The State of the MEMS Timing Industry

Time marches on. And so do advancements in the precision timing industry. Precision timing is important in many systems to accurately transmit and receive data, synchronize networks, determine GPS positioning, navigation and timing, and measure movements in the real work accurately. The last 15 years have seen dramatic improvements in precision timing derived from MEMS resonators – frequency stability improved 10,000 times, Allan deviation dropped 30,000 times and jitter reduced 800 times. This article reviews the state of MEMS timing today.

Stability

The most quoted oscillator specification is frequency over temperature stability (or, “stability” for short) in units of ppm or ppb. The best stability offered in MEMS technology today is ±5 ppb offered in both OCXO and TCXO classes of oscillator. However, other factors contributing to a device’s overall stability are often equally (or more important) for their impact on system performance. 

For example, network synchronization applications such as IEEE 1588 Precision Time Protocol (PTP) discipline an oscillator to network time with updates occurring several times a second. Here, the (10+ year) lifetime stability mentioned above isn’t relevant. What matters most is the oscillator’s short-term stability to minimize frequency drift between updates. Since this drift is dominated by thermal changes, the dominant specification impacting system performance is frequency (F) over temperature (T) slope, or dF/dT. Today’s MEMS-based TCXOs provide a dF/dT of ±0.3 ppb/C. And with stabilities down to ±5 ppb, these MEMS TCXOs have similar performance as entry-level quartz OCXOs.

Fig.1 Illustration of a Stratum 3E MEMS-based TCXO Specified for ±5 ppb stability and ±0.3 ppb/C dF/dT

Other applications require low aging to maximize the interval between equipment calibrations, which enables equipment to stay in the field longer. MEMS TCXOs are available with daily aging of ±0.2 ppb, and 20 years aging of ±150 ppb. What’s more remarkable is this aging is specified at 85C rather than the industry norm of 25C established by quartz oscillators. 

Aerospace and defense applications benefit from oscillators that not only survive but maintain stability when subjected to shock. One key metric is acceleration (or g) sensitivity, measured as a total gamma value over 3 axes per MIL-PRF-55310. Commercial MEMS TCXOs today achieve ±0.004 ppb/g typical and ±0.009 ppb/g max total gamma up to 105C ambient temperatures in a 5x3.2 mm² package – the best TCXO performance in the industry. For comparison, quartz oscillators optimized for shock range between 1 and 0.02 ppb/g.

Timing Noise 

Low-frequency noise is measured as close-in phase noise or Allan deviation. These metrics are important in RF, GNSS and other applications that observe target signals over relatively long intervals of time. They are primary metrics for precision timing. Today’s 10 MHz MEMS TCXO provides -80, -109, and -130 dBc/Hz phase noise at 1, 10 and 100 Hz offsets. Their Allan deviation measures an industry-leading 1.5e-11 over a 10 second averaging interval. And unlike quartz technology, this noise performance is maintained under shock and vibration due to MEMS inherent resistance. Further, MEMS resonators don’t suffer from traditional quartz-resonator artifacts such as activity dips and frequency jumps. Such artifacts can degrade system performance by causing downstream PLLs to unlock, and other intermittent errors that are difficult to troubleshoot.  

Precision timing also exhibits higher-frequency noise such as jitter, which is critical to minimize in high-speed wired interfaces. Traditionally low-jitter clock signals are sourced from XOs or clock generators. For example, today’s 400Gb Ethernet links require 100 fs rms or better RMS phase jitter. MEMS XOs sell today with 70 fs rms phase jitter (12 kHz to 20 MHz offset frequency range) using integer-N phase locked loops to minimize spurs. Equally important to high-speed designs is a device’s ability to reject system power supply noise from appearing at a clock output as jitter. MEMS XO performance here is among the best in the industry at 9 fs/mV. This means 10 mV pp of power supply ripple at 500 kHz, for example, would produce only 90 fs pp of jitter at 500 kHz on the clock output.

Fig. 2 SiT9501 Phase Noise Profile Illustrating a Phase Jitter of 70 fs rms, 12 kHz to 20 MHzPower

As MEMS timing is traditionally paired with a phase-locked loop to support a wide range of output frequencies, its power consumption can sometimes be higher than non-PLL based quartz solutions. But the last few years have seen much design creativity work around this limitation. The latest MEMS XOs feature a unique FlexSwing™ output buffer that performs like LVPECL but with independent control over VOH and VOL. Using a recommended optimized termination reduces the FlexSwing load current 3.5 times to only 7.5 mA. Also, support for 1.8V power-supply option reduces power 28% compared to 2.5V. If LVPECL is required, MEMS XOs have integrated source-bias resistors that reduce load currents by 16 mA. Alternatively, HCSL is offered in a dedicated low-power mode. Finally, since power can be traded off for jitter performance, today’s MEMS XOs minimize current consumption down to 1.4 μA in normal operation, important for IoT and mobile markets that rely on batteries.

Reliability 

Since MEMS timing devices are manufactured in traditional and ultra-clean semiconductor fabs, their quality and reliability are excellent. Published data shows MEMS oscillators have a mean-time between failure (MTBF) of 1,960M hours compared to quartz values of 38M (vendor A) and 28M (vendor B) hours. While one might think 28M hours, which is 3,194 years, should be sufficient for any application, note that the volume for any application is higher than a quantity of one. For example, a MTBF of 28M hours predicts 310 failures annually per million units shipped, versus 4 for a MEMS device. 

Diversity 

The last decade has seen MEMS timing expand into new classes of devices. Traditionally the domain of XOs, MEMS solutions today are also available for TCXOs, OCXOs, clock generators, jitter attenuators and network synchronizers. MEMS resonators can also get integrated with a 3^rd-party ASIC in the same package for the smallest footprint, fastest customer adoption and tamper-proof timing. Since MEMS uses a completely different manufacturing process than quartz-based devices, it presents a true second source for supply continuity. Its programmable, 100% silicon-supply chain is also extremely scalable, leading to excellent availability – shipping in 48 hours from distributors such as Digikey.

About the Author

Gary Giust, PhD, Senior Manager, Technical Marketing

Gary Giust, PhD, heads technical marketing at SiTime where he enjoys participating in industry standards, architecting timing solutions and educating customers. Prior to SiTime, Giust founded JitterLabs, and previously worked at Applied Micro, PhaseLink, Supertex, Cypress Semiconductor, and LSI Logic. Giust is an industry expert on timing, has co-authored a book, is an invited speaker, an internationally published author in trade and refereed journals, and a past Technical Chair for the Ethernet Alliance's backplane subcommittee. He holds 20 patents. Giust obtained a Doctorate from the Ira A. Fulton Schools of Engineering at Arizona State University, Tempe, a Master of Science from the College of Engineering & Applied Science at the University of Colorado Boulder, and a Bachelor of Science from the College of Engineering and Physical Sciences at the University of New Hampshire, Durham, all in electrical engineering.