Minshen Shares: The application of bone conduction sensors in TWS earphones

Compared with traditional earphones, TWS (True Wireless Stereo) truly wireless Bluetooth earphones have extremely high portability. With the addition of new functions such as active noise reduction and spatial audio, the functionality of TWS earphones is becoming richer, the user experience is constantly improving, and the rapid popularization of TWS earphones is further promoted.

In daily life, the application scenarios of voice calls are quite complex. For example, in public places such as subways, offices, and airports, there are high requirements for call quality. In upstream calls, the elimination of environmental noise has become a rigid demand. How to further improve the call noise reduction effect of TWS earphones, thereby highlighting the product's differentiation and enhancing the product's competitiveness, has become one of the key directions for many manufacturers to deploy products.

When a person speaks, the vibration signal generated by the vocal cords is transmitted in two ways: one is through the air medium to the outside, and the other is through the human skull bones and muscles to the outside, causing the auricle to vibrate. The bone conduction sensor uses the latter transmission method to detect the vibration signal of the auricle and pick up the wearer's voice information. The bone conduction sensor is not sensitive to airborne sound wave signals and has a natural inhibitory effect on airborne sound signals. Therefore, the call noise reduction algorithm is simpler, more natural, and noise suppression is more effective, providing better upstream call quality.

There are many types of TWS earphones on the market. In terms of ENC (Environmental Noise Cancellation) noise reduction, there are various product forms such as single-microphone, dual-microphone, three-microphone, and bone conduction, depending on different prices and performance positioning.

The single-microphone solution has a price advantage, lower requirements for the main control chip, and unrestricted structure. It uses neural network algorithms to identify and filter out noise signals. However, the noise reduction effect of the single-microphone solution is generally not very obvious in identifying and suppressing noise in complex scenarios. The dual-microphone can adapt to various forms and has a good experience in most high-noise scenarios. It has a high cost-effectiveness ratio and is currently the mainstream solution for ENC noise reduction.

Dual-microphone ENC noise reduction uses two microphones with a certain distance to identify sound signals from different directions. The minimum distance between the two microphones generally requires 10mm. The algorithm uses a relatively mature beamforming algorithm, pointing the waveform to the sound source where the wearer is speaking, enhancing the signal sensitivity in the sound source direction, and suppressing the signal sensitivity in other directions, thereby eliminating noise.

Three-microphone ENC noise reduction generally reuses the ANC active noise reduction feedback microphone (FB). The FB microphone is used to detect the sound signal transmitted from the vocal cords to the earphones through the cochlea. In addition, because the FB microphone is deep inside the ear canal, most active noise reduction earphones are in-ear type, and the rubber sleeves of the earphones have a good sealing effect, which can isolate external environmental noise. Three-microphone ENC noise reduction can be applied to most scenarios, has good wind noise resistance, but the algorithm is complex and the price is relatively high.

The combination of microphone and bone conduction provides better performance in resisting environmental human voice interference and less distortion of the caller's voice. Wind noise is a pain point for most earphones. When the wind speed is greater than 5m/s, the microphone will experience saturation distortion, at which point the microphone loses its ability to capture voice signals. The bone conduction sensor does not respond to air fluctuations, so even at high wind speeds, it can still accurately capture voice signals. In addition, for bean-shaped earphones, due to the compact space and small size, it is difficult to separate the two microphones, so the dual-microphone ENC noise reduction effect is poor in this case. A call microphone plus a bone conduction sensor can achieve a better call effect, strong wind noise resistance, and good noise suppression effect. The algorithm does not require complex beamforming, and simple voice algorithms can be implemented. The overall cost-effectiveness of the scheme is high, and it is adopted by more and more earphone manufacturers.

Based on MEMS (Micro-Electronic-Mechanical-System) micro-processing technology, a micrometer-sized elastic beam structure and a mass block of a certain weight are processed on a silicon substrate to form a vibration-sensitive mechanical structure. The capacitive plates on the mass block and the capacitive plates on the substrate form a comb-shaped movable capacitor. When vibration is applied to the device, the mass block vibrates, driving the capacitor comb to vibrate, and the distance between the two capacitor plates changes, thus changing the capacitance.

The internal ASIC (Application Specific Integrated Circuit) extracts, converts, and amplifies the weak capacitive signal, and outputs the vibration signal in the form of voltage, similar to the output signal of a traditional silicon microphone, which facilitates signal processing by the Bluetooth main control.

Using Metal-Lid LGA packaging, the overall appearance is a combination of a PCB substrate and a metal shell. The PCB substrate is located at the bottom, and various signal lines are distributed on the front and back of the substrate. The back is equipped with lead-out electrodes. Soldering grooves are distributed around the substrate, so that the top metal shell is connected to the substrate through solder to achieve complete electromagnetic signal shielding. The LGA mainly consists of a MEMS chip and an ASIC chip. The MEMS chip is mounted on the PCB substrate, and the ASIC chip is mounted on the MEMS chip, forming an up-and-down stacked form, making the structure more compact and the space utilization rate higher. Signal interconnection between the MEMS chip and the ASIC chip, and between the ASIC chip and the PCB substrate, is achieved through Wire-Bonding. The packaging size of 2.7x1.8x1.1 is small and compact, and is compatible with standard microphone sizes.

The device has a sensitivity of -26dBV/g @1kHz, a signal-to-noise ratio of 55dB, and a resonant frequency of 4kHz. Generally speaking, the bandwidth of voice signals is around 1kHz, so the device's frequency response is flat within 1.5kHz, which is more convenient for algorithm processing.