Microelectromechanical system (MEMS) microphones offer plenty of advantages typical of MEMS devices, including tiny size, low power usage, and consistent performance over time and temperature. But the audio specifications of these microphones have not been sufficient for certain designs - including ones where you need to capture sound from a distance or use multiple microphones. Now high-performance MEMS mics are changing what is possible, and many acoustics experts would say self noise is the first specification to consider.
Self noise - what you need to know
Any microphone produces some level of noise: through its electronics, its transducer element, its housing. This inherent noise is known as self noise. It’s a familiar sound to anyone using a cell phone. For example, self noise contributes to the hiss you hear when your mobile phone is on and nobody is talking.
For an electronics designer, microphone self noise is an ever-present constraint. The idea is to have the mic present as much signal as possible to the rest of the signal chain. But part of the signal you capture from the audio source will fall below the microphone’s inherent self noise, which is also called its noise floor.
The noisier a microphone is, the less signal you have available. A lower noise microphone will give you room to isolate the sound you want from the noise that you don’t. Then your processor - DSP or codec - has more signal to work with. As a result, the output from the signal chain sounds much better when you start with a quieter microphone.
A high signal-to-noise ratio (more about SNR in the next section) indicates a quiet microphone, and a lower SNR specification tells you the microphone has more self noise.
While a high SNR is always useful, you don’t depend on it as much when your audio source is very close to the microphone. There is usually enough signal for these near-field applications. In far-field applications where the microphone isn’t positioned next to the sound source, a noisy mic with a low SNR leaves you with a poor or unintelligible signal.
What the specs tell you
The self noise, or noise floor, of the microphone does a lot to define the quality of audio you are able to capture and pass onto the signal chain. Signal-to-noise ratio (SNR) and equivalent input noise (EIN) are two specifications that describe where that noise floor is.
And as we will soon discuss in more detail, the self noise of MEMS microphones has reached a level far better than past generations of the technology (Figure 1).
Figure 1: The self noise of MEMS microphones has reached a level far better than past generations of the technology.
Signal-to-noise ratio (SNR)
SNR is the ratio of a reference signal to the noise floor of the microphone. A microphone’s SNR is the difference between its inherent self noise level and a standard reference pressure, specifically 94 dB SPL (1 Pa) at 1 kHz.
You will typically see this specification presented as an A-weighted value (dBA) with a 20 kHz bandwidth. A-weighting means the SNR being presented includes a correction factor that corresponds to the human ear’s sensitivity to sound at different frequencies.
When you compare the SNR of different microphones, make sure they are based on the same weighting and bandwidth. A comparison will not be accurate if the measurements don’t use the same weighting and bandwidth.1
Equivalent input noise (EIN)
Equivalent input noise is the output noise level of the microphone represented as a theoretical acoustic noise source placed at the microphone’s input. The unit of measurement is sound pressure level, measured in decibels (dB SPL). SPLs less than the EIN level are below the noise floor of the microphone.
You can determine the EIN directly from the microphone’s SNR specification:2
= 94 dB - SNR
MEMS microphones double their SNR performance
The MEMS industry’s earlier generations of microphones offered an SNR around 58 to 60 dB, which did not equal the acoustic performance you could get from electret condenser microphones (ECMs). The situation is changing now that leading manufacturers are making dramatic improvements to the performance of MEMS microphones.
The ultra-low noise Analog Devices ADMP504 and ADMP521 MEMS microphones have lowered the noise floor of earlier MEMS mics by more than 2x. The ADMP504 and ADMP521 are the first MEMS microphones to reach the level of 65 dBA SNR (29 dBA EIN).
An SNR spec of 65 dBA is good even for an electret microphone, but ECMs tend to be much larger than MEMS mics with comparable SNRs. As the size of an ECM gets smaller, its SNR drops quickly (Figure 2). ECMs also don’t offer the other advantages that MEMS mics do, such as a consistent response to sound across all operating temperatures.
Figure 2: As the size of an electret condenser microphone gets smaller, its SNR drops quickly.