ACCERATION/VIBRATION
Table of Contents

Building a
Tiny Accelerometer
to Detect Very Small Signals

A new manufacturing process that permits integration of sensor and signal conditioning electronics on a single IC die can help accelerometers detect even smaller signals.

Harvey Weinberg, Analog Devices Inc.

Surface micromachining permits the moving acceleration sensor element and all the signal conditioning electronics to be integrated on the same IC die. The single-die approach yields small, low-cost sensors; tight integration of the sensor and signal conditioning is the key to high performance. The ADXL202E accelerometer, an inexpensive, surface micromachined 50 mm3 accelerometer, is capable of measuring both static (gravitational force) and dynamic (vibration and shock) accelerations.

Sensor Structure
The sensing element forms a differential capacitor whose output is proportional to acceleration. Because device performance is so dependent on sensor design, you need to understand the basic sensor structure.

The entire IC (mechanical and electronic elements) is built in a standard BiMOS IC process. The main difference between building a surface micromachined accelerometer and a conventional BiMOS op-amp is the number of steps performed in the manufacturing process.

figure
Figure 1. A cross-sectional view of a MEMS die shows the major process steps in building the suspended mechanical structure of the ADXL202.

Figure 1 shows the basic process steps for building the freestanding mechanical elements on the ADXL202.

figure
Figure 2. In the ADXL202 beam structure, note the four pairs of orthogonal serpentine springs at the corners of the proof mass. The entire mechanical structure is suspended from the four points where the springs meet.
The beam structure is shown in Figure 2. The proof mass (center portion of the beam) is attached to four pairs of serpentine polysilicon springs affixed to the substrate by four anchor points. It is free to move in the X and Y directions under the influence of static or dynamic acceleration. The proof mass has movable fingers extending radially in all four directions. These are interdigitated with the stationary fingers.

Look closely at the geometry of each set of fingers in Figure 3. The differential capacitance of each is proportional to the overlapping area between the fixed outer plates and the moving finger, and the displacement of the moving finger. Because these are very small capacitors, you need the largest practical differential capacitance to reduce noise and increase resolution.

figure
Figure 3. These are some of the key mechanical dimensions of the differential capacitors used to measure the deflection of the ADXL202 beam.

The finger height (3 microns for the example product) is fixed by process technology, while the overlap (125 microns here) is adjustable to some extent. Longer fingers are not desirable, however, because they are harder to manufacture and result in larger beam areas, which in turn translates to more expensive parts.

The movement of the beam is controlled by polysilicon springs, and obeys the basic laws of physics:

F = K • X (1)
F = M • A (2)

Therefore,

X = (M • A)/K (3)

where:

F= force

K= spring constant
X= displacement
= mass
A= acceleration

The only two parameters under your control are the spring constant and the mass. Reducing the spring constant seems like an easy way to improve beam sensitivity, but, as usual, nothing comes for free. The resonant frequency of the beam is proportional to the spring constant:

resonant frequency = K / (4M2 (4)

and the bandwidth of the accelerometer must not extend to the resonant frequency. In addition, higher spring constants make for more rugged beams that can survive greater shocks. So if the ideal is to keep the spring constant as high as possible, the only parameter left to change is the mass.

Adding mass normally implies a larger sensor area, which results in more expensive parts because the only way to add mass is to make the beam larger. The ADXL202’s beam structure, however, interleaves the X and Y axis beams, resulting in a reduction of the overall sensor area with added beam mass enhancing the unit’s resolution.

Signal Conditioning
In the signal conditioning scheme represented in Figure 4, only a single channel is shown for clarity. The fixed outer plates are driven with square waves that are 180° out of phase.

figure
Figure 4. A synchronous modulation/demodulation scheme is used in the ADXL202 circuit to convert differential capacitance into voltage.

When the movable fingers (and hence the beam) are centered between the fixed outer plates, both sides of the differential capacitor have equal capacitance and the AC voltage on the beam is zero. If the beam is displaced by an applied acceleration, however, the differential capacitance becomes unbalanced, resulting in an AC voltage with an amplitude proportional to the displacement of the beam. This AC voltage is amplified and then demodulated by a synchronous demodulator.

Figure 5 shows the overall circuit architecture.

figure
Figure 5. A block diagram of the overall ADXL202 circuit architecture illustrates the way the Xfilt and Yfilt pins provide the acceleration output to the user. The capacitors at these pins, along with the the internal 32 k ohm resistors, determine the bandwidth of the accelerometer.

Notice that the output of the demodulator drives the duty cycle modulator through a 32 k resistor. A pin is available on each channel to allow the user to set the bandwidth by adding two external capacitors (one per channel), creating a simple first order RC low-pass filter. The filtered signal is converted to a pulse width modulation (PWM) signal by the duty cycle modulator. The period of the PWM output may be set from 0.5 to 10 ms by a single resistor.

Performance and Applications
Because the signal conditioning is so tightly coupled to the sensing element, extremely small changes in differential capacitance and movements of the beam can be measured. The capacitance resolution of the accelerometer at low bandwidths is as low as 20 zeptofarads (20 3 10–21 F). In practice, this results in resolution of inclination of better than ±1° of tilt at bandwidths up to 50 Hz. At very low bandwidths, resolution of fractions of a degree of tilt are possible. The high-resolution (~14-bit) duty cycle modulator allows users to take advantage of the accelerometer’s capabilities in systems where a precision A/D converter would be too costly.

In fact, the ADXL202E’s combination of low cost and high performance acceleration/tilt sensing capabilities has opened the door to several nontraditional accelerometer applications. The unit is currently in use in applications such as car alarms (for jacking or towing sensing), pen-based handwriting recognition systems, portable electronic games, and wearable sports equipment.


Harvey Weinberg is an Applications Engineer, Analog Devices Inc., Micromachined Products Group, 21 Osborn St., Cambridge, MA 02139; 617-761-7644, fax 617-761-7607, harvey.weinberg@analog.com.

Home | About Us | Site Index | Privacy Policy
© 2006 Questex Media Group, Inc.. All rights reserved.
Reproduction in whole or in part is prohibited.
Please send any technical comments or questions to our webmaster.