MEMS Mirror Characterisation

There are now many different types of deformable mirrors available commercially, employing different technologies, having different numbers and arrangements of actuators, and capable of different strokes. Microelectromechanical systems (MEMS) is one of these emerging technologies which shows good promise for applications of adaptive optics (AO) to vision science - compact size, potential low cost, and large numbers of actuators. Here we model the ability of a MEMS mirror Boston Micromachines Corporation to perform a particular task: correction of the aberrations in healthy human eyes. The mirror has 140 actuators and a stroke of 3.5 µm with all actators at the maximum voltage. Full details of the fitting procedure used, and results obtained for other mirrors, are given in Eugénie Dalimier's research page.

Figure 1. Actuator layout of the MEMS mirror compared to other mirrors, and typical pupil sizes used.

The ability of all mirrors to generate the well-known Zernike polynomials was simulated first of all (see Figure 2). It is interesting to note that although the 37-channel membrane mirror and the MEMS mirror have the same stroke, the MEMS mirror is far superior in terms of PV deformation and fitting error. The greater number of actuators on the MEMS device appears to have a very positive effect here.

Figure 2. PV Zernike mode amplitudes for several mirrors.

Typical ocular wavefronts for well-corrected eyes were generated using a model developed by Thibos et al. [1]. The target wavefronts were samples of 100 eyes generated over 6 mm ocular pupils. The residual rms error after fitting to the same initial wavefront with each of the mirrors as a function of the number of modes used is shown in Figure 3. The fitting error for the BMC140 MEMS mirror also decreases as more modes are included in the calculation, although we found only negligible improvements on using more than 100 modes here. The minimum residual error is attained more gradually for this mirror, probably a result of the relatively localized influence functions. The OKO37 and BMC140 mirrors have practically identical strokes but differ in the number and arrangement of actuators, and this leads to very different results in terms of correcting the aberrations of the eye as simulated here. These results would lead us to conclude that it is the effectiveness of each actuator signal that is important, not the raw number of actuators. A detailed commentary on the performance of the other 3 mirrors can be found in [2].

Figure 3. Residual rms wavefront error after fitting with the four mirrors to the same initial ocular wavefront.

We also investigated the performance of the BMC140 mirror as a function of the beam diameter at the mirror, D/Dm, where Dm is 3.3 mm. We found that the lowest residual error was achieved on using the full diameter, and that it increased as less of the mirror was used.

Figure 4. Final residual rms wavefront error as a function of D/Dm on fitting with 100 modes of the BMC140 mirror. The line is a quadratic fit to the data.

We have simulated the ability of four different commercially-available deformable mirrors to fit typical ocular wavefronts. Only static aberrations are considered, and it is assumed that the actuator influence functions add in a linear fashion. The best performance was achieved with a 35-actuator bimorph deformable mirror, which had the largest stroke of all the mirrors tested. The MEMS device performed second best, despite having less or equal stroke than the two remaining mirrors. On the other hand it does have far more actuators than the other devices, so that it performs much better than, for example, the OKO37 device which has exactly the same stroke. What these calculations show is that it is both the raw number of actuators and the arrangement and effectiveness of each individual actuator which determine how well a mirror will perform a particular task.

References