Clock architectures and their impact on system performance and reliability
Yin-Chen Lu and Sassan Tabatabaei, SiTime Corporation - November 27, 2012
Systems designers need to consider many factors when selecting a timing device for their application. Selecting an appropriate oscillator and clock architecture can improve the system performance and reliability. It is important to consider parameters such as frequency stability, jitter and phase noise and the availability of oscillator features and architectures that can optimize timing for each specific application.
The clock signal is the heartbeat of every electronic system, making clock architecture critical to the performance of a wide range of products. A key building block for clock timing is the oscillator, consisting of a resonator and an oscillator circuit. Resonators have historically been manufactured from quartz, but advances in silicon processing have enabled resonators made from silicon. These resonators are sealed inside a silicon die and packaged using conventional plastic IC packaging .
In this paper we review the architecture of a MEMS-based oscillator and discuss the performance and reliability. We include availability of special features that add functionality to these devices. Next we analyze the key electrical performance specifications for digital applications and show that MEMS-based oscillators are easily capable of meeting these requirements.
MEMS-based Oscillator Architecture
Oscillators can be divided into several classes based on requirements for frequency stability over the operating temperature range. A basic oscillator (Figure 1a) has a simple design, where the resonator is connected to an oscillator circuit and vibrates at a specific frequency. Applications requiring more precise frequency stability use temperature compensated oscillators (TCXO, Figure 1b) that include a temperature sensor with a feedback loop that corrects the input voltage to the oscillator to account for variations in observed temperature. This can be accomplished in several ways. Temperature compensation circuits in commercially available MEMS-based oscillators are capable of achieving the TCXO requirements of ±1 to ±5 ppm.
Figure 1. Diagrams for several oscillator product categories: (a) XO, (b) TCXO, and (c) OCXO.
For even higher precision, on the order of ±0.1 ppm or less, oven controlled oscillators (OCXO, Figure 1) are an option . These devices include a heater that maintains a constant temperature, reducing temperature-related frequency instability to extremely low levels. Such devices are effective but usually need to be enclosed in a larger temperature-controlled, ovenized package, which adds cost and occupies valuable board space.
With advanced circuit design, MEMS-based oscillators have evolved significantly, reaching OCXO-level performance without a heater. MEMS-based oscillators include an analog CMOS oscillator IC that controls all functions needed to drive the resonator and adjust the resonating frequency. Resonators can be manufactured in a variety of configurations, as illustrated on the left side of Figure 2.
Figure 2. Architecture of SiTime MEMS-based oscillator.
Unlike quartz resonators, where the quartz is precision cut so it will vibrate at the specific frequency the application requires, the architecture in Figure 2 enables MEMS-based oscillators to generate a wide range of output frequencies, ranging from 1 MHz to 800 MHz, from a fixed or narrowly tuned MEMS resonator for many different applications.
The MEMS resonator is connected to MEMS-specific circuit blocks on the oscillator IC, which integrates a high resolution, high performance fractional-N PLL with digital temperature compensation, one-time-programming (OTP) non-volatile memories, and output driver circuits, as shown in Figure 2. A bias generator biases the electrostatic transducers that are built into the MEMS die. The resonator sustaining circuit brings the resonator into mechanical oscillation.