High-Order Wavefront Sensor
An image of the 2000 element deformable mirror used in SCExAO.
High-order wavefront sensing and correction is the key feature of the SCExAO Project. High order wavefront correction, or "extreme AO" requires a deformable mirror with many actuators across the pupil to correct for high spatial frequencies. In addition, it requires the control loop to run much faster than the coherence time of the atmosphere (~1 ms/1 kHz). For this reason we use the Boston MicroMachines 2000 element MEMS-based deformable mirror. For full details about the deformable mirror please see our publications.
One interesting feature of the DM is that it should not be operated in a high humidity environment as this causes advanced ageing. In order to accommodate this we have developed a dry air supply circuit. This circuit is interlocked such that the DM electronics are shut off if the humidity goes above 15% and or the pressure in the line deviates from the nominal setting of 0.4 psi. Please see this document for further details about the interlock system.
The other major aspect to high order wavefront correction is sensing. For this we utilize a pyramid wavefront sensor. This sensor operates in the visible between 600 and 900 nm. We have recently begun to use a pyramid prism which replaced the micro-lens arrays used in the past. The concept is to place the vertex of the pyramid/micro-lens array in the focal plane to create 4 distinct images of the pupil from each sector of the focal plane image. As each pupil image does not have all the Fourier information from the entire PSF, phase errors manifest as amplitude errors which are detectable by the sensor. Several typical images are shown below.
New approach: pyramid prism and OCAM2k
We determined that the PyWFS’s performance was being limited by hardware. Hence we undertook a number of hardware upgrades prior to the November run. These included
- A photon counting EMCCD (OCAM2k – FirstLight Imaging): binned frame rate = 3.5 kHz, negligible latency, 0.3 e- read-out noise at a gain of 600x.
- A fast tip/tilt mount to modulate the pyramid response: can be run up to 10 kHz but amplitude of modulation drops off at higher frequency.
- A pyramid optic: The pyramid optic was leant to us by MagAO (Laird Close and Jared Males). The optic consists of two pyramid-shaped optics back-to-back. The apex and vertices at the interface of the sides of the pyramid are extremely precise (less than 5 μm wide), which minimizes the affect of diffraction. The pyramid optic was needed as we determined micro-lenses were not ideal. Small micro-lenses had low diffraction but had a limited field of view, which meant it was not possible to modulate with them. Large micro-lenses allowed for modulation but the edge quality was poor, which lead to very high diffraction effects.
An image of the 4 pupils with the new optics is shown in the figure.
Pyramid wavefront sensor image with the Pyramid prism. The pupil image is now very clean and sharp. Note this is with 9 λ/D modulation radius.
Highlights of the key results to come from the November 2014 run:
- The PyWFS worked on-sky on a limited number of modes (up to 14) improving the PSF stability for science observations.
- After many days of experimenting, a robust approach to operating the PyWFS was determined. The PyWFS loop was closed on the internal laser source (with realistic turbulence, similar to post AO188 values) on up to 1030 modes! HiCAIO images with the loop open and closed clearly show the improvement in the PSF/speckle quality. The Strehl ratio improved from 23% to >90% on average. This indicates that the PyWFS will be able to take a post AO188 quality PSF and improve the Strehl to ~90% as designed in future!
To see full details please look at the reports on the publications page.
(Top) HiCAIO images (60 1.5s coadds) with the PyWFS loop open and closed on 1030 modes and 60 nm RMS turbulence (No modulation used). (Bottom) Strehl ratio when the PyWFS loop was open and closed on 830 modes with 300 nm RMS of turbulence and modulation (1.9 λ/D radius). Note internal broadband source used.
Old approach: micro-lens arrays
In early 2014 we have closed the loop with the high order wavefront sensor in the laboratory on the first 5 Zernike modes at 1.7 kHz and the first 900 Fourier modes at up to 300 Hz! The loop speed for the Fourier modes was limited by the matrix multiplication process. We moved the processing of the loop to GPUs. We used several GPUs (2 x Tesla K40, Tesla k20, 2 x GTX780) in an external PCI chassis to push the matrix multiplication speed to > 5 kHz.
Pyramid wavefront sensor images. The reference image is on the left and the resultant image from applying a sine wave to the DM is show on the right.
To test the loop performance we tested it on the vibrations induced by the cryo-cooler from HiCAIO. A movie of the PSF motion due to the HiCAIO Cryo-cooler was taken with the internal IR science camera and can be found here. A second movie with the cryo-cooler off clearly shows the affect of the cryo-cooler's pump on the instruments PSF. The Zernike loop performance in the presence of these vibrations is displayed below. The optical transfer function for the tip/tilt modes shows strong attenuation below 30 Hz. This was taken for one gain setting and hence the 0dB frequency could be tuned to some extent to attenuate higher frequencies.
The optical transfer function for the Zernike loop for the pyramid wavefront sensor. This was taken in the lab based
on the noise from the HiCIAO cryo-pump for the tip and tilt modes.
|HOWFS components||Camera||DM||Zernike loop||Fourier loop|
|Current speed (kHz)||1.7 (rolling shutter)||4.0||1.7||0.1-0.3|
|Future speed (kHz)||3.7||4.0||3.7||3.7|
The table summarises the current speed of various modules of the HOWFS and where we hope to get to by the end of 2014.