Success to report today:
In spite of very cranky computer hardware --
- a complete meltdown of the original RTComputer on Friday
- the heroic effort to rebuild everything onto a substitute machine Friday night and Saturday
- and it's still cantankerous: multiple crashes today...
We have closed the AO loop on the MEMS!
Explanation: the AO system is now assembled and aligned - this is a process that has taken the last 6 months. The final step is to insert the deformable mirrors. There are two of them: the "woofer" and the "tweeter". The tweeter is the 1000-element MEMS device that provides the fine wavefront control. The mirrors were tested and pre-calibrated in the lab, so, for example, we know the voltages that when applied to the mirrors, make them flat to a fraction of a wavelength.
We installed and tested both DMs last week, and poked their actuators to make sure they were working. Thursday, the woofer was taken out and replaced with a fixed mirror in preparation for testing just the tweeter.
On the left is the PSF of the beam when the tweeter is set to its flat-voltage. The steps are:
- With the mirror voltages set to the known flattening settings and light going through the system, collect reference data from the wavefront sensor.
- Relax the mirror to it's nominal "out-of-the-box" shape (believe me, this is far from flat)
- Close the AO control loop, now sending updates to the deformable mirror at the 1000 Hz frame rate of the wavefront sensor camera.
The PSF goes from a very blurry blob to the image shown on the right. The static flat and the closed loop PSFs have very close to the same full-width-half-max (about 2.8 pixels). Further analysis and testing is needed to get a reportable Strehl number, but it's looking very good.
The "ghost" image you see to the upper right is due to a UV leak in the dichroic splitter. This won't be there (will be filtered out) in the Sharcs camera images.
...
I experimented with various adjustments, feedback gain and things called "integrator leaks". The real system behaves surprisingly close to the simulations (I did a careful job writing a wave-optics simulator that the RTCode can wrap around).
Here are some pictures of the control screens:
In the foreground is the AO operator's console, showing (clockwise from upper left):
Hartmann wavefront sensor | Light intensity on the pupil (of the wfs)
Control signals to the tweeter | Control signals to the woofer
The display in the background shows the PSF camera image. This is a visible wavelength CCD camera.
 |
| MEMS flattening voltages applied, statically |
 |
| DM in "unpowered" state |
This is a cute movie. I boosted the feedback gain until the system was just on the verge of instability. A note of explanation: To have control sent to the right actuator, the system must know where that actuator is on the wavefront sensor plane. This "registration" calibration was done last Tuesday. The alignment has drifted ever so slightly since then (a few microns). The result is the waves you see drifting from lower right to upper left. This is Helmholtz equation in action! The wave velocity is the registration offset divided by the control cycle time.
Acronomia:
AO - adaptive optics
DM - deformable mirror
RT - real-time (anywhere from 50 to 1500 Hz control loop rates)
RTCode - the software that runs in real time
RTComputer - the hardware that runs the RTCode
