Redefining MEMS speakers: Q&A with xMEMS CEO Joseph Jiang

Nuying Huang, Taipei; Peng Chen, DIGITIMES Asia 0


US-headquartered xMEMS is on track to redefine Micro Electro Mechanical System (MEMS) speakers.

The company, which produces speaker ICs with MEMS, has released its third-generation product, Cypress, and impressed the global audio market. The low-frequency sound that Cypress offers creates a new auditory experience for human beings.

Taiwan's semiconductor supply chain fostered xMEMS. In an interview with DIGITIMES, Joseph Jiang, xMEMS's co-founder and CEO, said the company originated from the evolution of piezoelectric materials. TSMC was working on the R&D of these materials.

Jiang said he and other co-founders betted on TSMC's R&D success and decided to take a bold move to establish xMEMS. TSMC currently offers foundry services for xMEMS, also a customer of several Taiwanese semiconductor packaging and testing companies.

Below is a Q&A summary between DIGITIMES and Jiang.

Q: Does xMEMS hold a significant advantage as TSMC's solid material techniques back it?

A: xMEMS, established six years ago, saw its first challenge come from TSMC. xMEMS had to prove that it was qualified to be a TSMC partner to receive supplies.

Initially, xMEMS bought blank silicon wafers in the market and asked Japanese partners to put its piezoelectric materials on them. We kept testing in a lab rented from Stanford University in the US until the MEMS audio chip enabled sound. Then, we took the chip to TSMC, persuading the foundry to provide piezoelectric silicon wafers. It was the official start of our collaboration.

After securing supplies from TSMC, xMEMS began testing tens of thousands of dies to find and address issues. Finally, we made our first MEMS chip.

We initially estimated it would take 18 months to reach the mass-production level. In reality, we spent over 30 months. The yield rate started at 50% and has now achieved more than 90%.

MEMS requires high-level techniques. Although we had critical materials, we had to do many other things ourselves. For example, we design testing machines and ask equipment manufacturers to produce them. After that, we also need to operate the machine on our own.

In the case of sleeping earbuds, when xMEMS put its chip into regular earbuds, it did not see the expected result. Therefore, our engineers learned to do 3D printing and made the desired earbuds from scratch. Our customers recognized the move. They even use our prototype to develop new products.

Q: In xMEMS' early days, a mosquito buzz from a MEMS speaker prompted the company to celebrate its success with champagne. Why?

A: The sound created a wonderful memory for our team and a historical moment. It proved that xMEMS' theory can be realized.

From the perspective of MEMS, a product represents a production process. This explains why tailor-made solutions are critical. MEMS has shown its strengths in optics and other fields, including the automotive LiDAR segment.

Things work differently in the audio sector. Creating a vibration amplitude strong enough to be heard through a tiny IC is extremely difficult. It is why dynamic and balanced armature drivers have dominated the audio market for over 150 years.

Besides the amplitude hurdle, the IC's cost also must be competitive to encourage adoption. Therefore, the breakthrough of MEMS in the audio sector is historical. The mosquito buzz from the speaker represented the beginning of a new era.

Q: How does xMEMS position itself besides designing piezoelectric MEMS chips?

A: xMEMS is an IC design company with about 80 employees. The company has released three generations of chips. The first was Montara, which disrupted the speaker market dominated by traditional dynamic drivers for 150 years. On average, Bluetooth earbuds with the Montara cost US$1,500. The high price reflected the value of MEMS.

Cowell, the second-generation chip, has gained traction in the true wireless stereo (TWS) market with a price of US$120. Cowell holds 70% to 80% higher costs than a dynamic driver. The gap will be narrowed to within 30% when a new Cowell solution is launched in the second half of 2024.

Application-wise, xMEMS sees the most adoption in TWS, gaming headsets, personal sound amplification products, and sleep earbuds. Gaming headsets have the most potential for growth rate.

The cost of the MEMS is similar to or even lower than that of a traditional dynamic driver, and it weighs less. Moreover, MEMS can enable seven times better spatial reasoning, helping gamers precisely identify their enemies' locations.

Cypress, the third-generation chip released in October 2023, enables the low-frequency original sound. The product represents a historic breakthrough as it tackles the low-frequency issue that dynamic drivers cannot handle well. After trying and learning about Cypress, an analyst commented that if xMEMS did not launch the chip, the industry likely would have to wait for 10 to 20 years before another company takes on the challenge.

Q: What is the crucial factor for xMEMS to revolutionize human audio perception?

A: I think the key is our employees, which are mostly Taiwanese semiconductor engineers. Only two quit their jobs in the company's six years of history. Many of the staff asked to invest their money in xMEMS.

While Taiwanese engineers are generally recognized for their ability to lower costs, xMEMS witnesses their innovation and capability to build things from scratch, something many would neglect. It has something to do with enterprise culture and value.

I often joke that you can randomly pick several xMEMS employees and form a band because many are good at and enthusiastic about music. Therefore, they can conduct professional testing on the products they develop.

Q: How will xMEMS redefine the MEMS speaker market? Will other market players accept your plan?

A: We plan to define the so-called MEMS speaker more precisely as the SoC speaker. xMEMS has been the only one to realize the SoC speaker, holding a technique lead for at least three years. The company has also enhanced its cost competitiveness rapidly. This is why we can define the MEMS speaker.

For now, we plan to operate through different partnerships. Details will be more explicit in the second half of the year. Traditional dynamic driver makers will also join the effort. They have heard the bugle calls for revolution and are willing to accelerate transformation to avoid elimination. Iconic companies in the application market will also participate in the collaboration.

Q: It is said that Apple has been pushing xMEMS for the solution for Vision Pro. Is Cypress the solution? What other segments will xMEMS enter in the future?

A: xMEMS does not comment on any potential interactions with customers. However, customers have asked for small-sized MEMS speakers for future electronic products. Many have not set requirements for MEMS' audio frequency because they know the technique bar is high. The frequency is also why xMEMS can impress its customers.

The market has discussed Saratoga, a product xMEMS has not released. I am not sure when we will bring it to the market. It could be six months later, at the end of this year, 1.5 years later, or even longer than that.

Saratoga's biggest advantage is that it can handle high, medium, and low-frequency sounds together. Imagine that in the future, wearing an Augmented Reality (AR) or Virtual Reality (VR) headset will no longer feel like putting a TV on your head, something many have complained about. In addition, smartphones will easily become mobile speakers with Saratoga.

The solution will also reduce an EV's weight without effort. Saratoga's volume is only one-three thousandth of that of a traditional home speaker. Of course, presenting better original sound and providing a more enhanced user experience is the point.