BINTABLE extension with 3 columns that give, for each ring, the index number of the first sample, the ESA pointing ID of the ring, and the ring start time. All three are encoded as Int*8, and the time is given in units of 2-16 sec from TIMEZERO, 1958-01-01z00:00, as in the TOI files. Rings are numbered from 240 to 27005, as given in the BEGRING and ENDRING keywords
Destriping offsets
this ROI file contains a single BINTABLE extension with 3 columns that give, for each ring, the offsets to subtract from each bolometer signal timeline or vector. There are three offsets available for each bolometer: full-ring, first half-ring, and second half-ring, for building the full-ring or the two half-ring maps, respectively. The offsets must be subtracted from Index(N) to Index(N+1)-1 of the corresponding signal timeline, where N is the ring number, and the indices are given in the Global parameters file. The offsets are given in the same units as the signal vectors, that is Kcmb for the 100-353 GHz channels, and MJy/sr for the 545 and 857 GHz channels. All values are encoded as Real*4. The rings for which the offsets are set to 0.000 are badrings which are rejected in the mapmaking process. These rings are also flagged by the TotalFlag, as are the portions of rings longer than ~72.5 min, where the drift in the satellite's spin axis becomes important.

TBW: Note on half-ring offsets, which differ by 1-2% from full ring offsets

(Planck) High Frequency Instrument

(Planck) Low Frequency Instrument

Operation Day definition is geometric visibility driven as it runs from the start of a DTCP (satellite Acquisition Of Signal) to the start of the next DTCP. Given the different ground stations and spacecraft will takes which station for how long, the OD duration varies but it is basically once a day.

On-Board Time

Data Processing Center

Ring-Ordered Information (DMC group/object)

Flexible Image Transfer Specification

analog to digital converter

sudden change of the baseline level inside a ring

random telegraphic signal

Cosmic Microwave background

reduced IMO

Attitude History File

[ESA's] Mission Operation Center [Darmstadt, Germany]

Line Of Sight

LFI Radiometer Electronics Box Assembly

Angular momentum Control Mode

European Space Agency