For instance, in the airline industry the products are flights from point A to point B. The context is safety, on-time arrival, availability, and customer service. Unfortunately, many airlines don’t practice all of those. This video is light and fun, but its message is dead serious for airline execs. Harvard Business Review published this study on it.

In the semiconductor industry the context is quality, delivery, price, and support. Customers need chips with great specs, but they also require high reliability, on time delivery, and a compelling price. And if there is a problem, customers need it fixed. At SiTime we make great products, with functions and specifications that are valued by our customers. But just as important, we build them with exceptional quality, deliver them on time, supply them at the industry’s best prices, and fix any potential issues immediately and completely.

Building a great company sounds like a business-school mantra, and it is. Seems intuitive or even obvious, and it is. But it is difficult to get right. In fact, while it is difficult to build great products, it is even more difficult to build a great company. One strives to give customers the best possible experience. But if even a single part of the whole is missing or poorly done then that experience is not a great one.

Today I read an article that called SiTime “the leader in MEMS oscillators”. Not just the technology leader, or quality leader, or volume leader, or price leader, but the leader. That means leading in everything – technology, quality, production, delivery, distribution, price, support, and every other detail that matters to customers.

Striving to do all these things is one of the ways we are building a great company, and we feel it is critical to delivering great products.

The reasons crystals do this are not completely understood, but they include surface imperfections, contamination, material interfaces, anchor stress, and spurious resonances. With careful design and fabrication, hops and pops can be reduced but not removed.

The size of these sets a limit on how stable a frequency one can derive from a TCXO. There are other limiting factors in crystals, like compensation error, hysteresis, and retrace – but hops and pops cause abrupt frequency shifts, are unpredictable, and are particularly insidious. This means that a quartz oscillator can switch from one to another frequency suddenly. In high precision TCXOs these steps can be tens to hundreds of parts per billion. That may sound small, but many precision applications like GPS and timing references can fail when this happens.

Getting these hops and pops down to even to a few tens of parts per billion requires highly developed processes that only a handful of quartz companies have. High precision TCXOs with minimum hops and pops are rare and expensive, but still often not good enough. I know of one company that has been buying the world’s best quartz TCXOs and then testing and discarding half of them because of hops and pops. Why not just ask the quartz suppliers to provide oscillators pre-screened? It seems their suppliers couldn’t do that.

But SiTime’s MEMS oscillators don’t show hops and pops. Why is this?

First of all, we use a completely different material system. We build our resonators in silicon, not quartz, and billions of dollars have been invested in silicon to make it the purest and most defect free material in the world. Second, we have very clean resonators. Their surfaces are contamination-free, again because of the billions of dollars invested in silicon process and fabrication. Third, we have far less external interaction with our resonators. We don’t have metal on our resonators or material interfaces, and nothing touches them. Fourth, we build complex three dimensional resonators rather than the simple plates that quartz uses, and therefore we have the design freedom to avoid the spurious edge reflections that plague quartz.

In short, when new technologies replace old technologies, they don’t always just reduce the problems with the old technologies – sometimes they eliminate the problems entirely.  That is what we have done with hops and pops.

At the very highest data rates even slight noise in the transmit or receive clocks can degrade data fidelity. And because of this the cleanliness standards on these clocks are very tight. The primary spec is that the jitter integrated from 12 kHz to 20 MHz must be less than one picosecond (one trillionth of a second). SiTime has the only MEMS oscillators capable of meeting this spec. In fact, SiTime’s new oscillators beat this by a factor of two, delivering clocks with only a half-picosecond integrated jitter.

In quartz, oscillators that can beat a picosecond are often called “extremely low jitter”. The extremely is well deserved – it is a very low number. For example, in half a picosecond light travels only 150 microns in a vacuum. An electrical signal in a copper wire travels at about half the speed of light. Thus a clock signal traveling in a printed circuit board goes about 75 microns in a half picosecond. That is roughly the width of a human hair.

So think of what these oscillators are doing. They are producing electrical waves that travel out from their pins, across PC boards, and into application ICs. These waves move at half the speed of light while their edges are where they should be to the width of a human hair. Pretty amazing!

What is the connection to SiTime? It is that new ideas do not evolve incrementally; they burst on the scene exponentially. It took only ten years to develop an airline industry after this heroic pilot crashed his way across the country. Initially airlines delivered the mail, then soon people.

And who were the losers? The railroads. Before the airlines trains dominated mail delivery and transportation. Who today has letters delivered by train? Who travels across the US by train?

Was this foreseen by the railroad companies? Did they invest in airlines to maintain their control of transportation markets?  No. The airlines were started by entrepreneurs driven to do something new, not by railroad execs.

When industries change the incumbents rarely generate the change, and even less often manage the change. This is the case with MEMS oscillators. The companies driving this are not the incumbent quartz oscillator companies. The MEMS oscillator companies are started and run by entrepreneurs driven to do something new.

Now there are some exceptions to the rule. IBM and HP are examples of companies that have prospered across technology changes. But for every company that bridges the change there are many that don’t. And so time will tell how it works out in this case. It is likely that a few wise quartz companies will find ways to stay in the game, but most will not.

Encore is our newest technology platform. With Encore we can build oscillators with 40 dB lower phase noise, 20x better frequency accuracy, and 100x better short term stability. We can build XOs, TCXOs, VCXOs, DCXOs, and a bunch of special functions. For example, we are sampling SiT8208’s and 8209’s, the world’s only MEMS oscillators with sub-picosecond 12 kHz to 20 MHz integrated phase jitter; and we are sampling SiT5000, 5001, and 5002’s, the world’s only sub-ppm MEMS TCXOs. These parts don’t just squeak by with one picosecond integrated jitter or one ppm accuracy, they typically are delivering half that.

While these are the world’s only MEMS oscillators with these capabilities, their performance levels are also rare in the quartz field. Quartz oscillators with sub-picosecond integrated jitter are often advertised as “extremely low jitter”, and not everyone can build them. On the accuracy side, sub-ppm TCXOs are also rare, and not everyone can build them either.

How can we develop oscillators in just a couple of years that most quartz companies can’t develop after decades? I describe this in my last post, but here is another way to look at it: It is because our technology is based on modern semiconductors and is highly leveraged; while in comparison the quartz technology is specialized and far less leveraged. We design our products within software ecosystems we don’t have to write because others have come before us and already written and tested them. We build our parts in billion dollar fabs that we don’t need to own because we leverage the huge investments of semiconductor companies. We package and test our parts in standard IC packages with standard tooling in standardized factories that we don’t need to build because we leverage a standard IC production infrastructure.

None of this works for the quartz guys. When they need to design new parts they don’t have an industrial design infrastructure supporting them but are mostly on their own. When they need to build their quartz blanks they must build their own specialized fabs. When they need to package and test their parts they must buy specialized equipment for their own specialized facilities. This takes time, soaks up capital, and hinders innovation.

The net result is that we innovate faster and push our technology further that the quartz incumbents. We move in strides in an industry where progress has been measured in steps.

SiTime has just made such a quantum leap. We are now sampling 0.5 ppm TCXOs and we seem to have leaped from 10 ppm and skipped years of intermediate steps at 5 ppm, 2.5 ppm 2.0 ppm, and 1 ppm. How is this possible? It took quartz decades to step through these grades. What is behind this quantum leap?

Well, the answer is not a secret. It is the same as I have been writing and speaking about for years. It is simply that SiTime is an IC company, acts like an IC company, works like an IC company, and thinks like an IC company. That means we live under Moore’s law. Generation over generation we deliver more capability, more precision, and more value.

Yesterday I read an article in an electronics trade magazine in which a quartz oscillator exec was quoted as saying, “The rapid development of MEMS oscillators is also the choice of all new volume applications where a mere clock facility is required and the oscillator is fabricated into the silicon.” I’m happy enough with the first clause – we are the choice of all new volume applications. But the second clause – that MEMS in only suitable to supply a mere clock facility – is not right. SiTime now has high precision and also low jitter oscillators that are well beyond mere clocks.

SiTime has just made one of those rare quantum leaps.  We have appeared where we were not expected.

The incumbent quartz oscillator companies did not see this coming; they did not expect this could happen. I understand their thinking though, since it would seem highly unlikely that a strong industry with an entrenched technology could be displaced. But actually it is common; in fact it is the rule rather than the exception. It happens in every human endeavor, in every technology we invent, and in every industry we build.

One example I have recently been thinking about is the steam locomotive industry, and in particular Baldwin Locomotive Works. Baldwin was a hugely successful locomotive manufacturer founded in the 1830’s. In my opinion it built the most beautiful steam locomotives ever designed. Baldwin built the first locomotives in the United State and refined its technology to where it had an intrinsic beauty. This was a pinnacle of high technology and was responsible for tremendous socioeconomic change in the United States. Standing near these engines today I am always impressed, and to be near an operating one is an experience that one remembers.

But Baldwin didn’t understand how diesel locomotives could replace steam. In the early 1900′s it failed to grasp the key thing that made diesel important: Diesel was supported by a huge and growing industrial infrastructure forming around the internal combustion engine. This infrastructure included vast investments in engines and fuel. The steam infrastructure was dwarfed by comparison. Since Baldwin could not leverage the internal combustion engine it chose instead to emphasize powerful mainline steam engines, developing ever larger brutes to compete with the new yet still smaller diesel engines.

But within a few decades diesel engines grew in power, exceeded steam in durability, and became less expensive to operate. The story ended poorly for Baldwin when it eventually tried to transition to diesel but hedged too much and moved too late. It finally failed unceremoniously in the 1970’s. This happened to all the American steam locomotive manufacturers: Baldwin, Alco, Lima, Porter, and others, all gone or parted out.

A similar thing is now happening in the quartz oscillator industry. They can’t leverage the growing strength of CMOS and the silicon fab industry, and they are not making the transition. In one sense, what SiTime is doing is revolutionary – we are replacing an industry. In another sense what we are doing is common – we are turning the page on an old way of doing things and starting a new page with a better way of doing things.

SiTime introduced the world’s first MEMS oscillator, the SiT8002, in 2007. Since then, we developed and introduced two additional generations; the SiT8102, the first high performance programmable MEMS oscillator with 10 PPM stability and wide output frequency, and the SiT8003, the world’s lowest power programmable oscillator.

In the four years since 2007, we have shipped over 50 Million oscillators into consumer, computing, networking and industrial applications. Along the way, we developed a range of derivative products, such as the world’s lowest jitter spread spectrum parts (SiT9001), the world’s first spread spectrum differential oscillators (SiT9002), the world’s first differential MEMS oscillators (SiT9102 and SiT9107), the world’s lowest power high frequency oscillators (SiT8004), and the world’s thinnest oscillators (SiT8003XT). We also introduced the world’s first multi-output MEMS clock generators (SiT9103, SIT9104 and SiT9105).

And now, with the SiT8208 and SiT8209, we’re offering higher performance, more features, and revolutionary innovation. Our fourth generation in five years. These new oscillators have 12 kHz to 20 MHz integrated jitter of 600 femtoseconds, frequency stability of 10 ppm over -40 to +85C, and output frequencies from 1 to 220 MHz. They include a range of programmable features, for instance output drive strength and supply voltage, and are available in standard packages. This is a huge accomplishment and implies a whole list in signal quality and integrity successes. These replace high-end quartz oscillators for low frequencies, overtone oscillators for high frequencies, and SAW oscillators for low jitter. There are no quartz oscillators that can concurrently serve these needs or meet these specs.

These parts are targeted at applications that require exceptional quality clocks. Examples include multi-gigabit asynchronous transceivers such as 10GbE, GbE, SATA, SAS, Infiniband, Fibre Channel, PCI Express, and USB3.0; synchronous telecom applications such as SONET and Synchronous Ethernet (SynchE); wireline applications including xDSL, PLC, and DOCSIS; and wireless applications including WiFi and WiMax.

The point is that we are innovating faster than the incumbents in the quartz industry. Unlike quartz, our products incorporate highly evolved circuits and use semiconductor manufacturing techniques, allowing us to drive higher performance and improved features at a very fast rate.

Voltage controlled oscillators are commonly used where frequency or phase must be matched to an external clock. The most common example is in data receivers. When data is transmitted, perhaps on a wire, an optical fiber, or a wireless link, it is encoded at a rate set by the transmitter’s clock. The data must be parsed into the receiver at exactly the same rate and at the right phase, or it will be corrupted. To do this the receiver must match its clock to the transmitter’s clock.

The receiver compares clock information embedded in the signal against its own clock. It then fine tunes, or “pulls” its clock as needed. A typical receiver uses a quartz crystal to generate its clock. It pulls the clock by generating an analog voltage that it applies to varactor diodes that load its quartz crystal. Varying this voltage varies the capacitance on these diodes and tunes the frequency.

So far so good, but there are problems with these quartz VCXOs: One is that varactors are non-linear, so the oscillator’s voltage to frequency function is non-linear and varies from part to part. This makes it harder to build a good control circuit. Also, the varactors and quartz crystals have limited pull range. If a quartz oscillator were to be designed to pull more than a couple hundred parts per million (ppm) its performance would degrade dramatically. And sometimes a couple hundred ppm is not enough.

Worst of all, quartz VCXOs show frequency hops, pops, and jumps that are a function of the crystal’s temperature and the oscillator’s pull. These quick and sometimes retrograde frequency shifts often cause receivers to lose lock, which means that they stop following the transmitter’s clock. Losing lock is severe and results in data loss and dropped connections.

None of these are problems for SiTime’s MEMS VCXOs. This is because we do not pull our resonators; instead we digitize the analog control voltages and adjust the output frequencies with fractional-PLLs. Our voltage to frequency functions are highly linear and do not vary from part to part. Our fractional-PLLs adjust the output frequencies exquisitely smoothly. Our frequency adjustment ranges can be set from very small for fine control to very wide for robust tracking. And our MEMS resonators have no activity dips so they have no hops, pops, or jumps.

What are the benefits? The SiT380X family of VCXOs offers sub-picosecond integrated jitter, sub-percent pulling linearity, control ranges configurable from ±25 ppm to an astounding ±1600 ppm, and output frequencies from 1 to 220 MHz. This combination is physically impossible for varactor-pulled quartz and is an example of modern technology surpassing limits that were once thought to be insurmountable.

That is not the case for the quartz industry. The industry is concentrated in Japan and has seen extensive disruptions. Many of their factories were damaged and require repair.  One large quartz oscillator factory is in the nuclear exclusion zone and may not be reopened.  Thankfully there is little reported loss of life.

In addition to the direct damage from shaking, water, and fallout, the quartz industry is reliant on stable electrical power. Growing crystals requires months of stable autoclave time, ceramic packages require firing in kilns, and IC’s require precisely controlled fabs. Power shortages in Japan have thus caused widespread problems, far beyond the geographic limits of the physical damage.

How will this play out over the next year? The wise answer is that it is too early to know, but we have seen the start of it. The quartz industry has long lead times for some materials, and because of that they carry large strategic inventories. Therefore disruptions may not be evident for months. Customers buying quartz crystals and oscillators may think their supplies are assured, after all it has been two months since the earthquake, but shortages may still develop.

After learning that SiTime is unaffected by the earthquake, people often want to know if we are seeing a surge in orders. In the two weeks after the earthquake many customers wanted to know if we could cover for quartz shortages—which of course we can. Recently we have begun seeing increases in production orders. Our large customers have good visibility into their quartz suppliers and are shifting more of their consumption to MEMS.  It’s not a panic; it is more of a balancing of lead times and hedging of supply risk.

So how have SiTime and the quartz industry weathered the Japanese earthquake? SiTime is perfectly fine. The quartz industry has some problems, and the extent of those problems will not be known for a few more months. Large customers with supplier visibility are moving production incrementally from quartz to MEMS.