See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/269629228

"design and optimization of a micromachined thermal convective

accelerometer "

Conference Paper · January 2014

CITATIONS READS

0 170

3 authors, including:

Mohammad Maghsoudi Ganjeh Mohsen Hamedi

University of Texas at San Antonio University of Tehran
6 PUBLICATIONS   26 CITATIONS    91 PUBLICATIONS   621 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Multiscale Modeling of Ultrastructural Origins of Bone Fragility View project

Piezoelectric Energy Harvesting View project

All content following this page was uploaded by Mohsen Hamedi on 18 January 2015.

The user has requested enhancement of the downloaded file.

 The first international conference on MEMS and Microfabrication-ICMEMS2014
18-19 February 2014, New Technologies Research Center, Amirkabir University of Technology,
Tehran, Iran

ICMEMS2014-2382

Design and optimization of a micromachined thermal convective accelerometer

M.Maghsoudi1, M.Aryafar2, M.Hamedi3
1
University of Tehran, Mechanical Engineering Dept.; maghsoudi.mo@ut.ac.ir
2
University of Tehran, Mechanical Engineering Dept.; m.aryafar@ut.ac.ir
3
University of Tehran, Mechanical Engineering Dept.; mhamedi@ut.ac.ir

Abstract factors of thermal convective accelerometer with no

proof mass. In 2003 Mailly et al [4] did the thermal
This paper deals with design and optimization of a simulation of convective accelerometers.
micromachined thermal convective accelerometer. The This paper will present a powerful thermo-electro
acceleration induced thermal exchange due to mechanical coupled finite element model for computer
convection between hot bubble of air and sensor will simulation of the thermal accelerometer sensor and in
generate temperature gradient and consequently the spite of previous works, a kind of multi-objective
voltage signal is emerged. In this work a 2D FE-model optimization task utilizing from results of a full factorial
is provided for calculation of temperature difference of DOE approach is conducted . Here we use numerical
detectors. The most effective parameters on sensor method to build a capable model to predict the sensor
performance are identified and the sensor is optimized performance. Statistical graphs and charts are used to
considering multi objects (i.e. sensitivity and non- visualize the effect of each parameter and their
linearity) via DOE full factorial method. Novel interactions. FE model is validated by simulation of
Parametric design in COMSOL environment and sensor fabricated and characterized by Chaehoi et.al [6].
programming in MATLAB helped to perform numerous The employment of parametric design in COMSOL and
simulations needed for full factorial experiment which connecting it to MATLAB for automation of simulation
allows to profoundly understand the effect of each is introduced in this paper.
parameter and specially their integration. Coupled
thermo-electro mechanical simulation is conducted in 2. Sensor design and modeling
order to assume the power consumption of all 2.1 Sensor concept
understudy case consistent.
The optimum designed sensor has sensitivity as high as The accelerometer depicted in Figure 1 is composed of
0.48 (dT/g), non-linearity less than %10 in the range of three thin bridges suspended over a cavity (i.e.,
acceleration of [0-120] g with heater temperature of thermally isolated from the substrate) and oriented 45_
630ºK. from the cavity boundaries in order to speed-up the wet
etching process. polysilicon is used to implement
Keywords: Thermal sensor, Convective Accelerometer, resistors, for both heating (RH) and sensing (Rd1, Rd2)
MEMS, Design of Experiment
purposes. An alternative to polysilicon resistors is the
1. Introduction use of polysilicon/aluminum thermopiles which is more
efficient because doesn’t need power supply.
Accelerometers are inseparable and necessary in many The heater which is supplied by an electrical potential
fields of applications where motion control is required, creates a temperature distribution in the top and bottom
for instance in automotive industry, airplanes, cavities encompassed by the package and the etched
seismology, robotics, system automation etc. substrate. The temperature is therefore maximum
Nowadays, the ordinary piezoelectric or piezoresistive around the heater and minimum (i.e., at room
accelerometers are widely implemented in practice temperature) at the cavities boundaries. The initial
however they are very fragile resulting in the necessity distribution of the temperature in cavity is symmetrical
to apply highly sensitive sensors because of the thin and two detectors measure equal temperature.
beam and large mass in their structure [1]. Recently, a
new type of accelerometers without solid proof mass,
which operate based on thermal convection, has been
introduced [2]. Lau et al [3] studied some the key

1
 Nowadays finite element methods have found a lot of
applications in design and simulation of engineering
products. Herein we have utilized Finite Element (FE)
method to simulate and investigate the thermal
accelerometer response. The temperature distribution in
cavity is obtained via this 2D FE model conducted in
COMSOL Multiphysics non-isothermal flow (nitf)
environment. The acceleration is introduced by applying
volume force. The heat convention of the bridge and air
were separately calculated and added to model in order
to consider the temperature decrease effect and enhance
the simulation accuracy. FE Model is shown in Figure 3

Figure 1. Sensor configuration

Once acceleration is exerted in direction of the sensitive

axis (perpendicular to the bridges length) the
temperature gradient modifies due to natural convection
and a temperature difference dT is observed across the
detectors. This concept is illustrated in Figure 2.

Figure 3. FE Model of sensor

Figure 4 illustrates the temperature distribution due to

an acceleration of 10 g towards the left. The difference
of temperature on detectors is found from this model.

Figure 2. Principle of thermal accelerometer [5]

2.2 Governing equations

The thermal accelerometer is based on heat transfer and

Figure 4. 2D simulation of temperature distribution for 10g
especially on free convection heat transfer in a closed acceleration toward the left
cavity containing air. The Navier–Stokes equation (for a
steady-state flow with negligible viscous dissipation) The velocity contour of air flow in response to
acceleration toward left direction is plotted in Figure 5.
and Boussinesq approximation (which considers density
as constant with temperature except for buoyancy
forces) are the governing equation of the sensor.
𝜌(𝜗⃗. ∇
⃗⃗)𝜗⃗ = −∇
⃗⃗𝑝 − 𝜌𝛽(𝑇 − 𝑇𝐶 )𝛤⃗ + 𝜇∇2 𝜗⃗ (1)
Where 𝜗 is the flow velocity, ρ the gas density, p the
local pressure of gas, β the thermal gas expansion, T the
local gas temperature, TC the substrate temperature, μ
the dynamic viscosity, CP the specific heat and λ is the
thermal conductivity. Moreover the equation of
continuity and energy conservation are as: Figure 5. Velocity contour for accleration toward the left
⃗⃗. 𝜗⃗ = 0
∇ (2)
𝜌𝐶𝑝 𝜗⃗. ∇
⃗⃗𝑇 = 𝜆∇2 𝑇 (3) In order to verify the model, the sensor fabricated and
examined in Chaehoi et.al [6] research is simulated here.
The temperature difference of detectors for an
2.3 Finite Element Simulation
acceleration of 10 g versus wide range of heater

2 ICMEM2014, 18-19 February, 2014

temperature is calculated by 2D simulation and is programming in MATLAB helped to perform these
compared with experiments in Figure 6. Heater numerous simulations. Different design cases are
temperature in our sensor is about 630°C. As marked in supposed to consume equal electrical power. Therefore
figure 5 by dash line, the FEM and experiment results the heater temperature is varying from case to case. This
are well correlated at this temperature point. is achieved in simulation by applying electrical potential
on heater instead of imposing temperature boundary
condition.

3. Results and Discussion

The sensitivity and non-linearity are two objectives that

we are looking for their maximum and minimum
amount respectively in a proper design case. The scatter
plot of S and N.L of 243 analyzed case are depicted in
Figure 8 which reveals the profound effect of this
parameters on objectives variation. Regarding to desired
value of sensitivity and non-linearity appropriate design
Figure 6. FEM and experiment results comparison case could be selected among these 243 cases.

2.4 Optimization
Scatterplot of C8 vs C7
80
Different geometrical parameters influencing the 70
sensitivity and non-linearity of the sensor are studied 60
here. These include: cover size, cavity size and position 50

of detectors. Figure 7 presents these parameters. 40

N.L

30

20

10

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
S

Figure 8. Objectives distribution for 243 design case

The main effect of parameters on both the sensitivity

A
and non-linearity is depicted in Figure 9 and Figure 10.

Main Effects Plot for C7

Data Means

C1 C2 C3
0.4
Figure 7. key geometrical parameters
0.3

The assumed level of design variables are tabulated in 0.2

0.1
Table 1. In this case C1 = A/C5 represents the position
0.0
Mean

of the detectors and its distance from heater in the 1 2

C4
3 1 2
C5
3 1 2 3

0.4
cavity.
0.3
Table 1 design parameters levels
0.2

0.1
Parameter symbol lv.1 lv.2 lv.3 0.0
1 2 3 1 2 3

Detector position C1 0.4 0.5 0.6

Figure 9. Main effect of design parameters on mean value of the
Cover length C2 50 300 600 sensitivity

Cover width C3 300 600 900

Cavity depth C4 100 250 450
Cavity length C5 500 1100 1700

Full factorial 3K DOE approach is conducted to

optimize the sensor sensitivity, its non-linearity. The
number of full factorial simulation is 243 cases.
Parametric design in COMSOL environment and
3 ICMEM2014, 18-19 February, 2014
 Main Effects Plot for C8 Histogram of N.L
Data Means
100
C1 C2 C3
30

80
20

10

Frequency
60
0
Mean

1 2 3 1 2 3 1 2 3
C4 C5 40
30

20
20

10

0
0 0 10 20 30 40 50 60 70
1 2 3 1 2 3
N.L

Figure 10. Main effect of design parameters on mean value of the non- Figure 13. Histogram plot of the objective N.L
linearity
We put a constraint on object N.L and limit our search
In investigated range of parameters (table.1) the results
domain for 157 cases which satisfy the condition of
show that decrease of C1 and increase of C4 enhances
N.L<%10. As a result the scatter plot of distribution of
the sensitivity while don’t adversely affect the non-
objectives for each design case is as follows in Figure
linearity. Also compared with other parameters, C2 is
14. The cases with maximum sensitivity, minimum non-
not strongly effective however level 2 is its optimum
linearity and optimum case with N.L<%10 constraint
level. Parameter C3 dramatically influences both
are specified in Table 2 and Figure 15.
objectives. Middle level of parameter C5 is preferred
Scatterplot of C12 vs C11
from higher sensitivity point of view.
10
Figure 11 plots the interaction of parameters for
objective S. 8

6
N.L

Interaction Plot for C6

C12

Data Means
4
1 2 3 1 2 3 1 2 3 1 2 3

C1
0.50
1
C1 0.25
2 2
3
0.00
0.50 C2
1 0
2
C2 0.25
3 0.0 0.1 0.2 0.3 0.4 0.5
0.00
0.50 C3
SC11
1
2
C3 0.25
3
Figure 14. Objectives distribution with condition of N.L<%10
0.00
C4
0.50
1
2
Table 2 Optimum design specification
C4 0.25
3
0.00

Case C1 C2 C3 C4 C5 S N.L
C5

S max 1 1 3 3 2 0.9 47%

Figure 11. Interaction plot of design parameters
N min 3 3 1 1 1 0.016 0.04
Histogram plot of sensitivity and non-linearity are Opt. Level 1 2 2 3 2
0.48 9.5
shown in Figure 12 and Figure 13. It can be concluded Opt. Value 0.4 300 600 450 1100
that more than %75 of the examined case have
sensitivity less than 0.3(dT/g) and non-linearity less
50
than %15 in acceleration range of [0-120g]. SMax
40
Histogram of S
50
30
N.L

40
20

30
Frequency

10 Optimum design
20
0 N.L Min
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
10 S

0
Figure 15. Location of Smax , N.Lmin and optimum design
0.00 0.15 0.30 0.45 0.60 0.75 0.90
S

Figure 12. Histogram plot of the objective S

4 ICMEM2014, 18-19 February, 2014

 The dT of optimum designed sensor for acceleration 625
range of [0-120] g is depicted in Figure 16.
620

615
50 TH(°K)
610
40

605
30
dT

600
20 0 20 40 60 80 100 120
a(g)
10 Figure 18. Heater temperature as a function of acceleration

0 Conclusions
0 20 40 60 80 100 120
a(g)
A thermal convective accelerometer is studied and
Figure 16. Temperature difference of detectors versus acceleration numerically simulated. The temperature distribution and
velocity contour of hot bubble due to acceleration is
The effect of [0-100g] Off-axis acceleration on dT of
extracted through a 2-D CFD model entailing the forced
optimum sensor is investigated and Figure 17 reveals
convection equation. In the next step, important design
that applying 30g acceleration in Z direction will change
parameters and key factors of sensor configuration
dT about %10 in condition of 120g acceleration along
which influence the sensor performance are determined.
the main sensitive direction.
Full factorial 3K DOE approach is implemented to plan
the experiments and attain an optimum sensor enjoying
60
good sensitivity and low non-linearity.
dT variation
50 Parametric design and connecting COMSOL and
MATLAB made it possible to carry out 243 simulation
40 case and make a comprehensive interpretation of the
results. The optimum designed sensor has sensitivity as
dT(ºC) 30 high as 0.48 (dT/g), non-linearity less than %10 in the
20 range of acceleration of [0-120] g with heater
temperature of 630ºK.
10

0 References
0 20 40 60a(g) 80 100 120
0g 30 g 60 g 100 g
[1] W. Kuehnel, S. Sherman, “A surface micromachined silicon
Figure 17. Off-axis effect
accelerometer with on-chip detection circuitry”, Sens. Actuators
A 45 (1994) 7–17.
In FE-Model, the heater temperature is not considered
[2] X.B. Luo, Z.X. Li, Z.Y. Guo, Y.J. Yang, “Study on linearity of
consistent and the 5V applied voltage ruled by joule
micromachined convective accelerometer”, Microelectron. Eng.
heat effect determines its temperature (Figure 18). The 65 (2003) 87–101.
variation is negligible and the temperature is around [3] X B Luo, Y J Yang, F Zheng, Z X Li1 and Z Y Guo1, “An
620ºC. optimized micromachined convective accelerometer with no
proof mass”, Jorunal of Micromechanic and Microengineering
.11 (2001)504
[4] F. Mailly, A. Martinez, A. Giani, F. Pascal-Delannoy, A. Boyer,
“Design of a micromachined thermal accelerometer: thermal
simulation and experimental results”, Microelectronics Journal.
34 (2003), 275-280
[5] A. Garraud, A. Giani, P. Combette, B. Charlot, and M. Richard,
“A dual axis CMOS micromachined convective thermal
accelerometer,” Sensors Actuators Phys., vol. 170, no. 1–2, pp.
44–50, Nov. 2011.
[6] A. Chaehoi, F. Mailly, L. Latorre, and P. Nouet, “Experimental
and finite-element study of convective accelerometer on
CMOS,” Sensors Actuators Phys., vol. 132, no. 1, pp. 78–84,
Nov. 2006.

5 ICMEM2014, 18-19 February, 2014

View publication stats