Silicon Labs CMEMs

Enabling Superior Frequency Control with the Integration of CMOS and MEMS Technologies

Since the mid-1900s, the frequency control market has been dominated by quartz crystal resonators and quartz-based oscillators. And even today, virtually all electronic equipment depends in some way upon a machined quartz crystal to generate at least one of many potential operating frequencies. Crystals or crystal oscillators (XOs) are used in the great majority of billions of electronic devices in the market today, ranging from guitar amplifiers to wrist watches and from smart phones to forklifts.

By leveraging the enormous economies of scale from the billions of crystals used in the electronics market each year, the manufacturing of quartz crystals and quartz-based oscillators has reached new levels of sophistication and has delivered smaller, thinner and higher-frequency solutions. And up until recently there have been few viable alternatives because of the well-understood characteristics and stability of quartz-based piezoelectric resonators, which have made it easy to build a crystal oscillator (XO) with proven performance.

However, over the past few years MEMS (Micro-Electro-Mechanical Systems) based oscillators that mimic the XO architecture by having two components (resonator and amplifier) have made inroads into the frequency control market by offering superior reliability coupled with cost benefits in all package sizes, and particularly in small package footprints where XOs have a high cost structure. Further, with the introduction of single-chip MEMS solutions which layer the MEMS resonator directly on top of the CMOS amplifier base, additional advancements in MEMS oscillator reliability, programmability, temperature stability and cost levels are now possible.

 

Crystal Oscillators

Crystal oscillators are designed to operate over a wide range of frequencies from a few kilohertz up to several hundred megahertz. Crystal oscillators combine a quartz resonator with an amplifier circuit in a hermetically sealed ceramic package with a metal lid. The ceramic package and metal lid provide a highly protective casing for the very fragile crystal and prevent damage to the assembled components. In general, the amplifier circuit leverages the piezoelectric properties of the crystal by using electrical feedback to create a resonance or oscillation at a specific frequency controlled by the size, cut and plating of the crystal resonator. To support the wide range of frequencies required by the electronics industry, frequency control suppliers must design, stock and manufacture hundreds or even thousands of different custom crystal resonators.

Crystal-based solutions face manufacturing challenges in addition to custom crystal resonators. Portable devices represent a huge percentage of the overall crystal market. Thinner and smaller portable devices require increasingly smaller form factor components from all suppliers. This requirement presents problems for crystal-based oscillators as shrinking the size of the quartz resonator at all the desired frequencies creates manufacturing complexity and reliability challenges from smaller, more fragile crystals. In addition, in every market, a major issue crystal-based solutions face is their inherent sensitivity to environmental factors such as shock, vibration, thermal stress and manufacturing variability from lot-to-lot, which can lead to start-up issues and subsequent field failures.

 

MEMS Resonators

Over the past few years, MEMS-based oscillators have become a viable alternative to quartz-based solutions for a number of reasons. First, MEMS oscillators are manufactured in silicon-based processes, subject to extremely high quality control, and thus produce very reliable performance over many billions of units, provided they are properly designed, guaranteed and characterized by the supplier.

Second, as a direct result of their silicon-based processes, they are subject to Moore’s Law, which dictates an ever-increasing amount of processing power for an ever-decreasing cost. In other words, smaller, more advanced silicon-based devices will inevitably cost less over time. Crystal-based solutions unfortunately suffer from the inverse of this law, dictating that their material becomes more expensive as it grows smaller due to the manufacturing difficulties mentioned above. In addition, as crystal manufacturing becomes more difficult and costly, the yield of crystals decreases due to more and more fragile, smaller devices.

Again, the third advantage is based in the silicon-based process. As a silicon solution, MEMS oscillators can be designed to be inherently more robust against environmental factors. This is not to say that all MEMS-based solutions are equally good in this respect. The product design greatly determines how well one MEMS oscillator performs versus another. The fact remains, however, that a silicon solution can be designed to be more robust against shock and vibration than a crystal, especially a small crystal.

 

First-Generation MEMS Oscillators

The first generation of MEMS oscillators is similar to quartz oscillator architectures in that they combine two physically distinct components, the resonator and the amplifier IC/base die, which compensates for any drift in the resonator frequency. The use of MEMS introduced a major improvement in the manufacture of oscillators by eliminating the complicated materials processing techniques required by quartz-based oscillators, as well as replacing the more costly ceramic package and metal lid used in quartz oscillators with a more economical plastic package.

However, this first-generation approach is still inherently limited by the two-component architecture also used with crystal-based oscillators. These limitations begin with a complex package with two components joined by at least double the number of wires required in a similar monolithic assembly. This causes the package to be more expensive, and also to have more points of potential failure than similar monolithic die/assembly processes.

Another limitation is two-component solutions lack the ability to effectively compensate for all temperature changes, a problem that crystal-based solutions also face. This stems from the fact that the two components (resonator and amplifier/base) form a system that must move in unison. The base compensates for the resonator’s frequency change with temperature. Because the two devices are not integrated, but rather separate and joined by multiple bond wires, they do not move in unison as temperature changes. This lack of direct correlation causes these systems to deviate in a changing temperature scenario. In fact, crystal oscillators still outperform multi-chip MEMS devices in these variable temperature situations.

 

Second-Generation MEMS Oscillators

Recently, new advancements in process technology have allowed MEMS resonators to be manufactured directly on top of the CMOS base die. This is a significant step forward for several reasons. First, when manufacturing a single die in a standard foundry, cost will be lower than creating a two-die solution with components from multiple foundries or manufacturing a crystal-based solution. Second, in single-die architectures, the resonator is directly integrated with the base compensation and amplification die, and as such is a single, unified system providing excellent stability over shock, vibration, aging and fluctuating temperatures, where crystals and first generation MEMS struggle. And finally, as Moore’s Law dictates, the single die solution offers more flexibility and features than its predecessors, and generally can be had for a lower price point.

The single-die integration of MEMS and CMOS has been realized with the advent of CMEMS technology -- the contraction of two acronyms, CMOS + MEMS. Silicon Laboratories developed this unique fabrication technology and process, working in conjunction with other market leaders in foundry services. CMEMS is the first process of its kind that enables direct post-processing of high-quality MEMS layers on top of advanced CMOS technology as a single monolithic die. The company’s first CMEMS products are MEMS-based oscillators designed to provide 10-year guaranteed reliability, immunity to shock and vibration, extensive programmability for many target applications, and unmatched performance in changing temperature environments.

 

Exploring the Monolithic Die Advantages

First-generation two-die MEMS oscillator architectures required wire bonding between the MEMS resonator die and oscillator die, adding cost, complexity, and many failure points in the multi-chip module (MCM) design. Furthermore, in first-generation two-die solutions, the MEMS resonators are fabricated in specialized MEMS foundries, located in Europe, where manufacturing costs tend to be higher than in Asia. These MEMS resonator wafers are singulated and then co-packaged with standard CMOS die from more cost-effective foundries located in the Far East.

CMEMS oscillators are fabricated in standard CMOS in the second largest foundry in the world, Semiconductor Manufacturing International Corporation (SMIC). The CMEMS resonator is fabricated directly on top of CMOS using silicon germanium (SiGe), a widely used and therefore economical material. This processing innovation allows CMEMS solutions to benefit from lower cost wafers from a single source, eliminating margin stacking and complex packaging from multiple die and high-bond wire counts.

 

Temperature Compensation

There are additional performance benefits from single-chip CMEMS oscillators relative to two-die architectures. For the reasons above, two-die architectures provide an output frequency that can be negatively affected by temperature changes.

As shown in Figure 1, leading crystal oscillators, first-generation MEMS oscillators and CMEMS oscillators from Silicon Labs were exposed to rapid temperature changes to measure their output frequency stability. The ideal result is to show no change from the 0 value on the y-axis, representing no change in output frequency from temperature changes.

 Figure 1: Rapid Cold Temperature Change Experiment on XO, First-Generation MEMS and CMEMS

While the crystal oscillators (XO) show frequency deviations up to two times their 20 ppm guarantee, the first-generation MEMS solutions show deviations up to eight times their 20 ppm specification. These two solutions are in stark contrast to the CMEMS solution, which changed less than 1 ppm from its target frequency.

 

Si501 CMEMS Oscillator Overview

Figure 2 shows the single-chip architecture of Silicon Labs’ Si50x CMEMS oscillators.

Figure 2: Silicon Labs Monolithic CMEMS Resonator Circuitry

The Si50x oscillator family supports any frequency up to six-digit resolution between 32kHz and 100MHz. The oscillator family includes four base devices, which are highly configurable according to supply voltage, output rise and fall time, frequency stability, temperature support and so on. The devices are functionally compatible to many XOs and first-generation MEMS solutions, and they are offered in pin-compatible 4-pin packages (2 x 2.5mm, 2.5 x 3.2 mm and 3.2 x 5 mm).

Each Si50x oscillator supports a single clock output frequency at any one time. The oscillators are segmented according to the number of clock frequencies they store in on-chip memory. The Si501 oscillator supports a single stored frequency, enabled with the output-enable (OE) functionality. The Si502 stores two frequencies which can be selected with frequency-select (FS) and enabled/disabled with OE functionality. The Si503 stores four frequencies, selected with FS functionality. The Si503 does not support OE functionality. And the Si504 is a programmable oscillator, controlled through a single pin interface (C1D). For more information about the Si50x oscillators including product data sheets, visit www.silabs.com/CMEMS.

For added flexibility, the Si50x CMEMS devices can be ordered with a low-cost programming board and configured immediately to any of the supported settings. This enables ultra-fast prototyping or testing for the CMEMS devices in a developer’s system.

 

Summary

The Si50x CMEMS oscillator solutions from Silicon Labs provide an outstanding alternative to XOs and first-generation MEMS oscillators. CMEMS oscillators provide superior stability over time, guaranteed for 10 years of operation, and also offer the benefits of an all-silicon solution including high levels of immunity to shock and vibration. As all-silicon devices, the CMEMS oscillators are also offered at aggressive price points, adding lower cost to their already impressive set of features.