High SNR Is Not the Whole Story: What Really Defines MEMS Microphone Performance

1.Introduction
2.Step 1
3.Step 2
4.Step 3
5.Bestar: Full-Spectrum Support
6.Conclusion

Introduction
The MEMS microphone is a tiny yet crucial piece found in every voice assistant, wireless earbud and smart speaker. It has a simple job, to record sound correctly and abhor all other things. In reality, that is much more challenging than it seems.
The point is noise. Any electronic component produces a certain amount of internal noise. In a microphone, this noise exists below the audio signal and it determines how quietly the microphone can be heard. The noise floor of the microphone itself makes the difference between capturing a signal you want, such as a spoken command in a low whisper, a using a speaker at the other end of the line, an outer sound that is poorly recorded, etc., and losing the signal.
This relationship is measured by signal-to-noise ratio (SNR). The difference in sensitivity of a microphone to sound and the internal noise level of the microphone is measured and in a unit of decibel. The greater the SNR, the better the microphone is capable of hearing the lower the sound is. SNR is not only a specification in the modern audio applications, it is the heart of the user experience.
However, SNR is not all that. A noise performance microphone can easily be clipped in very noisy situations, or have problems with some PCB layout, and add latency that can cause real-time processing problems. Choosing a good microphone is a planned task, and it must start with setting performance goals and conclude with the verification of system-level performance.
It is a three-step process that is discussed in this blog, along with an explanation of what makes a component supplier a true acoustic engineering partner.

Step 1
Determine how high an SNR floor you may need to satisfy your application before comparing datasheets. The number varies greatly depending on the categories of products.
In case of high-end TWS earbuds, high-end smart speakers, automotive voice control systems, and similar flagship consumer audio products, the SNR value is usually between 68 dB and 72 dB or more. The microphone is easy to use and does not need any postprocessing to work effectively at these levels; it can pick finer details of human speech, remains natural during ambient sound surveillance, and it can even enable far-field accuracy of wake-word recognition at a range of two meters or more. It is possible to measure a one-decibel improvement at this distance. It manifests itself to users as decreased listening fatigue, less noisy call audio, and a reduction in misrecognitions.
In the case of the standard communications equipment: voice call handsets, entry-level TWS, video conferencing peripherals, an SNR of 62 dB to 66 dB would be adequate. Such applications focus on cost-effectiveness, and good quality of speech recognition when there is average acoustic performance. The microphone is not required to hear whispers; it must be able to deal with normal conversational voice at all times.
The equivalent noise input (EIN) figure is important along with SNR. EIN is an absolute description of the noise floor of the microphone. An ambient micro-phone picks up everything in the background without coloring it with an electronic noise. This is essential to active noise cancellation (ANC) feedback microphones that have to effectively sample the acoustic field to compute the appropriate cancellation signal. When the feedforward microphone contributes its own noise to that sample, the ANC system will not be able to differentiate between environmental noise and artifacts caused by itself, the ANC system will become unable to silence the noise of the environment and its own artifacts, and the loss in the quality of the cancellation.
The general guideline is as follows: plan your SNR and EIN needs to meet the acoustic conditions of the environment into which your device will be used, rather than to the maximum. Unnecessarily narrow specifications are unnecessary expenses. Not specifying adequately results in a product that does not work out in the field. First define threshold and then choose.

Step 2
MEMS microphones are available in 2 port and a variety of signal output formats. They influence board layout, acoustic path design, system integration, and interoperate with SNR, in ways that are all too simple to find.
Port Configuration: Top-Port vs. Bottom-Port
Sound is introduced into a top-port microphone by a hole on the top of the package. An opening in the PCB side of the device (sound goes through an acoustic inlet) introduces a bottom-port microphone into the device, with the acoustic inlet touching the rear of the board and sound making contact with the PCB through a hole.
Top-port designs make industrial design easier, in the situation where the microphone is near the exterior surface of a product, such as near a speaker grille or a housing vent. Bottom-port designs permit the placement of the microphone not on the surface, but on acoustic passage through the PCB which provides industrial designers with more flexibility in location of the acoustic opening.
Layout is not the only impact of the choice. Essentially frequency response, and in certain arrangements sensitivity to mechanical vibration, is directly proportional to the length and shape of the acoustic path between the source of the sound and the diaphragm of the microphone. One acoustic path poorly-designed may cause resonances that compromises SNR at certain frequencies. 
Signal Interface: I2S, PDM and Analog
Microphones that provide analog output are a voltage signal that is proportional to the acoustic input. They have low latency, and simple signal chains, so are useful in applications where an independent audio codec already exists within the system. The trade off is it becomes vulnerable to electromagnetic interference (EMI) on signal line between the microphone and the codec. This interference may directly reduce SNR in most modern wireless consumer devices, in compact PCB layouts, with RF components close to each other.
PDM ( Pulse Density Modulation ) and I2S digital output microphones encode the acoustic signal within, and provide a digital bitstream. The digital signal is also immune to the nature of the conducted and radiated EMI that impairs analog paths. PDM especially is a suitable candidate in the case of multi-microphone array because clock-phase multiplexing allows multiple microphones to be connected to a single data line. This eases routing PCBs in devices with 2,3, or 4 microphones significantly.
The choice between digital and analog is seldom related to acoustical performance. It relates to the microphone system architecture. Determine the audio processing path first, followed by the choice of the interface that fits in such path without adding other noise coupling.
Stability and Uniformity
MEMS microphones can be used in many temperature ranges, inside a car in summer or in the elements with a wearable. Both sensitivity and noise performance are temperature dependent to a certain extent. When the SNR over the operating temperature range is important, typically in automotive voice control, industrial control, and outdoor consumer devices, at least check the temperature coefficient of sensitivity (TCS) and ensure that the SNR specification is valid throughout the operating range. A microphone that has a high SNR at room temperature, and high thermal drift, might not work in practice.

Step 3
SNR is a measure of the low noise performance of a microphone. The ability of acoustic overload point (AOP) to deal with high levels. Maximization of either of these without the other results in a microphone that works well in the laboratory but poorly in the field.
AOP is the level of the sound pressure where the distortion of the microphone output is beyond a preset level, which is usually 10% total harmonic distortion (THD). The typical values of AOP are 120-135 dB SPL or greater. Low AOP microphones are clipped off in a noisy environment: a concert, a car engine at full throttle, wind sound against an outside housing or just the user using the microphone into the mouth and close-up.
The combination of limits is known as dynamic range. The distance between the noise floor and the overload point, the range of operation of the microphone which is clean is its span. With a 65dB SNR and 130dB SPL AOP, a 65dB (approximately 65 dB or a bit more) dynamic range will be obtained (from about 29dB SPL to 130dB SPL), when standard sensitivity is used.
The most important thing to note is that high SNR and high AOP are somewhat incompatible. Attaining a very low noise floor can in some cases necessitate design tradeoffs that means sacrificing the capability of the diaphragm to react to high motion at loud levels. The manufacturers balance this trade-off using the diaphragm geometry, back-volume design, and ASIC architecture. What this means is that not all high-SNR microphones are performanceable in high-SPL situations, and not all high-AOP microphones have SN performance suitable to quiet situations.
The engineering is to match the dynamic range to the acoustic environment. The office voice assistant requires a high SNR and can tolerate a moderate AOP, as it is unlikely to encounter loud acoustics. A microphone on an outdoor smart speaker or in a conference room is expected to have the ability to both deal with silent input and the occasional loud occurrences. An industrial monitoring microphone or concert recording device must have a high AOP most of all. Establish the acoustic environment, i.e., the environment you are operating in, then determine the dynamic range specification which is appropriate to that environment, not aiming to maximize the SNR number used in any situation.

Bestar: Full-Spectrum Support
It is required to pick MEMS microphone with appropriate specifications. Enhancing that microphone to sing at specification in an finished item is another challenge and this is where friendship towards a supplier comes in.
Bestar is an acoustic engineering partner, not a component provider, in terms of customer interactions. The difference is utilitarian. One of the parts suppliers provides an element to a specification and responds to datasheet inquiries. Acoustic engineering partner collaborates during pre-designbased product development prior to the hardware being completed and helps to identify risks during integration early, and supports the process technically during production.

Conclusion
One systems engineering problem is high-performance audio capture. One of the most important aspects of a larger network is the microphone which encompasses housing acoustics, PCB layout, signal processing, and firmware. Choosing the appropriate microphone to use in the application in terms of SNR is the beginning of the journey, and not the end.
Drawing on a systematic methodology detailed here, establish your acoustic threshold, overcome the physical and interface trade-offs, trade off dynamic range versus AOP which provides the engineering teams with an effective pattern of obtaining an informed decision about the choice they have instead of falling into the top headline specification given.
Selecting an appropriate supplier relationship magnifies the worth of such a structure. A supplier supplying with acoustic simulation, PCB layout, custom test and standard production ensures that the technical risk is reduced at all levels throughout the elaboration process.
Welcome contact the Bestar acoustic engineering team to discuss your target SNR and noise floor requirements, get specific guidance to select the best options to meet your goals.