comment on this article

Will MEMS devices overtake crystal technology in ultra low jitter applications?

Despite the continuing development of silicon based technology, quartz crystals remain the only approach which can provide the performance required by high speed communications systems.

James Wilson, director of marketing for Silicon Labs' timing products, noted: "We're seeing a lot of interest in timing products, particularly from those building web infrastructure. Here, there's a strong trend towards higher capacity, higher speed networks driven by the explosion in mobile and video streaming. All these things need higher performance timing – and that's coming from the crystal end of the spectrum."

And that's a view shared by Prescott Sakai, director of On Semiconductor's clock and data products business. "As protocols and data rates increase, there is the need for lower phase noise and higher frequency clocks. Traditionally, crystal oscillators (XOs) have ruled in this segment."

But there is a growing belief that the days of the quartz crystal are coming to an end. The reason? MEMS technology is getting more robust and developers believe that it won't be too long before all silicon solutions are being deployed in the most demanding applications. And MEMS developers are looking for a larger share of the $5billion timing market.
Sakai noted: "We're now seeing the lower end of the market being served by MEMS and all silicon oscillators."

Rami Kanama, general manager of Micrel's timing and communications group, noted: "MEMS is a perfect fit for mid range applications and brings flexibility and robustness to the consumer market. Several crystals can be replaced by one MEMS package, with space and power consumption benefits."

However, for the moment, quartz remains the 'go to' solution at the top end of the market, where designers are looking in particular for ultra low jitter. Amongst Micrel's offerings there is the MX55/MX57 XO family, designed to provide high performance in networking, storage, server and telecommunications equipment. Typical phase jitter is said to range from 160fs to 230fs, improving bit error rates in high line rate applications. But where these parts differ from traditional XOs is in the packaging.

"For traditional XOs," Kanama noted, "the package is the most expensive element and only comes from a couple of suppliers. So let's move it away to a semiconductor assembly process with a plastic package."

Wilson added: "Because XOs come in a sealed package, if there's any contamination, it could affect the crystal and its ability to start up reliably."



The other issue with quartz is its temperature stability. Kanama said: "The higher the temperature, the lower the stability and the more drift. This means you need to correct this with either a more expensive package or compensation circuitry."

Sakai added: "As you move to higher frequencies, the quartz crystal needs to be thinner and this brings reliability problems. The consequence is that semiconductor based options are moving up the food chain and taking sockets traditionally held by XOs."

Wilson noted that Silicon Labs sees a similar trend. "MEMS has a presence at the lower end of the market, but the technology has the potential to be used more broadly. However, it comes down to a couple of factors.

"One is technology. As MEMS technology improves in terms of jitter and stability, we will see more market penetration. But the other factor is customer adoption. While some customers are excited about MEMS, others are more conservative. They will adopt MEMS more slowly because they are more risk averse and will be looking for solutions with proven reliability."

According to Sakai, On Semi isn't pursuing MEMS as yet, but it is pushing silicon alternatives. "While there are some more exotic ways of generating clean and accurate clocks for networking applications using XOs, the semiconductor industry has improved PLL technology significantly. This ability to produce high frequency, low phase noise clocks is bringing a move from quartz to silicon. But that route needs a high precision PLL or an LC tank circuit."

Wilson suggested that LC tanks are the way to go. "In the optical networking market, data rates are moving from 10G to 40G and beyond. That needs a high performance frequency source and, more often than not, low jitter clock sources. When you talk about these speeds and of clocking FPGA transceivers at 32G, you need an LC tank based approach – it's the only way to hit the jitter requirement."

An example of how MEMS is making its way up the food chain comes from Micrel and the DSC400 range – the MEMS technology coming from Micrel's acquisition in 2013 of MEMS timing pioneer Discera. With a wide range of applications, DSC400 parts address frequencies of up to 460MHz with a jitter of around 1ps and high stability. But Micrel's MX55/57 XO devices offer 170fs jitter and Kanama said the goal is to get MEMS to that level.

Wilson believes high performance clocking starts at 1ps. "Gigabit Ethernet typically uses a 1ps reference clock; 40G and 100G will need less than 500fs. We're already seeing system requirements from equipment builders that call for jitter of less than 300fs. That's becoming a key figure of merit and our latest products will have a jitter performance of around 150fs."

Sakai said that, while the consumer market needed programmable solutions at low cost, the networking sector was pushing towards jitter of less than 500fs. "And we have to do that at frequencies of up to 700MHz; sometimes more. Cost isn't always the main concern here, but a good high performance clock is. That includes a range of interfaces and driving into an ASIC. That's the advantage of silicon; meeting those needs is getting harder for XO companies."

Tantalisingly, he added: "We've just got a part back from the fab that will address the high frequency, low jitter space; something that's very interesting."

Wilson said Silicon Labs has a new range of clock products optimised for high speed, high demand networks. "Customers want more frequency flexibility and better jitter performance. But they also want good power consumption and cost."

Sakai noted: "While requirements are appearing for 300fs jitter, which we think can be achieved, some users are looking to drive this even lower. I don't think there's a solution for that at the moment, but the solutions will blend analogue design, digital PLLs and LC tanks for more flexibility."

So are we seeing the end of quartz? Sakai said: "Some thought MEMS would revolutionise the industry, but it hasn't as yet. The number of apps which MEMS services will continue to grow, but I don't think its market share will be as large as anticipated."

Wilson: "I'm confident MEMS will replace a significant amount of the market catered for by quartz."

"No," Kanama concluded. "That is still a long way off, but one way to get MEMS more widely accepted is to get SoC and MCU companies to embed the technology within their devices – and you can't do that with a crystal."

Author
Graham Pitcher

Comment on this article


This material is protected by MA Business copyright See Terms and Conditions. One-off usage is permitted but bulk copying is not. For multiple copies contact the sales team.

What you think about this article:


Add your comments

Name
 
Email
 
Comments
 

Your comments/feedback may be edited prior to publishing. Not all entries will be published.
Please view our Terms and Conditions before leaving a comment.

Related Articles