Silicon Micromachined Microphones


Project Background:

Microphones are widely employed in a variety of consumer products and specialty instruments such as telephone sets, tape-recorders, video cameras, speech amplifiers and hearing aids. Currently, microphone sales world wide represent a 400 million dollar consumer market which is expected to hold its steady growth rate for many years to come. Most of the microphones available on the market are condenser type microphones which are based on the principle of a variable capacitance, where one electrode of the capacitor is on a flexible diaphragm and moves in response to an acoustic signal. These transducers are either self-biased (also known as electret microphones) or externally biased. According to the equation Q=CV for a capacitor, a change of capacitance C will result in a change in voltage V if the charge Q of the capacitor is kept constant. In this way, an acoustic signal (pressure wave) is converted to change of capacitance via the deflection of the diaphragm which in turn is transferred to an electrical signal. Currently, all of the commercially available microphones are fabricated from separate metal parts and polymer foils and much of the fabrication process is done by hand. Even in a highly automated production-line, the fabrication is still piece-wise. As a result, these microphones suffer severely from lack of reproducibility and performance degradation induced by variations in mechanical assembly.

Motivated by the need of an alternative type of transducer technology which can address the problems outlined above, AMC decided to take a new approach and develop a Silicon Micromachined Microphone (SMM) using the newly emerged silicon micromachining technology. Silicon micromachining technology provides the opportunity for batch fabrication of micro-electro-mechanical devices such as microphones. Standard integrated circuit (IC) processing technology utilized in silicon micromachining allows excellent process control, and therefore offers the potential of improved reproducibility and high-degree of miniaturization and integration which will help overcome many of the shortcomings presently experienced by conventional microphone technologies.

SMM Technology

A. Capacitive microphone

The technology developed at AMC for fabricating capacitive silicon micromachined microphone or SMM utilizes both surface and bulk micromachining techniques. The microphone design is illustrated schematically in Figure 1. The support structure of the device is made of silicon wafer, and the acoustic window is shaped with bulk silicon micromachining process. The two parallel capacitor plates are made of metalized silicon nitride films, and the cavity is created using sacrificial etching technique. The back plate is perforated with square holes to reduce acoustic damping and to provide access to the sacrificial layer to allow its removal. The diaphragm material is low stress silicon nitride film, typically 0.5 m m in thickness, with a tensile stress of 100 MPa.. The back plate, which is also made of silicon nitride, has higher tensile stress (by a factor of 2) and larger thickness of 1.5 m m, in order to achieve the desired mechanical stiffness. Both plates are metalized with aluminum. These aluminum layers are patterned with standard IC photolithography steps and bonding pads are formed for connections to the outside world. A sacrificial material is etched away to form the air gap between the two plates. The diaphragm and back plate are about 2mm by 2mm square, and the air-gap is typically 3-4 m m in height. The overall die dimension of the current design is approximately 3.5mm by 3.5mm by 0.4mm.

B Piezoelectric Microphone

An alternative approach to the capacitive transducer is the use of a piezoelectric material to convert an acoustic signal to an electrical one. This technology was investigated by AMC in parallel with the capacitive microphone development, however development was concentrated on the capacitive devices due to excellent initial results. Several possible advantages of the piezoelectric transducer including ease of processing and a robust design have maintained AMC's interest in this technology.

The piezoelectric transducer depends on a characteristic of the crystal structure of certain materials. These structures exhibit a separation of charge across the crystal when it is deformed by an external force. A thin layer of piezoelectric material on a microphone diaphragm can then be used to generate a charge or voltage signal when the diaphragm deflects and the piezoelectric is mechanically stressed.

The microphone structure is similar to the capacitive device in its use of a low stress PECVD silicon nitride diaphragm. Sputtered aluminum electrodes are patterned to optimize the output voltage by accounting for the stress distribution in the diaphragm. Sputtered zinc oxide or aluminum nitride is used as a piezoelectric layer. A cross section of the structure is shown in Figure 2. Advantages of this design are the possibility of encapsulating the transducer to eliminate performance variations due to humidity, the absence of bias circuitry, and simpler processing compared with the capacitive design. SMM Performance

I) Sensitivity and frequency response

A. Capacitive microphone

The microphone design has gone through a number of iterations since the fabrication of the first batch of working devices. The most notable changes have come from variations of the diaphragm and air-gap thicknesses. Numerous devices have been tested and evaluated. The best performing microphones came from the most recent design variation that has a nitride diaphragm thickness of 0.5 m m and an air-gap thickness of 3.0 m m. This set of devices has an average sensitivity of 3 mV/Pa or -50 dB (referred to 1V/Pa) at 1 kHz and 6V bias, or about 60% of the maximum allowable bias. The signal to noise ratio was about 30 dB and frequency response extends from 100 Hz to 10 kHz (± 5 dB) as shown in Figure 3. The corresponding open circuit sensitivity (zero loading capacitance) is about 6 mV/Pa, comparable with values reported in literature for micromachined microphones of similar dimensions. In terms of commercial applications, both the sensitivity and frequency response of the microphone are considered adequate in meeting the specifications of cartridges used in telephones, hearing aids and a variety of other consumer products. Table 1 summarizes in detail the microphone parameters and performance of some of the early designs along with the last design for comparison.

B. Piezoelectric microphone

The design for this microphone has undergone less development than the capacitive design due to inferior sensitivity and frequency response of the initial design. A typical sensitivity is 0.6 mV/Pa or -65 dB (re 1V/Pa) at 1kHz with a signal to noise ratio of 20 dB and frequency response (± 5 dB) from 300 Hz to 4 kHz. These results are competitive with values reported for microphones with similar design. With further process development using advanced piezoelectric materials and an increased resonant frequency significantly improved performance is anticipated.

II) Noise, power dissipation, and low voltage operation

Both noise level and power dissipation for a capacitive microphone will be determined almost entirely by the pre-amplifier. It should be possible to design a single transistor amplifier based on a source follower. An analysis of noise in microphone pre-amplifiers was presented by van der Donk et.al. in Sensors and Actuators A 25-27 (1991), pp. 515-520. The conclusions were that there is very little difference between using a JFET and a p-channel MOSFET. Thermal noise due to the frequency dependent leakage resistance of the package was found to dominate. If the pre-amplifier is integrated with the microphone this is eliminated and the A-weighted equivalent input noise level should be about 5 m V for a MOSFET. In this case the noise from the bias element dominates.

For a single transistor source follower pre-amplifier, power dissipation will ideally be determined by the source resistor. The resistor value will be determined by the current requirements for the next stage of electronics. For operation at 1.2 V with a resistor Rs = 120 kW the power dissipated will be about 12 m W at a current of 10 m A.

Sensitivity for this type of microphone is theoretically linear with bias voltage. With the current geometry, which gives about -50 dB re 1V/Pa at 6 V bias, operation at 1.2 V would reduce the sensitivity by 14 dB, and operation at 2.4 V would reduce it by 8 dB. However, reducing the operating voltage means a smaller gap can be used without danger of the capacitor plates pulling together. If the design is optimized for low voltage operation by decreasing the air gap, the sensitivity loss in dropping from 6 V bias to 1.2 V bias is only 4.6 dB and in dropping to 2.4 V bias is only 2.6 dB. By decreasing the diaphragm compliance, a device that operates at 1.2 V with no loss of performance is possible but reliability may suffer.

Summary

Silicon micromachined microphones have been successfully developed and tested. The fabricated microphones have high sensitivity, good signal to noise ratio and excellent frequency response. These microphones, when properly packaged, should be suitable for applications in certain consumer products such as telephone sets, hearing aids and camcorders. Since no plastic material is involved in the construction, SMM is also suited for high temperature (200-300 oC) applications such as heavy machinery monitoring and car muffler noise cancellation.