•  
• EDN Europe MAGAZINE
•  
• Free SUBSCRIPTION
• Subscription Renewal
• View Media Kit

• EDN
• EDN Australia
• EDN Asia
• EDN China
• EDN Japan

MEMS enables multi-microphone consumer products

IN RECENT WEEKS AND MONTHS, THE LIST OF COMPANIES OFFERING ALL-SILICON, MINIATURE MICROPHONE PRODUCTS TO THE MARKETPLACE HAS GROWN WITH THE ENTRY OF SOME WELL-KNOWN NAMES. MICROPHONES USING MEMS (MICRO-ELECTRO-MECHANICAL-SYSTEM) TECHNOLOGY HAVE BEEN AVAILABLE FOR SEVERAL YEARS, AND THEIR—MAINLY SPECIALIST—MAKERS HAVE CLAIMED A GROWING SHARE OF THE MARKET.

 

MEMS-based microphones, as a business sector, appears to have now grown to a size that makes a market entry attractive for broadrange semiconductor suppliers that also have expertise in MEMS design, such as Analog Devices. In the audio-signal-processing chain, it is the transducers at each end of the process that can pose some of the most intractable problems for designers. While you can readily achieve distortion figures of tiny fractions of a percent and correspondingly low noise levels in the electronic domain, you must still capture sounds with a microphone and reproduce them with a loudspeaker or earphone. Even the best of those devices can introduce errors that are orders of magnitude greater than those of the electronics; designing for the transition from acoustic to electrical and back again still demands more than a little art along with the science.

Today, increasing numbers of designers of handheld and portable products have to deal with this problem. In addition to products that deal with speech as their primary function—cell phones, headsets and voice recorders—other classes of product, such as digital still cameras, offer audio input as a secondary function. Others will add it as a means of control, as voice-activation becomes a feature that users expect to find in their consumer gadgetry.

There is still only a short list of MEMS-microphone manufacturers— all of them attempting to displace a single established technology: the ECM (electret capacitor microphone). (In the field of microphones, for reasons that are not clear, you often find the historical term “condenser” instead of “capacitor”, hence, “EMC” also stands for “electret condenser microphone”.)

FLEXIBLE-PLATE CAPACITORS PREDOMINATE

The sound-detecting element of almost all miniature microphones in today’s consumer products is a parallel-plate capacitor. Other types of microphone, of which there are many, are almost all significantly physically larger—and more expensive—and do not find use in portable consumer appliances. One plate of the capacitor is fixed and rigid; the other is a flexible membrane that is free to move in response to the changing air-pressure patterns of incoming sounds. When the capacitor carries a charge, the varying distance between the plates yields a voltage signal. To make a simple capacitor function as a microphone, you need to apply a bias voltage; the term electret denotes the use—for the purpose of forming the flexible diaphragm—of a polymer material that carries a permanent charge trapped in its structure, removing the need for that bias. The electret microphone “capsule” (Figure 1), which is the structure that the Japanese company Hosiden uses, is therefore electrically very simple: it comprises the capacitor, plus an FET that provides first-stage amplification. Nevertheless, in mechanical terms the ECM is a precision fabrication in multiple materials, and only the fact that they are the outcome of a mature manufacturing process—in other words, once they reach production volumes of millions—means that you can buy them for under €1 each.

The MEMS microphone is an all-silicon device. Using the techniques of micro-fabrication, their makers produce a membrane that is thin enough to provide the necessary flexibility to be the moving part of the sensing capacitor. Perhaps surprisingly, different manufacturers use different aspects of the semiconductor fabrication to make the membrane: they variously employ the native silicon of the substrate, or a polysilicon layer, or a deposited metal layer.

Given that most of the suppliers of MEMS microphones say they have developed micro-fabrication processes that are compatible with standard CMOS, you might suppose that their products would be single-chip devices with the sense capacitor, plus bias generation and preamplifier, all produced in the same production flow. In fact, the most common construction is a two-chip configuration with the MEMS elements on one device and the electronic circuitry on another, the two being mounted in a miniature multi-chip module. At present, only one maker uses monolithic construction.

But, detail construction differences apart, all challenge the ECM on a number of common points. Depending on construction, the MEMS devices are smaller—or considerably smaller—than all except the smallest ECMs, which appear to have reached a minimum size at about 3 or 4 mm in diameter and about 1 mm in height. As all-silicon devices, they claim greater robustness compared to the ECM, particularly in tolerance of temperature. Until recently, the ECM could not tolerate the temperature of a solder-reflow process; when you specified the component, you had to attach it to the circuit board in a separate process, often by hand. Laterally, makers of ECMs have produced versions that will withstand the reflow process but, according to the competitive statements of several MEMS-microphone manufacturers, even those ECMs will suffer a shift in their operating characteristics if you do so, due to the organic materials employed in their construction. And the higher temperatures of lead-free solder processing may still be out of reach.

What can you do with a MEMS microphone that would not be possible with previous devices? Some concepts are enabled by size alone; using multiple microphones rather than just one may allow you to provide ambient noise cancellation in products such as Bluetooth headsets or cellular handsets, for example. Several manufacturers make the point that they can supply microphones with key parameters such as sensitivity and phase response closely matched. The siliconfabrication process produces devices that have a much smaller spread of production tolerances than earlier microphone types; also, manufacturers are selecting from production lots to produce precisely matched units. With these units, you can build arrays of two or more microphones that will give your product directional sensitivity—all of the MEMS microphones, as individual items, are omnidirectional in response. You might use this feature to isolate voice input from a user in noisy surroundings; examples that MEMS makers suggest include hands-free operation of communication or navigation devices—as an OEM fitment or an after-market addition—in the automotive environment.

An option that becomes possible with the all-silicon microphone is digital output; more than once source offers—or promises—the addition of an analogueto- digital converter within the microphone package. In an electrically noisy environment, this feature can provide a measure of immunity to interference that may be hard to achieve with millivoltsignal- level analogue lines.

However, because you are working at the interface between the acoustic and electrical worlds, the data sheets will never tell the complete story. Manufacturers will specify high sensitivity enabling socalled “far-field” uses—for example, voiceoperated portable-navigation devices, or desktop speakerphones. And they will specify linearity up to high sound-pressure levels—the cellphone user speaking loudly into his phone, perhaps only millimetres from the microphone, might generate very high pressure peaks, which is a “nearfield” example. But you will also have to take into account the acoustic properties of the casing or enclosure of your product, which may have as great, or greater, an effect on the perceived sound quality. At all scales—not just the miniature—there are subtle aspects of microphone performance that affect the user’s experience. For example, how does the microphone handle “plosives” (“p” and “t” sounds when the speaker expels a pulse of air); or sibilance (“s” sounds)? If your application demands good perceived sound quality, there may be no alternative other than prototyping the product with different microphone types and conducting listening tests.

A significant recent announcement in the field comes from Analog Devices, entering the market with the ADMP401-1 and ADMP421 (Figure 2). ADI has a considerable history in fabricating MEMS chips, having supplied high volumes of airbag accelerometer sensors to the automotive industry for several years. Now, according to Christoph Wagner, the company’s MEMSfield- applications engineer, consumer-product makers are asking for large volumes of silicon microphones, prompting the company to introduce both analogue- and digital-output models. Reporting that engineers are asking for better signal/noise ratio and frequency response, Wagner notes that improved audio quality improves not only the consumer experience but can lead to more accurate voice recognition. The parts “address most of the issues of ECMs,” Wagner says; you can use them in a fully automated assembly process, and the integrated device replaces the ECM plus an amplifier, an A/D converter—if you require digital output—and a voltage regulator. You get greatly improved immunity to RF noise, and the device—in analogue form—uses under 200 A, “almost 50% less than the best ECM,” according to Wagner. ADI assembles a two-chip package, which consists of its own MEMS sensor and ASIC, the latter having the function of a preamp-and-bias generator. With this statement, the company—and other vendors—hints that not all suppliers fabricate both elements of their products, and that the amplifier ASIC may be common to several designs; however, none of the companies interviewed was forthcoming on this point. Analog Devices builds a small substrate with a metal can enclosing the chips, a commonly used construction. Sound enters through a hole in the host circuit board and a matching hole in the microphone substrate. The microphone sensor has an etched cavity; entering this space, the sound first encounters the rigid backplate of the capacitor, micro-machined from the silicon. This backplate is micro-perforated to permit the sound to pass through to the flexible membrane formed above it. The company claims that this perforated structure also acts as a particle filter to protect the membrane. The air sealed within the can acts as a “backvolume”—cushioning or damping the membrane but avoiding, ADI says, resonances that can occur when the effective volume of this backvolume is smaller. The response is flat to 15 kHz. The analogue-output part has a sensitivity of -37 dB for an SPL (sound pressure level) of 94 dB. Total harmonic distortion at 104 dB SPL is a maximum of 3%, signal/noise ratio is 61 dB (A-weighted) and PSRR (power-supply-rejection ratio) is over 50 dBV. The package measures 4.72x3.76x1.0 mm; a 3.3x2.5x1.0-mm version sacrifices some low-frequency response, lifting the lower end of the flat band from 100 to 280 Hz. The digital version, in a package 3x4x1 mm in size, has -26 dB full-scale sensitivity and 80 dBFS PSRR.

MONOLITHIC-MICROPHONE DESIGN

Start-up company Akustica has recently broadened its microphone range with the AKU1126 analogue microphone. Akustica entered the market in 2006 as a niche supplier specifically targeting the laptop- and notebook-computer segment and offered those designers a digital microphone from the outset. Vice president of marketing and product management DavinYuknis says that in the space of some 30 months, 30 to 40% of such PCs now have MEMS digital microphones, and the company is branching out—specifically looking at the cell-phone market, where analogue microphones still predominate. Akustica is almost the only company offering a monolithic solution (Figure 3), with sensor and electronic circuitry fabricated on the same substrate. It also has, Yuknis insists, the only process that is 100% fabricated in standard CMOS— others, he says, may be CMOS-compatible to a greater or lesser extent but still depend on some form of special processing. The company forms its device’s membrane in deposited metal, which is very flexible and therefore sensitive. Together with the CMOS processing, Yuknis says, this technology allows the company to scale the device size down; its complete chip measures 11 mm, with a membrane 100 µm in diameter. However, it still needs a package with a “confining volume”; consequently, the overall size is 2x2x1.25 mm. Sound enters through a hole in the top of the package, routes via a channel to the bottom of the package, and then passes through a perforated backplate to meet the flexible membrane. Sensitivity is expressed as -42 dBV/Pa with 58 dB SNR. THD is described as “50% better than [ECM products]” at 115 dB SPL—or 3%, which is similar to the ADI offering. PSRR is 40 dB, and current consumption is 140 µA from 2V. The device has a usable response to 20 kHz, Yuknis says. He adds that you can house three of the devices in the space that an ECM would occupy, and four in the space of some competitive MEMS microphones, so finding space for multimicrophone arrays becomes simpler.

In mid-2006, Infineon moved to production with a microphone concept that had started as a research project as long ago as 1998 and the company now manufactures as the SMM310. Infineon has formed an alliance with Hosiden in the microphone field and is accessing some of that company’s expertise in the area of acoustics. Infineon builds a microphone with separate sensor and amplifier chips. The sensor chip measures 1.3x1.3 mm and has a 1-mm diameter membrane, which Infineon’s process fabricates in polysilicon and is 300 nm thick. In Infineon’s design, etching from the back of the silicon substrate forms a cavity with the membrane at the top; the rigid backplate sits above that membrane, and sound impinges on the top of the backplate—which is perforated, as in other designs (Figure 4). Again, an underlying substrate carries both sensor and amplifier chips, and a metal can covers the assembly. Package height is 1.25 mm with a reduction to 1 mm envisaged; Infineon plans a digitaloutput variant with an on-board ADC in 2009. The company has designs that can tune the sensitivity of the membrane by employing different etched “spring” mounting points around its edge or by forming circular corrugations around the perimeter of the membrane to increase its movement for a given sound pressure. The company claims that in certain circumstances— a big over-pressure air pulse, perhaps—the membrane of a silicon sensor may deflect far enough to touch the backplate. Due to surface-charge effects, it may stick there. Infineon has a system of tiny spikes or bumps, formed on the backplate, that prevent this phenomenon. (In other designs the microphone ASIC may detect this occurrence and drive the plates of the capacitor apart, but this does not appear in data sheets.) Infineon says that a combination of electromechanical and EDA CAD tools enables the product-design team to accurately model and predict the behaviour of the capacitor-plate-suspension designs and will allow the company to develop families of devices with customised specifications. The SMM 310, with a sound hole in the top of the can, has a sensitivity of -42 dBV/Pa and a 59 dB(A) SNR. THD is, again, under 3%; PSRR is 55 dB. A “hole-down” version, the SMM350, is in production via Hosiden, whose catalogue lists it as the KRM5000. In this configuration, sound enters through a substrate hole and travels within the package to the sensor; a later version, KRM5100, will mount the sensor directly above the substrate hole to increase sentitivity. However, says Infineon’s senior product marketing manager Dr Roland Helm, the former arrangement yields better SNR.

RF IMMUNITY

Helm notes that designers often need support in areas such as acoustics, mechanical mounting—especially if a design requires vibration-isolated mounting— and RF suppression. Reference 1 deals with using the product in a high-RF-noise environment. Helm notes that this is increasingly the case in handsets; to minimise RF exposure for the user, designers increasingly locate the antenna at the bottom of the case, in close proximity to the microphone.

Approaching the subject from a longestablished position as a supplier of many different types of microphone, Knowles Acoustics was an early entrant into MEMS products. The company originated the substrate-plus-metal-can package footprint that several other suppliers— explicitly or not—adopt in their product lines. It offers a number of formats, including types “hardened” against GSM-RF signal-burst noise intrusion; and, for designers embarking on a multimicrophone- array project, its IntelliSonic software embodies algorithms that can achieve 16 dB noise suppression, 30 dB interference cancellation though the use of beam arrays, and 27 dB acoustic-echo cancellation.

CHIP-SCALE PACKAGING

Pulse Engineering (Technitrol) acquired Danish company Sonion early in 2008, which became its MEMS division. In contrast to multi- and single-chip offerings that other suppliers build inside larger housings, Pulse uses a two-chip approach but delivers it as a CSP (chipscale package) with no outer housing. Analogue and digital version are in current production under the names SiMic and DigiSiMic—their small size suit them to applications in headsets, as well as handsets and other portables. A new version of the company’s digital product delivers devices matched to ±1 dB for use in arrays. Division president Dr Jacob Philipsen anticipates a growing demand for beam-steered applications in automotive designs—the MEMS devices exhibit close matching in amplitude and phase response and stay that way over time. The company’s approach to its sensor employs reverse etching to define a membrane with a well through the substrate beneath it. The process creates the rigid plate of the capacitor on top of the membrane. The construction proceeds by inverting the die and flip-chip, mounting it on a CSP substrate along with the ASIC. This means that the membrane of Pulse’s device is directly exposed to incoming sound, and the atmosphere: Philipsen says it is robust enough to withstand normal usage, but a custommoulded rubber boot is available to add protection and simplify building-in the product in certain applications. The entire package measures 2.3x1.6x0.9 mm, or 2.6x1.6x0.9 mm for the digitaloutput version. The backchamber of this construction is much smaller than that of other designs, which contributes to the extended frequency range up to 80 kHz. It is completely flat, Philipsen says, from 20 Hz to 20 kHz. Its SNR is 60 dB or better. In its digital form, it outputs—as other digital products in the market also do—a pulse-density-modulated waveform, and you can multiplex two microphone signals on a single wire.

The most recent entrant to the sector is fabless semiconductor house Wolfson Microelectronics, building on its market position in supplying codecs into a number of handheld consumer-product designs. Wolfson acquired MEMS startup Oligon in 2007, and its True Mics introduction is the first result of this acquisition. Once again, the sensor employs a perforated, rigid backplate atop a flexible membrane—this time, built in silicon nitride—with the cavity resulting from back-etching the silicon wafer forming a chamber behind the capacitor element. And, once again, a two-chip design resides under a metal can with a sound hole on top—also called a “hole-up” model (Figure 5). The company asserts that both silicon elements are “100% Wolfson” and that it can fabricate the sensor in a standard CMOS foundry. The first part released is an analogue device; a digital version will follow. WM7110 occupies a 4.72x3.76x1.25-mm outline, while WM7120 is 3.76x2.95x1.10 mm; both come in a selected version that yields ±1-dB matching. Some of the operating parameters include 62 dB SNR, 160 µA current and -42 dBV/Pa sensitivity. THD is, the company says, bestin- class at 0.5% at 100 dB SPL; however, Wolfson is emphatic that audio quality is not defined by any one parameter but is an aggregate of the specification sheet and of the environment into which the designer builds the part. Wolfson quotes a price for the device of $1.64 (1000). But the company joins its competitors in urging users to consider the system cost, plus the reduced parts-count that the integrated device enables.