How SiTime is Redefining the “Heartbeat” of Electronics

A Peek into the Process of Launching the “Precision Timing” Category

Over the past few months, I’ve told the stories of various companies that have successfully launched new categories that fundamentally improved how industries operate and the world works. I thought it might be interesting to take a look at a company in the midst of launching a new category today.

The Connected Intelligence revolution is built upon three core elements: networked computing infrastructure, correlated data, and intelligent software. None of those work without a healthy “heartbeat”. That heartbeat is in the form of clock signals that tell processors when to execute the next instruction, tell networking equipment when to send the next packet, and place highly precise timestamps on data so that it can be correctly interpreted and correlated by intelligent software.

In our bodies, the healthier our heart is, the better our brain, lungs, and muscles work. The same is true in electronics. The healthier the timing heartbeat, the better everything works — there are fewer errors to correct, everything is synchronized and operating at peak efficiency, and the device works better for longer. A “healthy” electronic heartbeat is one that is accurate and reliable over time in real-world operating conditions.

This is nothing new. The electronics industry dates back to at least the 1920s. Clock signals became increasingly important with the advent of transistors in 1947, integrated circuits in 1960, and microprocessors in 1971. A basic timing circuit can be built with an inductor and a capacitor, regularly producing a pulse based on the inductance and capacitance values, but it won’t be very precise, accurate, or stable. For anything requiring a reliable clock signal, engineers soon turned to quartz crystal oscillators.

Crystal Oscillators

In the late 19th century scientists discovered an amazing property of quartz crystals called piezoelectricity. The word comes from the Greek words piezo (meaning to press) and elektron (representing electrical current). If you stress certain crystals in a certain way, they will produce an electrical charge, and inversely, if you apply an electrical charge, they will distort. When the charge is removed, the crystal returns to its original state and the process can repeat.

In the 1920s scientists used this property to create the first crystal resonators. They learned how to carefully cut a crystal to a specific size and shape so that it would naturally vibrate at a desired frequency when a given electrical voltage is applied. This is called the crystal’s resonant frequency and is specified as the number of vibrations per second, with the unit being hertz (1 Hz = 1 cycle per second). Scientists also found that quartz crystals could be cut to resonate at the frequencies used in everything from clocks and watches (typically 10s of thousands of cycles per second — most modern quartz watches use a 32 kHz crystal) to radios (10s of millions of cycles per second — for example FM radio bands are between 87.5 MHz and 108.0 MHz).

Engineers build a useable oscillator component by creating an electronic circuit around the crystal resonator. The circuit takes the generated electrical pulse from the quartz crystal and feeds it back into the crystal so that it oscillates at the resonant frequency. The circuitry also turns the resonant pulses into an output that can be used by the rest of the electronic device. Since the science involved in getting a crystal to vibrate is a very different discipline from that of designing useful electronic signals, these functions are typically performed by separate teams, often at different companies.

The end result has generally worked well for the needs of the electronics industry for decades. Over the years engineers learned how to compensate for some of the shortcomings of quartz crystals.

One of the challenges with using quartz is that a crystal’s resonant frequency varies with temperature. Oscillator manufacturers compensate for this in a couple of ways. The simplest, smallest, and least expensive approach is to have a sensor that measures the temperature around the crystal and then have the circuitry adjust the frequency of the clock signal produced to correct for the error. A more expensive approach that provides a more temperature-stable clock is to actually create a tiny oven, keep the oven at a constant known temperature, place the quartz crystal inside this oven, and then have the circuitry produce the correct clock frequency based on this known temperature. The industry uses the acronym XO to generically represent crystal oscillators (xtal oscillator); TCXOs are the temperature-compensated variant; and OCXOs are oven-controlled crystal oscillators.

Design engineers also compensate for shortcomings in all flavors of crystal oscillators (XOs, TCXOs, OCXOs, and even more variants). Vibration and shock can disrupt the signal from a crystal oscillator, so for applications likely to encounter these issues (e.g. in cars, avionics, cellphones, etc.), they build damping mechanisms around the oscillator to minimize the impact, or place the oscillator in an area of the system less sensitive to these stressors. The electronic “noise” produced by power supplies and other electronic components can also alter the frequency of an oscillator’s clock signal, so engineers often have to make specific design and layout changes — such as placing the oscillator on a “quiet” part of the circuit board, even if that is a long ways from where the clock signal is needed (an approach which creates its own integrity challenges).

Up until recent years, these approaches have been sufficient to keep crystal oscillators working well enough for the vast majority of electronic products.

What’s Changed?

Most new categories involve a new approach to solving an old problem. The world embraces the new category because the new approach dramatically improves on old approaches, is radically different — not just better (and therefore is defensible), and can easily be adopted by the target users.

Even then, most innovations fail to ever reach broad adoption because humans are naturally change resistant. Unless we are forced to, we want to keep doing what we’ve always done, how we’ve always done it. Sometimes we are forced to change because a competitor (often a new entrant) adopts the new way and we have to follow to remain competitive. Other times, the world changes and the old approach simply no longer works.

The Connected Intelligence revolution has placed tremendous new demands on electronics, demands that quartz crystals are struggling to meet. For simplicity’s sake, I’ll talk about these changes in three buckets: speed, environment, and synchronization.

Everything is getting faster. Computer processors run at faster clock speeds (up to about 5 GHz are becoming common) and networks send data at faster speeds (mobile data speeds now exceed 100 Mbps and Ethernet now operates at 100s of Gbps). The increased speed translates directly into less room for error — the clock signal has to be perfect for everything to keep working. Engineers pay close attention to key parameters such as Jitter, Frequency Stability, Phase Noise, and Allan Deviation, all of which need to be within increasingly challenging bounds for equipment to operate as expected. As a simple example, 5G wireless networking demands a clock signal that is 10x better than was required for 4G.

The second challenge is the environment in which electronics are expected to operate. Until recently, most electronics sat in a stable, controlled environment — whether that be a living room, an office, or a data center. Today, as connected intelligence is built into every type of product, processors and networking (with their clock sources) are increasingly expected to operate reliably in harsh environments, with wide and rapid temperature swings, constant vibrations, and regular shocks. For example, smart doorbells and security cameras are mounted outside our homes; radio equipment no longer just sits in climate controlled cabinets at the base of solid steel towers, but is mounted on utility poles, traffic lights, roof tops, and throughout stadiums; IoT sensors are everywhere (e.g. from factory floors to farmlands); and we carry sophisticated computers everywhere we go — in our purses, pockets, and attached to our wrists.

Just think of the modern automobile. Deloitte estimates that electronics represent more than 40% of a new car’s cost. Our lives depend on those electronics operating reliably, even when driving 70 MPH over pot-holes and changing road conditions, with rapidly changing temperatures in the engine compartment and widely varying outside temperatures.

The third challenge is the need for precise synchronization. Increasingly Artificial Intelligence-driven systems are making life-and-death decisions based on data collected from a variety of distributed systems. The software has to be able to correlate all of that information, which requires very granular, reliable, and consistent timestamps on all of the data. Many systems rely on signals from GPS satellites to provide that precise synchronization, but a GPS signal isn’t always available (e.g. indoors, when blocked by a tall building, during severe weather, etc.). During these gaps, the internal clocking systems have to be precise enough to maintain the reliability of the data timestamps. Even worse, GPS signals are not secure and can easily be spoofed. Mission critical systems need to be prepared for a hostile attack against the GPS system.

SiTime’s Innovation

In 2003 Aaron Partridge was a researcher at Stanford University and Markus Lutz was a senior engineer for German technology powerhouse Bosch. Together they were working to improve MEMS technology and find new applications of the technology.

MEMS stands for Micro-Electronic Mechanical Systems. MEMS devices are typically created using traditional semiconductor processes. They are tiny machines whose moving parts are so small they can only be seen with a microscope.

You probably use MEMS technology everyday without even realizing it. For example, your smartphone uses MEMS for the microphone, for the accelerometer that detects when you rotate the phone, the image stabilization that makes the videos you record watchable, for radio wave filters so that the phone can use the best spectrum to connect to your wireless provider, and even sensors to protect the battery from overcharging. Because MEMS are so small, they help many products (such as smartphones) incorporate powerful features in a small package.

MEMS are also essential for many life-saving and life-changing applications. Just consider some of the ways that MEMS devices are used in modern cars. MEMS-based products are used to prevent accidents, such as in anti-lock breaking systems and in yaw rate sensors that detect when your car is starting to spin or might be about to roll and automatically take corrective action. In the midst of an accident, MEMS devices protect occupants in the car, for example determining when a car airbag should deploy, and locking the seatbelts to keep passengers in a safe position for impact. MEMS based sensors also help identify when an accident has occurred so that a call for help can automatically be placed. The tiny size of MEMS devices have enabled applications that previously couldn’t even be considered (for example airbags in smaller spaces). Their small size and power efficiency also help these devices take action faster than their pre-MEMS equivalents at times when micro-seconds matter.

MEMS-based devices have also been essential to the development of new, previously unimaginable features that make cars easier and safer to drive including adaptive cruise control, lane keeping assistance, and automatic emergency braking. While these features primarily rely on optical systems like LiDAR, those optical systems rely on optical MEMS systems including micromirrors.

One of the challenges facing MEMS devices is that even microscopic particles can keep the machines from working correctly. In February 2003, Lutz and Partridge filed a patent for a new method to solve that problem by perfectly sealing a MEMS device. The two started working to identify applications that would benefit most from this improvement.

The pair saw the opportunity to build a better resonator and timing oscillator using MEMS technology. Since Bosch owned the patent, Lutz and Partridge took the idea to the company. Bosch agreed that there was market potential, but didn’t see it as aligned with the focus of the company. With Bosch’s blessing, the two decided to pursue it outside of the company. They sought funding and began building a team. They decided to name their company SiTime because they were going to build timing products entirely using silicon (Si is the scientific symbol for silicon) and semiconductor processes.

The company’s first efforts to build a MEMS-based resonator confirmed the benefits of using MEMS for timing. Over the coming years, the company learned how to use the technology to overcome many of the challenges faced by quartz crystals. The machines are so tiny and lightweight that they are not impacted much by vibrations, shock, and electrical noise. Since they are perfectly sealed they aren’t impacted by air flow or air pressure and they don’t “age” like crystals (which change resonant frequency over time as their mass changes). They can be produced in a much smaller package and require much less power than crystal resonators, making them particularly well suited for mobile and IoT applications.

However, those early efforts also uncovered that it was really hard to produce a resonator that produced exactly the right frequency. The new company realized that what was needed was a more systems-level approach to timing — excellence not only in the component parts of the MEMS resonator, electronic circuitry around it, and even the packaging — but most importantly that all of it be designed together as a complete system. They would need to develop all of it in a much more integrated fashion than the timing industry typically had and with the ability to easily fine-tune/program the performance. This made sense for SiTime since all the parts were Silicon-based and manufactured using standard semiconductor processes.

The company released its first product in 2007. Like most disruptive technologies, that first iteration wasn’t nearly as good as the best oscillators on the market. But, while quartz-based oscillators were barely improving (constrained by the industrial-age processes around physically cutting crystals), SiTime’s products were improving much more rapidly (benefitting from the computer-age/Moore’s Law advances in semiconductor technology).

To drive adoption, many of the company’s products were designed as direct pin-level replacements for traditional oscillators. A customer could pull a crystal-based product out of the circuit design and drop in a SiTime product and know that everything would continue to work.

As the technology continued to improve, some customers recognized that only SiTime’s products could meet their unique needs — whether that be a component small enough, low power enough, and precise enough to fit into an attractive smartwatch product, or one that was resilient enough to keep operating well through the violent launch of a rocket and the extreme temperature variations of a low-earth orbiting satellite. These customers were happy to work with SiTime and buy premium products uniquely produced by the company.

There are many different parameters that can be used to evaluate different timing products, and different ones matter to different customers and for different applications. In 2011, the company decided to shift away from just making pin compatible products to focusing on products that uniquely leveraged the advantages of MEMS that customers were looking for. The first products from this strategy came to market in 2013 and were 85% smaller than any quartz timing device.

Year-by-year SiTime products closed the gap and caught quartz technology on other performance metrics so that by 2022 SiTime’s products had clearly reached at least parity on all the factors that had traditionally mattered most to engineers. While century-old quartz technology is improving very slowly, MEMS-based technology is still on a steep improvement curve, continuing to widen the gap with crystal timing devices.

Furthermore, with the changing demands driven by the Connected Intelligence revolution, new factors were commonly becoming critical to engineers, including shock, vibration, temperature resilience, and airflow. SiTime’s products were already dramatically outperforming what crystal-based timing devices could ever hope to achieve on these factors.

Creating a New Category

While Lutz and Partridge continued to drive technology innovation within the company, they quickly brought in experienced management to lead the company. Rajesh Vashist has been CEO since 2007 and the company has grown to over 350 employees. In 2019 SiTime had $84M in revenue and completed its initial public offering (IPO) on Nasdaq. The IPO was priced at $13/share and closed trading the first day at $18.65. By 2022 revenue had more than tripled to $284M. Today the stock is trading at about 10 times the IPO price.

The company knew that what it had developed wasn’t just better than traditional oscillators, it was something altogether different. Being considered “just like” crystal oscillators had helped SiTime products gain acceptance with customers, but as the company pulled away from legacy products along important dimensions, being in the wrong category was hurting their ability to help customers in entirely new ways.

New categories enable new possibilities. Because engineers didn’t see SiTime’s products as a new category, most of them failed to see how crystal’s constraints that had held them back for decades were now being broken, enabling entirely new types of products that they could design.

New categories also directly address how the world is changing. Engineers were like the “frog in the kettle” not recognizing the significant new demands increasingly being placed on timing by the Connected Intelligence revolution. They were struggling to accommodate new requirements with old technology rather than jumping beyond crystal’s constraints by embracing the possibilities enabled by SiTime.

SiTime knew they had to create a new category and started using the name “Precision Timing”.

Category creation typically follows three major steps: 1) Define the problem, the solution, and the category. 2) Launch the category, often with a “lightning strike”. 3) Own the category by continuing to lead in capability, mind-share, and market-share.

The biggest immediate need was to clearly define the category.

While it would be easy to define the category based on technology (e.g. “MOs” — MEMS-based oscillators), SiTime has become more passionate about improving timing than about just advancing MEMS technology. In this article, I have focused on the largest product segment of the market, which has historically been well served by crystal oscillators but now is better served by MEMS-based oscillators. However, there are other product segments that could also benefit from SiTime’s focus on precision, resilience, scalability, programmability, power, size, tight integration, and value.

As an example, the company recently announced the acquisition of the clock product portfolio of Aura Semiconductor. Unlike oscillators, a clock can provide multiple different clock signals at different frequencies for different needs within an electronic device. These products will clearly benefit from SiTime’s unique capabilities. Most clock products don’t have an integrated oscillator, but instead need to be connected to some kind of timing source, whether that be an existing SiTime product or another source. SiTime’s focus on systems-level integration will bring a number of benefits to this product segment including simplified design, smaller size, improved performance, and overall efficiency.

The SiTime team worked to nail down a category definition that would capture what was truly different and that would redefine the industry. Often times, a good category definition captures that one thing that sets the category apart — any company/product with that one thing would be in the category and any company/product lacking that one thing would not be in the category. But what makes SiTime’s innovation special is that “good” timing is not about a single “thing”.

Different customers care about different aspects of timing depending on the nature of their application. All care about having an acceptably stable clock signal at the right frequency. Some care about very high stability; others specifically care about stability in the face of temperature change; others care about vibration and shock; still others care about size and power. Historically, engineers had to choose where to compromise to accommodate what mattered most to them. If SiTime picked one criteria for the category definition, many competing products could meet that one criteria while failing on several other criteria the team didn’t pick but that still matter to many customers. Unlike traditional crystal oscillators, SiTime’s products don’t force engineers to make compromises, providing a high level of tunability to meet the needs of almost any specific application.

The team ended up defining the category this way: “In today’s intelligent, connected world, Precision Timing uniquely enables products that are smarter, faster, and safer by delivering precise, reliable clock signals that always work in every environment. To provide these benefits, Precision Timing uses semiconductor technology, systems innovation, and programmability.” This definition clearly includes the company’s current products but also could include the clock products acquired by Aura and a variety of other potential products and services.

With the category defined, the company could focus on an impactful launch of the category.

The challenge that any category creator faces is convincing customers to embrace a totally different way of working. The easy and safe approach is to keep using the old category. For engineers, they knew how to design quartz-based timing into their products. They’d learned how to work around the limitations and constraints imposed by crystals and most of the time these legacy devices worked as expected. How could SiTime’s new category break through?

For some customers, that breakthrough happened organically. Innovative product companies push the boundaries of how small, how accurate, and how mobile their products can be. Some of these product designers discovered that only SiTime’s MEMS-based timing devices could meet their design requirements. Over the years Precision Timing has been built into many iconic products that you see and use everyday.

Other customers needed a little more encouragement. In October 2020 a fire broke out at a manufacturing plant in Japan that makes a critical component for most of the largest producers of crystal oscillators. Because of the unique manufacturing required for quartz crystal oscillators, providers couldn’t quickly switch to another factory. The shutdown dramatically impacted the supply chain for XOs, TCXOs, and OCXOs. Thankfully, SiTime was not impacted by the fire, but because it uses standard MEMS and semiconductor processes, even if it had been, the company could’ve quickly and easily adapted around any single supply chain failure. (The company’s primary manufacturing partners operate many facilities around the world, providing SiTime with a robust supply chain.)

As a result of the fire, many electronics companies who previously had been hesitant to switch technologies gave MEMS-based oscillators a try. Those companies shipped billions of dollars of their products that otherwise would’ve been unable to reach their customers. SiTime gained many new customers and the vast majority of those companies had great experiences and have continued as loyal SiTime clients. This has been great for the company, but it didn’t provide the best opportunity to explain to the world how Precision Timing is different and enables engineers to take advantage of all the benefits offered by the new category.

Helping customers realize the radical nature of a new category often requires making a major splash with an event that is sometimes called a “lightning strike”. In their book Play Bigger, the authors define a lightning strike as “an event meant to explode onto the market, [and] grab the attention of customers, investors, analysts, and media”. A lightning strike takes preparation so that the full category story can be compellingly told in context.

This past September SiTime announced Epoch, a new product line providing twice the accuracy of traditional OCXO products in a package 1/25th the size and consuming 1/3 the power. The company focused on making this launch the “lightning strike” that would establish the category.

Source: SiTime

In a podcast at the time, Executive Vice President of Marketing Piyush Sevalia explained “Everything is getting siliconized.” He used the example of vacuum tubes moving to transistors, incandescent bulbs moving to LED bulbs, and hard disk drives moving to solid state drives. “We think timing is getting siliconized in the same way. Because silicon is one of the most unique technologies on the planet providing more features, higher performance, better reliability, lower cost over the long term; it just gives you many benefits that you can’t get with the incumbent technology.” In short, the company was positioning this moment in time as the transition from the old world of timing to the Precision Timing future.

Understanding why that transition matters can be hard for design engineers, much less for those of us with limited understanding of what matters when designing an electronic product. Let me walk through a couple of examples to help explain.

Consider how watches have changed. I’ve had a watch on my wrist for most waking hours of my adult life. The mechanical watches of the past are amazing in their own way with finely crafted mechanisms, often relying on a quartz crystal to keep the watch accurate within a few seconds each month. Today we expect our watches to do so much more — to notify us of messages, to tell us the weather, to track our activity, to monitor our health and safety, and yes to tell us the time. While an accuracy of a few seconds a month is probably all I need to make it to my appointments, much greater precision is required by the device itself to stay connected and to act in synch with other systems around the world. For everything that I expect of it, the product designer needs to squeeze an amazing amount of technology into a tiny space, and design all of it to maximize battery life. Like any watch, my smartwatch will need to endure occasionally accidentally getting banged against things, being swung around as I participate in sports, and surviving the unexpected rainstorm. Given everything I’ve described in this article, I hope you can see how modern smartwatches would be very challenging to design using traditional quartz crystals. Precision Timing makes possible the smartwatches that previously were considered science fiction.

Also consider how automobiles are changing. A typical car from even just a decade ago was little more than a powerful gas-powered engine attached to a strong steel frame, and surrounded with materials to make it beautiful and comfortable. Sure there was usually a simple radio/CD player to keep us informed and entertained on our daily commute and headlights to light our way, but the electronics were very basic. SiTime recently identified four current automotive trends in a recent SiTime blog post: electrification, shared automobile models, active safety systems, and driving automation. The net result is that today’s car has become a rolling data center with high speed data networks and constant connectivity to the digital world around it. Not only do precision, size, and energy efficiency matter (as with smartwatches), but the environmental requirements are taken to an even higher level. Intense shocks and vibrations, engine compartments that exceed 100 degrees celsius, and the ability to stay synchronized even sitting through the coldest winter night. Designing tomorrows cars will become almost impossible without Precision Timing.

SiTime will need to continue to work Precision Timing into how electronics design engineers think about building the heartbeat into their products. This likely will require an ongoing string of impactful announcements that tangibly improve the design experience and help engineers be successful in producing great products. I call these ongoing announcements “thunder claps” that echo the core message of the lightning strike while catching the attention of the audience and helping them experience the change that is happening.

There are already thousands of satellites in low earth orbit, millions of electric vehicles on the road, and hundreds of millions of smartwatches all relying on Precision Timing. Before long, Precision Timing will become the standard practice for engineers designing almost any kind of electronic product.

Let me know if I can ever help you with launching your new category to the world!

Full disclosure: I have done consulting work for SiTime in the past but am not currently engaged by the company.