ELSEVIER
Sensors and Actuators A 44 (1994) 1-11
Review Paper
A review of silicon microphones
P.R. Scheepera,
A.G.H. van der Donkb, W. Olthuis”, P. Bergveld”
‘Briicl and Kjm, Skodrborgvj 307, DK-2850 Namm, Denmark
bTtxm I- Holland B.k$ PO Box 43, 76OMA Almelo, Nethcria~
‘MESA Research Imtitute, Uniwmly of Twmte, PO Bar 217, 75WAE Emchede, Netherlands
Received 20 July 1993; in revised form 14 January 1994, accepted 22 February 1994
Abstract
Silicon micromachining has successfully been applied to fabricate piezoelectric, piezoresistive and capacitive microphones.
The use of silicon has allowed the fabrication of microphones with integrated electronic circuitry and the development of the
new FET microphone. The introduction of lithographic techniques has resulted in microphones with very small (1 mm’)
diaphragms and with specially shaped backplates. The application of corrugated diaphragms seems a promising future development
for silicon microphones. It is concluded from a noise consideration that the FET microphone shows a high noise level, which
is mainly due to the small sensor capacitance. From this noise consideration, it can be shown that integration of a capacitive
microphone and a preamplifier wig result in a further reduction of the noise.
1. IntroducGon
Microphones are transducers that convert acoustic
energy into electrical energy. Many transduction prin-
ciples have been developed, including the electrodyn-
amic, the piezoelectric, the piezoresistive, the capacitive
and the contact microphone (carbon microphone) [l].
These transduction principles were already known at
the end of the nineteenth century and the beginning
of the twentieth century. In 1978 a new type of mi-
crophone was introduced that was based on optical
waveguides [2].
Royer et al. presented the first microphone to be
fabricated using silicon micromachining techniques in
1983 [3]. The introduction of silicon technology allows
accurate control of the dimensions, a high degree of
miniaturization, and batch fabrication of microphones
at low cost and with good reproducibility [4]. Fur-
thermore, integration of electronic circuiby with the
microphone is possible, as in, for example, the piezo-
electric microphone of Royer et al. [3] and Kim et al.
[5] and the FET microphone (condenser microphone
with integrated field-effect transistor) of Kiihnel [6]
and Graf et al. [7].
The transduction principles that are most convenient
for realization with micromachining techniques are the
piezoelectric, piezoresistive and condenser micro-
phones.
In this paper, a review of silicon microphones is
given. The piezoelectric and piezoresistive microphones
will be discussed briefly in Sections 2 and 3. Since most
of the silicon microphones are based on the capacitive
principle, a more extensive review on this type of
microphone is given in Section 4. The effect of the
parameters on the performance of the capacitive mi-
crophones is discussed in Section 4.1. The development
of the silicon capacitive microphones is given in Section
4.2. In Section 4.3, the FET microphone is discussed
separately. The expected future developments of silicon
microphones are discussed in Section 5. Conclusions
are drawn in Section 6.
2. Piezoelectric microphones
A piezoelectric microphone consists of a thin dia-
phragm that is either provided with a piezoelectric
material or mechanically connected to a bimorph bender,
which is a cantilever beam of two layers of piezoelectric
material having opposite polarizations. The movement
of the diaphragm causes stress in the piezoelectric
material, which generates an electric voltage.
Royer et al. presented the first piezoelectric micro-
phone realized using silicon micromachining techniques
in 1983 [3]. The microphone consists of a 30 pm thick
silicon diaphragm, having a diameter of 3 mm, with a
0924-4247/!34/s07.00 63 1994 Elsevier Science S.A. Au rights reserved
rcnr rNJ,4_4747/04\“o,Cln.*
P.R. Scheeper et al. I Sensors and Actuators A 44 (1994) l-11
Al upper electrode
polysilicon
piezoresistor
metallizotion
I
Al concentric
lower electrodes
Fig. 1. Schematic cross-sectional view of the piezoelectric microphone
of Royer et al. [3].
3-5 pm ZnO layer on top, sandwiched between two
SiOz layers which contain the electrodes (see Fig. 1).
Some microphones were provided with an integrated
MOSFET preamplifier. A sensitivity of Xl-250 PV Pa-’
and a frequency response of 10 Hz to 10 MHZ, which
was flat within 5 dB, were measured.
Other authors also presented piezoelectric silicon
microphones [5,8,9]. The sensitivity of these micro-
phones was between 0.025 mV Pa-l [2,8] and 1 mV
Pa-’ [5]. A disadvantage of the piezoelectric micro-
phones is the relatively high noise level, which is between
50 dBA SPL’ [5] and 72 dBA SPL [2].
3. Piezoresistive microphones
A piezoresistive microphone consists of a diaphragm
that is provided with four piezoresistors in a Wheatstone
bridge configuration. Two resistors are placed in the
middle and two are placed at the edge of the diaphragm
for polycrystalline silicon. If the diaphragm deflects,
the strains at the middle and the edge of the diaphragm
have opposite signs, thus causing an opposite resistance
change of the piezoresistors. An advantage of the
piezoresistive microphone is the relatively low output
impedance.
Schellin and Hess presented a piezoresistive silicon-
based microphone in 1992 [lo]. The microphone dia-
phragm is 1 pm thick highly boron-doped silicon with
an area of 1 mm2. The diaphragm is provided with
250 nm thick p-type polysilicon resistors, which are
isolated from the diaphragm by a 60 nm silicon diozide
layer (see Fig. 2).
Using a supply voltage of 6 V, the microphone showed
a sensitivity of 25 PV Pa-’ and a frequency response
of 100 Hz to 5 kHz (f3 dB). The sensitivity was lower
than expected by more than a factor of 10, which was
explained by the initial (static) stress in the highly
boron-doped silicon diaphragm.
‘This is the noise level that is measured using an A-weighing filter,
in dBs relative to 2x10-’ Pa, which is the lowest sound level
detectable by the human ear. The A-weighing filter corrects for the
frequency characteristic of the human ear, thus providing a measure
of the audibility of noise.
502 +
Si -
SO*-
boron ‘doped
diophrogm
Fig. 2. Schematic cross-sectional view of the piezoresistive microphone
of Schellin and Hess [lo].
The piezoresistive coupling factor of the polysilicon,
to which the microphone sensitivity is proportional, has
been optimized by the authors. The only remaining
design factor that enhances the sensitivity of this type
of microphone is increased diaphragm area.
4. Capacitive microphones
The majority of silicon microphones are based on
the capacitive detection principle. Therefore, this type
of microphone will be treated in more detail. In Section
4.1 the principle of operation of the capacitive micro-
phone is discussed, in order to understand the effect
of the parameters of the microphone and its preamplifier
which determine the microphone performance. In Sec-
tion 4.2 an overview of capacitive microphones in silicon
is given. In Section 4.3 the recently developed capacitive
microphone with an integrated field-effect transistor is
discussed.
4.1. Principre of operation
Microphone sensitivity
The open-circuit sensitivity of a condenser micro-
phone, as shown schematically in Fig. 3(a), is considered
to consist of two components, S, and S,, i.e., the
mechanical and electrical sensitivity of the microphone,
respectively [11,12].
The mechanical sensitivity is defined as the increase
in the deflection of the microphone diaphragm, dw,
resulting from an increase in the pressure, dP, acting
on the diaphragm:
s =!!!?
m dP
(1)
Note that dw= -ds,,,, where s,,, is the thickness of the
air gap between the two plates. It is assumed that the
diaphragm moves like a rigid piston, with the average
movement of an actual microphone diaphragm. If the
microphone is provided with a circular diaphragm with
a large initial tensile stress, the mechanical sensitivity
of the piston diaphragm is equal to [13]
P.R Scheeper ef aL f Sensors and Actuators A 44 (1!794) l-11
3
c diaphragm
- spacer
- bockplote
(o) b ,,,,,,,,,,,,,, // ,,,, Jj ;e;2alizotion
microphone 1
I
1
h
(b) voltage source ! parasitic !
copocitonce
amplifier
Fig. 3. (a) Schematic cross-sectional view of a condenser microphone.
(b) The condenser microphone, connected to an external d.c. bias
voltage source, loaded by a parasitic capacitance C, a bias resistor
Rb and a preamplifier with an input capacitance Ci.
R2
s,= -
84,
(2)
where R is the radius of the diaphragm, o,, the diaphragm
stress and’h, the diaphragm thickness. In Eq. (2) it is
assumed that the compression of air in the backchamber
does not intluence the diaphragm movement.
The relation between a change in the thickness of
the air gap, &,, and the resulting change in the voltage
across the air gap, dV, is given by the electrical sensitivity
of the microphone:
se= p
acl
Note that in the case of a piston diaphragm, the electric
field strength E, in the air gap is homogeneous. For
fast diaphragm movements, the charge on the plates
of the microphone remains constant. Consequently, the
electric field strength between the plates remains con-
stant. The electrical sensitivity is then given by
S,=E.= 2
where Vb is the d.c. bias voltage of the microphone.
The electric field in the air gap can also be supplied
by built-in charge. In this case, the backplate is provided
with a thin dielectric layer, which is charged. The
charged layer is the so-called electret. The electric field
strength in the air gap is then given by
(5)
where s, and l = are the thickness and the relative
dielectric constant of the electret, u, is the charge
density and co is the dielectric constant of a vacuum.
The voltage V, is the resulting electret voltage of the
charge Us.
The quasi-static open-circuit sensitivity SW, of a
condenser microphone is defined as
S
open =
- SJ, (6)
Note that a minus sign has been introduced in Eq.
(6), because dsaO = - dw. Thus the microphone sensitivity
has a negative value. For clarity, if we compare mi-
crophone sensitivities, we will compare the absolute
values of the sensitivities.
The open-circuit microphone signal, e,, can be de-
scribed as
c, = S,SJJ
where p is the sound pressure.
(7)
When the sensitivity is measured, the microphone is
connected to a preamplifier, which acts as an impedance
converter (see Fig. 3(b)). The source follower is a
commonly used preamplifier. It has a gain H,, which
is close to unity, and an input capacitance C,. The
measured microphone sensitivity, S,,,, is equal to
where H, is the capacitive signal attenuation due to
the input capacitance of the preamplifier and the par-
asitic capacitance C,:
H,=
Gl
cm+cj+cp
(9)
where C, is the microphone capacitance.
For a source follower, the gain is given by [11,14,15]
Ha= ++
m s
where g,,, is the transconductance of the field-effect
transistor (FET) and R, is the value of the source
resistor.
Thus the measured value of the microphone sensitivity
is also determined by the gain and the input capacitance
of the preamplifier. Therefore, a more reasonable com-
parison of microphones can be made if open-circuit
sensitivities are compared. However, in that case the
microphone capacitance must also be given, so that
every author can calculate the capacitive signal atten-
uation for his specific preamplifier.
As can be seen from Eq. (4), the electrical sensitivity,
and thus the open-circuit microphone sensitivity, in-
creases if the bias voltage is increased. However, the
d.c. bias voltage cannot be increased without limit. At
a certain bias voltage, the microphone diaphragm col-
lapses to the backplate. For a piston diaphragm with
4
P.R Scheeper et al. I Sensors and Actuators A 44 (1994) 1-11
a mechanical sensitivity given by Eq. (2) the collapse
voltage is equal to [16]
(11)
Note that the collapse voltage is inversely proportional
to the square root of the mechanical sensitivity. Thus,
improving the sensitivity of a condenser microphone
by increasing the mechanical sensitivity of the diaphragm
is limited, because the collapse voltage is decreased.
If the d.c. bias voltage is always kept at a tied fraction
of the collapse voltage, it can be concluded from Eqs.
(2), (4), (6) and (11) that the open-circuit sensitivity
of a condenser microphone is proportional to the square
root of its mechanical sensitivity.
Note that in Eq. (11) it is assumed that the backplate
is rigid. For microphones with a thin backplate [17-191
that is not perfectly rigid, the collapse voltage is lower
than that predicted by Eq. (11).
Eq. (11) is an approximation that is valid for a piston
diaphragm. Warren [20] has presented more accurate
calculations for microphone diaphragms. Warren as-
sumed that the diaphragm is very thin and has a high
initial stress. The fact that the electrostatic force is
not distributed uniformly across the diaphragm was
taken into account. The collapse voltage that was nu-
merically calculated by Warren is 18% lower than the
value given by Eq. (11).
Frequency response of [I cupacitive microphone
A well-known method for calculating the frequency
response of a mechanical-acoustic system is to describe
it using an analogous electrical circuit. Current is then
analogous to volume flow and voltage is analogous to
pressure. Properties such as mass, friction and com-
pliance (the inverse of the mechanical sensitivity) are
represented by their electrical equivalents: inductance,
resistance and capacitance, respectively [21].
Fig. 4 gives the equivalent circuit of the microphone
shown in Fig. 3. The resistanceR, describes the frictional
force due to radiation of sound back into the surrounding
medium. The inductance L, describes the mass of the
air close to the diaphragm that is vibrating in phase
with the diaphragm [21]. C, and C,, represent the
compliance of the diaphragm and the backplate, re-
BACKPLATE
Fig. 4. Analogous electrical circuit of a condenser microphone with
a thin backplate, mounted on a backchamber.
spectively. Note that the thin backplate is modelled as
a thin diaphragm with a large initial stress. Therefore,
the mechanical sensitivity of the backplate is also given
by Es. (2).
The air-streaming resistance of the air gap, R,, has
a large effect on the frequency response of the mi-
crophone. The value of R, has been calculated by Skvor
[22]:
R= -&WI
with
(12)
(13)
where q7, is the viscosity of air (17.1~10-~ Pa s), n
is the number of acoustic holes per unit area (acoustic
hole density) and A is the ratio of the area of the
acoustic holes to the total backplate area (0 <A < 1).
Noise performance of a capacitive microphone
Another important property of condenser micro-
phones is the noise performance. The most significant
noise is produced in the bias resistor, the preamplifier
and the impedance of the package (housing) of the
microphone and the preamplifier [12,14,15]. A more
detailed diagram of the condenser microphone and the
source follower, extended with the noise sources, is
shown in Fig. 5. The open-circuit output voltage of the
microphone is given by e,, as described by Eq. (7).
SUPPlY
Fig. 5. (a) General model of an FET preamplifier, connected to a
capacitive microphone. (b) Small-signal circuit with noise sources.
P.R Scheqw et al. / Sensors and Achwtors A 44 (1994) 1-11
5
The noise in the bias resistor and the package gate
impedance can be modelled by the current sources i,,
and i,, respectively, and the noise of the FET by a
current source i,. The admittances in Fig. 5 can be
described in more detail as follows:
Y,=joC,
(output capacitance microphone)
Y,=jwC,
(parasitic gate-to-ground capacitance)
Yb = l/I&, + joC, (bias element)
Y,=joC,
(gate-to-source capacitance)
Ygd =jd,
(gate-to-drain capacitance)
The admittances Y, and Yd are resistive. The effect
of all noise sources on the output of the preamplifier
has been calculated by van der Donk et al. [14]. An
expression for the squared equivalent input noise density
NF was obtained, which can be considered as a noise
generator with a value of Ni V Hz-in in series with
the signal source e,. NF is given by [14]
No= (C,+C,+C,+C,+C,)*
I
G
+ 4kT( l/R, + l/R,)
02C;
(14)
where K is the flicker noise coefficient, which strongly
depends on the technology used to fabricate MOSFETs
[14], f is the frequency, k=1.38x10TU J K-’ (the
Boltzmann constant) and T is the absolute temperature.
The first term in Eq. (14) represents the contribution
of the channel noise of the FET. The second term
represents the thermal noise of the bias element and
the packaging leakage resistance. Eq. (14) is valid if
either the source or the drain is used as the output.
It can be concluded from Eq. (14) that the ratio of
noise to signal can be minimized by choosing C,, Rb
and R, as large as possible. Note that R, is not a
component, but an unwanted effect of the packaging
of the preamplifier and the microphone. The noise
caused by this resistance may be considerable [14]. The
design and fabrication of the MOSFETs can be op-
timized to reduce the channel noise [14].
4.2. Capacitive microphones in silicon
An overview of the most sign&ant dimensions, the
capacitance, the measured sensitivity, the noise level
and the frequency range of capacitive silicon micro-
phones (if mentioned), as reported by several authors,
is given in Table 1.
Hohm and Gerhard-Multhaupt [23] in 1984 presented
the first electret microphone that was based on silicon
technology (see Fig. 6). The backplate was 1 cm X 1
cm silicon and was provided with one circular acoustic
hole, with a diameter of 1 mm, which was fabricated
by sand blasting. A 2 pm thick SiO, layer was used
as electret and was charged to about -350 V. The
diaphragm was a metallized 13 pm thick Mylar foil,
with a diameter of 8 mm. A 30 pm Mylar foil was
used as a spacer, thus yielding an air gap with the
same thickness.
Polymer foil diaphragms were also applied by Spren-
kels [ll] in 1988 and by Murphy et al. [24] in 1989.
The microphone fabrication process was made more
compatible with standard thin-film technology by using
anisotropic etching with KOH for fabrication of the
backplate. In the microphone of Sprenkels [ll] the
Mylar foil was glued directly on the backplate wafer
(see Fig. 7). In the microphone of Murphy et al. [24]
the foil diaphragm was mounted on a support wafer
for the microphone assembly. Sprenkels showed that
silicon microphones with Mylar diaphragms can show
very high sensitivities of 19 mV Pa-l (25 mV Pa-’
open-circuit) [25,26].
Hohm improved the microphone fabrication process
in 1986 [27,28] by using anisotropic etching and replacing
the Mylar foil diaphragm by a 150 nm thick low pressure
chemical vapour deposited (LPCVD) silicon nitride
film. The microphone design is shown in Fig. 8. The
diaphragm and the backplate were fabricated on sep-
arate wafers. The diaphragm stress was controlled by
implantation with nitrogen ions. Owing to the smaller
size of the microphone diaphragm (0.8 mmX0.8 mm),
the open-circuit sensitivity and the capacitance de-
creased from 8.8 mV Pa-’ [23] to 4.3 mV Pa-’ [27,28]
and from 9 to 1.4 pF, respectively.
Bergqvist and Rudolf [29] showed in 1990 that silicon-
based solid-state microphones also can show a high
sensitivity. Microphones with a 2 mm x 2 mm diaphragm
showed an open-circuit sensittity of 1.4-13 mV Pa-l.
The 5-8 pm thick silicon diaphragm was fabricated
using anisotropic etching in a KOH solution and applying
an electrochemical etch-stop. The backplate was a glass
plate, which was thinned by mechanical polishing (see
Fig. 9). The backplate wafer was bonded together with
the diaphragm wafer and another silicon and glass
wafer. This sandwich of four wafers forms a condenser
microphone with a backchamber.
A disadvantage of many silicon condenser micro-
phones appeared to be the decreased sensitivity for
high frequencies due to the air-streaming resistance of
the narrow air gap. The first microphones with polymer
diaphragms were provided with relatively thick air gaps
of 20 to 95 pm [11,23,24]. The microphones with silicon
nitride and silicon diaphragms were provided with air
gaps of 2 pm [27,28] or 4 pm [29]. According to Eq.
(12), the air-streaming resistance increases considerably
if the air-gap thickness is reduced. Consequently, the
frequency response of these microphones was limited
to 2 kHz [27,28] and 4 kHz [29] for the most sensitive
types, in contrast to flat frequency responses up to 15
kHz for microphones with thicker air gaps [11,24].
![Fig. 9. Schematic moss-sectional view of the silicon condenser microphone of Bergqvist and Rudolf [29].](/figures/fig-9-schematic-moss-sectional-view-of-the-silicon-condenser-3amrwlgh.webp)
![Fig. 13. Schematic cross-sectional view of the single-wafer fabricated silicon condenser microphone of Scheeper et al. [19].](/figures/fig-13-schematic-cross-sectional-view-of-the-single-wafer-1ug6700a.webp)
