Acoustic Performance Design of High-Bandwidth Accelerometer Based on MEMS Microphone

MEMS microphones capture wide bandwidth sound waves in the air, are small, have a high signal-to-noise ratio, high sensitivity, a flat frequency response, and a high acoustic overload point. While these microphones are excellent, audio quality can be further improved with data from specialized accelerometers that can capture low-frequency vibrations less than 2kHz through solid materials.

This is particularly important when using accelerometers to measure skull vibrations caused by air and bone conduction stimuli [1, 2], which can be used to detect skull vibrations caused by a speaker's voice. Accelerometers are not affected by airborne noise. This article reviews the key specifications of such audio accelerometers, presents their performance data, and describes the working principle of a fused MEMS microphone and audio accelerometer, and how this combination helps achieve better audio quality within the accelerometer frequency range.

Although audio accelerometers and standard accelerometers have the same basic operating principles, audio accelerometers have specific features tailored for voice/sound applications. It has a higher sampling and output data rate (8kHz/16kHz/24kHz) to match most digital audio systems, and utilizes a time division multiplexing (TDM) interface instead of I2C for data transmission.

Device Description

The accelerometer can capture vibration signals up to 2.4KHz, and using a MEMS microphone with noise reduction can significantly improve the audio quality of ear-worn devices or smartphones. The self-test function embedded in the accelerometer eliminates the need for mechanical testing of each product PCB [3].

Unlike traditional accelerometers, you need an accelerometer that utilizes a TDM interface to transmit accelerometer data. The TDM bus is a common multi-slot data transmission interface for audio systems and is present on some accelerometers. It allows multiple data frames to share a single digital interface [4]. Figure 1 shows a typical TDM clock generation using a microprocessor/DSP and an external oscillator. The accelerometer is a secondary-only TDM device. It relies on the processor to generate the TDM clock signal.

Figure 1: Example of a common microprocessor/DSP configuration.

Equipment performance

You can test the audio performance of an accelerometer using a reference microphone 6 inches from the center of the speaker, mounted directly to the outer support ring of the speaker cone with the accelerometer. The speaker output can then be set to -6dBFS (114dbSPL) as measured by this reference microphone. Capturing the accelerometer output via recording software and analyzing it using the loopback feature of Audio Precision can provide detailed information on the total harmonic distortion (THD) ratio of the reference microphone and the accelerometer. The results are shown in Table 1.

Table 1: Total harmonic distortion ratio of reference microphone and accelerometer at different frequencies.

The data collected showed that the distortion performance of the accelerometer was comparable to that of the reference microphone. The signals analyzed in this experiment included distortion components from the power amplifier and speaker, as well as the microphone/accelerometer. Figure 2 shows the Fast Fourier Transform of the accelerometer output, where a 500Hz excitation was applied to the sensor.

Figure 2: 500Hz vibration test results.

Figure 3 shows the noise performance of the accelerometer without excitation. Figure 3a is the X-axis noise curve, and Figure 3b is the Z-axis noise curve. For simplicity, we omit the Y-axis noise curve because it is similar to the X-axis curve. On the other hand, compared with the X/Y axes, the Z-axis profile is different due to the difference in the sensing elements.

Figure 3a: Accelerometer X-axis noise density.

Figure 3b: Accelerometer Z-axis noise density.

Accelerometer and microphone fusion

Audio accelerometers improve microphone performance in a range of applications including headphones/earphones, smart speakers capable of identifying and amplifying low-volume sounds, voice-activated devices, bone conduction earphones, applications requiring beamforming, and speech enhancement applications.

These results show that designers can fuse data from accelerometers and microphones to achieve superior audio system performance. For example, fusion can improve the performance of mobile phones during phone calls, improve accuracy with voice-activated devices, provide a better online gaming experience, and reduce keyboard noise during video calls.

Figures 4 and 5 illustrate the primary application of accelerometer and MEMS microphone fusion, which is wind noise reduction in hearables or headphones. Wind noise is typically below 1kHz and has a random spectrum, making it difficult to remove without affecting the overall audio quality. Traditional non-mechanical wind noise reduction requires filtering, which also attenuates the audio content of the sounds you want to hear. Our proposed microphone/accelerometer fusion leverages the accelerometer’s ability to resist wind noise to maintain audio fidelity.

图4显示了麦克风和加速度计的示例信号链及算法。该算法将加速度计的低频范围与麦克风的高频范围相结合，从而产生高质量的宽带宽音频。这种方法可以显著提高多风环境中的音频质量。虽然仅麦克风输出中的音频信号以风噪声分量为主，但麦克风和加速度计解决方案消除了风噪声，因此它只输出所需的音频。

Figure 4: Microphone and accelerometer signal chain example.

Figure 5: Pre-processing and post-processing of audio signals.

in conclusion

Combining an accelerometer with advanced specifications and features with a microphone can help system designers improve the sound quality of their systems. Compared to traditional accelerometers, audio accelerometers have a 2.4kHz bandwidth, digital audio compatible output data rate, and TDM interface, making them easy to integrate with common digital audio systems.

Key to its value is the accelerometer’s immunity to environmental noise sources, which provides solutions to issues such as wind noise and improves audio quality without the drawbacks of traditional technologies.