Case Study 5 Capacitive Accelerometer

Download as pdf or txt
Download as pdf or txt
You are on page 1of 4

Case Study 5

Capacitive accelerometer

Figure 1: Top view of the accelerometer ADXL150 developed by Analog Devices


(http://microblog.routed.net/2007/04/06/ic-friday-analog-devices-adxl150/).

In this case study, you will investigate a capacitive accelerometer similar to the ADXL150
developed by Analog Devices (Fig. 1). Schematics are given in Fig. 2(a). A shuttle mass is
attached between two folded springs. It is free to move along the axis joining these springs. All
the moving parts of the ADXL150 are made of H = 2µm-thick polysilicon (density 2400kg/m3 ,
Young’s modulus E = 160GPa). Structures are released by etching a sacrificial oxide of thickness
Hs = 1.5µm. The sensor comprises N = 54 cells. Each cell is a differential capacitor that consists
in one electrode (fixed to the shuttle) that moves in between two other electrodes (fixed to the
underlying substrate) - cf. figure 2(b). The voltage is imposed to V1 and V2 for both fixed electrodes,
and V3 for the moving electrode. The electrode length and width are Le = 100µm and We = 4µm,
respectively. Each moving electrode is separated from the fixed electrodes by a nominal air gap
d = 1.5µm of permittivity ε ' 9×10−12 F/m. When the shuttle is accelerated, the middle electrode
of each cell moves a distance x towards a fixed electrode.

Warm-up
Find C1 and C2 as functions of x (fig. 2b). Write the Taylor series of these expressions around
x = 0 (keep terms up to x3 included), and use this approximation here after.

1
(a) (b)

Figure 2: (a) Schematics of the accelerometer ADXL150 (Senturia, Microsystem Design, p.514).
(b) Schematic zoom on one cell (one electrode moving between two fixed electrodes).

F/2"
L"
x"

Figure 3: One of the two spring systems of the ADXL150. (left) Undeformed. (right) Deformed.
Here the red hatched parts are much stiffer than the blue ones, and their deformation can be
considered as negligible.

2
Mechanical response and noise
The mass of the shuttle (+ moving electrodes) is m = 2.4 × 10−10 kg. The central part has a length
Ls = 500µm and a width Ws = 50µm. The spring mechanism consists of two identical sets of beams
of length L = 100µm and width W = 2.5µm distributed on both sides of the shuttle (Fig. 3).

1. Calculate the mechanical stiffness k = F/x of the accelerometer, as a function of the beam
dimensions. [Hint: consider first the bending of a cantilever beam of length L in response to
both a force and a moment at its tip.]

2. Signals V1 and V2 are chosen as sinusoids of amplitude VS = 2.5 V. V3 is connected to the


ground. These voltages generate an additional electrostatic force on the shuttle that partially
balances the elastic force (cf. electrostatic actuator). This results in a decreased effective
stiffness ke of the accelerometer. Calculate ke in the limit of displacements x much smaller
than the pull-in position. [Hint: calculate the second derivative of the electrostatic energy]
Deduce the resonant angular frequency ω0 of the shuttle.

3. What is the maximum acceleration that can be experienced without going beyond 50% of the
pull-in position?

4. Calculate the damping associated to the Couette flow between the shuttle and the underlying
substrate. Then calculate the damping from the squeezed-film in between the electrodes.
Estimate the quality factor Q resulting from fluid damping (the air viscosity is µa ' 18 ×
10−6 Pa.s). The measured quality factor is Q = 5; compare this value to your estimation and
conclude.

5. Fluid damping is usually associated to a thermal noise, as well as any other dissipative
process (e.g. resistors) in thermal equilibrium. This white noise generates a random force Fn
of variance Z +∞
¯2
F = Sn (f )df (1)
n
0
where the spectral density Sn (f ) is

Sn (f ) = 4kB T b, (2)

and the thermal energy kB T ' 4 × 10−21 J at ambient temperature. Suppose that the ac-
celerometer only picks up frequencies in a bandwidth ∆f = 1kHz. Calculate the noise (root
mean square) on the acceleration, based on the experimental quality factor Q = 5.

6. Estimate the squeezed-film damping of a motion perpendicular to the plane of the shuttle
(i.e. along H). Conclude about the intrinsic cross-axis sensitivity of the device.

Signal processing of the capacitive accelerometer


1. Find i3 (defined in Fig.4) as a function of V1 , V2 and V3 .

2. If the operational amplifier is ideal, how much is V3 ?

3
C3" i4"
V1"
i5"
N.C1(x)" R3"
i1" i3" V3"
+"
i2"
???"
N.C2(x)" +" V4" V o"

V2"

Figure 4: Electrical interface of the differential capacitors.

3. We would like the sensor to be as linear (according to x) and sensitive as possible. For a
given V1 , how would you choose V2 to achieve this goal ? Once this choice made, rewrite i3
as a function of V1 , x and their time derivatives.

4. If V1 was chosen independent of time, what physical quantity would the sensor measure ?

5. We want the sensor to measure the position x, and nothing else. For this reason, we choose
V1 (t) = VS sin(ωS t). If you know that the signal x(t) contains frequencies up to ωx , how
would you choose ωS ? Once this choice made, rewrite i3 as a function of V1 , x and their
time derivatives.

6. Calculate V4 as a function of both x and V1 in the limit where R3 C3 ωS  1.

7. Sketch V4 (t) for a low-frequency x(t). The process by which x(t) is turned into V4 (t) is well-
known in signal processing. Could you find the name of this process ? (hint: think about
radio broadcasting).

8. In order to have an output voltage Vo proportional to x, what kind of electrical circuit would
you suggest between V4 and Vo ? In the following, we will assume that this additional circuit
has a gain equal to unity, so Vo has the same magnitude as V4 .

9. The input voltage is VS = 2.5V, and the capacitance C3 is 2.4 fF. Calculate the displacement
sensitivity Vo /x. The resonant angular frequency of the structure is 2×105 rad/s. Deduce the
acceleration sensitivity Vo /a, where a is the acceleration of the shuttle. Express it in mV/g
(millivolts per g). What is the maximum acceleration that can be detected if the amplifier
delivers a maximum output of 5V ?

You might also like