Electronics for MEMS: design issues and tradeoffs

While many focus on mechanical aspects of MEMS, designers are coming to realize that ASICs are key to MEMS success

BY AMR HAFEZ
Si-Ware Systems
Cairo, Eqypt
http://www.si-ware.com

Foreword

by Roger Grace
Roger Grace Associates
Naples, FL
http://www.rgrace.com
This is the second of a series of articles related to MEMS-based systems solutions; the first appeared in the February issue of Electronic Products. It continues to explore the themes discussed in the previous article, where I presented several case studies on these such solutions. Since writing that article, I’ve had the chance to attend the Photonics West Conference in San Francisco, the Pittcon Conference in Orlando, and most recently, the Smart Systems Integration Conference in Como, Italy, where I’ve made presentations and conducted market research on this subject. From the many discussions I’ve had at these events, it is very clear that this concept of a systems approach to MEMS-based solutions is being widely accepted in all areas of application, including the portable instrumentation and mobile phone and game sectors.

I hope you’ll want to continue to follow this series to obtain better insight and appreciation of how these systems are created and applied. In the event you wish to know even more on the subject of MEMS-based systems solutions, there are a number of upcoming events at which I will speak and, in some cases, have created all-day sessions to explore the topic. The events include:

MEMS Industry Group’s annual METRIC event , May 18 and19, San Jose, CA (www.memsmetric.com). There will be a panel discussion on MEMS integration in which I will participates on May 18.Sensors Expo June 6, 7, and 8, Rosemount, IL (www.sensorsexpo.com). The program for the June 6 all-day MEMS-Based Systems Solutions seminar I organized has been finalized. I will chair this event and make a presentation.MEPTEC’s 8th. Annual MEMS Symposium on MEMS and IC Systems Integration, May 20, San Jose, CA. (www.meptec8thannualm.html) I will make a presentation at the event.Microtech 2010, June 21 through 24, Anaheim, CA. The program for the June 24 all-day MEMS-Based Systems Solutions Special Program (www.techconnectworld.com//Microtech2010/program/RGrace.html) that I organized and will chair has been finalized, and I will make a presentation.

MEMS developers are increasingly coming to the realization that electronics play a fundamental role in any microelectromechanical system, and are a key element in realizing many benefits of MEMS, such as cost reduction and miniaturization. A good electronics design will showcase the real performance of a MEMS part, while a mediocre electronics design will limit the performance or cause unacceptable behaviors.

In this article, we touch upon some of the design considerations for an electronic system that interfaces with a MEMS device. We illustrate our conclusions with a real life example of a MEMS-based FTIR spectroscopy system.

Accurate modelling

First of all, it is important to emphasize that a successful electronic interface requires an accurate model for the MEMS part. Fundamental parameters of the device, such as resonance frequency, quality factor, and capacitance/position curves, have to be established. Tolerances should also be reliably estimated. Furthermore, unwanted behaviors such as parasitic resonance modes or harmful stiction have to be accurately modeled.

Three-D finite element simulators are usually necessary to establish the accurate MEMS model. This behavior then has to be converted into a form that is compatible with ASIC-design CAD tools.

The ASIC’s key roles

The main element that interfaces with the MEMS device is usually an ASIC, which performs many functions, the foremost of which is signal conditioning. This block usually needs a large amount of gain, highlighting a basic tradeoff in MEMS design: die size and signal-to-noise ratio. For example, in optical MEMS, the miniature MEMS die leads to optical structures with small spot sizes, leading in turn to a weak optical signal coming out. The small signal being processed requires a large gain and also a very low noise level to maintain a good SNR. This also leads to high resolution ADC.

In many cases, for the MEMS part to do its function, it needs to be actuated. One example is the actuation of comb drivers that move optical mirrors. The mechanism for actuating a MEMS part entails many design issues and tradeoffs.

The initial decision is usually whether the actuator will be operated in its fundamental resonance mode, or in a damped controlled motion profile. In many applications, the resonance motion provides many benefits, such as extended travel range and more immunity to stiction. It also achieves the fastest possible speed. This leads, however, to a wide system bandwidth that may limit system noise performance and/or lead to onerous conditions on ADC and signal processing speed.

Another important design parameter that relates to the actuator is the required actuation voltage. Keeping the actuation voltage low makes the system more compatible with IC technology, enabling more integration and potentially better performance, while a high actuation voltage may limit technology selection options. The tradeoff here is with MEMS die size (using a large actuator, hence reducing actuation voltage needed) and/or travel range (reducing comb finger spacing which may cause stiction at high travel ranges).

Other critical considerations

Another fundamental function of the ASIC is usually capacitive sensing, necessary for positioning or for force-feedback loops used in accelerometers and gyros. This function presents a host of issues and tradeoffs.

The first issue is the interaction with the actuation, and whether the two functions will be done on the same ports or separate ones. Using the same ports, with time or frequency division multiplexing, is beneficial in reducing the maximum actuation voltage necessary and increasing the sensed capacitance. However, this also leads to potentially harmful interaction between the sense and actuation circuitry. In addition, a very low-noise capacitance-to-voltage converter is essential to perform this function.

Discrete-time and continuous-time implementations both have their merits and drawbacks and the selection between the two should be made in conjunction with the architecture of the ADC used and whether the C/V is part of the conversion loop or not.

Digital signal processing is also a critical part of the system. Integrating the DSP on the ASIC yields an attractive, self-contained solution that can be easily integrated in a larger system. However, this imposes more restriction on the ASIC technology selection and may result in harmful interference between the digital part and the sensitive analog front end. Another alternative is to have a multiple ASIC solution, each optimized to its specific function, or integrate the DSP in a larger SoC.

The process and temperature variations as well as drifts of the MEMS sensors usually require temperature sensors and calibration circuits within the ASIC. It may also require the inclusion of OTPs on the ASIC, another factor in technology selection.

Finally, packaging is an important issue in MEMS-based system, both in terms of packaging of the MEMS part or the integration of the sensor and ASIC in one package. MEMS packaging under vacuum leads to very high mechanical quality factors, a beneficial aspect in resonators, but potentially problematic in system that adopt damped motion. Multichip packaging of MEMS and ASIC (Fig. 1) is very beneficial in reducing parasitics that directly impact the capacitance sensor as well as reducing harmful interference from the rest of the system. This however may lead to requiring wafer-level testing.

Fig. 1. Multichip packaging of MEMS devices can reduce parasitics and external interference.

A system example

A system that comprehensively illustrates the above mentioned design issues and tradeoffs is a MEMS FTIR spectrometer system (see Fig. 2) that uses a general-purpose MEMS-interface CMOS ASIC developed by Si-Ware Systems (.si-ware.com).

Fig. 2. As seen in this diagram of a optical-MEMS FTIR spectrometer system, the general-purpose MEMS-interface CMOS ASIC plays a crucial role.

The system consists of a photodetector, a fixed mirror, and a moving mirror. With the motion of the mirror, the photodetector captures the optical interference pattern. Spectrum analysis of the pattern can identify the light wavelength and the spectral print of any material in the light path. The highly integrated ASIC includes a low-noise signal conditioning path that amplifies the signal, removes the dc offsets and provides necessary anti-aliasing filtering. The signal conditioning is done in a highly linear manner to reduce any spurious tones in the final output spectrum.

The ASIC also includes a generic dual-channel actuation and capacitive sensing paths. The actuation path consists of a DAC that can support any arbitrary actuation profile. The DAC is of very high resolution in order to reduce actuation noise and enjoys a high spurious-free dynamic range to make sure that unwanted resonance modes are not excited. The ‘positioning’ path consists of a very low noise capacitance to voltage converter with all the supporting circuitry such as carrier generation, amplification and filtering. The C/V has a wide range of gain and d.c. offset removal to support wide capacitance ranges superimposed on various fixed capacitances. Very low noise levels are necessary for this application as the mirror position inaccuracy directly impacts the system SNR. The ASIC exhibits a very low voltage and current noise levels that allow for resolutions in excess of 18 bits. The ASIC also features an on-chip digital-output temperature sensor, a capacitance calibration circuitry as well as a precision reference generator and clock generation circuit for the external ADC. For the next generation of this ASIC, Si-Ware plans to integrate two high resolution ADCs. ■

Source:

http://www2.electronicproducts.com/Electronics_for_MEMS_design_issues_and_tradeoffs-article-farc_grace_apr2010-html.aspx