Thoughts on noise in LISA What do we learn from LPF? M Hewitson LISA Consortium Meeting Paris 13th October 2015

Transcription

1 Thoughts on noise in LISA What do we learn from LPF? M Hewitson LISA Consortium Meeting Paris 13th October 2015

2 About the noise Noise in LISA has some sources similar to LPF, and some new sources Local instrument/readout noise Link noise Residual forces (environmental noise) Likely to be time varying by how much, and which time-scales, we don t know we may learn something about this from LPF why? temperature, dynamics, life-time of components Where does the noise appear? phase meter outputs, GRS readouts, environmental sensors, noise in pre-processing and TDI application We care about the total noise in TDI variables (?) 2

3 Why do we care about the noise? Do we just need to know the noise is low enough to allow the signals to be recovered? Do we need a rough or an accurate model of the noise (a few power-laws) for data analysis? how accurate? 1%, 10%, factor 2? Over whole band? Do we need to model non-stationarities in DA? Do we need to model noise at TDI outputs? Also Sangac combination? what does that tell us about noise in X? Do we need a noise model at the start, or can we use iterative methods in DA to recover the noise? 3

4 Building a noise model Let s suppose we do need an accurate noise model for matched filtering LISA will be signal dominated How do we measure the noise? How do we do our LPF-like characterisation and system identification? We get a comprehensive physical model of local force noises from LPF use this to project the noise in LISA needs many measurements to assess the key parameters (see later) is this accurate enough? We get some information about readout noise of GRS and IFO might be difficult at lower frequencies where TM motion dominates 4

5 Noise model 5

6 Parameters of the model Noise Sources Parameters OMS sensing electrostatic actuation shot noise, readout electronics, phase meter distortions, frequency noise, etc suspension loops on certain degrees of freedom, voltage stability, cross-talk coefficients temperature, laser power, alignment, etc actuation authority, temperature, component lifetime, alignment and working point of TMs gas damping residual gas around TM temperature, pressure magnetic laser radiation pressure thermal interplanetary, sc fields, gradient fluctuations momentum transfer to TM radiometer, thermal radiation, outgassing TM susceptibility, power switching around SC TM reflectivity, laser power stability at mhz at TM pressure, temperature, temperature stability, etc and so on 6

7 Operation of LISA All those model parameters will need to be measured in flight if we need an accurate model we need to build in the capabilities for this when the instrument is designed How often these need to be remeasured will depend on: each parameter, and how it changes; how accurate the model needs to be; the stability of the underlying noise sources; some we may have sensors for others will require more invasive measurement procedures 7

8 An example: radiation pressure noise Only one beam hits the TM We need to measure the power fluctuations in that beam as near to the TM as possible On LPF we have a fairly poor photodiode readout for the individual beams need to do this better on LISA We need a measurement of the TM reflectivity measure on ground (we didn t do this for LPF) measure in flight, because it may change with time outgassing, alignment (scratches, etc), cosmic rays colour centres in-flight measurement on LPF requires switching beams on/off and looking at ratios of powers does this scheme work with the proposed optical layout on LISA? We need the ability to modulate the measurement beam difficult on LPF because the system wasn t designed for this operation mode how does such a force noise modulation appear in X? 8

9 Direct force noise measurement On LPF we have a witness TM to estimate the force noise on drag-free mass actually we measure the differential force noise On LISA we don t have this could consider witness TMs behind each measurement TM major instrumental change Other (crazy) ideas? form a differential IFO between two TMs in one SC? what else? 9

10 Building the model If direct measurement is unavailable Use Sagnac to estimate noise in X? to be studied Measure all parameters Form individual instrument models Propagate these models to TDI variables how does each noise source propagate and accumulate? studies beginning on this Repeat when needed Or directly establish the noise budget in terms of TDI variables multiple temperatures, multiple pressures, multiple laser powers, etc 10

11 Optimising the noise in flight From LPF we will learn a lot about how to optimise the working point of such an instrument which parameters matter which knobs should we turn which knobs are missing what are the operational constraints, and what should we design for in future? what is the impact of these measurement campaigns in flight? How long do they take? How often do they need to be repeated? 11

12 Early (morning) thoughts on calibration So far we typically do forward modelling of sources and compare to TDI output how accurate is this? Sufficient? Does this set the accuracy we need on noise modelling and system calibration? What can we say about un-modelled sources? calibrate TDI to incident gravitational wave strain how? in which reference frame? Calibrate TDI variables in terms of force on TMs we have ways to impose a known force on TMs Express waveforms in terms of stress tensor at the constellation We have found in LPF that working in acceleration eases the treatment of forces and system dynamics noise subtractions, force calibrations, etc 12

13 Summary Much to think about Some may say it s too early After LPF we will know much better the capabilities we require to operate, characterise and maintain such an instrument Build LISA with all the features we need to use it easy to use -> minimise impact on science output (downtime) 13

14 Open questions How good does the noise model need to be? Are iterative methods good enough? How do we calibrate in general for noise subtraction for un-modelled sources How do different noise sources propagate through TDI? What can we learn about instrument noise from Sagnac combination? 14