Quartz crystals have dominated accurate frequency generation since the mid-1940s. That is about to change with the fruition of 40 years of research in MEMS resonators.
Quartz crystals have remarkable mechanical and piezoelectric properties that have made them the fundamental timing component of choice since the mid-1940s. Despite 60 years of research into ceramic, silicon, and RLC circuits, no material or technology has been able to replace the quartz crystal because of its exceptional temperature stability and phase noise performance. In 2006, an estimated 10 billion quartz crystals and oscillators will be manufactured and placed in automobiles, digital cameras, industrial equipment, gaming devices, broad band interfaces, cell phones, and virtually every digital product produced. More quartz crystals are made each year than there are people on earth!
However, quartz crystals have drawbacks including their inability to integrate onto silicon wafers, their relative higher cost when shrunk to smaller sizes, their non-industry standard manufacturing and packaging methods, and their sensitivity to heat, shock, and vibration. Thus the electronics industry has sought in vain for a technology to overcome these deficiencies while not sacrificing performance.
MEMS resonators: a history of unrealized promise
![]() Click here to enlarge |
The technology with the most potential to replace quartz crystals has long been the micromechanical resonator. As early as 1965, H.C. Nathanson published a paper (IEEE Applied. Physics. Letters, v.7 pp.84-86) describing micromachined resonators made with fine metal wire, and in 1967, with released aluminum traces. In 1982, Kurt Petersen published a paper titled “Silicon as a mechanical material (Proc. of the IEEE v.70, n.5, pp. 420-457).” This silicon mechanical technology, now one of the foundations of the multi-billion dollar MEMS (microelectromechanical systems) industry, advanced MEMS resonator design with a material suitable to replace quartz crystals. MEMS resonators and quartz crystals are entirely distinct—they rely on different mechanical properties, different electrical properties, different fabrication technologies, and different drive circuitry; even the scale is different with common MEMS dimensions an order of magnitude smaller than those generally used in quartz crystals. The relative simplicity of manufacturing silicon vibrating beams, the high unit volume of the quartz market, and the potential cost advantages of building resonators on 100 mm (4”), then 150 mm (6”), and now 200 mm (8”) wafers instead of sub-100 mm rectangular quartz wafers has been compelling.
![]() Click here to enlarge |
Unfortunately, the promise of cheap, high quality, fully integrated resonators ran into harsh realities Early researchers, including Petersen, made many advances but also discovered many difficult technical issues. These included silicon’s 30 ppm/ºC frequency temperature coefficient, aging from polysilicon fatigue, and drift from packaging contamination. This drift was one of the most intractable problems because the resonant elements were so small they were especially sensitive to mass loading and therefore surface contamination. A single atomic layer of contaminant could shift a MEMS resonator’s frequency out of typical quartz crystal specifications. In addition, the technology had cost problems—the available packaging technology was similar to that used for quartz crystals, and since the packaging dominated the
costs of the finished components they could not be attractively priced. These technical and cost limitations prevented MEMS resonators from reaching the market as a quartz replacement technology.
![]() Click here to enlarge |
The rise of the MEMS oscillator
MEMS oscillators, however, are now a technical reality, extremely cost-effective, and very small. In Q1 2006, SiTime Corp., Sunnyvale, Calif., began sampling a mass market MEMS oscillator to directly compete with quartz crystals and oscillators. The key to this breakthrough was discovered by Robert Bosch scientists, Markus Lutz and Aaron Partridge, at the Bosch Research and Technology Center, Palo Alto, Calif., now at SiTime Corp.
In a counter-intuitive discovery, it was silicon packaging inventions not silicon resonator improvements that opened the flood gates to MEMS oscillators and extreme PCB real-estate reductions. The MEMS-first and EpiSeal packaging discoveries allowed resonator die to be assembled in industry-standard and low-cost plastic packages, solved the problems of maintaining clean vacuums, virtually eliminated cavity contamination and aging effects, and mitigated the complexities of temperature compensation and drift.
Click here to enlarge |
MEMS-first: better than quartz
The image at right illustrates a series of MEMS fabrication cross sections. The process begins with etching 0.4 µm trenches down to the glass insulation layer of 10 µm SOI wafers to form resonator structures. In operation, these resonators will vibrate horizontally to the surface of the wafer. The trenches are covered with silicon oxide (glass) and the wafers are put into an epitaxial reactor to grow a thin layer of silicon and polysilicon over the surface. Vents are etched in this layer through which some of the glass is removed to release the resonator. The wafers are again placed in an epitaxial reactor and cleaned at over 1,000ºC to burn off contaminants, the vents are sealed shut, and thick silicon and polysilicon caps are grown. The high temperature anneals the resonators, removes micro-cracks, pits, and scallops, and leaves the MEMS resonators permanently sealed within extremely clean vacuum cavities. The thick polysilicon caps are mechanically strong and withstand over a 100-atmospheres pressure during plastic package molding. After polishing, the wafers appear new and unprocessed despite containing tens of thousands of resonators hidden below their surfaces. Vias are then cut through the cap silicon to form electric contacts to the resonator’s drive and sense electrodes.
The wafers are finished with simple metal traces and bondpads for multi-chip or system-on-chip packaged oscillators, or CMOS circuitry for integrated oscillators. Standard CMOS circuits may be built providing that care is taken not to place transistors in the polysilicon cap above the MEMS resonators.
EpiSeal: it’s about cleanliness
For a technology to be introduced successfully it often must bridge the performance standards set by the incumbent technology before its new capabilities can be fully utilized. The early SiTime products will replace quartz oscillators in fit and function, later the full advantages of the MEMS technology will be brought to bear. The following sub-sections list some of the performance advantages of MEMS-first oscillators:
Anti-aging resonators– Both MEMS resonators and quartz crystals have three primary frequency error terms that make up the total frequency tolerance measured in parts per million (ppm). The three error terms are initial frequency offset, temperature coefficient, and aging. Controlling aging has been the difficult problem with MEMS resonators, but this is now solved.
MEMS-first silicon resonators are made solely from annealed silicon and silicon dioxide, which results in a nearly ideal resonating beam material. Subsequent tests conducted at Stanford Univ., Calif., demonstrated a resonator frequency drift over one year of less than one part per million, which is the limit of the measurement accuracy. Measurements at SiTime show less than 0.05 ppm (parts per million) over two weeks at elevated temperature, which is at the limit of our test capability. These were measured without pre-annealing or pre-aging of the resonators. These results can be attributed to the extreme cleanliness of the resonator chamber and the stability of the high-temperature annealed resonator material.
On the other hand, quartz crystals cannot be annealed at high temperature because the crystal undergoes a lattice change at 573ºC. This limits the maximum anneal possible with quartz, and disallows the surface reformation that SiTime performs on silicon. As a result, aging related frequency drift in quartz crystals is partially caused by changes in the mechanical properties of the crystal itself.
When a quartz crystal blank is ground to the proper thickness, the grinding process creates micro-cracks and pits in the surface of the crystal. These lattice anomalies relax as they are repetitively flexed and temperature cycled resulting in subtle frequency changes. Additional aging sources in quartz crystals include changes in the anchor springs, stress in the packaging, changes in the die attach and contact epoxies, and outgassing of various materials. First year aging effects are typically +/- 5 ppm for small surface mount quartz products down to +/- 1 ppm for the larger metal can quartz crystals. Much of that appears in the first weeks, especially at elevated temperature. Over 10 years the aging is commonly specified from +/-10 to +/-15 ppm. These errors can be mitigated in precision quartz products, but this requires extra processing, long anneals, and is very expensive. Laboratory grade quartz oscillators can age less than 0.1 ppm/yr, but these can cost hundreds of dollars each.
![]() |
Low thermal hysteresis-The ultra-clean MEMS-first resonators have been tested for more than 300 temperature cycles from -50 to +80C with no discernable frequency shift or thermal hysteresis. Precision laboratory tests show the MEMS resonators have intrinsic hysteresis of less than +/- 0.2 ppm.
In the case of quartz crystals, thermal hysteresis is caused by vacuum cavity contamination, support stress, and various poorly understood intrinsic effects. Contaminating material condenses and evaporates off the crystals at temperature extremes or at start up and shut down causing frequency hysteresis, and the crystals themselves show a memory of their recent operation. Common quartz AT-cut crystals in small packages typically show 0.1 to 1.0 ppm hysteresis.
EpiSeal magic
One method to test for leaks in the MEMS vacuum cavities is to directly measure the resonant Q of the enclosed resonators. If the cavity leaks, then the Q will drop as energy is transferred to gas molecules in the cavity. If the cavity remains sealed, then the Q should remain the same. However, this was not the case. At Stanford, data was taken on the Q to test for cavity leaks and the Q was shown to improve over time and thus the vacuum had to be improving as well. The improving Q and vacuum results have been explained in the following manner: during the EpiSeal process the only gas of significant quantity that remains in the vacuum cavity that surrounds the resonator is hydrogen. Hydrogen is known to diffuse directly through silicon as opposed to nitrogen, oxygen, and water vapor, which make up the majority of our atmosphere and do not readily diffuse through silicon. Thus, as hydrogen seeks equilibrium it diffuses out of the resonator cavity but is not replaced by other gasses.
Just one word: “plastics”
As stated above, the MEMS-first encapsulation process enables the use of any standard IC packaging including: SOIC, SSOP, BGA, CSP and QFN. SiTime chose QFN-type plastic injection molded packaging for high reliability, low lead inductance, good thermal performance, flexible pad layout, and low cost. Quartz crystals in comparison require expensive special purpose ceramic or metal packages.
The first generation SiTime oscillators are supplied in 2.0×2.5 mm, 2.5×3.2 mm, and 3.2×5.0 mm packages with 0.85 mm height for PCB compatibility with existing quartz oscillators. Two product families include the SiT8002 programmable oscillators and the SiT11xx fixed frequency oscillators. These have identical specifications and performance so that users can design and test with the programmable versions and manufacture with the fixed frequency versions. They have a frequency range of 1 to 125 MHz and a frequency error specification of +/- 100 and +/-50 ppm over temperature, power supply, and aging. These performance specifications are similar to common consumer, industrial, and computation quartz products.
MEMS oscillators and a new silicon industry
The electronics industry is now enabled with a new technology that allows the integration of small, high Q, low ppm, single or multiple resonators, at a cost below quartz crystal products. Some of the targeted value-added applications are:
Consumer and computational products-Notebook computers, digital cameras, gaming boxes, video recorders, portable media players, set top boxes, high definition televisions, and printers all consume quartz products. For example, PC motherboards require numerous quartz crystals, quartz oscillators, VCXOs, and CMOS PLL chips.
MEMS-first resonators are CMOS compatible and may be integrated with PLLs, Logic, and analog circuitry to produce products that reduce EMI, routing complexity, and timing circuitry area by up to 70%. As these MEMS-first resonators are only sold in oscillator form, the customary crystal load capacitors and resistor are removed saving additional PCB board space and freeing the system designer from crystal start up concerns and layout interference.
Automotive applications- For automotive applications, the new MEMS-first silicon resonators are superior to quartz crystals by physics, by design, and by manufacturing process.
As described earlier, the silicon resonators are annealed at over 1,000°C. Therefore, normal operating temperatures have virtually no effect on them. The rest of the components are standard and have well understood temperature and reliability limits. In fact, the operating temperature for the finished oscillators is limited not by the resonators, but by the standard CMOS circuitry and packaging.
Shock and vibration insensitivity are improved as compared to AT-cut quartz because the resonators themselves have vastly higher fundamental resonant mode frequencies than mounted quartz crystals.
![]() |
Wireless applications– One of the early application targets for this technology is in ultra-compact wireless nodes which integrate one or more resonators. Radio nodes in the 315, 433, 868, and 915 MHz range may all benefit from the <20 ps RMS jitter and +/- 50 ppm performance of the first generation oscillators and achieve node space saving of greater than 50%. One, two, or more MEMS-first resonators may be integrated on a single die for wireless applications that require a 32.768 kHz or equivalent oscillator for a low-power wake up and real-time clock, as well as a high-frequency oscillator for transmit, receive, and processing functions.
Nod to Moore’s Law
One of the more exciting features of the new resonator technology is that it scales with process geometry. All too often MEMS products are relegated to older generation fabs. However, in the case of SiTime’s resonator technology, the performance benefits that are provided by shrinking CMOS geometries also improve the performance of the resonators as they get smaller. SiTime’s SiT8002 MEMS resonators have electrode spacing of 0.4µm, which sets a limit on the amount of signal presented to the CMOS oscillator circuits. In future generations, reducing this electrode gap will increase the sense signals, improve the signal to noise ratios, and give better phase noise and jitter specifications for the oscillators.
Next generation
Going forward, MEMS oscillators will be brought to market at higher frequencies and lower phase noise while maintaining the size and cost benefits associated with MEMS-first encapsulation technology. The cell phone is one such application that will benefit from second generation products. Research has shown that even the tough GSM and CDMA cell phone TCXO standards are possible as this technology is matured in 2008 and beyond.
—Aaron Partridge, Chief Technical Officer and
John McDonald, VP
SiTime Corporation