10 January 2011
Are mems based oscillators a viable alternative to the quartz crystal?
Traditionally, quartz crystal oscillators have been the preferred choice for clock generation in consumer, computing and communication applications. But the quest for a replacement technology is on the increase – partly due to manufacturing complexity, partly because of reliability issues.
MEMS based oscillators have emerged as a possible alternative but such devices are not constructed to support the wide range of formats required by the electronics industry. Coupled with the issue of narrower frequency ranges, can silicon really take on quartz?
One company that believes it can is eoSemi, a privately funded start up based in South Yorkshire. The firm is developing a new timing device based entirely on silicon circuitry which it claims can replace quartz crystal mechanisms widely used in consumer and industrial devices.
Ian Macbeth, eoSemi's ceo, says the technology will enable far more compact designs and addresses the limitations of physically vibrating quartz devices. "Traditional quartz crystals have high manufacturing costs, can be bulky and are susceptible to shock damage," he observed. "eoSemi's new silicon approach allows a timing reference to be placed directly onto the existing silicon of the device."
This, claims Macbeth, reduces the number of parts required for each device and therefore cost and size.
The technology is entirely based on standard cmos silicon circuit design and uses a tuning mechanism for compensating for the characterised natural variations the oscillator exhibits under varying environmental conditions. "This relies on a proprietary linear fine tuning mechanism that delivers the necessary part per million (ppm) tuning of the oscillator itself," says Macbeth. "This precision tuning expertise is a speciality of the eoSemi engineering team and I believe the technology could replace quartz crystals in a range of consumer goods, as well as in industrial and automotive systems and wireless applications. 2011 will be a very interesting year."
According to Piyush Sevalia, vp of marketing at SiTime, mems based silicon timing solutions offer a range of benefits not available from legacy quartz products. "As we have seen in the past, silicon always replaces legacy technology," he observed. "Therefore, we fully expect that a vast majority of timing devices, if not all, will be based on silicon mems technology."
It's a long term commitment. By the end of 2010, SiTime will have shipped 35million units and more than 95% of these shipments are replacing quartz technology in various applications. "That trend is only going to accelerate," added Sevalia.
Dean Miles, EMEA technical marketing manager at Tektronix, agrees there are an assortment of medium to lower stability timing applications where a mems device is viable as a replacement.
"As the mems technology matures, the ultimate question will be what 'Q' is possible with the physics involved in the mems technology," Miles asserted. "This will determine which applications can utilise such timing devices. For devices such as garage door openers, calculators, cd players, ad infinitum, the timing stability required is already now often met with non quartz resonators."
However, the stability needed for high performance frequency conversion devices – such as spectrum analysers or rf signal generators – includes long term and exceedingly short term stabilities. "For the frequency measurements that such devices perform, long term stability is measured in years (and integrated over maybe a few milliseconds), while the 'phase noise' requirements are expressed in stability measured in microseconds or even nanoseconds," Miles said. "On the other hand, a precision reference for a clock may have long term requirements that are also high, but it usually has no need for stability much less than 1ms at best."
Over the years, a number of frequency references have solved the problem of stability over temperature. Often, these require that frequency variations over temperature be known or measured, then a compensation circuit simply measures the temperature and applies the necessary compensation.
"Even when temperature variations may not be so predictable," said Miles, "the method of mounting the oscillator in proximity to a variable source of heat can simply stabilise the temperature of the timing device. Hence 'ovenised oscillators'."
Sevalia concurred: "For a silicon mems based timing device, the biggest hurdles have been to achieve the needed long term stability and to demonstrate stability across temperature changes. SiTime has patented core technology that addresses the long term stability for low aging resonators. For temperature compensation, a silicon mems device has its intelligence in the analogue circuits that go along with the resonator and these circuits must be designed to that the output remains perfectly stable over temperature and fabrication tolerances."
For some applications, even higher levels of temperature stability are required and the oven controlled crystal oscillator (ocxo) is often required.
Peter Sinclair, application support manager at IQD Frequency Products, doesn't envisage a time when silicon or mems technology will be used in ocxo applications. "Silicon solutions are being looked at for low grade temperature compensated crystal oscillators (tcxos), which have stabilities of around 2.5ppm. But even there, the technology is not good enough to replicate the performance of crystal products, which have been optimally designed for performance over temperature."
OCXOs are physically larger than a simple crystal oscillator and, as well as the crystal oscillator, they also need to incorporate a heater, control circuitry and thermal insulation. IQD recently launched the IQOV-40 series, which it believes to be the world's smallest surface mount ocxo, with stabilities down to ±20ppb over an operating temperature range of -40 to 85°C.
With such extreme levels of temperature stability required, specialised optimisation was required to enable this performance in a 9.7 x 7.5 x 4.1mm package. "Basically, a crystal has a lower and an upper inversion point," Sinclair explained. "The lower is around -20°C and the upper is around 85°C. There's no point in using the lower inversion point with an ocxo because it would mean if the ambient temperature rises to more than -20°C – which it's going to – then you're back on the curve of the crystal and getting no compensation.
"Obviously, you then use the upper inversion point because, at that point, the frequency shift of the crystal is minimised because it's almost on a flat plateau." Because of this, a few degrees in either direction has a relatively minor effect on the frequency shift of the crystal.
"By setting the oven within the oscillator, it replicates the exact temperature of the crystal inversion, but a batch of 100 crystals will still be fractionally different," Sinclair noted. "Each crystal manufactured at the same time, by the same people, using the same process, will be slightly different. And that's what the control circuit in an ocxo is designed to do – to hit the exact inversion point to minimise the frequency shift or any ambient temperature shift."
These ocxo units are more expensive than crystals on their own, but the performance is considerably enhanced on that of a simple crystal in an unregulated electrical and physical environment.
Sevalia concluded: "Aging and temperature stability are engineering problems with engineering solutions. The cost is an operations problem that is solved with massive industrial leverage. The reason that silicon always wins is that one silicon application leverages off the others and the combined industrial infrastructure is overwhelming. Cost and integration are two prime drivers for mems oscillators becoming 'replacement' technologies for conventional quartz crystal oscillators."