MEMS Microphones

MEMS Microphones

Analog & Digital Microphone Interface Techniques

There’s a popular belief that once a digital version of a long-standing analog device is created, the writing is on the wall for the older part. That belief was once widely held in the Human Interface Device world where voice has become the fastest growing method to send instructions to TVs, intelligent helpers like Alexa and Siri and of course, automotive GPS systems. A great many saw the advent of the Digital MEMS microphone as the death knell for its venerable analog counterpart, but fortunately for many voice applications, the news of its passing has been greatly exaggerated. In this article we’re going to show why analog microphones are not going anywhere despite the revolution in digital MEMS mics and why understanding when to use the right output can save lots of grief in design.

We’re going to discuss…

  • What is a MEMS Microphone?
  • 1st Order design considerations
  • Microphone applications
  • Electronic Interfaces to the Digital MEMS
  • Electronic Interfaces to the Analog MEMS

To keep things straightforward, we’re going to restrict this blog to analog and digital microphones with a MEMS construction. There are other considerations to be made when considering analog Electret Microphones, but we’ll leave those for another day.

What is a MEMS Microphone?

A MEMS, or Micro-Mechanical Structure is a device made using semiconductor techniques that allows the physical property of pressure to be translated to voltage or current with appropriate tethering and electrical bias. These movements generate sufficiently large analog signals that they can be used to form extremely small sensors.

MEMS microphone
Fig. 1: Typical floating metal structure of a MEMS device.

Figure 2 shows how the MEMS device is used in the case of a microphone. A second chip is assembled in the same package as the MEMS in what is known as a Multi-Chip Module (MCM). Gold wire bonds are used to connect the chips electrically and it is the chip assembled with the MEMS device that determines the analog or digital nature of the finished device. The second chip in the MCM of an Analog MEMS mic is almost always an amplifier with voltage output, while a Digital MEMS mic will have a sampling device like a 1-bit Sigma-Delta ADC and a digital PDM output.

The analog microphone structure is common to both MEMS device formats.

MEMS device is used in a microphone
Fig. 2

1st Order design Considerations

One of the most important determinations designers must make in any microphone application is whether there is sufficient dynamic range in the proposed forward path to faithfully represent the sound presented without clipping.

The maximum voltage swing of a microphone (including the analog portion of a digital microphone) can be estimated quickly by examining two very important specifications found on every microphone data sheet.
Sensitivity, the microphone’s output voltage with a 94dB SPL (Sound Pressure Level) and AOP, the Acoustic Overload Point when Total Harmonic Distortion rises to 10%. Let’s do a little math.

The Sensitivity Voltage is more generally expressed as;

The rms output voltage can therefore be calculated at the 94dBSPLrms reference sound input.

If the AOP is specified as a peak voltage, the rms output voltage at the AOL can be found by calculating the decibel difference between the AOP and the reference output as shown;

Rewriting this in an easy to remember format using the spec terms from typical data sheets;

Volts pk/pk

Now we have an easy check on the microphone’s suitability for the application using dynamic range. Table 1 shows an example of how this test could be applied.

ParameterTypical dBMin dBMax dB
Dynamic Range2.2V3.6V1.12V
Table 1: Dynamic Range Check

The Golden Rule from Table 1 is don’t design microphone circuits using datasheet Typical values. If this design was to run on a +/- 1.5V power supply with likely peak to peak voltages of 2.8V, the microphone would clip the input signal when the device had the Sensitivity parameter closer to the minimum value it could be shipped with.

Clipping due to design error
Fig. 3: Clipping due to design error

Microphone Applications

The complexity of the sound an audio design is trying to capture is the major limiting factor on how simple the design process is to achieve the right result. Let’s assume the target is a headset mic with all of the required sound coming from directly in front within a few inches of the wearer. This is known as a Near Field sound application and is described in Figure 4.

Near Field / Far Field sound projection

Fig. 4: Near Field / Far Field sound projection

The definition of Field in sonic terms is that the sound is heard differently the further a listener is from the sound source. In near field applications, dynamic sound qualities like attack and decay and the energy associated with them are easier to capture. However, as much as 6dB of sound pressure is lost on each doubling of wavelength, making it harder to capture tone, sharpness and content as the receiver is moved further away from the sound source.

The headset application is therefore relatively simple; near field dynamics mean the shape of the sound emitted will be very close to the shape of the sound captured. The biggest factors in determining the quality and suitability of the design will likely be made in the determination of dynamic range of the signal path which we have just shown can be easily maximized by over-specifying components in the forward path.

Far Field Microphone Applications

The specteral range of the average human voice used in a control application is between 1kHz and 8kHz. The wavelength of these sounds is determined by dividing the speed of sound in air by its frequency, so in most voice apps we have wavelengths of 342mm to 42mm to work with. Far field considerations come into play at 2 wavelengths, so any voice application where the sound source is outside of the 84mm to 684mm window likely cannot be designed solely in consideration of dynamic path variables. The majority of voice control applications in the home fall into this category.

Quantization Noise

Almost all digital MEMS microphones use a 1-bit PDM converter to digitize the analog mic signal. This very convenient format requires only two control signals, clock and data, to transfer digital data to a microprocessor. By contrast most analog mics are digitized using higher bit ADCs such as 16-bit and 24-bit devices. In Far Field applications, this can be a significant advantage because of the different levels of quantization noise each application generates.

Figure 5 shows the relative noise levels between each converter.

Quantization Noise in 1-bit and 16-bit converters
Fig. 5: Quantization Noise in 1-bit and 16-bit Converters (courtesy of Texas Instruments)

Pass band noise levels are typically an order of magnitude lower in multi-bit converters which makes accurate sound capture in noisy environments particularly challenging. For this reason, most digital MEMS mics find less use in far field applications such as voice control than they do in near field apps like smartphones, headsets and wearables.

Digital MEMS Interfaces

Digital MEMS microphones make it extremely easy to connect to a microcontroller. Figure 6 shows RDI’s RMM-3526DB mic connected to a typical micro. The PDM data stream is clocked through to the micro on the leading edge of the clock thanks to the high state on the L/R control. The 1uF decoupling cap is essential to provide the dynamic current required by the continuously running PDM modulator.

In addition to a very simple circuit design, Digital MEMS microphones provide excellent noise immunity and are a great choice when there is lots of electrical transient activity in the environment.

The L/R logic line on the microphone allows simple stereo coding based on the edge of the clock. When L/R is set high, data is valid on the rising edge of the clock. When L/R is low, data is valid on the falling edge. Two digital MEMS microphones are shown connected to a stereo controller in Figure 7. The timing diagram shows when data is valid on each clock cycle. Power lines and decoupling caps were left off for clarity but are of course required here too.

Digital MEMS Interface Circuit
Fig. 6: Digital MEMS Interface Circuit
Stereo Implementation with Digital MEMS
Fig. 7: Stereo Implementation with Digital MEMS

Non-Inverting Analog MEMS Interface

Figure 8 shows the Analog MEMS mic in a non-inverting op-amp circuit. RDI’s RMM-2718AB is shown. Note that bypass caps are not shown on the op-amp or microphone and the power supply lines have been removed for clarity on the op-amp.

Analog MEMS mic with non-inverting amplifier
Fig. 8: Analog MEMS mic with Non-Inverting Amplifier

Almost any low noise op-amp can be used here because of the low gain needed to fully saturate the mic output, but care should be taken to select an op-amp with good Common Mode Rejection (CMR) in non-inverting mode configurations since common mode signals will be present on the input pins. These then sum (with gain) and distort the output signal if not rejected by the op-amp. This amplifier configuration is used when the mic output simply cannot be inverted.

Inverting MEMS Interface

In the more common inverting mode, both input pins of the op-amp are held at ground (virtually) eliminating the CMR problem.

Analog MEMS mic with inverting amplifier
Fig. 9: Analog MEMS mic with Inverting Amplifier

The output is of course inverted, but the most common reason not to use this mode is the design tradeoff forced because the input resistor on the inverting input forms a potential divider with the output resistance of the microphone, typically 200 to 300. The input resistance must be chosen so that it is high enough to prevent too much loss of the mic signal, but low enough to prevent the addition of resistance noise in the forward path of the op-amp. A 2k input resistor can cause as much as 12% signal loss.

Analog MEMS Microphone Corner Frequency

The final design consideration for the standard op-amp interfaces is the determination of corner frequency. A High Pass filter is created by capacitor C1 and R1 in the circuits of Figures 8 and 9. Use the desired roll off frequency to determine the value of C if R is fixed for the reasons just discussed.

Far Field Sound Capture Solutions

The most common solution to Far Field sonic capture problems is to use a combination of a high AOL, high SNR analog MEMS mic and a high-speed multi-bit sampling ADC to improve quantization noise. In Far Field, a high dynamic range PGA (programable gain amplifier) is used with a multi-bit delta-sigma converter to produce a very wide dynamic range microphone response able to meet the far field challenge of hearing quiet sounds in a quiet room as well as normal sounds when high levels of background noise from other sources are present.

Several semiconductor manufacturers offer integrated versions of this analog front end with digital conversion.

Connections Newsletter

Sign up for email copies and never miss another blog.

Scroll to Top