Free-flight experiments
in LISA Pathfinder
C. Cutler, A. Gryngier, M. Hewitson, P. Maghami, S.Vitale, I. Thorpe, W. Weber
Qualitative Motivation
LISA
•
•
•
LPF
Actuate SC1 to maintain TM1 free-fall
Actuate SC2 to maintain TM2 free-fall
J.I. Thorpe - NASA/GSFC
•
•
•
•
Actuate SC to maintain TM1 free-fall
Actuate TM2 to follow TM1 (or SC)
LISA X - Gainesville, FL - 2014.05.20
2
Quantitative Motivation
Design Requirements
Best Estimate
nsi
ng
100
Fourier Frequency [mHz]
•
For
ces
OM
SS
100
Acceleration Noise
ru
Th
10
ect
Total
Electrostatic
Actuation
1
0.1
ker
1
Dir
Trac
ker
Trac
1
0.1
en
sin
g
Se
st
er
s
Electrostatic
Actuation
10
Star
Star
Acceleration Noise
r
ula
ng
Direct Force
s
Requirements
OM
S
Total
SA
OM
10
100
Thrusters
1
10
Fourier Frequency [mHz]
100
Electrostatic actuation noise may limit our ability to measure forces on the TM
J.I. Thorpe - NASA/GSFC
LISA X - Gainesville, FL - 2014.05.20
3
Drift Mode Concept
Standard Science Mode
SC gravity
Applied Force
“Constant” Applied Force
Drift or Free-Flight Mode
SC gravity
SC gravity
Applied Force
Applied Force
Discrete “Kick”
J.I. Thorpe - NASA/GSFC
SC gravity
Free Fall
LISA X - Gainesville, FL - 2014.05.20
Discrete “Kick”
4
Example Data
Free-Flight Segments ~300s
Kicks (truncated)
J.I. Thorpe - NASA/GSFC
LISA X - Gainesville, FL - 2014.05.20
5
Data Analysis Challenges
•
•
•
Excise kicks
Estimate acceleration noise
below kick frequency
•
longest free-flight is ~300s
or 1/(3mHz)
•
Want acceleration noise at
1mHz or even 0.1mHz
Remove large free-fall trajectory
(~10um) with sufficient
accuracy
•
small residuals are aliased by
“gap transfer function”
J.I. Thorpe - NASA/GSFC
LISA X - Gainesville, FL - 2014.05.20
6
Windowed Approach
•
Approach
•
Build a customized window that smoothly
goes to zero at each gap location
(suppress gap transfer function)
•
Apply normal PSD estimation techniques
J.I. Thorpe - NASA/GSFC
LISA X - Gainesville, FL - 2014.05.20
7
Windowed Approach Results
•
results from STOC Sim #3:
Able to recover performance
near science mode
performance and perhaps
below.
•
Not clear if noise floors in
OSE simulator were
configured properly, may not
have been actuation noise
limited
Credit: M. Hewitson
J.I. Thorpe - NASA/GSFC
LISA X - Gainesville, FL - 2014.05.20
8
0 Elsewhere
!4 if i ! j = 1
Spectral Leakage Suppression Method
e. R d2 i, j =
1 if i ! j = 2
6 if i ! j = 0
the autocorrelation of second derivative of DT white noise
0 Elsewhere
All these autocorrelation functions where convoluted with the low-pass filter described in sect. 4.1,
except for the 1/f2 and the 1/f4 terms where the filter had very little effect.
It must be stressed that, despite the chosen model may suggest some physical interpretation of the
various contributions, for what is concerned here, this model only constitutes a smooth function of
frequency for the purpose of fitting the data.
•
•
4.2.2 Fitting
With the model in eq. 22 we have performed a standard least square fit to the data of Figure 4.
Approach
The results of this best fit procedure are reported in Figure 5.
•
apply low-pass filter and detrend data between
gaps
•
•
•
Fit analytic model to resulting spectrum
Compute transfer function of gap process Apply “inverse gap transfer function” to model
spectrum
Results (noise only)
•
•
Significant improvement over simple gap filling
(e.g. straing lines, splines, etc.)
Some bias introduced at low frequencies
Figure 5 The results of the fitting procedure with the spectral leakage
model.
Black line: the PSD
Credit:
S.Vitale
data form Figure 4, Blue line: the best fit prediction. Gray line: the native PSD that, through the
action of the leakage and that of the low-pass filter converts into the blue line. The gray band
represents the 2-σ uncertainty area. For comparison the red line is PSD calculated on raw data
after de-trending taken from Figure 2, left.
Doc: S2-UTN-TN-3091 1.0 April 11, 2012
J.I. Thorpe - NASA/GSFC
LISA X - Gainesville, FL - 2014.05.20
9
9
Gap-filling Approach
•
Approach
•
•
Fill gaps with random data generated from an
assumption of an underlying noise spectrum
Algorithm
•
•
Create two-point function from model spectrum •
use data adjacent to gaps and two-point function to
adjust mean of random samples
•
Update spectral model and iterate if necessary
draw zero-mean random samples matching that
distribution
J.I. Thorpe - NASA/GSFC
LISA X - Gainesville, FL - 2014.05.20
10
Gap-Filling Results
Noise Only Test
Complete Problem
LTPDA 2.8.dev (R2014a Prerelease)
2014−01−21 11:06:56.514 EST
ltpda: 4634eac
iplot
−8
10
Science Mode
Drift Mode
−9
10
−10
[m s−2 Hz−1/2]
10
−11
10
−12
10
−13
10
−14
10
−15
10
−5
10
•
•
−4
10
−3
−2
10
10
Frequency [Hz]
−1
10
0
10
Gap filling technique works well, but is slow
Residuals from imperfect subtraction of parabolic flight corrupt spectrum in full
problem
J.I. Thorpe - NASA/GSFC
LISA X - Gainesville, FL - 2014.05.20
11
Drift Mode for ST7?
•
Alternate Control Design Select mode based on instrument state
variables
•
Automatic compensation of changes in
gravity gradient, noise levels, etc.
•
irregular gap intervals (possible advantage
for data analysis)
J.I. Thorpe - NASA/GSFC
LISA X - Gainesville, FL - 2014.05.20
Suspension
Controller
all
Features
e-f
•
Use existing validated controller modes - fre
•
•
12
Drift Mode in the Lab
•
•
U. Trento Torsion Pendulum Experiment
•
Apply external torque to mimic LPF gravity
gradient
•
“Suspend” test mass with kick-controlled
torque to recover nominal angular position
•
Measure noise in kicked case as compared to
zero-gradient case
•
accelerations of 6x10-10m/s2, free flights of
~225s, impulses of ~25s
AEI Interferometry Readout
•
•
Mimic TM motions in drift mode with piezo
mirrors, check how IFO tracks motion
UF Torsion Pendulum Experiments
J.I. Thorpe - NASA/GSFC
LISA X - Gainesville, FL - 2014.05.20
Credit: W. Weber
13
Conclusions
•
Free-flight experiments allow validation of drag-free flight in the presence of
actuation noise
•
•
Multiple data analysis strategies have been developed, all with some success
•
Experimental work will generate data with more realism, providing further
tests for the DA pipelines
J.I. Thorpe - NASA/GSFC
Work remains to finalize and validate analysis pipelines in preparation for
operations
LISA X - Gainesville, FL - 2014.05.20
14