HI all
sorry I can't make the telecon this week. I will have an update on the WFS optical design next time.
I have read carefully the preliminary Marcos report and include some comments and questions here. I thought iwas really nice to have this document. thanks for arranging this theo. Perhaps you all can discuss this today on the phone.. I'd specially like to hear Mike's comments regarding how important it is for Marcos to use a proper model of the camera (currently it just wrong).
most of my comments and questions in this txt file are for our group .. i don't expect this to be sent to marcos for answers, but just for discussion amongst ourselves. maybe a subset of these questions can be passed onto Marcos.
this report reminds of some critical open questions taht affect my optical design:
a) number of shack hartmann spots -- hexagonal or rectangular
b) how well-sampeld each spot should be and what total FOV we need to maintain.
c) do we need space in the optical design for
> ND filter
> red-pass filter
> possibly leave room for possible ADC upgrade some day?
I will write some code so that we can play with these numbers and will start to work towards fiducial numbers in the next weeks...
COMMENTS/FEEDBACK/INPUT WELCOME!
-John
jdm_comments_simulation.txt
JDM comments on Marcos van Dam Memo
"Monte Carlo simulations of an AO system for CHARA"
v1. 8 Aug 2012
Section 2.
2.1 Why does M3 limit our beam? shouldn't it be M2? can be this corrected?
2.2
Main problem is that the camera model is not sufficient. A photon counting camera is not well modeled as a camera with 0.5 e- read noise. The clock induced charge is only .0005 events/pixel, which means that in each 128x128 readout there will be about 9 false events. This needs to be properly modeled if we are to believe the limiting magnitude estimates.
2.3. Our current plans od not include an atmospheric dispersion corrector -- and so I thought we decided we need to filter the light going to camera. the illations assume we can use all the light short of 1 micron filtered only by the intrinsic CCD response curve. how much sensitivity do we lose by throwing out the blue? presumably not so much for red stars?
Q: what does "a figure of 30% is used to account for transmission losses in the WFS? this means the total transmission goes from 41% before WFS to 30% at the detector? so the transmission is about 75% through WFS?
2.5. we might want to spread out light ove rmore than 6x6 pixels in order to keep detector in photon counting mode over a wider range of conditions.
Also -- not sure about if we need wider FOV in case of open loop operations, esp. fixin static aberrations in telescopes.
TBD. having a little high angular resoluiont might be a good idea since even a single photon can yield a fairly a precise centroid.
should we use hexagonal grid or rectangular? i thought the actual DM we might use is hexagonal? is it important to match the pitch of the DM (if possible)?
2.7. we had a particular DM in mind.. I don't thin kit was 7x7.
actually we had 2 priced out and we had mostly thought the cheaper one was ok.. can we dig this info out?
2.8. With the EMCCD, the frame rate and expire controls are more complicated. for instance, we probably always run at max speed, but the amount of EM gain will change depending on the brigthness. Not sure we need Marcos to model this, but maybe Mike can comment this.
now -- even though theta FRAME rate is always high to make best use of high gain mode, we will be doing coadding in software, and so the 'effective' update rate for AO control will be much lower.
3.1. I don' think the camera model is correct. hopefully we can have better performance than this. I'm pretty sure we get improvement using Tiptilt down below mag 12.
also -- Strehl ratio ios probably NOT the right thing for the tipt-tilt only calculation. here we are interested in how faint we can do any trackking -- if at rates slower than 30 HzÉ so the real calculation is when can we even track a star at all say at 1 Hz?
3.2. I don't understand how there can be a good estimate of the open loop/ closed loop performance. this seems only to deal with the linearity of the wavefront tilt estimators in shack hartmann when far from zero ?
3.3. From memory,aren't these numbers asignificantly worse than our simulation from proposal? I presume this is because the camera noise model?
Figure 20.. so it seems a pyramid sensor gives the same strehl but down to 0.5 mag fainter. not altogether insignificant, but must be traded against complexity. Would a proper noise model change this calculation in relative terms?