LISA Data Issues Taskforce Report 2005 December 10 Pete Bender Neil Cornish Lee Samuel Finn Guido Mu

1
LISA Data Issues Taskforce Report2005 December
10Pete BenderNeil CornishLee Samuel
FinnGuido MuellerMarcello SallustiBonny
SchumakerJean-Yves Vinet

• Presented to the LIST by
• Bonny Schumaker
• NASA LISA Deputy Mission Scientist
• Bonny.Schumaker_at_jpl.nasa.gov

2
Charter of the Data Issues Task Force

• Created at the Jul 2005 LIST meeting in Bern,
  Switzerland
• Objective (excerpted)
• Achieve in a short timeframe a better
  understanding of the most important data
  availability requirements driving the design of
  both the LISA instrument (S/C,P/L, and ground
  infrastructure) and the data analysis systems
  design.
• Charter (excerpted)
• A. Identify sciencekeeping data needed to
  support processing and analysis of the science
  data. Items to be clarified include
• Estimation of instrument noise
• Assessment of data quality and construction of
  vetoes
• Calibration modes
• B. Identify procedures and parameters needed for
  near-real-time assessment of data quality,
  constellation health, and science performance
• Onboard real-time critical parameters
• Downlinked parameters
• C. Identify requirements on data gaps,
  disturbances, degradations, and protected
  observing periods

3
Contents of Data Issues Report

• 1 Introduction and Objectives
• 2 Sciencekeeping data
• 2a Data characteristics
• 2b Instrument noise characterization
• 2b.1 Four-link configurations (two-
  or
• three-arm)
• 2b.2 Five-link configurations
• 2b.3 Baseline six-link
  configuration
• 2c Primary sciencekeeping data
• 2d Secondary sciencekeeping data
• 2e Sciencekeeping data rates

3 Data interruptions 3a Data disturbances 3b
Data degradations 3c Data gaps 3d Protected
observing periods 4 References
4
1. Introduction and Objectives

• Clarify critical requirements on LISA data and
  data-taking imposed by needs of science data
  analysis
• 1 Sciencekeeping data
• Nature and amount needed to support processing
  and analysis of LISA science data
• Emphasis on characterizing instrument noise,
  monitoring instrument science performance, and
  assessing data quality within 2 to 3 days of
  real-time
• Consider only data that directly impacts science
  data analysis
• Laser received light intensity, beam steering,
  freq intensity noise PM charge
  SC temp variations, some station-keeping and
  attitude control etc.
• No housekeeping no SC power generation, comm
  systems, detailed ACS.
• 2 Data interruptions Causes, ramifications,
  design requirements
• Disturbances (caused by disturbance to PMs)
• Degradation (drop in quality caused by mild
  excursion in instrument noise)
• Gap (absence of valid data, but no disturbance or
  degradation)
• Protected observing periods Durations, advance
  notice, frequency, duty cycle, restriction
  against scheduled disruptions

5
2a. Sciencekeeping Data Definitions

• Science data Phase-meter outputs sent to ground
  at 3Hz, from which TDI observables are formed.
• Assume no other science-driven data is downlinked
  at this high a rate.
• N.b. Science data provide the most sensitive
  measure of instrument performance!
• Sciencekeeping (SK) data Auxiliary data to
  monitor instrument health as it affects science
  performance. Includes onboard parameters that
• Affect precision or continuity of phase
  measurements or
• Contribute in-band noise not calibrated from
  science data alone.
• Two major functions of sciencekeeping
• Characterize instrument noise to help calibrate
  science performance
• Assess data quality and inform of changes in
  instrument performance
• Primary SK (PSK) data Downlinked at much slower
  rate than science data
• PM charges, temperatures (critical locations),
  laser intensity fluctuations, outputs of critical
  servo loops (e.g., laser phase-lock loops)
• Secondary SK (SSK) data Used onboard to monitor
  instrument health and raise flags. Downlinked as
  requested (or in some cases automatically).
• Not routinely necessary for science data
  processing or analysis, under normal operations
• E.g., error signal from laser pre-stabilization
  reference cavity, spacecraft DRS parameters
  (temperatures, voltages, etc.).
• Stored as required by sources e.g., months to
  accommodate EMRI data-filtering needs.

6
2a. (cont) Sciencekeeping Data Latency

• We specify whether SK data require formation of
  TDI observables (may contribute to critical
  design parameters, such as latency). Ex
• Monitoring relative motion between local optical
  benches, or motion of SC (TDI)
• Emergency flags, e.g., drift from sweet spot
  of phase front (no TDI)
• Latency Elapsed time between onboard
  acquisition of data, and availability of data
  product suitable for analysis for GW sources.
  (Includes downlink, processing to form TDI
  observables, data validation, noise
  characterization, etc.)
• Long latency affects both operations and science
  analysis
• Delay between onset of problem and its
  recognition and diagnosis means potential loss or
  degradation of science data
• Loss of ability to warn and point other detectors
  toward imminent events
• ? Recommend careful review of acceptable latency
  based on science analysis to meet science
  requirements, and trade studies among
  implementation approaches.

7
2b. Instrument noise characterization -1

• Must distinguish source signals from instrument
  effects and source background
• Link combined phasemeter outputs from 2 laser
  heterodyne measurements local distant beams
  and local local beams (to probe local PM
  motion).
• Expected approaches for worst to best scenarios
  4, 5, or 6 links operational
• Four links (worst-case) Only one TDI observable
  (X,B,M, or R)
• No simultaneous GW polarizations constellation
  motion only (for long-lived sources)
• Remove thousands of resolved Galactic binaries
  5mHz possible by virtue of high SNRs (few
  weeks of integration), and low number per
  frequency bin
• Solve for and remove MBH binaries ? Initial
  estimates of instrument noise levels spectra
• Use this info to estimate unresolved WDB
  backround
• Iteratively compare data with source catalog to
  remove estimated known sources and check for
  stationarity in spectrum of the residuals
• Could identify external or DRS disturbances (by
  comparing with expected instrument noise)
• Use signals from verification binaries with known
  locations, frequencies, and amplitudes (x2) to
  verify instrument science performance

8
2b. Instrument noise characterization - 2

• Five links Five TDI observables (X, B, M,
  R1,R2)
• Each sees different combinations of instrument
  noise, and of unresolved-source noise (because of
  the source-noise dependence on spatial
  orientation of each link) ?
  Use to identify components producing excess
  instrument noise, and to weight different
  observables optimally for particular sources
• N.b. Science data alone (even six links) cannot
  localize all instrument noise.
• For ex, cannot tell which of 2 PMs (in one arm)
  has excess accel noise or with only one link,
  may not identify whether photodetector has excess
  noise.
• Additional SK data could identify such problems
  unambiguously e.g., more sensitive DRS
  readouts, light-level monitors, even local
  bench-bench picometer IFOs to measure
  differential PM motion (Guido Mueller 2005)

9
2b. Instrument noise characterization - 3

• All six links All TDI observables, plus ? and
  ZSS (Armstrong, Estrabrook, Tinto)
• Powerful way to characterize low- and medium-freq
  instrument noise
• Symmetrized Sagnac ?, insensitive to GWs MhZ
• Compare with related T, whose noise transfer
  fcn differs slightly
• High-frequency instrument noise characterization
• After subtraction of strong resolvable sources,
  compare the different TDI channels, which reflect
  noise from different optical elements on
  different SC.
• Even without source subtraction, can isolate some
  noise sources by tracking differing Doppler
  records of the different TDI channels (John
  Armstrong 2005)
• Discriminate spatially-localized sources from
  instrument noise across full band
• Zero-signal solution (ZSS) TDI combination
  cancels GWs from two opposite directions in sky
• Good for spatial regions increasingly smaller at
  higher frequencies
• Improvement over ? at low frequencies (for
  localized sources)
• Above 20mHz, cannot distinguish instrument noise
  from unmodeled isotropic sources, because there
  is only one LISA! (cf. LIGOs cross-correlated
  outputs).
• HW-injected signals probably cannot improve on
  estimates using all TDI observables. (TBD
  whether it can help in the 4- or 5-link
  configurations.)

10
2c. Primary sciencekeeping data data rates

• Inter-SC ranges Either from inter-SC (science)
  links or TDI-R.
• 10m, low-freq variations. ? 0.5 bps
• Interferometer parameters
• Laser intensity noise 15 bands in 0.00110Hz,
  every 1000 sec ? 5 bps
• Phase-meter calibration (beat-signal amplitudes)
  comes from science data. Averaged over 1000
  sec. If needed, ? 5 bps.
• Temperatures mean temps on benches (variations
  too small) (see below)
• PM and DRS parameters
• PM charge a few bits/PM every 100 sec
  (electrical torque around sensitive axis or
  actuate along non-sensitive axes). Takes 10 hrs
  to measure charge. ? 2bps
• Environmental data local and interplanetary (too
  weak 3nT) magnetic field, solar wind and
  energetic particles. TBD whether possible, but
  low BW. ?
• Temperatures 20 sensors around the SC (thermal
  distortions drive gravl perturbations to PMs).
  Easily measured variations, every 1000 s ? 5 bps
• PM actuation along non-sensitive axes 5dofs/PM,
  readout every 300 sec? 10 bps
• Thruster actuation signals If thrust levels
  read every 30 sec ? 30 bps
• Total Primary Sciencekeeping data rate bps/SC (10-IFO, 50-PM/DRS)
• N.b. Baseline science data rate 400 bps/SC. ?
  SK wont affect downlink design.

11
2d. Secondary sciencekeeping data data rates

• Interferometer parameters
• Laser freq pre-stabilization cavity transmitted
  reflected laser powers, cos and sin components
  of demodulated signal ? 300 bps
• Laser polarization sensor 10 bps
• Spatial modes (out of backside fibers on optical
  benches) 10 bps
• GRS parameters (other than DRS housekeeping data)
• Power-supply voltages (1-mHz BW) 30 bps
• Antenna orientations 0.1 bps
• SC power systems 10 bps
• Transfer functions between objects in measurement
  train (e.g., actuators and other PSK inputs) and
  TDI observables
• Measured (from TDI observables)and minimized
  during commissioning, then monitored and
  recalibrated due to long-term drifts, possibly
  via HW injections
• Electro-optic Laser intensity noise, freq
  stabilization modulation, phase-lock and
  arm-locking control loops
• Electro-mechanical All mecanical actuators PM
  (linear angular) , telescope look-behind,
  high-freq response of thrusters
• Total Secondary Sciencekeeping data rate bps/SC (320-IFO, 40- GRS)

12
3. Data interruptions

• Data Disturbance Interruption caused by a
  temporally isolated acceleration, velocity, or
  displacement offset to one or more PMs
• Data Degradation Drop in data quality caused by
  mild excursion in
• instrument noise
• Data Gap Continuous time interval in which
  valid data are not available, but no
  disturbances or degradations are present
• N.b. All data interruptions reduce signal power
  and must be accounted for in meeting the SRD
  stipulation for a duty cycle ? 80
• (? fraction of total observing time during
  which valid data are available).

13
3a. Data Disturbances

• LISA will suffer scheduled and unscheduled PM
  disturbances Antenna rotations, telescope
  articulation, etc. (if done episodically) solar
  flares.
• Large unscheduled ones may require solving for PM
  offsets, not just downweighting selected data.
• They appear in the phase record as offsets or
  linear or quadratic ramps easily confused with
  low-frequency (2004)
• Must solve for at least 3 variables after each
  disturbance (pos, vel, accel) (more if
  unscheduled and start time unknown).
• Recovery time shorter for higher frequencies
  (1 period to distinguish between sinusoid of GW
  Fourier component and ramp of a disturbance)
• Ultimately, fit previous disturbances and use
  most data coherently.
• Might compromise identification of low-frequency
  events needing prompt (
• Science duty cycle is frequency-dependent, driven
  by time distribution of disturbances
• Low-frequency science may require longer
  disturbance-free intervals.
• Constrain mean disturbance rate e.g., such that
  low-freq data loss ? ? 75
• 2TDf data points in time T, BW Df , 2 channels ?
  2/day for Df 0.01mHz .
• For max of 25 loss in science data, aim for
  

14
3b. Data Degradations

• Data valid but of lesser quality, caused by noise
  excursion in detector components or control
  systems.
• Need careful study of how degraded data can be
  used in science data analysis.
• Effect depends on timescales of excursions
• short (
• long ( 1/f) treat as stationary noise with
  higher rms value
• intermediate or combination (both short and
  long, for different GW signal contribution) ?
  Needs further study
• Equivalent loss in duty cycle due to solar
  energetic particle events could reach -5 in some
  years.
• ? Estimate a 5 loss in duty cycle due to data
  degradation (of the allowable 20 loss for an
  overall 80 duty cycle)

15
3c. Data Gaps

• Gaps involve no disturbance to PM inertial
  trajectories
• They may cause algorithmic inconvenience for
  continuous sources, but bandwidth remains 2/?t,
  where ?t is smallest gap.
• ? Primary consequence is loss of cumulative
  signal power, i.e., decreased effective duty
  cycle.
• This assumes that data segments are coherently
  re-connected, e.g., using a verification binary
  signal.
• Possible using well-known data processing methods
  for dealing with gap-containing data (not
  discrete Fourier transforms, which led to
  previously concluded restrictions).

16
3d. Protected Observing Periods (POPs)

• Timing of data disturbances or gaps is not
  critical for quasi-continuous sources (Galactic
  binaries or EMRIs years before plunge), but is
  critical for burst-like events, such as SMBH
  binary mergers where SNR increases steeply in the
  final days or weeks and the systems become fully
  relativistic, or the interesting final zoom and
  whirl of EMRIs.
• ? Protected Observing Periods must be planned.
• Possible to give 2 wk notice for sources out to
  Z5, Mchirp
• SRD specifies 4-day period w/ 2-wk notice.
  
• Propose
• 10-20 day protected period in order to see last
  100 or more cycles (lower mass systems would
  need proportionally shorter periods).
• Specified at least 1 wk in advance
• No more frequently than once every 45 days.
• ALSO
• During a protected observing period, duty cycle
  should be 90
• The last 3 days (3e5 sec) of a POP should be
  free of any scheduled disruption (6days?)

17
Tasks not yet addressed (mentioned in Jul 2005)

• Instrument noise how estimated, requirements
  to be imposed on it (e.g., stationarity)
• Specific tests during commissioning
• Data architecture tasks quality assurance,
  testing, synthesizing data, estimates of
  computing resources, data products and
  communication pipelines, data acquisition,
  operations
• Onboard data storage desire for and practical
  limits on how much, how often, and how fast for
  downlink/uplink
• Quantitative effects of and methods to cope with
  loss of links or other performance degradation