ME 7203 107

Chapter 5

Sensor Dynamics
• Thus far we were only considering static or slow varying measurements, i.e., the
reading from the sensor is recorded after it has reached a steady state value.

• This topic focuses on dynamic characteristics of a measurement systems, which con-

siders the transient behaviour of a sensor and its implication on measurement system
design.

• The completion of this topic will help you answer the following questions:

a. What happens to a measurand physical signal when it passes through a mea-

surement system?
b. How to select the bandwidth of a sensor given an application?
c. How to select a sampling speed for a sensor given an application?
ME 7203 108

5.1 Modelling Dynamic Sensor Behaviour

• The dynamic behaviour of a sensor (linear or nonlinear) can be modelled using dif-
ferential equations.

Example 5.1: Transfer function

A temperature probe with a sensitivity of α V /◦ C is used to measure the temperature
T (t) ◦ C of a liquid bath. The temperature of the probe is denoted by Ts (t). The
voltage output of the sensor Vo (t) = αTs (t). Find the transfer function VTo(s)
(s)
of the
temperature sensor. The probe mass is m, the specific heat constant of the probe
is C, the convection coefficient between probe and the liquid is h, and the area of
contact between the probe and the liquid is A.

V0(t)

Vs

T(t)
Ts
ME 7203 109

Example 5.2: Op-Amp Filters

Find the transfer functions of the following op amp circuits. Comment on what type
of filtering the circuits perform.

R2

+Vs
Vin R1
- Vo
+
-Vs
ME 7203 110

R2

C +Vs
Vin R1
- Vo
+
-Vs

R2

C2

C1 +Vs
Vin R1
- Vo
+
-Vs
ME 7203 111

• However for dynamic measurement applications, linear sensors (more specifically sen-
sors having linear time invariant (LTI) models) are highly recommended. Reasons
being:

a. The system input x(t) and output y(t) relationship can be described using a
linear ordinary differential equation.

dn y dy dx d2 x dm x
an + · · · + a 1 + a 0 y = b0 x + b1 + b2 + · · · + bm (49)
dtn dt dt dt2 dtm
b. As a result the system can be described mathematically as a transfer function
G(s).

an sn + · · · + a1 s + a0 Y (s) = b0 + b1 s + b2 s2 + · · · + bm sm X(s)

Y (s) b0 + b1 s + b2 s2 + · · · + bm sm (50)
G(s) = = 2 n
X(s) a0 + a1 s + a2 s + · · · + an s

The order of the system is the highest power of s of the denominator polynomial.
c. Input and output waveforms have the same frequency content.
d. The system can be graphically represented as a bode plot (frequency response).
e. Variety of mathematical tools from control theory are available for design and
analysis. e.g., root locus, bode diagram, transient response, Fourier transforms
etc.
ME 7203 112

5.1.1 How to identify a linear sensor?

• In the static case a sensor should have a linear calibration plot with constant slope
for it to be considered linear.

• For the dynamic case a sensor should satisfy the condition of superposition for it to
be considered as a linear sensor.

x1(t) y1(t)

LTI system
t t

y2(t)
x2(t)

LTI system t
t

ax1(t)+ bx2(t) LTI system ay1(t)+ by2(t)

• A real life sensor or system is never linear. It can only be approximated to a linear
system for a given operating range.
ME 7203 113

5.1.2 What happens to an input sine signal?

• The input sine signal can be defined using its frequency ω, magnitudes A and phases
φ0 .
x(t) = A sin(ωt + φ0 ) (51)

• For a linear sensor the output can be easily found using the sensor‘s frequency re-
sponse.

• The frequency response of a sensor includes the magnitude response M (ω), and the
phase response φ(ω). These dictate what happens to the amplitude A and phase φ0
of the signal at the output respectively. i.e.,

y(t) = M (ω)A sin(ωt + φ0 + φ(ω)) (52)

ME 7203 114

5.1.3 What happens to an input complex signal?

• Any periodic integrable complex signal x(t) can be decomposed in to a sum of pure
sine waves with different frequencies ω, magnitudes An and phases φn .
∞
X
x(t) = An sin(nωt + φn ) (53)
n=0

• The Fourier transform is the mathematical process of decomposing complex signals to

its constituent sine waves. A close physical analogy to this process is light dispersion
using a prism.

• The frequency response of a sensor dictates what happens to each of these constituent
signals (An sin(nωt + φn )) as a result of the sensor.
∞
X
y(t) = M (nω)An sin(nωt + φn + φ(nω)) (54)
n=0

• The individual frequency responses add together to form the output of the sensor y(t).
This combination is mathematically equivalent to the inverse Fourier transform.
ME 7203 115

5.1.4 Sensor dynamic models

• Zeroth order sensors : A system which instantly changes its response to an input
signal. The transfer function has a zero degree denominator polynomial.

Transfer Function Step response Sine response

G(s) X(s) = 1s x(t) = ASin(ωt)

y = Kx y(t) = K y(t) = KASin(ωt)

Y (s)
=K
X(s)

x(t)
y(t) 20 log M(ω)
y(t)
K

1 20 log K

t ω
τ
System step response System Bode Magnitude plot

y(t) ɸ(ω)
+900
KA
A

t ω

-900
System Sinosoidal response System Bode phase plot
ME 7203 116

• First order sensors : The systems response is characterized by its time constant τ .
The transfer function has a denominator polynomial with s terms with highest power
1.

Transfer Function Step response Sine response

G(s) X(s) = 1s x(t) = ASin(ωt)

−t
τ ẏ + y = Kx y(t) = K(1 − e τ ) y(t) = T.R. + M ASin(ωt + φ)
Y (s) K K
= M (ω) =  1
X(s) τs + 1
1 + (ωτ )2
2

φ(ω) = −T an−1 (ωτ )

x(t)
y(t) 20 log M(ω)
y(t)
K
63%
1 20 log K

t 1/τ
ω
τ
System step response System Bode Magnitude plot

y(t) ɸ(ω)
+900
AM(ω)
A
ϕ
t 1/τ ω

-900
System Sinosoidal response System Bode phase plot
ME 7203 117

Example 5.3: Transient response

10
The transfer function of a sensor is given as G(s) = s+10 . Draw the output waveform
of the sensor for the following input signals.
y(t)

.5 1 1.5
t(s)
y(t)

250 500 750 t(ms)

ME 7203 118

• Second order sensors : The systems response is characterized by its natural frequncy
ωn and the damping ratio ζ. The transfer function has a denominator polynomial
with s terms with highest power 2.

Transfer Function Step response Sine response

G(s) X(s) = 1s x(t) = ASin(ωt)

1 2ζ T hree solutions y(t) = T.R. + M ASin(ωt + φ)

ÿ + ẏ + y = Kx
ωn 2 ωn K
M (ω) = 
Y (s) Kωn2  2 2 
1
2 2
= 2
!
X(s) s +p 2ζωn s + ωn2  1− ω +
2ζω 
ωd := ωn 1 − ζ 2 ωn ωn
ω
!
2ζ ω
φ(ω) = −T an−1 n
1 − ( ωωn )2

x(t)
y(t) ζ=.25
ζ=0.5
20 log M(ω)
y(t)
K
ζ=.25
ζ=1

1 20 log K ζ=.5
ζ=1

t ωd ω
τ
System step response System Bode Magnitude plot

y(t) ɸ(ω)
+1800
AM(ω)
A
ϕ
t ωd ω
ζ=.25

-1800 ζ=1

System Sinosoidal response System Bode phase plot

ME 7203 119

Example 5.4: Frequency response

Find the output of a sensor with K = 1 and τ = 0.0318s for an input signal
x(t) = 5 sin(10πt + π/2).

5.2 Dynamic errors and Bandwidth

20 log M(ω)

Magnitude Error

20 log K

1/τ ω

• The amplitudes of frequencies of the input signal x(t) are amplified or attenuated at
the output of a sensor.

• The sensor should ideally produce a flat frequency response.

ME 7203 120

• Amplitude distortion occurs when some frequencies are amplified at a different sen-
sitivity than of the flat region.
• Flat frequency response is necessary for frequency reproduction. e.g. For a vibration
measurement application, a flat response ensures that an accelerometer capture all
modes of vibration and their amplitudes without distortion.
• For first order systems the amplitude distortion is close to −30% for an input signal
of frequency τ1 rad/s. This is also termed as dynamic errors.

Example 5.5: Dynamic errors

You are required to design the time constant of a temperature sensor which is mod-
elled as a first order system. The sensor is intended for applications where frequencies
from 1 − 5 Hz are expected. The sensitivity of the device K = 1 and the acceptable
dynamic error is ±2%.
ME 7203 121

Example 5.6: Frequency response

A pressure transducer has a damping ratio of ζ = 0.5, a ringing frequency of 1200 Hz,
and a DC gain of 1. Denote the input to the pressure transducer as x(t) and the
output as y(t).

1. Determine the transfer function of the sensor.

2. What is the expected maximum dynamic error of the sensor for an input
sinusoidal signal with a frequency of 500 Hz.

3. The input pressure signal x(t) to the sensor can be expressed as x(t) =
sin(200πt) + sin(1000πt). Compute the output signal y(t) of the sensor for
the above signal.
ME 7203 122
ME 7203 123

5.3 Sampling Theorem

• Consider a sinusoidal signal y(t) having a frequency of 10 Hz. The signal y(t) is sam-
pled using a sampling speed of fs . The figure below illustrates the resulting waveform
after sampling at three different sampling speeds (100Hz, 27Hz and 12Hz).

• Sampling theorem states that; in order to reconstruct the frequency content of a

measured signal, the sampling frequency fs should be more than twice the highest
frequency fm contained in the measured signal. i.e. fs should be larger than 20 Hz
for the above example.

fs ≥ 2fm (55)

• The maximum frequency that can be measured using a signal which is sampled at a
frequency of fs is defined as the Nyquist frequency fN

fs
fN = (56)
2
• All frequencies above fN appear as aliases in the measured signal. The folding dia-
gram is used to identify the aliasing frequencies.
ME 7203 124

• In order to minimize the aliasing effect, we should low pass filter the measured signal
such that signals above fN are attenuated.

• A filter used to minimize aliasing effects is termed an anti-aliasing filter.

ME 7203 125

5.4 Guidelines for selecting sensors for dynamic measurement applica-

tions
• Select a linear sensor : Dynamic modelling and analysis of linear sensors can be
conveniently performed using a transfer function. Limit application of nonlinear
sensors to slowly varying processes (Quasi static applications with dynamics in the
order of seconds).
• Select a high enough bandwidth: Select bandwidth such that the sensor can capture
the desired frequency content of the measured input signal. The dynamic error of
the sensor should meet design specifications (e.g. < 1% for 0 − 500Hz).
• Select a high enough sampling rate with anti aliasing filters: Sampling rate should
be set to 2 times the highest frequency fm we intend to measure. A filter should
attenuate frequencies higher than fm .

Example 5.7: Sampling

The dynamic response of an accelerometer can be modelled as a second order system
with a damping ratio ζ = 0.7 and a natural frequency ωn = 3000Hz. What is the
maximum dynamic error expected from the accelerometer when measuring input
signals within the frequency range of 0 to 1000 Hz. Suggest a sampling speed for
this application.
ME 7203 126

Example 5.8: Anti aliasing filter

A signal from an accelerometer is sampled at 100Hz in order to measure vibrations
of a structure. The measurement engineer is interested in measuring vibrations up
to 40Hz.
1. It is suspected that the 40Hz signal measurements are aliased with high fre-
quency signals. Find two possible signal frequencies that can appear aliased
as 40Hz.
2. To reduce this effect a first order low pass filter is used. The transfer function
1
of the filter is G(s) = τ s+1 . Find the time constant τ of the filter if the
acceptable dynamic error for the 40Hz signal is ±5%.
3. What is the attenuation of the two high frequency aliases you found in part
(a) after introducing the low pass filter.
 References
[1] Figliola, R.S. and Beasley, D., Theory and Design for Mechanical Measurements, 6th
Edition, Wiley, 2015.

[2] Statease design expert tutorials, Statease, 2018. URL: www.statease.com/docs

[3] e-Handbook of Statistical Methods, NIST/SEMATECH, 2012. URL:

http://www.itl.nist.gov/div898/handbook/

[4] Lye, L., Design and Analysis of Experiments, DOE Course notes, 2015.

[5] NI USB-6008/6009 - User Guide, National Instruments, 2015. URL: 371303n.pdf

[6] Analog Devices 335 3 axis accelerometer datasheet, Analog Devices, 2015. URL:
adxl335.pdf

[7] Temperature probes: How to choose the right temperature sensor type, Omega Inc. 2018.
URL: thermocouple-RTD.html