Physical Model Driven
Calibration
Paul Bristow (INSY/SED/ESO)
Thanks to:
Michael Rosa, Yves Jung, Florian Kerber, Andrea
Modigliani, Sabine Moehler (ESO)
Data Simulation Workshop – ESO Garching – April 2016
Brief history of IPMG
 1999: Michael Rosa founded the Instrument
Physical Modelling Group inside the Space
Telescope – European Co-ordinating Facility
 Florian Kerber, Anastasia Alexov (later Mauro
Fiorentino) & Paul Bristow
 ST-ECF was wound up in 2006
 Legacy at ESO
 Physical modelling and reference data for:
 CRIRES
 X-shooter
Brief history of IPMG cont.
 Projects included:
 FOS
 Wavelength scale (geomagnetic environment of HST)
 Scattered light (grating analysis)
 STIS
 Wavelength calibration (discussed in detail below)
 Simulated readout (CTE)
 Reference data
 HCL characterisation (spectral atlases, operational behaviour
and ageing)
 Material characterisation (e.g. refractive index data)
 Recognised by a NASA group achievement award shared with
NIST colleagues
Astronet Roadmap 2009
“As a core fundamental element, and as
a guide, it is recommended that funding
provision for laboratory astrophysics be
included in the planning of all
astronomical and space mission
research programmes at a level of the
order of 2% of overall budget, with
each programme taking “ownership”
and peer-review of this part of the
project.”
Astronet Infrastructure Roadmap, p.132
F. Kerber
Precision Radial Velocity, PSU 2010
4
Calibration Reference Data
Traceable to laboratory standards >>
“ground truth”
For wavelength calibration reference data
link to frequency/time
Meta Data describing “what and how”
Error information
Documented by original data provider
Published
F. Kerber
Precision Radial Velocity, PSU
2010
5
Physical Model Based
Wavelength Calibration
Matrix representation of
optical components
 ME is the matrix representation of the
order m transformation performed by an
Echelle grating with E at off-blaze angle
. This operates on a 4D vector with
components (wavelength, x, y, z).
Physical Model Optimisation
Default
configuration file
cal’ lamp line list
wavelengths
Simulated Annealing
X-Shooter
Physical Model
Output
Predicted positions
of spectral features
(pixels on detector
array)
Change configuration
No
Measured line
positions (pixels)
from cal’ exposure
Compare lists and
compute metric which
describes how well
they match
Satisfactory
match?
Yes
Simulated Annealing
Optimal
XSPM
parameters
for this cal’
exposure
Applications
Wavelength calibration
Simulations
Early DRS development
Effects of modifications/upgrades
Instrument monitoring/QC
Advanced ETC?
A Precautionary Note
Regarding the CRIRES
Implementation
Potentially more useful for CRIRES
because of moving components
Never clear why it didn’t work
properly
Erratic behaviour of the instrument
did not help, but the problem was
almost certainly with the model.
X-Shooter (300nm-2.5m)
Commissioned 2009
Vernet et al. 2011.
A & A. in press
Model for UVB, VIS
& NIR arms
Same model kernel
Independent
configuration files
Cross dispersed, medium res’n, single slit
Single mode (no moving components)
Cassegrain & heavy => Flexure
NIR Th-Ar HCL full slit
Solar like stellar point source and sky
Physical Model Optimisation
Default
configuration file
cal’ lamp line list
wavelengths
Simulated Annealing
X-Shooter
Physical Model
Output
Predicted positions
of spectral features
(pixels on detector
array)
Change configuration
No
Measured line
positions (pixels)
from cal’ exposure
Compare lists and
compute metric which
describes how well
they match
Satisfactory
match?
Yes
Simulated Annealing
Optimal
XSPM
parameters
for this cal’
exposure
Effective camera focal length (mm)
Effective camera focal length (mm)
UVB Camera temperature sensor reading (°C)
VIS Camera temperature sensor reading (°C)
Modified Julian Date (days)
Effective camera
focal length (mm)
Detector tip (°)
Detector tilt (°)
X-shooter
Flexure
 Backbone flexure
 Causes movement of target
on spectrograph slits
 Corrected with Automatic
Flexure Compensation
exposures
 Spectrograph flexure
 Flexing of spectrograph
optical bench
 Can also be measured in
AFC exposures
 First order translation
automatically removed by
pipeline
UVB
VIS
NIR
Lab Measurements
• NIR arm
• Multi-pinhole
• Translational & higher order distortions
AFC Exposures
• Obtained with every science obs =>
large dataset ~300 exp from Jan –
May 2011
• Single pinhole, Pen-ray lamp
• Window:
• 1000x1000 win (UVB 12/VIS 14 lines)
• Entire array (NIR 160 lines)
NIR
UVB
VIS
Choosing “open” parameters
All parameters open
Slow
Optimal result
Degeneracy
Physically motivated:
Related to flexure
Constrained by data
In these results:
Prism orientation; Grating Orientation;
Grating constant; Camera focal length;
Detector position and orientation
NIR
NIR
NIR
NIR
(Product moment correlation)
VIS
VIS
Improvements and
Updates
 Full ray trace
 computationally possible now
 integration of eg. a Zemax library
 Removes modelling assumptions (would be much easier to
diagnose problems such as occurred for CRIRES)
 No direct implementation in CPL
 Despite fantastic support from Andrea and Yves…
 It was still a huge overhead
 Begin the development much earlier in the project
lifetime
 STIS: during operations
 CRIRES: end of MAIT
 X-Shooter: middle of MAIT
 Ideal: During preliminary design phase
Summary
Model based calibration is not a new
concept
It has been applied with some success
to HST and ESO instruments
Needs to be supported by appropriate
reference data
In 2016 this is computationally
inexpensive while the development
represents a small investment relative
to EELT instrument resources