How to select the right timing device--A case for MEMS—Part I

When selecting components for a system, what is the first thing you think of? Chances are it is the microprocessor or another element that is central to the operation of your device. A timing device may be the last thing on your mind. Selecting timing components that operate at correct frequencies for the application may appear to be a straightforward process, however there are a number of factors to consider that affect system performance.

The clock signal is the heartbeat upon which all signals in the system are dependent. Timing components impact the performance of the entire system, so it is important to select them carefully, considering the requirements of each specific application. When selecting timing devices, and particularly when changing them in an existing design to improve performance, it is also important to consider the effect on overall design of the product and on the supply chain.

One of the primary building blocks for clock timing is the oscillator, consisting of a resonator and an oscillator circuit. Historically, resonators were made from quartz crystals connected to an analog oscillator circuit that drives the quartz to vibrate at the resonating frequency. Oscillators using micro-electro-mechanical system (MEMS) technology, whereby the mechanical resonator is built into a silicon wafer and packaged together with the oscillator circuit, have come into the market in just the past decade. MEMS-based oscillators offer many benefits in performance, reliability, and flexibility when compared to quartz.

Frequency and frequency stability

The basic parameter for any oscillator is its frequency. Oscillators are commonly available in frequencies ranging from kHz to several GHz.  In the case of quartz oscillators, the quartz needs to be cut to precisely the correct size and thickness for a given frequency. Quartz oscillators are available in a limited number of standard frequencies and have to be produced to order if a non-standard frequency is necessary. This usually requires substantial volume, upfront NRE, and approximately a 16-week lead-time. In some cases, a quartz supplier may refuse to produce a non-standard device.

For MEMS oscillators, the same resonator is used for every frequency, and the specific frequency is determined by the electronics in the oscillator circuit. Since the circuits are programmable, any desired frequency is readily available. The frequency is independent of component size, an advantage that allows greater flexibility in board design.

Another important specification is frequency stability, or the variation from nominal clock frequency. Overall frequency stability, expressed in parts per million (ppm), consists of several components: initial tolerance, variation over temperature, variation with power supply voltage, variation with loading, and aging over time. The initial tolerance and temperature variation are the largest contributors to frequency stability. Table 1 shows frequency stability versus temperature for quartz and MEMS oscillators for a variety of frequencies. While some quartz oscillators exhibit excellent stability, tolerance for the MEMS oscillator is much tighter across all frequencies.

Table 1. Components of frequency stability for quartz crystal and MEMS oscillators measured across extended commercial temperature range (-20C - +70C).

Single-ended vs. differential

Chipset vendors may specify the required signal mode for timing chips, or the system designer may have some leeway. Single-ended oscillators are lower cost and easier to implement than those with differential outputs, but they have limitations. They are somewhat sensitive to board noise and are therefore typically best suited to frequencies below 166 MHz. Output types include LVCMOS, CMOS and LVTTL.

Differential signaling is a more expensive option, but it allows for more precisely defined timing and is preferred for higher frequency applications. Since any noise common to both differential traces will be zeroed out, this mode is less sensitive to external noise and generates lower levels of jitter and EMI. Figure 1 illustrates the power supply rejection ratio performance of several differential oscillators.  Applications that benefit from differential signaling run at frequencies typically above 100 MHz and include servers, storage, telecom equipment, high-end printers, industrial equipment and FPGA-based systems. Outputs are usually LVPECL, but LVDS is an option for lower power applications.

Figure 1. Power supply rejection ratio (PSRR) of SiT9121 MEMS differential LVPECL clock compared to quartz oscillators. Source: SiTime measured data.

Jitter

Jitter, the deviation from an ideal clock signal, is one of the main contributors to system timing errors, so it is critical to account for the oscillator’s jitter when evaluating total timing budget. This is not necessarily a simple matter. Jitter requirements vary by application, and oscillator manufacturers do not all specify jitter in the same way.

RMS phase jitter is historically integrated over 12 kHz to 20 MHz offset from the carrier frequency, but this range is not appropriate in all cases. Most serial IO applications, for example, use a band-pass jitter filter to compute effective jitter. The lower cutoff frequencies of such filters range from 300 kHz to 6 MHz, while the upper cutoff frequencies range from 2.5 MHz to 20 MHz. Choosing a clock to meet the jitter requirements for these applications requires integrating phase noise over the appropriate range.

Historically, phase-locked loops (PLLs) used in oscillator circuitry are another source of jitter in the system. However, some MEMS oscillator circuits include high performance fractional-N PLLs designed to output the precise frequency required with extremely low levels of jitter, comparable to or better than that of fixed-frequency quartz oscillators. Programmable, ring-based quartz oscillators exhibit greater levels of jitter, making them suitable only for low-precision applications.

Reliability

Integrated circuits are graded for reliability in units of mean time between failure (MTBF), expressed in hours. The equivalent reliability grade is called failure in time (FIT). FIT rate is simply a mathematical equivalent of MTBF, expressed as the product of 1x109/MTBF. Since MEMS oscillators use a silicon-based manufacturing process, they exhibit semiconductor-level reliability at around 500 Million-hours MTBF. 500M hours MTBF translates to a FIT rate of 2. Quartz oscillators are often rated at around 30 Million-hours MTBF. Oscillators that are designed by silicon manufacturers tend to have better reliability, as shown in Figure 2.  Furthermore, 100% silicon oscillators tend to have the best reliability.

Figure 2. Quality and reliability (FIT rate comparison) of quartz verses silicon products. Source: Company quality-reliability FIT/MTBF data.

Oscillators are designed to operate reliably within a specified temperature range, typically commercial (-20 to +70°C) or industrial (-40 to +85°C). A few are rated for an extended range as large as -55 to +125°C. It is important to select an oscillator rated for the appropriate temperature range for the application. When exposed to changes in temperature, Quartz oscillators can exhibit activity dips, in which the oscillator stops functioning. This can be a serious problem in high precision applications. Therefore, if the customer is concerned about activity dips, it is necessary to test the components over the expected operating range, a step that adds cost. MEMS oscillators do not experience activity dips.

Figure 3. Vibration tolerance of MEMS verses quartz oscillators.

Vibration, especially near the resonating frequency, can also introduce phase noise and slightly worsen frequency stability as shown in figure 3. Proper damping or mounting can help limit the effects of vibration on oscillator performance. However specific MEMS resonators are an order of magnitude more resistant in shock and vibration due to their unique design. 

Classes of oscillators

Several classes of oscillators are available to meet the specific needs of different applications. What follows is an overview of the different options available. We explain specific applications in more detail in Part II of this article. Figure 4 shows the various classes of oscillator arranged by frequency stability and price. Standard oscillators (XO) cover a wide range of frequencies and are suitable for some industrial, consumer electronics and computing applications, where precise frequency stability is not required.

Figure 4. Various classes of oscillator by frequency stability and price.

For applications such as networking, GPS, and mobile devices with internet connectivity, improved frequency stability across the operating temperature range is important. For these applications, it is best to use temperature compensated oscillators (TCXO) to minimize the variation in frequency stability due to temperature shifts. Fixed-frequency quartz TCXOs below 50 MHz can achieve good temperature stability across -20 to +70 °C, but MEMS TCXOs can cost-effectively offer frequency stability down to ±0.5 ppm across the entire industrial temperature range (-40 to +85°C).

Voltage controlled oscillators (VCXO) are especially useful for clock synchronization for telecom and broadband applications. The output frequency is "pulled," or fine-tuned by up to ±200 ppm for quartz or up to ±1600 ppm for MEMS. VCTCXO oscillators incorporate both voltage and temperature control for telecom, networking and wireless products.

Spread spectrum oscillators (SSXO) may be helpful in cases where EMI is a major concern. Spreading out the oscillator frequency over a wider range reduces the peak of radiated energy while not adversely affecting performance. This technique is commonly used in processor, memory and some serial I/O clocking applications.

Sometimes it may be necessary to have a clock with changeable outputs for multi-protocol systems. MEMS oscillators can incorporate multiple frequency selection within a single device, which is not possible with quartz.

In Part I of this article we have discussed the key parameters affecting oscillator performance, including frequency, frequency stability, signal mode, jitter and reliability. We have explained that while both MEMS and quartz oscillators can be designed to meet required specifications for some applications, the programmability of MEMS oscillators enables greater flexibility and the addition of features and classes of oscillators not available in quartz. Part II will delve deeply into application-specific design criteria and important practical aspects to consider when selecting timing devices.

About the Authors

Steve Pratt holds the position of Director of Marketing at SiTime. Steve has been in the semiconductor business for more than 20 years and prior to SiTime, has worked for various analog semiconductor companies including Maxim, Micrel, and Monolithic Power Systems. Steve holds a BS degree in Industrial Engineering & Technology.

Mehdi Behnami, Director of Product Marketing at SiTime, has over 20 years experience in the semiconductor and electronics industry. Prior to SiTime, Mehdi held management and engineering positions spanning marketing, applications and research. Mehdi received his Master’s degree in Electrical and Computer Engineering from the University of Iowa.