Dear Don and other adaptive secondary friends,

here is a summary of our recent progress with the 30 actuator adaptive mirror prototype (P30). I am sending copy of this message to my entire list of "adaptive secondary friends" in Arizona + Thermotrex, Germany, Ohio and Italy. I have just created a very simple web page (http://www.arcetri.astro.it/~salinari/proto/p30_news.htm) where you will find a copy of this message with links to the figures and, in future, more information. In future you will only receive a message when the web page will have important changes. Additions to (and subtractions from) the mailing list are easy, just let me know.

P30 arrived in Arcetri at the end of July already preliminarily tested by Media Lario and Microgate (i.e., Daniele, Roberto and collaborators) and therefore with a working version of the electronics and basic software to control the 15 DSP. The fast connection between the DSPs and the main computer was (and still is) not available and we are using the existing DSP software for the present phase. We therefore are still using the very first control law we received from Roberto. We will acquire the capability of changing it when the software for programming the DSP will be available, we hope next week. Most of the time during August was spent in solving an "optical" problem and in writing high level software routines (using available DSP low level routines) to execute the sequences of operations needed to perform preliminary measurements.

The following figures show the main pieces of P30:

STAND.JPG shows the entire unit (upside down) including mirror+backplate+Aluminium plate, "MMT type attachment flange", laboratory stand and electronics.

MIRROR.JPG shows the aluminised mirror and the central membrane.

MAGNETS.JPG shows the back side of the mirror with the magnets.

CAPACIT.JPG shows the capacitive ring armatures on the backplate and, inside the holes, the tip of the coils facing the magnets.

BACKPL.JPG shows the perforated backplate.

AL-PLATE.JPG shows the Al plate supporting the actuators and the cooling lines. In its centre is the central support shaft used to support also the glass backplate and the membrane.

1. Invar ring and mirror figure

The first time we tried to make an interferogram of the mirror it was really horrible! It took some time, and the hint of the previous similar experience of P60 (thank you Don) to understand that most of the problem was caused by the procedure used to attach the inner invar ring. After un-gluing and re-gluing with RTV the figure was much better, although still not really good, but at least the entire mirror was well measurable with our interferometer. Having obtained what we wanted for the moment (a measurable mirror) we didn't insist in improving the figure. We had further indication (see the point about FEA in the following) that simple RTV gluing of the ring is likely to be the right thing. What seems to be important is tightening the Mylard annular membrane BEFORE gluing the ring to the glass. This means that the replacement of a damaged Mylard membrane requires a new gluing. We are therefore considering "disposable" membrane rings, made of much lighter and less rigid plastics, pre-glued to the membrane, that we can remove with a razor, if necessary, and re-glue with a suitable jig.

b) Intrinsic mirror figure

During the above "fix" (after having removed the invar ring) we have measured the mirror figure by floating it on "nominal" forces (loop closed on tree points, theoretical forces as derived from FEA simulations on all the others). After removing tilt and focus the "free" mirror had 0.24 micron RMS (1.6 mu PtV) surface error, mainly astigmatism. Taking into account possible errors in the forces and the effect of next point c), the mirror itself is more than good for our purposes (thank you, Steve and Buddy).

c) Chipping

We noticed that a very small chipping on the outer edge (less than a sq mm in area, probably .1 mm at the deepest point) can cause a fairly extended (~2 sq cm) fairly significant (~1 mu) local error. We will have to be careful with the edges!

We can study mirror dynamic response even with non-well calibrated sensors and forces and with a modest mirror figure. We basically confirmed that the air trapped between the backplate and the mirror introduces a very high damping in the system over an useful range of distances (up to about 90 micron separation). Qualitatively we expected this effect but, honestly, it is larger than we thought . During the last two weeks we collected hundreds of megabytes of data in a broad spectral range (0 to >2 kHz) with a variety of excitation modes. The general pattern is the following:

• all the mirror vibration modes are more than critically damped, or about critically damped for distances between mirror and backplate <~ 70 micron.
• In the region between about 70 and about 90 the damping becomes less than critical and at about 90 microns typical values of Q are close to 3 (i.e. still very damped!).
• Even the present (essentially random) PID control law works for distances < 90 micron ( while it produces oscillations at larger distances).

TREFOIL.gif shows the response in amplitude (i.e. ratio between exciting force and displacement) and phase of the trefoil mode of the mirror at about 540 Hz, the least damped mode we found until now, when it is excited by an appropriate set of actuators (three with 0 phase, three with 180 deg phase, respectively at the pick and valley points of the mode). The various curves refer to the following "nominal" distances from the backplate:

(At distance larger than 90 mu, as said, our control oscillates, while we will have to change the DSP "linearization" routine to access distances less than ~50 mu. The fact that everything occurs in the above range is therefore partly due to physics and partly to luck!)

A more global picture can be seen in S_ACT_14.gif and the similar files S_ACT_20 and S_ACT_26 where we used excitation by a single actuator to excite all the anti-symmetric mirror modes. Measurements in the above plots refer to a distance of ~ 90, 70, 50 micron from the backplate respectively. We are going to produce similar plots with symmetric excitation in the next few days and these will be added in the web page without further notice.

In parallel with the dynamic measurement we tuned up the FEA model of the mirror to get the best possible agreement. In doing this we learned something interesting on the influence of the rigidity of the inner invar ring on the mirror: basically no influence. In fact, the following table shows a summary of measured and calculated frequencies; the fairly good match is obtained by taking into account the invar ring mass, but not its rigidity. We will check and refine this "tuning" with optical influence function measurements, especially for actuators on the inner circle, but already this result tells us that the RTV glue is VERY compliant. With a rigid mirror-invar connection all the frequencies go up by 15-30 %.

The column "Zernike Polinom." reports the number and the radial and azimutal order of the polinomial resembling the mode.

* hro = higher radial order (i.e.: looks like the mode with that name in azimuth, but it has a different radial dependence with more radial slope changes)
** typical error for measured frequencies is + - 5 Hz

Given the high damping (broadening and shifting of the peak amplitude) and the number of modes (accumulation of phase shift) the best way we found of deriving a precise frequency for each mode from the data is that of using phase plots at various damping values (separations), like TREFOIL.JPG, and finding the point of equal phase change. That identifies the -90 deg phase shift, i.e. the resonant frequency. Long but safe. It will take some time to fill up the third column of the above table, but it is worth the effort, because with a well tuned Finite Element model and accurate damping coefficients we can use more simulations and fewer experiments for tuning up the control system. When this process will be well tuned for prototype(s), we will be able to extend it to the final systems, MMT and LBT, that will have different, but calculable, resonant modes, predictable damping etc..

Sorry for the length.

Ciao,

Piero