In the 1986 film, Star Trek IV – The Voyage Home, Scottie tried to talk to the computer but was told to use the keyboard. “How quaint” is his comment. He had time-traveled back from several hundred years in the future yet it has only taken us a couple of decades to achieve increasingly widespread use of voice user interfaces (VUIs) such as Siri and Alexa.
For many years, microphones used the change in capacitance between a thin film that moved in response to sound waves and a fixed surface to generate an electrical signal. So called Electret Condenser Microphones (ECMs), the performance characteristics of each microphone can vary from one to another, even off the same production run.
The big problem with ECMs is that they are too large to easily put into portable devices. The breakthrough came from using the manufacturing techniques for electronic chips and adapting them to make MEMS microphones. Unlike ECMs, where you can see the component parts with the naked eye, you need a powerful microscope to see these new microphone structures at 800 µm x 800 µm (with a packaged size of 2.75 mm x 1.85 mm) at their smallest. They are an example of a Micro-Electro-Mechanical System or MEMS. Effectively the ECM’s design of two surfaces, one fixed and one moving in response to sound and causing a change in capacitance between the two charged surfaces, has been shrunk so that the MEMS microphone is several orders of magnitude smaller than an ECM.
You might think that these microscopic microphones are much more delicate than their bigger predecessors but actually they are quite robust. The robustness comes from them having incredibly tiny moving parts with hardly any mass so impacts have little effect. The tiny moving mass also means that the MEMS microphone can match the performance of ECM designs and yet demand less power: Infineon offers a MEMS microphone via its packaging partner that needs only 200 microWatts (µW) whereas ECMs typically need at least 800 µW.
So, much smaller and much less power hungry. And there are more ways in which their superior characteristics are making them the microphone solution of choice and ECMs a thing of the past. Let’s have a look at some of them.
- Thermal stable so they can operate over a wide temperature range of -20 °C to 85 °C.It also means that they can go through a standard PCB reflow process so that they can be incorporated onto a PCB during manufacturing and not added later as with ECMs, thus reducing device production costs.
- As they are mass produced with high yields using similar automated techniques to those perfected to make billions of electronic chips in fabs, MEMS microphones are much less expensive than ECMs that are made on a factory assembly line.
- Virtually identical performances for samples of a model of MEMS microphone due to the automated manufacturing just mentioned.
- MEMS microphones have consistent performance over their lifetime whereas ECMs can drift from their original specifications over time.
- The higher the Signal-to-Noise Ratio (SNR), the better and MEMS microphones can have SNRs as high as 80 dBA (Decibel A-wt.). The noise in this case is the noise-floor generated in the microphone and its control electronics, also called self-noise that you would hear as a hiss. Any audio signal below the noise-floor of the mic will be lost in the noise and cannot be recorded. So expressing the ratio of unwanted noise to wanted signal gives the SNR, which you want to be as high as possible. The upcoming Infineon MEMS microphones will have a SNR of 73.5 dBA which means a very low noise-floor to capture the tiniest of sounds such as whispers at 25 dBSPL (Decibel Sound Pressure Level).
- You might have thought that such sensitivity means that the MEMS design cannot cope with very loud noises. The limit for loud noises is known as the Acoustic Overload Point (AOP) and defined as being where the distortion of the captured signal has exceeded 10%. This is when you would hear cracking, clipping or distortion. MEMS microphones can actually go higher than the 130 dBSPL found at a rock concert which is similar to high performance, expensive and larger ECMs.
- An important thing to check when looking at the AOP is the amount of distortion that occurs along the journey to reach this point. You can see from the graph below that some MEMS microphones can have notable distortion occurring before the AOP point is reached, which is usually to be avoided if you want a good clear signal across the whole range of sound input levels.
The first application area for MEMS microphone was in smartphones where their tiny size and low power consumption made them the solution of choice. Their tiny size also means that Active Noise Cancellation, which was previously only possible in large headphones, can be shrunk into earbuds with up to six microphones in each earbud.
The year-on-year improving quality of the audio signals makes voice recognition more and more accurate for better voice commands of consumer devices such as smart speakers or infotainment systems in cars. Now, for many metrics, the performance of MEMS microphones is getting close to that of human ears so there will be many new possible application areas especially those using VUIs to give natural, intuitive interactions with devices. Humans have been communicating with speech for millennia so MEMS microphones will enable VUI to become the standard and keyboards to become a ‘quaint’ memory.
So far, the comparison has been with the conventional ECM and shows why MEMS microphone are replacing them in many applications and made possible new applications. So which model of MEMS microphone to choose? Infineon has been perfecting its MEMS microphones for over 15 years and offers the best-in-class solutions with its XENSIV™ products. For example, based on an internal study, a high SNR microphone results in up to 40% better performance than its nearest, competitive, standard SNR MEMS microphone for word recognition and capture at low input sounds such as whispers. This high SNR also means that it is up to 25% better than competitors in a ‘cross room scenario’ where the speaker is across a room or even in a next-door room making it the ideal choice for integrating voice controls into almost any smart home device.
If you are by now itching to get your hands on a MEMS microphone to put it through its paces, the good news is that you don’t need to reach for your trusty soldering iron as Infineon has done the hard work and made one of the few, off-the-shelf, MEMS microphone evaluation boards in the industry.
The Infineon Audiohub Nano has two XENSIV™ IM69D130 MEMS microphones so you can capture 24-bit audio, 48 kHz sampling rate, mono or stereo signals, which are output to a computer via a micro USB interface that also provides the power.
It really is out-of-the-box, plug-and-play simple. Connect the Audiohub Nano to the computer using the supplied micro-USB cable and then just use your preferred audio recording software or load the freely available Audacity software onto your PC, which enables you to record and edit the audio signals to evaluate the microphones.
The supplied user manual walks you through the set up and can be download here.
If you are using Audacity, select the audio source as being Audiohub Nano and choose mono or stereo. Now you are ready to record using the intuitive interface.
One feature that is very useful is the onboard LEDs that light up in sequence to show the volume (dBSPL) when streaming audio data not only by the LED illuminated but also by the color as per this table.
Full details of the EVAK AHNB IM69D130V01 evaluation board can be found at the product page here. The board can be obtained from leading distributors.
1) The simplest project is to capture audio across a range of conditions to see how well the high SNR enables the MEMS microphone to pick up really quiet sounds as well as really loud ones with minimal distortion across this range and world-class clarity.
2) Using the two, on-board microphones, they can be implemented to sort out wanted from unwanted sounds in a similar way that your two ears do – even managing challenges such as echoes and reverberation – as well as determining the direction of a sound source with the help of right algorithms.
3) Voice User Interface (VUI). Now that your computer can listen to you, you can start giving it commands and even train it to recognise different voices. The only extra item required is to select the appropriate software from places such as GitHub or Infineon’s software partners.
4) Binaural recording for 3D audio. At CES 2020, Infineon showcased a demo that used a model of the human head with a Hub inside and the two microphones arranged to pick up the sounds from either side where the ears are located. This perfectly mimics the way that humans hear and so, by playing the recorded sound back through headphones, a three-dimensional sound stage is created where the location of sounds are easily determined. This demo is easy for you to repeat using a simple, rough model of a human head and well worth it as the results are dramatically superior to the usual, two-dimensional stereo sound. For the demo, Infineon made a wearable recording device using a Hub with a microphone on either side of a head. This was used to record a walk through some woods and the position of birds in the audio landscape of the recording was very clear on playback. With earbuds becoming increasingly popular, binaural audio could become the new immersive way to listen and ‘stereo’ viewed as quaint.
Experience it yourself. Watch the video – please note for the best result use headphones.
5) Smart speaker. This is where the virtually identical performance characteristics of MEMS microphones plays an enabling role, as you can compare the difference in the audio signals received to determine where the signal is coming from. Smart speakers such as Amazon Echo use an array of several microphones to provide increased accuracy – a technique called beam forming. This enables the device to home in on the source of the voice giving the commands and supress the unwanted background noise, such as music, by using acoustic echo cancellation and advanced noise suppression techniques. This is known as ‘barge-in’, i.e. detecting a command when music is playing. You can easily build a smart speaker by using a couple of Hubs connected to your computer, provided that it will accept two different USB inputs, along with software that is readily available. Further reading on this can be found on this page.
For these more advanced projects, Infineon has a network of partners who can supply software for voice recognition, Voice User Interfaces, Audio Noise Cancellation, etc.
If you are looking for information on mechanical and acoustic implementation tips, please see the Infineon application note: MEMS microphone mechanical and acoustical implementation
For further information on Infineon’s range of MEMS microphones, please visit the product page.
If the Infineon Audiohub Nano has wetted your appetite for seeing what can be done with this next generation of small, high performance MEMS microphones, Infineon has a more advanced development board that is part of its Shield2Go range. This range has many different sensor boards that can be simply clicked into an adapter board with no need to solder. Up to three sensor boards can fit into the My IoT Adapter board in different combination and can also be stacked upon one another vertically to combine even more sensor boards in one system.
It’s really easy to try out different combinations of sensors to make your project smart. For example, a smart home alarm using a microphone and a pressure sensor board.
The range of boards includes sensors for pressure, magnetic fields, and current detection so that, with the microphone version, a wide range of configurations for different smart home applications can be assembled and tested very quickly. It’s plug-and-play so it’s really easy to try out different combinations of sensors to make your project smart. For example, a smart home alarm using a microphone and a pressure sensor board.
The My IoT Adapter then connects to a compatible hardware module with Arduino Uno form factor such as Infineon’s Arduino compatible XMC1100 boot kit. The system can be programmed in minutes using the Arduino IDE and the Shield2Go Arduino libraries that are available on GitHub.
Other hardware platforms such as Raspberry Pi and Espressif Wi-Fi chips can also be used.
Again, there is a wealth of software already available from Infineon partners and on GitHub to help the rapid development of your smart project.
For example, you could experiment with the sensors that you might put into an earbud, which can be packed with many different sensors depending on what features you want it to have. Noise cancellation needs several microphones in each earbud to detect the ambient noise so that an inverse signal can be generated to cancel this out. Known as ANC (Acoustic Noise Cancellation), it can improve the listening experience of both music and incoming speech on a phone call as well as improving the clarity of outgoing speech. The click-and-connect feature makes it easy to experiment with several different combinations of microphones to achieve the desired results.
Further details can be found in this white paper along with a list of stockists.