Документ взят из кэша поисковой машины. Адрес оригинального документа : http://www.eso.org/~lelouarn/minutes.txt Дата изменения: Fri Dec 1 19:23:42 2000 Дата индексирования: Mon Oct 1 23:45:56 2012 Кодировка: Поисковые слова: lightning

Minutes of the MCAO @ CfAO meeting
M. Le Louarn

The first hours of the meeting were devoted to explain the goals and means of
MCAO. We introduced the problems specific to Laser Guide Stars (LGSs): the
tilt problem and the cone effect. Then we went on to explain more in detail
how to build an interaction matrix in MCAO.

There are two main goals for MCAO:
- Solve the cone effect (1 DM + several LGS)
- Increase the corrected FOV (several DMs, several LGSs)

The problem with the first goal is whether or not the wavefront sensing can be
done in closed loop, because the wavefront sensor looks at the LGS cone, when
the correction is made for a parallel propagation.

We then went on presentation of current and future MCAO projects by different
groups. Presentations were given by (in chronological order):

M. Le Louarn
- Numerical simulations of MCAO
- Measure 5 modes from a NGS
-> See presentation on this web page

B. Ellerbroek
- Numerical simulations of MCAO
- Measure 2 modes from 3 NGSs
- Construction of a real MCAO system
-> See presentation on this web page

R. Dekany
- Simulations of isoplanatic effects
- Upgrade of the Palomar AO system to a MCAO system
-> See presentation on this web page

B. Bauman
- Possible MCAO demonstrator on the Lick AO system. Needs only a WFS camera

J.-P. Veran
- Lab experiment to demonstrate MCAO at UVic
-> See presentation on this web page

M. Schoek

- Do we need to repeat Ragazzoni's experiment w/ higher accuracy and better
signal to noise ratio ?

- A larger telescope is needed to increase the S/N to get shorter
integration times -> Keck time required.
- Use of Keck Guider ? -> Need fast camera to freeze the turbulence
- This experiment could be made to assess the variability of the
tomographic matrix (several nights, reconstruct this matrix each time
and see how much it changes with time)
- Investigate single curvature WFS method more thoroughly -> Test it on
the sky with a curvature sensor ?

A. Quirrenbach
- D. Kirkman at UCSD has worked with Ragazzoni on the optimal conjugation
height of DMs, use of turbulence profiles.
- Tests on the Keck to make an MCAO experiment: failed for technical
reasons (optical aberrations).

Complete lab experiments are under way at:
- UVic
- Durham
- Lund
They all want to demonstrate the feasability of MCAO, with DMs, WFSs and phase
screens. The money is there, in most of the cases. There seems to be a lack of
manpower to conduct the experiments (except at UVic, where the project just
started). The experiment at Uvic should be over at the end of 2002.

- Discussion of Fresnel propagation effects:

The issue is to know if propagation effects can significantly reduce the
performance of MCAO systems with respect to simulations. This could impact on
the positionning of the DMs in the optical path (Currently, Gemini has a
reversed order for the DMs compared to the order which compensates propagation
effects).
- Brent: all these effects can be simulated in the computer (Fresnel
propagation code being implemented)
- Brent will make simulation of this effect with his new code
- The effects at large zenithal angles remain to be investigated by Gemini.
At shorter wavelengths, this will probably become a significant effect.
- Miska has done an experiment in the lab with SLM phase screens (liquid
crystal). Tomography was demonstrated, but the limiting factor was the
additivity of the two phase screens (additivity verified to 2-3 %). This
could have been due to strong scintillation (phase screens placed too
high), but it is not sure.

- Discussion on comparing simulations

- Brent's code: ratfor for the analysis code, matlab for the numerical
simulation
- Miska's code: IDL, should be made more user friendly
- Uses a "real" Shack-Hartmann (i.e. computes the spot shapes in each
sub-aperture), no optimal estimation
- Miska also has an analytical code (Tokovinin et al.), similar to
Brent's analysis code, except it is not "closed loop" (no temporal
behaviour...)
- General agreement in the behaviour of MCAO
- One discrepancy seems to be the aliasing/fitting error (higher peak
Strehl for Brent, even with a single NGS)
- Transfer of phase screens between simulators to get closer results
- Agreed to Establish test procedures to test results
- Rich could write modules to Brent's simulation
- Parallelization doesn't seem possible if the codes are written in a high
level language (IDL, matlab...). Parallelization could be desirable for
ELT simulations

- Collaborations possible on realization of phase screens for experiments / AO systems:
- Palomar (?)
- UVic
- Lick observatory

There are two main possibilities to make phase perturbations for a lab
experiment:
- The "hair dryer" approach, consists of blowing hot air in the optical
path. The pros are that it is fairly easy to do, and the turbulence is
continuous. The cons are that the phase is not Kolmogorov (lack of low
spatial frequencies) and the phase is not repeatable.
- Phase plates (either etched or liquid crystals): exactly the opposite from
the other method

- Almost everybody would like to get some results out of the simulations:
- Simulate lab experiments
- Simulate future MCAO systems
So in this area, many collaborations are possible

- Discussion on inversion of the interaction matrix on ELTs (~30m diameter
telescopes)
- These are large matrices (~25000 x 8000)
- Need special techniques to be inverted
- Use Sparse techniques
- Use special properties of these matrices (symmetries...)
- Problem is to filter out only a small number of eigenvalues
- Brent+ Miska are going to generate such matrices and send them to Rasmus
+ Curt


- Discussion on the effect of LGS projector-jitter on MCAO: changes the
interaction matrix
- If the LGS projector jitters (by say 1''), the LGSs will not be at the
locations where they are supposed to be -> They won't probe all right
turbulence -> source of error
- Need input on the tolerancing of a closed loop system maintaining the LGSs
in position
- Possible to use "total least squares" on the matrix inversion ? This
algorithm takes into account the fact both the model (the matrix) and the
data can be noisy. The problem is likely to implement this in real
time. The modal filtering can also be a problem (the change in LGS
position induces a change in the shape of the eigenmodes, so modal
filtering is likely to be tricky in this approach). The efficiency has
also not been tested.


We all agreed that such a meeting on MCAO should be held every 6 months or so.