MEMS Reliability
Will Their Hinges Break?
We have come a long way in the last decade in the design of large MEMS switches for telecom use. Many techniques have been tried, most didn’t work. During the boom there were over 30 companies trying to build large MEMS switches, all but two failed. The current generation of MEMS switches are finally in operation carrying real world traffic. However, with everything that has changed, one thing has remained constant, concern with potential MEMS reliability problems. After all, MEMS switches are, by definition, mechanical and everyone knows that mechanical switches wear and are unreliable. Let’s put MEMS switch reliability under the microscope and examine the high reliability of today’s MEMS switches.
Potential concerns fall into 3 catagories:
- Wear over time effecting performance
- Stiction or other switching failures
- Hinge reliability
Wear
Anytime you have a mechanical system, there is always a concern about wear. Indeed, some of the early MEMS switch designs utilized sliders as shutters to redirect the lightbeams. There was a potential for wear problems over time with these sliders. The current generation of 3D MEMS for large switches have no mechanical contact points. Each mirror is like a trampoline, supported by multiple “springs”. The mirror is positioned using electrostatic forces instead of mechanical levers.
Figure 1. 3D MEMs
Furthermore, dynamic feedback circuits are used to guarantee that the mirrors are in the optimal position for a low loss connection. Thus, not only is there no opportunity for wear, but any drift over time would be monitored and corrected by the dynamic feedback circuits.
Figure 2. Dynamic Feedback Control
Stiction
Stiction has been the bane of many technologies including early hard disk drives and high speed tape drives where, under certain conditions, read/write heads would literally “stick” to the disk platters or magnetic tape causing damage. Stiction is a particularly vexing concern for telecom operators. While many telecom circuits may need to be changed often, many others will be setup once and remain unchanged for years until they are suddenly required to reroute. A fast, low-loss switch is of no use if, over time, its elements will get stuck either open or closed. Again, in the early planar designs, the “flip up” mirrors would be positioned by a mechanical stop in the up position.
The mirrors would occasionally get stuck against this stop and refuse to go back down. In this case, the simplicity of the on-off control of a 2D switch does not out weigh the potential for stiction problems. Since 3D MEMS based switches have no mechanical contact points, stiction is not an issue. Not only is mechanical stiction not an issue, but the dynamic feedback circuits continually exercise everything required to move the mirrors, further reducing concerns about stuck open or stuck closed switches.
Hinge Reliability
The concern about hinge reliability also comes from the earliest MEMs switch designs. They were simple planar designs with “flip up” mirrors. In these designs, the mirrors move 90 degrees, putting a lot of stress on the hinge. However, these designs were shown to be unsuitable for large, low-loss telecom switches and have been superseded by 3D designs. In a 3D design, each mirror directs the lightbeam to a target of a large number of destinations. Mirror movement is slight, only a few degrees maximum. This requires a much more sophisticated mirror positioning system, but results in extremely low stress on the hinges. Furthermore, since the mirror positioning is 3D the mirror is connected by multiple “hinges” or gimbals and each is more like the spring of a trampoline than the hinge of old. This further distributes the stress, avoiding a single pivot point. Experimentation indicates that the stress on Glimmerglass hinges is 1000 times less than the level required to cause fatigue ( C. L. Muhlstein et. al, J. MEMS, vol. 10, pp.593-600, 2001).
Theoretical and logical reasoning is fine, but the proof is in the testing and real world usage. Glimmerglass has shipped over 20,000 ports of 3D MEMS based large optical switches. Through extensive testing of each switch before it’s shipped and a separate program of long term testing, we have amassed a mountain of data on MEMS switch reliability. They don’t break.
For example, for the test shown below 56 mirrors were tested with each performing a maximum swing 650,000 times over 36 days. No mirror failed and there was no change in the resonant frequency. In other words, the mechanical characteristics of the hinges were unchanged.
