Inquiry Regarding SOapy Adaptive Optics Configuration Parameters: Pupil Size, Simulation Performance, and Strehl Ratio Behaviour

Question

Inquiry Regarding SOapy Adaptive Optics Configuration Parameters: Pupil Size, Simulation Performance, and Strehl Ratio Behaviour

gurwindersinghhundal opened this issue 5 months ago · comments

gurwindersinghhundal commented 5 months ago

Hi, I'd like to gather some information regarding parameters defined in YAML-based configuration files for different adaptive optics systems, specifically focusing on SOapy AO.

What does "pupil size" mean, as indicated by something like "pupilSize: 100" in the configuration file?
For a particular pupil size, the AO simulations perform well. For instance, with a 2m class telescope, considering a 17x17 Shack-Hartmann wavefront sensor (SHWS) lenslet array and an 18x18 deformable mirror, selecting a pupil size of 100 leads to successful NGS (Natural Guide Star) AO simulations. However, when I increase the pupil size to 264, especially for off-axis fields of view (FOV) of the detector, the AO simulation either stops working or freezes intermittently. Could you please explain why this happens?
I've observed that the Strehl ratio remains constant regardless of the magnitude of the guide star. Ideally, it should decrease as the magnitude of the star increases. Can you help identify the reason for this or suggest any changes needed to achieve this variation?

Ollie Farley · Answer 1 · Wed Feb 28 2024 19:23:14 GMT+0800 (China Standard Time)

Hi @gurwindersinghhundal, thanks for your questions.

Pupil size is the number of simulation elements (pixels) across the telescope pupil. So larger numbers increase the resolution of the simulation in the pupil (aperture) plane.
Could you share your config file? Increasing the pupilSize should not break the simulation in this way
By default no noise is included in either the WFS or science cameras. You need to set photonNoise: True in your WFS/science camera for the guide star magnitude to make a difference. You can also add read noise in electrons with the eReadNoise parameter.

gurwindersinghhundal · Answer 2 · Tue Mar 05 2024 13:28:15 GMT+0800 (China Standard Time)

I have the following questions:

How is pupil size related to wavefront sensor subapertures and
deformable mirror? I mean, how does pupil size affect the wavefront
sensor?
I added photon noise to the wavefront sensor (WFS) in the
configuration file (for NGS). It gives good Strehl ratios until the
magnitude of 10/11. However, suddenly, it decreases to 1% when I choose
a guide star magnitude of 12 with the AO loop on. Ideally, it should
decrease almost linearly until the magnitude of 18 with the AO loop on.
It gives the good strehl with AO loop off, by setting the gain of DM and
WFS on GUI to zero.
I have performed above simulations in visible band. The same
phenomenon occurs across different spectral bands (J, H, K bands).

I have attached the following configuration files below:
a) sh_8x8.yaml (default conf file provided by soapy)
b) sh_8x8_IIA.yaml (my configuration file)

gurwindersinghhundal · Answer 3 · Tue Mar 05 2024 13:29:58 GMT+0800 (China Standard Time)

this is default configuration file provided by the soapy : sh_8x8.yaml (NGS)
simName:
pupilSize: 100
nGS: 1
nDM: 2
nSci: 1
nIters: 5000
loopTime: 0.0025
threads: 4

verbosity: 2

saveCMat: False
saveSlopes: True
saveDmCommands: False
saveLgsPsf: False
saveSciPsf: True

Atmosphere:
scrnNo: 4
scrnHeights: [0, 5000, 10000, 15000]
scrnStrengths: [0.5, 0.3, 0.1, 0.1]
windDirs: [0, 45, 90, 135]
windSpeeds: [10, 10, 15, 20]
wholeScrnSize: 2048
r0: 0.16
L0: [20, 20, 20, 20]
infinite: True

Telescope:
telDiam: 8.
obsDiam: 1.1
mask: circle

WFS:
0:
type: ShackHartmann
GSPosition: [0, 0]
GSHeight: 0
GSMag: 8
nxSubaps: 8
pxlsPerSubap: 10
subapFOV: 2.5
wavelength: 600e-9

DM:
0:
type: TT
closed: True
iMatValue: 0.25

1:
type: FastPiezo
closed: True
nxActuators: 9
iMatValue: 500

Reconstructor:
type: MVM
svdConditioning: 0.03
gain: 0.6

Science:
0:
position: [0, 0]
FOV: 1
wavelength: 1.65e-6
pxls: 64

gurwindersinghhundal · Answer 4 · Tue Mar 05 2024 13:31:31 GMT+0800 (China Standard Time)

This is my configuration file: sh_8x8_IIA.yaml

simName:
pupilSize: 100
nGS: 1
nDM: 2
nSci: 1
nIters: 500
loopTime: 0.0025
threads: 4

verbosity: 2

saveCMat: False
saveSlopes: True
saveDmCommands: False
saveLgsPsf: False
saveSciPsf: True

Atmosphere:
scrnNo: 4
scrnHeights: [0, 5000, 10000, 15000]
scrnStrengths: [0.5, 0.3, 0.1, 0.1]
windDirs: [0, 45, 90, 135]
windSpeeds: [10, 10, 15, 20]
wholeScrnSize: 2048
r0: 0.13
L0: [20, 20, 20, 20]
infinite: True

Telescope:
telDiam: 2.01
obsDiam: 0.337
mask: circle

WFS:
0:
type: ShackHartmann
GSPosition: [0, 0]
GSHeight: 0
GSMag: 8
nxSubaps: 11
pxlsPerSubap: 10
subapFOV: 2.5
wavelength: 600e-9
photonNoise: True
eReadNoise: 0
DM:
0:
type: TT
closed: True
iMatValue: 0.25

1:
type: FastPiezo
closed: True
nxActuators: 12
iMatValue: 500

Reconstructor:
type: MVM
svdConditioning: 0.03
gain: 0.6

Science:
0:
position: [0, 0]
FOV: 60
wavelength: 0.65e-6
pxls: 256

gurwindersinghhundal · Answer 5 · Fri Mar 08 2024 12:56:10 GMT+0800 (China Standard Time)

I have the following questions:

How is pupil size related to wavefront sensor subapertures and
deformable mirror? I mean, how does pupil size affect the wavefront
sensor?

I added photon noise to the wavefront sensor (WFS) in the
configuration file (for NGS). It gives good Strehl ratios until the
magnitude of 10/11. However, suddenly, it decreases to 1% when I choose
a guide star magnitude of 12 with the AO loop on. Ideally, it should
decrease almost linearly until the magnitude of 18 with the AO loop on.
It gives the good strehl with AO loop off, by setting the gain of DM and
WFS on GUI to zero.

I have performed above simulations in visible band. The same
phenomenon occurs across different spectral bands (J, H, K bands).

I have attached the following configuration files below:
a) sh_8x8.yaml (default conf file provided by soapy)
b) sh_8x8_IIA.yaml (my configuration file)

Ollie Farley · Answer 6 · Wed Mar 13 2024 22:27:30 GMT+0800 (China Standard Time)

Pupil size defines the resolution of the turbulent phase screens that are the input to the WFS. There are some rules of thumb around this, usually you want to at least Nyquist sample the WFS subaperture size and/or r0, whichever is smaller. A pupilsize of 100 should accomplish this (100/17=5.8 pixels per subap). I tried your config file and the "freezes" you are getting are because of errors in the code that are shown in the terminal (not the GUI). When using 264 pixels and looking off-axis you get

numpy.linalg.LinAlgError: Could not invert Covariance Matrix to for A and B Matrices. Try with a larger pixel scale or smaller L0

This means your pixel scale is too small (equivalently, pupilSize too large) to compute the phase screens. This is a separate issue but can easily be solved by using a smaller pupilSize.

I don't believe the strehl ratio will necessarily decrease linearly with increasing noise. What you describe sounds like the loop collapsing at higher noise. This usually means you need to adjust the gain values for the DM(s) until you get a stable loop. You will need to perform this adjustment for each different noise value.