Transcript A.Freise: News from Finesse

FINESSE +
future plans and work in progress
Andreas Freise 22nd July 2006
A. Freise 22.07.2006
FINESSE
http://www.rzg.mpg.de/~adf/
A. Freise 22.07.2006
Which Version to Use?
0.99.3 -> current (22.07.2006)
fixed bugs:
o several memory/pointer errors but none showed any difference
when running test examples
A. Freise 22.07.2006
Which Version to Use?
0.99.2 -> 0.99.3 (25.02.2006)
fixed bugs:
o 'cav' tracing did not find its way through a BS when approaching
it from node 4.
A. Freise 22.07.2006
Which Version to Use?
0.99.1 -> 0.99.2 (26.10.2005)
fixed bugs:
o a 'retrace' on an unstable cavity or 'retrace' with no 'cav'
nor 'gauss' command caused a "bug found" message.
o the 'func' command could not parse function strings with
multiple occurences of the same variable name.
o the phase delay on propagation through a 'space' component
had the wrong sign. This basically flipped the frequency
dependence, e.g. upper sidebands were behaving like
lower sidebands and vice-versa. This must have been
wrong since the beginning of Finesse. Should affect
almost all results but mostly by only switching frequency signs.
A. Freise 22.07.2006
Which Version to Use?
0.99 -> 0.99.1 (26.08.2005)
fixed bugs:
o inconsistency in definition of demodulation phase
o a pointer error in the function call for the evaluation
of 'func' commands caused some arbitrary results when
more than one argument was used
o a memory allocation error occurred when 'bp' detectors were used
together with other detector types
o demodulating at a signal frequency introduced a factor of 0.5
for all demodulations except for the standard transfer function
with a pd2.
A. Freise 22.07.2006
Ways to Improve FINESSE
 collect simualtion examples for validation/verification
 develop and collect tools for automation and data


post-processing
improve performance, flexibility of the FINESSE binary
add new components, mainly for non-classic interferometry
 gratings
 quantum noise
 polarisation
A. Freise 22.07.2006
The FINESSE Test Suite
 Nightly run of the latest version
against several hundret example
files.
 What is missing, is a set of files
which have been checked by an
independent user against
calculation
and/or experimental data
Testing:
-------Thu Jul 20 10:46:08 BST 2006
Entering physics ...
OK: darkf-comp.kat
OK: grating1.kat
Entering random ...
OK: 00all_gouy.kat
OK: 00first.kat
OK: 01cavity-mode.kat
OK: 01resoances.kat
OK: 01tilt.kat
OK: 02fringes-pm.kat
OK: 02fringes-untuned.kat
OK: 02fringes.kat
OK: 03mode-separation.kat
OK: 04modulation-b.kat
OK: 04modulation-c.kat
OK: 04modulation.kat
OK: 05IB.kat
OK: 06matching.kat
OK: 07CCD.kat
OK: 08telescope.kat
….
A. Freise 22.07.2006
Data Post-Processing
 Automate certain simulation tasks
 Perform data mining, data preparation
 Example 1: complex control matrices
 Example 2: „tolerancing“
A. Freise 22.07.2006
Getting More out of FINESSE
 Change FINESSE input/output so that the binary can be called
as a function from other programs, e.g. Matlab/Octave
 started by Martin Hewitson, copying the idea of his
data server:
Client1
Client2
TCP/IP
Threaded server
‘katd’
Client3
Server ‘katd’ started with input file, without ‘xaxis’
Clients send commands like: component parameter value
and receive one result (data point) per command
A. Freise 22.07.2006
Going Parallel
 Beowulf Cluster:
 Python script reads input file, „understand“ the dimension
of the parameter space (I.e. xaxis, x2axis) and creates
N new input files, submits them via Condor to the cluster
and then merges the N output files
(student project, first examples work!)
 MPI Cluster (true parallel computing)
 FINESSE spends 90% of its time in the Sparse routines
„solving“ the set of linear equations. This package can be
replaced by a modern one which can be optimised for
MPI compilers. Then each data point is computed on
aparallel computer.
(another student project?)
A. Freise 22.07.2006
Quantum Noise in FINESSE
 Quantum noise can be computed after the standard

calculation has been completed.
An almost identical procedure is used for building the
the quantum noise matrix
A. Freise 22.07.2006
Quantum Noise in FINESSE
 Quantum noise can be computed after the standard

calculation has been completed.
An almost identical procedure is used for building the
the quantum noise matrix
A. Freise 22.07.2006
Quantum Noise in FINESSE
 Prepared a second interferometer matrix to compute
quantum noise couplings
 use field amplitudes and mirror masses as input


parameters
treat every port as a potential input of vacuum
vacuum can be squeezed
 Missing:
 coupling equations (exist)
 some understanding of how to work with multiple fields
 testing!
A. Freise 22.07.2006
Conclusion
 FINESSE is alive and changing
 A communty effort would make it better (especially
with respect to add-on tools)
 Simulation work for commissioning might be the
best way to document and preserve knowledge.
Don‘t build an interferometer that way next time!
A. Freise 22.07.2006
End
A. Freise 22.07.2006