(Read Part 1 of this article by clicking here)
General problems with quartz crystals
Despite quartz's remarkable properties, in today's highly integrated electronic systems, quartz-based crystal oscillators present challenges in three primary areas:
1. Fundamental size and frequency limitations of quartz resonators;
2. Broad set of considerations for optimal oscillator design (Reference 8);
3. Long manufacturing cycles (particularly when compared to mask-programmable semiconductor devices) (Reference 9)
One can list a host of other issues such as crystal aging, sensitivity to shock and vibration, and dependency on external passive components, among others (Reference 10). But the three issues cited above are fundamental problems that impact development schedules, and increase the complexity of product design.
While high-order resonators or clock multipliers help to overcome the frequency limit problem, other limitations with quartz crystals remain. Along with the many system design and manufacturability considerations for quartz crystals, particular attention needs to be paid to proper oscillator and PCB design because incorrect designs may lead to power-up failures that may seem random in nature. Quartz crystals are shown to exhibit discontinuities in the frequency vs. temperature, or reactance vs. temperature characteristics of a crystal which may cause the frequency of the oscillation to shift or stop completely under a particular combination of temperature and voltage, in otherwise normally functioning units.
Defined as an "activity dip" (Reference 11) this phenomenon can lead to unexpected system failures. A properly manufactured and tested crystal which connects to a manufacturer recommended oscillator circuit should not exhibit any activity dips within the specified environmental conditions.
While improving in efficiency, the overall manufacturing process for crystal oscillators has not changed significantly since the introduction of quartz frequency generators many decades ago. The process involves quartz blank fabrication, blank processing, assembly, and seal and test steps. The process could take 8-12 weeks, but may be accelerated considerably for common crystals. The rapid deployment of consumer electronics devices has put tremendous pressure on crystal manufacturers and procurement professionals to manage the crystal inventories and order lead-times effectively. Still, crystals remain one of the potential bottlenecks on the electronic supply chain as manufacturing cycles and order lead-times shrink.
Furthermore, quartz-based crystals and crystal oscillators are subject to a number of other failure mechanisms including aging, load-capacitance variances, shock/vibration, and radiation. As mentioned previously, aging is inevitable in mechanical structures that vibrate under electrical stress for millions of cycles per second. A typical crystal would age most in the first year of operation, around +5 ppm, and then +1 to +3 ppm in subsequent years.
Additionally, since crystals require one or two load capacitors for stable operation, depending on the cut of the crystal, tolerances on the absolute value of the load capacitors and their aging affect the frequency accuracy of the crystal directly.
Time-dependent acceleration, such as vibration, can cause a large increase in the noise level of a quartz crystal oscillator. In fact, in frequency sources operating on mobile platforms, the vibration-induced phase noise dominates the phase-noise response (Reference 6). In an effort to improve quality, and shock and vibration immunity, crystal manufacturers employ passive and active acceleration-compensation schemes (Reference 12). Similarly, board designers frequently alter component placements, include pads, and choose more-expensive crystals or more-appropriate oscillator packages. Unfortunately, these necessary precautions are often contrasted by aggressive board-space constraints and cost pressures.
Despite the widespread use of crystal oscillators, there has been continual interest in replacing them due to the issues outlined above. For low-precision designs, ceramic resonators have been a successful replacement due to their lower cost, smaller surface-mount packages, and mechanical robustness. For higher-precision designs, crystals remain the predominant solution, while new approaches have begun to address its various limitations.
High-frequency applications led to the proliferation of surface acoustic wave (SAW) oscillators. While they initially gained popularity in filter applications, SAW devices are now used extensively as frequency sources in high-frequency wireless applications, from garage-door openers to advanced telemetry systems. In principal, as the name implies, SAW oscillators rely on the propagation of a wave along the surface of a piezoelectric substrate between two inter-digital transducers.
The early SAW oscillators had inferior aging and temperature stability compared with quartz oscillators. Nonetheless, they gained market traction because they offered better far-from-carrier phase noise (f >1 kHz) and can be manufactured to resonate up to and even beyond 1 GHz. Today, there are SAW resonators in the market with a total frequency variation of ±50 ppm.
In contrast to high-frequency SAW resonators, the industry has also studied low-frequency resonators as alternatives to crystals. Some of the proposed low-frequency alternatives offer distinct advantages in manufacturing cycle times, size and scalability. Architecturally, however, they are at a disadvantage from a phase-noise perspective:
In the lower-frequency resonator approach, the frequency needs to be multiplied up with a PLL to obtain the desired frequencies of today's systems. As showed in Table 1, the drawback of the frequency-multiplication topology is that the phase noise of the resultant output is increased by +20 logM, where M is the multiplication ratio of the PLL.
(Click to enlarge)
Table 1: Phase noise impact of frequency multiplication and division
As discussed previously, the increase in phase noise leads to inaccuracies in the clock domain. Today, a typical 24 MHz fundamental-mode crystal oscillator can have phase noise as low as -120 dBc/Hz at 100 kHz offset from the carrier. In contrast, a similar-quality resonator with a fundamental mode of 2.4 MHz will yield 120 dBc/Hz + 20 log(10) = -100 dBc/Hz phase noise, due to the 10x PLL multiplication required to generate the 24 MHz output.
An optimal frequency source
To graphically summarize the discussion, Figure 5 depicts the common versions of quartz crystals, their derivative products, and some alternatives on three primary axes, which designers may consider when making a selection.
(Click to enlarge)
Figure 5: An overview of common frequency sources
Certainly, it's impossible to clearly convey all alternatives on a single graph, or cover all aspects of a product in only three dimensions. The graph is provided only as a guideline of common design practices in the industry, and average selling prices for the highest volume customers.
For instance, it is not intended to rule out using of a ceramic resonator in a USB 2.0 high-speed application, or dictate that only TCXOs be used in cell phones. There are certainly exceptional products in each category which transcend the boundaries depicted, design work-arounds, and innovative approaches to utilizing what's available, given a project's specific needs and budget.
So, what is the optimal frequency source, given the discussion above? There are noteworthy trends that will dictate the direction of the market in the near future. First, it's reasonable to assume that the projected increase in the bandwidth requirements of consumer, storage and communication systems will reduce the tolerance for timing jitter in designs. Consequently, the growing need for higher throughput and lower BER in interface links will make it harder for designers to compromise on jitter and phase noise.
Additionally, the consumer electronics segment is solidifying its position as the primary stimulant of electronics growth in the coming years, fueling interest to work around the long order lead-times and limited scalability of quartz crystals, both in physical dimensions and in volume manufacturing.
An optimal frequency source should match the phase noise and jitter performance of quartz crystals upon which the next interface standards are being developed. It should also be able to maintain its mean frequency within the specified ppm accuracy budgets of these interfaces over environmental conditions such as temperature, power supply, and vibration. Perhaps, most importantly, it should improve on the cost, reliability and lead-time of quartz crystals.
No technology would better suit the challenge than the technology which itself that allowed for the impressive growth of electronics: standard CMOS. While the need for improving the accuracy of the existing self-compensation schemes in CMOS resonators remains valid, as with all other aspects of the electronics space, a pure CMOS semiconductor approach holds the true potential to displace legacy non-semiconductor components. Much exciting and needed work is being carried out in this area today, and will arrive in the market in the very near future. But that is the subject for another paper!
8. M.Finkelstein, "Inverted Mesa Crystals Carry Oscillators into the Internet Age," Electronic Design
9. "Programmable Crystal Oscillators with Sub-ps Jitter and Multiple Frequency Capability", Silicon Lab, whitepaper.
10. J.R. Vig, A. Ballato "Frequency Control devices", 1999.
11. Introduction to Quartz Frequency Standards-Static Frequency versus Temperature Stability, JR Vig, IEEE uFFC tutorial.
12. Raymond L. Filler, "The Acceleration Sensitivity of Quartz Crystal Oscillators", IEEE, uFFC, May 1987.
About the author
Tunc Cenger, director of product marketing at Mobius Microsystems, has held project lead, marketing and business management positions at Cypress Semiconductor, and most recently at Maxim where he was business manager for audio products. He holds a BSEE in microelectronics from Istanbul Technical University. Mobius Microsystems, based in Sunnyvale, CA, is a fabless start-up developing all-CMOS oscillators.