IInertial Sensor

Chapter 4
Steady linear motion is Linear acceleration Steady rotation is
not detected is detected detected
4.2.2 Sensor Principles and Resonance
– A seismic mass, which generates a Newton force when
accelerated.
– A spring with spring constant k, which transduces this
force in a measurable extension. In this way,
measurement of acceleration is reduced to the
measurement of elastic deformation, which is known
from force sensors.
– We have to measure the elastic deformation by
measuring the movement amplitude of the seismic
mass relative to the accelerometer x(t).
– A spring mass system has a time behavior of degree 2,
which means it can overshoot. To prevent this, we need
a damping b.
 d2 x dx iωt iωt
m + b + k x = F ex t e
= maexte dt2 dt
(1) (2) (3)
(4)

Using the ansatz x = x 0 e i ω t e i ϕ , we can find the amplitude:

dx d x
2
2
= iωx and = −ω
x dt dt2
 For the amplitude
we find:

𝑎𝑒𝑥𝑡 𝑚
𝑥 𝑜=
√𝑚 (𝑤 2 2
𝑜 −𝑤 ) +𝑏 𝑤
2 2 2
𝑏=2 √ 𝑚𝑘
or Q = 0.5, using the quality factor of a resonator

Damping
System answer

Low

Critical

High
Active
control
 Accelerometer Electronics:

– Like every spring, also the spring of the accelerometer will

show non- linearity at higher movement amplitude.
– The sensor can be faster. When we measure amplitude, after a
step in signal, we have to wait until the spring mass system finds
a new place, as shown in Figure .
– The active control starts as soon as the mass wants to start
moving. Regulating back to zero is much faster than waiting for
a new equilibrium.
 Resolution and Noise

 How can we estimate the noise of an accelerometer?

 Nyquist resistor noise cannot be applied, there is no resistor.

 But the basic idea of noise due to thermal movement still

holds.

 we find a force proportional to the velocity and the damping b of

the resonator.

F N = bvN
Force and velocity result in power4
2
1 𝐹𝑁
𝑃 𝑁= 𝐹 𝑁 𝑣 𝑁= P N = k B Tf.
4 4𝑏
F 2
= 4k B bT ∆ f
 Sensor Idea and Coriolis Force

y = v t a n t = Ωrt
The radius at time t is r = v r a d t, therefore

y = v rad Ωt 2 .
The rotating observer explains this with the Coriolis acceleration a C acting
on the mass:
y = v r a d Ωt 2 → y˙ = 2v r a d Ωt → y¨ = 2v r a d Ω = a C

The Coriolis force F C is the acceleration multiplied by the mass:

F C = ma C = 2mv r a d Ω

To allow for movements in all directions, we have to write a vector equation

−→ →−
F C = 2m→−v × Ω

(4.20)
How can we make a micro sensor for angular rate? We
need three movements in three perpendicular axes:
 The external rotation Ω we want to measure.
 The linear movement v (vrad). We can realize this with a spring mass resonator, but
we have to drive this motion actively. This motion is called the primary mode.
 Coriolis force is going to drive a secondary mode, which is perpendicular to 1 and 2.

Now we will analyze the double function of comb drives as actuators and
as sensors.
First, we discuss an electrostatic actuator at the instance of a normal plate capacity
(Figure 4.15).
When a capacity C is loaded to a voltage U, the stored charge Q is

Q = CU

(4.21)

And the stored energy E is 2

1 2 1 𝑄
𝐸= 𝐶 𝑈 =
2 2 𝐶

Conservation of energy tells us that

F d x + dE = 0
Spring
for x–y
Move
ment
Mass

Sensor
moving
finger

Sensor
fixed
finger

Actuat
or

Figure 4.14 Schematic of an angular rate sensor.

A:
Comb drive actuator.
With respect to just two
capacitor plates, the comb drive
actuator allows longer path and
it has a linear characteristics.
Movement is in–out

B:
Comb drive displacement sensor
Movement is up–down
To realize a differential capacity,
for each moving finger, we need
two fixed fingers on different
capacities.
Needs two separate layers of
interconnects to allow the
crossing of the lines.
For the plate capacity

d: distance between the fingers

h: height of the fingers = thickness of the
device layer
n: number of gaps
4.3.5 Movement and Noise
To analyze the motion of a gyro, we write the Coriolis force. The
primary motion is denoted by x, and the secondary motion by y.
Primary movement is x = x 0 sinω 0 t.
Assuming a driving voltage of 2 V, a resonance frequency of 10 kHz
and a quality factor of 1000, we can realize a primary amplitude of x 0
= 5µm.
This shows us that in the case of the gyro, we must have a high
quality factor, i.e. a low damping. Critical damping would stall the
primary motion.
The secondary motion y is driven by the Coriolis acceleration

a C = 2x˙Ω = 2Ωω0x0cosωt = aC0cosωt

a C 0 = 2Ωω0x0
4.3.6 Inertial Measurement Units (IMU) and Sensor Fusion
For inertial navigation, we need six degrees of freedom: three axes of
acceleration and three axes of rotation. By integration, the speed vector,
the orientation and the location in space can be found. Small hybrid IMUs
(inertial measurement unit), which integrate the sensors and the electronics,
are commercially available.

Figure 4.21 Typical noise spectrum of a measurement system. At low frequencies, 1/f noise
is dominant, and at high frequencies, thermal (Nyquist) noise is dominant. The 1/f corner is
typically at 10 Hz . . . 100 Hz.

16
Per Bak: How nature works. Copernicus ed. 1996.
Figure 4.22 Measured noise spectrum of voltage measurement using an ADC
converter17. The signals at 50, 100 and 150 Hz stem from the electric power grid. 1/f
noise is clearly visible, the 1/f corner is at 140 Hz. Figure by Paul Meilahn.

– The gyro is good in detecting rotations, but it has no reference to the true
north, and due to integrating baseline error, it will lose direction after some
hours.
– The magnetometer sees the earth magnetic field and, in this way, has
a long-term reference. On the other hand, it can be irritated by metal
objects or electric currents.
ANY Questions?