In this section, the characteristic behavior of different vacuum packaged resonant 1D- and 2D-MEMS scanning mirrors is presented.
6.1. Qualitative Result
Prior to a detailed presentation of measurement data, a simple qualitative result shall be presented in order to give an impression of the benefit of vacuum encapsulated MEMS mirrors. As
Figure 7 shows, the oscillation amplitude of the biaxial vacuum packaged MEMS scanning mirror is roughly 10 times larger in each axis than a MEMS mirror of identical design actuated at atmosphere. The bright spot in the center of the projection area is produced by direct reflection of the incoming laser beam at the glass cover of the vacuum packaged mirror.
Figure 7.
(a) Biaxially scanned laser beam based on a MEMS mirror actuated at atmosphere with 40 V applied to fast axis (17.8 kHz), 5 V applied to slow axis (0.5 kHz); (b) Biaxially scanned laser beam based on identical but vacuum packaged MEMS mirror design at identical driving voltages of 40 V applied to fast axis, 5 V applied to slow axis.
Figure 7.
(a) Biaxially scanned laser beam based on a MEMS mirror actuated at atmosphere with 40 V applied to fast axis (17.8 kHz), 5 V applied to slow axis (0.5 kHz); (b) Biaxially scanned laser beam based on identical but vacuum packaged MEMS mirror design at identical driving voltages of 40 V applied to fast axis, 5 V applied to slow axis.
6.3. Characterization of 1D-Fast Axis Scanners
In this section measurement data of different vacuum encapsulated fast-axis MEMS mirror resonator designs is presented. Pictures of the two chips incorporating four different MEMS mirrors are shown in
Figure 9.
Table 3 provides the related dimensions of mirror geometry and torsional suspensions. The four MEMS mirrors exhibit resonant frequencies between 16 kHz and 38 kHz. For each of these MEMS scanners the decay of the resonance amplitude was recorded after switching off the driving signal for determination of the MEMS scanner’s Q-factors.
Figure 9.
Single-axis high-Q resonators. (a) Design #1 (b) Designs #2, #3, #4.
Figure 9.
Single-axis high-Q resonators. (a) Design #1 (b) Designs #2, #3, #4.
Table 3.
Geometrical design data and measurement data of four different fast axis mirror resonators.
Table 3.
Geometrical design data and measurement data of four different fast axis mirror resonators.
| Design #1 | Design #2 | Design #3 | Design #4 |
---|
Mirror aperture size | 0.8 mm | 1.0 mm | 1.0 mm | 1.0 mm |
Mirror thickness | 60 µm | 60 µm | 60 µm | 60 µm |
Width of beams | 47 µm | 44 µm | 65 µm | 65 µm |
Length of beams | 1,440 µm | 1,450 µm | 1,450 µm | 600 µm |
Thickness of beams | 60 µm | 60 µm | 60 µm | 60 µm |
Resonant frequency | 30.8 kHz | 16.5 kHz | 25.1 kHz | 38.5 kHz |
Q-factor | 26,800 | 75,500 | 70,000 | 49,300 |
Theta-D-product | 17,3 mm × deg | 13.8 mm × deg | 13.8 mm × deg | 10.8 mm × deg |
Figure 10.
Resonance amplitude decay plots of different high-speed MEMS mirror resonators. (a) Design #1 (b) Design #2.
Figure 10.
Resonance amplitude decay plots of different high-speed MEMS mirror resonators. (a) Design #1 (b) Design #2.
Figure 11.
Resonance amplitude decay plots of (a) Design #3 (b) Design #4.
Figure 11.
Resonance amplitude decay plots of (a) Design #3 (b) Design #4.
Figure 12.
Frequency response. Blue curves represent downward sweep, black curves represent upward sweep. (a) Design#1 (b) Design#2 (c) Design#3 (d) Design#4.
Figure 12.
Frequency response. Blue curves represent downward sweep, black curves represent upward sweep. (a) Design#1 (b) Design#2 (c) Design#3 (d) Design#4.
As
Figure 10 and
Figure 11 show, damping inside the MEMS package is so low that the high frequency scanning mirrors continue to oscillate for many seconds without excitation signal. For the three resonator designs #2, #3 and #4, which are implemented on a single chip, Q-values of 75,500, 70,000 and 49,300, were measured, respectively. The highest resonant frequency coincides with the lowest measured Q-value and
vice versa. The corresponding frequency response plots of these four different MEMS scanners are shown in
Figure 12, presenting nonlinear resonance characteristics. All scanners exhibit spring stiffening effects, making it necessary to drive the oscillators into resonance from lower frequencies to higher frequencies. However, designs #2 and #3 show a considerable lowering of that stiffening effect at total optical scan angles larger than 20 degrees, corresponding to mechanical tilt angle amplitudes larger than five degrees. Design #1 principally differs from the other three designs with respect to the comb electrode arrangement. In design #1 the electrodes are placed close to the torsional axis which enables overlap with the stator electrodes over a large angular range. The electrode fingers of the other three resonators are located along the mirror edges at a larger distance to the torsional axis, enabling only a short angular range of electrode finger overlap. Design #1 achieves a total optical scan angle of 86 degrees at 30.8 kHz. Because of a mirror aperture size of 0.8 mm the theta-D-product, which is the product of mirror size and mechanical tilt angle amplitude, is calculated to be 17.3 mm × degrees. That is sufficient for a raster-scanning display with HD720 resolution. A comparison of the fast axis MEMS scanning mirror presented in this work with other published developments is given in
Figure 13.
Figure 13.
Comparison of fast axis MEMS scanning mirrors. The reference numbers of the corresponding published articles are listed in the caption below.
Figure 13.
Comparison of fast axis MEMS scanning mirrors. The reference numbers of the corresponding published articles are listed in the caption below.
(A) = [11] | (B) = [12] | (C) = [13] | (D) = [14] | (E) = [15] | (F) = [16] | (G) = [17] | (H) = [18] |
(I) = [19] | (J) = this work design #3 | (K) = [20] | (L) = [20] | (M) = this work design #1 | (N) = [21] | (O) = [22] | (P) = this work design #4 |
During oscillation a resonant MEMS scanning mirror does not remain perfectly flat. It always experiences dynamic deformation. Maximum deformation occurs in the turning points of the oscillation where the angular acceleration
is:
(2)
where
θ is the mirror tilt angle amplitude and
ω is the angular frequency. Because mirror deformation contributes to beam divergence, it reduces the optical resolution of the projection display. It can be shown that the dynamic mirror deformation
δ scales proportional to the fifth power of mirror size
L [
23]:
(3)
where
f is the scan frequency,
E is Young’s modulus and
t denotes the mirror thickness. The dynamic mirror deformation that has to be expected for the four different single axis MEMS mirrors has been calculated by Finite Element Analysis applying the acceleration given by Equation (2). In order to minimize loss of resolution, dynamic deformation should not exceed one tenth of the wavelength. That means that the deformation should be less than 50 nm. In
Figure 14, the calculated dynamic deformations are plotted as a function of the scan angle. For design #1 and design #2 the deformation is within the limits of lambda/10 up to a total optical scan angle of 65 degrees. For designs #3 and #4 the deformation increases much faster because of the mirror diameter and a higher scan frequency. Mirror flatness can be improved by either increasing the thickness of the mirror plate or by using a more sophisticated mirror mounting [
24].
Figure 14.
Finite Element Analysis based calculation of dynamic mirror deformation for all four designs.
Figure 14.
Finite Element Analysis based calculation of dynamic mirror deformation for all four designs.
6.4. Characterization of a 2D-Gimbal Mounted Scanner
In this section a biaxial resonant gimbal mounted MEMS mirror for Lissajous laser projection is presented (
Figure 15). The principal mirror properties are:
Mirror aperture size: 1 mm
Fast axis resonant frequency: 17.8 kHz
Slow axis resonant frequency: 0.5 kHz
Figure 15.
Biaxial gimbal mounted MEMS scanning mirror for Lissajous laser projection.
Figure 15.
Biaxial gimbal mounted MEMS scanning mirror for Lissajous laser projection.
This MEMS scanner features separated pairs of driving and sensing comb electrodes for each axis. While driving electrodes are placed distant to the axis, the sensing electrodes for capacitive phase feedback are arranged close to the respective axis. Because of the very low gas damping, the vacuum packaged biaxial MEMS scanning mirror accumulates energy over several thousand oscillations.
Figure 16 shows the slow decay of the resonance amplitude after switching off the excitation signal. It takes about 90 s until the resonance amplitude of the gimbal reduces by a factor of two. The Q-factors were calculated to be 145,000 and 60,000 for slow axis and fast axis, respectively. The benefit of this energy conserving MEMS scanner becomes obvious when analyzing the deflection angle as a function of voltage, shown in
Figure 17. An excitation pulse height of only 10 V is necessary to drive the slow axis to a total optical scan angle of more than 70 degrees, whereas the fast axis needs a pulse height of only 60 V to achieve a total optical scan angle of 60 degrees. Due to the low driving voltages the power consumption of the 2D-scanning mirror is less than one milliwatt.
Figure 16.
Amplitude decay over time of gimbal (a) and mirror (b).
Figure 16.
Amplitude decay over time of gimbal (a) and mirror (b).
Figure 17.
Total optical scan angle as function of driving voltage for gimbal (a) and mirror (b).
Figure 17.
Total optical scan angle as function of driving voltage for gimbal (a) and mirror (b).
6.5. Capacitive Phase Feedback and Closed Loop Control
Because of the high Q-factors, the frequency response of the vacuum encapsulated 2D-MEMS scanning mirror exhibits extremely narrow resonance peaks and steep phase-frequency dependence. This leads to a high sensitivity of the oscillation amplitude to changes of temperature. Since in a laser video projection the changing laser power permanently causes temperature changes of the mirror and its suspensions, it is necessary to control each scan axis by an individual and independent phase control loop (PLL) in order to maintain a constant scan amplitude. For that reason, dedicated sensing comb electrodes provide capacitive phase feedback and thereby enable monitoring and stabilizing the scanning amplitude at both axes (
Figure 18).
Figure 19 shows the transient signals of capacitive feedback, optical position feedback and applied driving signal. The phase control of the biaxial resonant scanning mirror results in a Lissajous scanning mode with variable frequencies of the fast and slow axes. An additional control loop is therefore required to control the density of projected lines, which can change depending on the frequency ratio of the two oscillating axes.
Figure 18.
Capacitive phase feedback.
Figure 18.
Capacitive phase feedback.
Figure 19.
.Measured capacitive feedback signal over time in comparison with optically measured PSD output signal and driving pulse signal applied on the gimbal comb electrodes
Figure 19.
.Measured capacitive feedback signal over time in comparison with optically measured PSD output signal and driving pulse signal applied on the gimbal comb electrodes
6.6. Lissajous Laser Projection Based on High-Q Scanning Mirrors
Synchronization of MEMS mirror and RGB laser sources is performed on a FPGA platform (
Figure 20). Because of the high Q-factor the mirror oscillation is almost a perfect sine. That enables the accurate calculation of the instantaneous amplitude based on the measured phase and frequency. This xy-position information is used to read out the corresponding RGB pixel data, which then is used for modulating the light sources. According to the large scan angles of more than 60 degrees in both axes, laser projection based on vacuum packaged MEMS scanning mirrors enables high-resolution image and video projections (
Figure 21).
Figure 20.
Analog and digital control electronics for MEMS scanning mirror based laser display
Figure 20.
Analog and digital control electronics for MEMS scanning mirror based laser display
Figure 21.
Lissajous-laser-projection of a 512 × 1,024 pixel image based on a biaxial scanner with a plane glass window. The bright spot in the center of the image is caused by a direct reflex of the incoming laser beam at the glass lid of the MEMS mirror.
Figure 21.
Lissajous-laser-projection of a 512 × 1,024 pixel image based on a biaxial scanner with a plane glass window. The bright spot in the center of the image is caused by a direct reflex of the incoming laser beam at the glass lid of the MEMS mirror.
6.7. Ongoing Improvements
The non-homogeneous line density of the Lissajous trajectory causes one existing disturbing artifact that the picture of
Figure 21 does not demonstrate. Within one period of the slow axis oscillation at 500 Hz, two interlace images of 17.8 lines each are projected. Thus, projection of the whole image of 512 lines requires at least 14.3 oscillations of the gimbal axis. That means the complete image is projected after 28.8 ms, corresponding to a frame rate of 35 Hz. However, to achieve a standard video frame rate of 60 Hz the whole number of lines needs to be projected within 16.7 ms. In addition to the low frame rate, the ratio of fast axis and slow axis frequency is disadvantageous since it leads to a scan pattern of varying line density. In a simulation of the image projection, it can be shown what the scan pattern looks like after 1/60 s (
Figure 22(a)). The projected image is severely disrupted by dark patterns. In a live video demonstration with a real Lissajous laser projector, those patterns become visible as disturbing flicker effects. Hence, in order to increase the line density and in order to stabilize it even under the impact of rapidly changing laser power, there are two principle options. The first option is shown in
Figure 22(b). The frequency ratio of the fast axis and the slow axis can either be chosen as high as possible, meaning that the fast axis frequency of the biaxial scanner needs to be increased and the slow axis should be decreased to the image refresh rate of 60 Hz. As an alternative, the frequency ratio can also be chosen as low as possible, meaning that the fast axis frequency and the slow axis frequency differ only by the image refresh rate of 60 Hz (
Figure 22(c)). The best coverage is achieved with the high frequency ratio. However, the drawback is that the MEMS device becomes more fragile and more sensitive to vibrations by such a low resonant frequency of the slow axis. Thus, for automotive display applications the second concept based on a MEMS mirror with two fast axes of almost identical resonance frequencies could be a very robust alternative.
Figure 22.
Simulation of Lissajous laser projected images based on three different scanning mirrors. The integration time is 1/60 s and the source image has a resolution of HD720.
Figure 22.
Simulation of Lissajous laser projected images based on three different scanning mirrors. The integration time is 1/60 s and the source image has a resolution of HD720.
Another necessary and already achieved improvement of the display quality refers to the suppression of the bright spot which is caused by direct reflection of the incoming laser beam at the glass window in front of the oscillating mirror (
Figure 21). To avoid this disturbing bright spot, there are two alternatives: Either the MEMS mirror is permanently tilted inside the cavity with respect to the plane window, or the glass cap is tilted with respect to the mirror in its non-deflected position. Both solutions spatially separate the projected image from the bright spot, provided that the angle of inclination is larger than the mechanical tilt angle amplitude.
Figure 23 shows the schematic cross section of a packaged MEMS mirror with a tilted glass window. Part of the incoming laser beam does not enter the MEMS mirror cavity because it is reflected at one of the two glass interfaces. As the picture shows, this parasitic beam is reflected into a direction that is spatially separated from the angular range that can be used for laser image projection. The glass forming technology has been further developed to enable fabrication of glass wafers containing hundreds of tilted optical windows (
Figure 24). Such glass cap wafers have been successfully applied for wafer level vacuum packaging. Successful suppression of the bright spot is demonstrated by projection (
Figure 25).
Figure 23.
Schematic cross-section of packaged MEMS mirror with tilted optical window.
Figure 23.
Schematic cross-section of packaged MEMS mirror with tilted optical window.
Figure 24.
Wafer level vacuum packaged MEMS mirrors with tilted glass windows.
Figure 24.
Wafer level vacuum packaged MEMS mirrors with tilted glass windows.
Figure 25.
Suppressed reflex by use of a vacuum packaged mirror with tilted window.
Figure 25.
Suppressed reflex by use of a vacuum packaged mirror with tilted window.