Could this new approach to monolithic RF filters help pave the way to practical tri-band 5G handsets?
While cellular carriers would like us to all believe that we’re only a few steps away from a shining 5G future, filled with new services and seemingly-limitless bandwidth, the designers charged with navigating the road to that future know otherwise.
For one thing, they’ve got to come up with radios that can efficiently navigate the complex patchwork of frequencies spread across three widely-spaced frequency bands. As a former EE, I know it is quite possible to create signal chains capable of operating in the fragmented slices of bandwidth strewn between 600 MHz and 42 GHz. That said, my limited experience with RF design had me pretty much convinced that the cost and complexity required to achieve those capabilities could be a significant stumbling block for the industry.
I’m a bit less pessimistic now, after a rather lengthy conversation with Dylan Kelly, Chief Operating Officer at Resonant Inc., about the company’s new monolithic RF filter technology. It changed my understanding of what is possible for modern radios and, if Resonant can deliver on its claims, their technology should enable the creation of simpler, compact, and cost-effective RF front ends that don’t compromise on performance.
To understand why I’m so intrigued, let’s quickly review how most of today’s cellular signal chains go about their business.
Cellular radios have operated across multiple frequency bands for many years, using multiple Tx and Rx chains, each with its own string of amplifiers, switches, and filters. Each signal chain relies on a set of filters to eliminate unwanted interference and keep the energy transmitted by the radio within the channel’s tightly-defined range. Nearly all of these filters are piezoelectric devices, manufactured using photolithographic processes to create either a surface acoustic wave (SAW), bulk acoustic wave (BAW) or acoustic resonator (AR) structure.
Figure 1 These acoustic wave filter technologies are manufactured using photolithographic processes. Source: Resonant
SAWs and temperature-controlled SAWs have been traditional favorites as the technology is very mature and common enough to keep component costs down. While adequate for many applications, their moderate selectivity and relatively high signal losses can be a problem for others. In addition, SAWs have difficulty operating much above 2.5 GHz.
BAW filters, which propagate the signal through the filter material, instead of across its surface, are a more recent introduction. They offer much lower insertion losses (under 2 dB), better filtering performance, and recent developments that should enable them to support 5G frequencies. Unfortunately, producing them involves more complex manufacturing processes that make them cost-prohibitive for some applications. This is especially true when you consider how many of these devices will be required to build a 5G “world phone.”
Today’s top-tier 4G/LTE smartphones require 50-90 of these of these tiny filters but 5G phones will operate on far more frequencies and also use MIMO radios that support multiple Rx/Tx chains. This could swell the number of filters by 5-10x, or possibly more. In addition, the filters used in 5G radios will need to have sharper, more precise cutoff characteristics to enable narrow guard bands surrounding the channels that avoid wasting too much useable bandwidth.
Resonant may have solved many of these problems with a structure they call an IDT-membrane resonator. It consists of a single crystal, piezoelectric membrane, with a metal interdigital transducer (IDT) on the top surface of the membrane. “The metal traces excite a bulk acoustic wave within the membrane,” Kelly explained. “The wave’s primary frequency and coupling characteristics are determined by the physical dimensions and properties of the piezoelectric.”
According to Kelly, their IDT BAW devices can be manufactured with the same fabrication facilities and processes used to produce SAW filters, making them significantly less costly to produce than traditional BAW devices. In addition, the low-capacity structures have demonstrated the ability to support operating frequencies in excess of 40 GHz, more than necessary to support 5G applications.
Figure 2 This cross section of the IDT-membrane resonator shows the basic structure and bulk wave excited by IDT fingers. Source: Resonant
Resonant’s other “secret super power” is that their process can be used to produce multiple BAW filters on a common substrate that can be packaged as a single component. Kelly was hesitant to discuss precisely how many filters could be put into a single substrate, but did say that the technology might be able to keep the complexity of a 5G radio close to that of today’s premium 4G/LTE radios.
When I asked why other manufacturers were not offering a similar solution, Kelly said that, in addition to any IP they’d developed, the challenges involved with designing and accurately simulating the unique structures that support IDT-based filter devices remained a significant barrier to entry. He attributed much of the company’s success to its home-grown advanced modeling software, which supports highly- accurate simulation of filter structures. He explained that the software’s ability to predict a filter’s performance before it is built enables them to bring products into production quickly, without the guesswork usually involved with developing SAW/BAW devices that can require up to 20 time-consuming spins in the fab.
I still suspect that the road to a 5G future will be a bumpy one, but designers may have one less obstacle to negotiate along the way.
Lee Goldberg is a self-identified “recovering engineer,” who has worked designing microprocessors and embedded systems, and is now a maker/hacker, a geek dad, and a tech journalist.