FBAR filtering 的更多创新– 访谈Broadcom

5G mobile communications is enabled by the opening of higher frequencybands as well as band re-farming. This is leading to a need for new components which must handle more stringent requirements on insertion loss, widerbandwidth and steeper out of band rejection, while always ensuring a smaller footprint at a reduced cost. Even though many technologies have been tried forfiltering, the future is through the continuous incremental innovation ofcurrent SAW, BAW and FBAR filtering technologies. Broadcom is a market leaderin filters for mobile phone RF front ends.

Cédric Malaquin, Technology & Market Analyst at YoleDéveloppement (Yole) discussed with Rich Ruby, Director of FBARTechnology, and William Muller, PrincipalTechnology Strategist, at Broadcom.

They agreed to tell us their view of the future of filtering technologiesfor 5G applications which has been addressed at the market, technology andintellectual property level in Yole, System Plus Consulting and KnowMadereports, respectively.

i-Micronewsinvites you to share the discussion:

CédricMalaquin (CM): Could you please describe your position and mission at Broadcomto our readers?

WilliamMuller (WM): Iuse the title Principal Technology Strategist. My brief is to stay awareof developments in the handset RF Front End (RFFE) space to help direct wheremy division assigns resources. To this end I interface with serviceproviders and regulators on matters of direction (wants on their end,capabilities on ours) as well as serving as a delegate to standards bodiesincluding 3GPP.

Rich Ruby(RR): I am theDirector of Technology for FBAR. FBAR (Film Bulk Acoustic Resonator) isBroadcom’s version of Bulk Acoustic Wave (BAW) piezoelectric filtering. Iconstantly look at ‘disruptive’ ways to improve our product (size, cost,performance) and also keep an eye out for orthogonal applications of thetechnology that we develop for the handset markets.

CM: Pleaseintroduce Broadcom, its product offerings and the markets that the companytargets?

WM: Broadcom is a diverse company,generally focusing on technology for Communications markets, EnterpriseStorage, Industrial and Enterprise Software. The Wireless SemiconductorDivision, where Rich and I work, focuses on developing multi-technology modulesfor the RFFE. Our customers are leading handset manufacturers.

Broadcom RFFE Module – © 2016 Broadcom Limited. All rights reserved

CM: Couldyou describe how the previous year (2018) has been for Broadcom, in general?

WM: Broadcom is a publicly tradedcompany. You can check the filed financial results. I would say “good”.

CM: Inyour opinion, what is the impact of 5G in Non-Standalone mode, meaning a dualconnectivity of both LTE and 5G?

WM: It appears the majority of early5G rollouts will be Non-Standalone (NSA). Largely this is driven byeconomics, as conversion to a 5G core can take significant investment. NSA allows the data to be sent over an NR link (NR is the 5G air interface),while network control is done over an LTE link (LTE is the 4G airinterface). This Dual Connectivity (DC, or EN-DC in 3GPP speak) allowsthe existing 4G network control to function, yet still makes use of thespectral advantages of 5G for the data transmission. Without thiscapability, deployment of 5G would be much slower.

NSA bydefinition means two links have to be maintained to make a call. Twotransmissions simultaneously can lead to significant operational challenges asthe signals can interact in ways that de-sense or even block reception. So NSA comes with many more challenges in linearity, and a stronger need toisolate signals. This increases the utility of high rejection filtering, tominimize interactions, and also of multiplexing, to share antenna ports.Sharing antennas becomes more critical under 5G as 4×4 MIMO (Multiple InMultiple Out) is required by standard for some bands, meaning 4 receiveantennas have to be used.

CM: DoesBroadcom expect a disruptive innovation in filtering technology to address thecontinuously higher number of filters in phones without additional space andwithout a price premium?

WM: By definition disruptive changesare hard to anticipate. However, I believe continued evolution of presenttechnologies is more likely than any disruptive change. The installedbase of technologies appears to be very capable of supporting the requirementsthat have so far been identified for 5G. So far, we have been able to shrink thefootprint required per band by about 15-20% per year, compensating for addedfunctionality. Maturehigh-volume processes are also typically less costly than radically newprocesses. The need for multiple simultaneous connections means tunablefiltering is not the answer, due to reduced selectivity inherent in tunableschemes. I would say the biggest challenges of 5G come not from thefilters, where we have answers, but from the antennas, where in many cases wedon’t.

RR: I agree tunable filters are not the answer. Due to the existence ofcarrier aggregated bands, multiple tunable filters would have to be switched inparallel between various ‘approved’ CA bands (ex. CA B1/B3/B7 à CAB25.B66,B30).

To date,tunable filters have not even demonstrated decent performance switching fromjust one band to another. We have also studied the new guided wave SAW(Surface Acoustic Wave) devices (sometimes called IHP SAWs). Although exciting from thepoint-of-view that they give new life to the SAW community, given a lack ofadvantage to bulk acoustics in size or performance we don’t expect thistechnology to be disruptive.

PeriodicallyA/D and D/A technologies are put forward as a way to replace filtersaltogether. To achieve the wide dynamic range and the linearity and powerhandling needed to replace filters (and at a reasonable power budget) will beextremely challenging. I would say that for the next 5 to 10 years, youwill see more and more ‘polishing’ of the piezoelectric filter technologies that do get intophones.

We alsolooked at novel circulators using FBAR, but do not see an intersection pointfor this technology and cell phones.

Broadcom FBAR filterdie – © 2016 Broadcom Limited. All rights reserved

CM: DoesBroadcom expect also to enter the infrastructure market once power levels andrequirements of massive MIMO active antennas reduce enough?

WM: It takes a fairly large teamto make the kind of product we supply. So far, we find we do better focusing on integrated solutions forhandsets rather than on other applications. It’s hard to compete with the volumesrequired in the handset space. As to powers in 5G infrastructurereducing: maybe, but early interest in 128×128 element base stations hashit the reality of cost of network deployment, and nowadays there is much moreinterest in 32×32 or even 16×16 solutions. More than a few operators areanticipating a roll out of macro coverage as opposed to the earlier envisionedsmall cell networks. So the volume of handsets is even more compelling incomparison.

CM: Do yousee the industry evolving to 12 inch wafers for FBAR technology? What is theposition of Broadcom on that topic?

RR: We constantly evaluate thepossibility of moving to 12 inch wafers. As MIMO adds more filters to thephone, the number of filters could eventually justify more than one vendorgoing to 12 inch.However, having personally been involved with multiple conversions, startingwith a 3 inch to 4 inch conversion, I can say that it will not be easy toachieve. For now, we haveplenty of capacity to support our customers’ needs.

CM:According to you, will FBAR technology handle the above 3.5 GHz 5G NR bands(N77, N78, N79) which will be the primary bands of 5G?

WM: First, a minor difference ofopinion. In my view it is an oversimplification to claim that these areuniversally the “primary bands” for 5G. n41, for example, is likely to bemore significant worldwide than either n77 or n79. In the US many peoplebelieve the new millimeter wave bands will be the “primary bands”. To me“primary” is a reflection of deployment, and my belief is 5G will get deployedon whatever unused spectrum an operator has available. I do agree,however, that in many parts of the world, including in Europe, this is likelyto be on n78. However,we should not forget that the legacy bands remain important to 5G, especiallypotentially underutilized ones such as n28, n71, or even n5.

But toaddress the gist of your question, FBAR can indeed serve these new higherfrequency bands. We can support either n78 or n79 with a single acousticfilter. We can alsosupport n77 with hybrid structures incorporating L-C structures as well as FBARresonators. SuchFBAR-based filters can solve coexistence issues with WiFi spectrum, enable asynchronous operation on n78 andn79, and provide amore protected environment for n78 operation. We believe such solutionswill be used in some phones where they are seen as adding significantcapability. We don’t expect such solutions to be universally adopted, at least in early implementations.

RR: I argue that all filters usedbelow 6 GHz will use either piezoelectric technologies or integrated passivesdevices (IPD). Piezoelectrics will be used where steep roll-off andcomplex multiplexing are needed. Where performance is not so important, you will see IPD technology.

But aninteresting question is at what frequency does BAW (Bulk Acoustic Wave) filtertechnology begin to stumble? We have already demonstrated technologiesthat can address applications up to 10 GHz. If bands in the 10 to 20 GHz regionwere to open up, we could support this region as well. However, at somepoint adequate filters can be patterned on a low loss substrate or on a die,and also the parasitic losses going in and out of packages become excessive. So, the frequencyis not so much the limiter to BAW devices as are alternate simpler technologiesor integration scenarios.

CM: Isthere still room to integrate more FBAR filters in already highly densifiedPower Amplifier modules?

WM: Briefly, yes. There arecoming technology elements that allow for a smaller filter area in theRFFE. Broadcom has significant effort in that direction. Also fortechnology that allows for denser assembly, e.g. double-sided assembly, smallerkeep-out regions, denser ball grid arrays, etc. We have several years ofclear roadmap to keep on track with the demands that have been identified todate.

RR: I agree. We are working onreducing the areas of all the components that go into the front end module.Given that filters take up the largest area in today’s modules, we are drivenhard to innovate in that direction.

CM: Basedon your technical expertise and your vision of the RF Electronics industry, doyou foresee a restriction of the number of filtering functions in LTE phone RFfront ends?

WM: The restrictions come fromcost and size, but as mentioned, for flagship phones at least, we have to datebeen able to keep pace with industry demands. From what we see so far, expected future demands are not aconcern for us.

RR: I once wrote an Op Ed piece(in 2016) where I predicted ~100 filters per phone. Today, the flagship phones have 60+filters (driven by diversity and MIMO modules as well as the primary front endmodule). I stronglybelieve that ~100 filters/phone will happen. And, I don’t see that number beinga big issue. Should it become clear that much higher numbers of filtersare needed, say 300+ filters/phone, then we will see a bifurcation of phonestargeting geographical regions (e.g. Asia, Americas, Europe etc.…); supportingthe primary region fully for data rates and less so for the other geographicalregions.

CM: Oncethe 5G network becomes standalone, do you expect the filtering function todecrease in the handset?

WM: In a word, “no”. Legacybands still need to be supported. 4G will remain active in most networksfor an extended horizon, long past the first SA networks. 5G also makesuse of Carrier Aggregation (CA). So I do not foresee a decrease in filteringrequirements in any visible future.

RR: Besides the legacy issues, theidea of designing a single phone board that is qualified for a large swath ofcountries and service providers will always keep the filter count high. Thealternative of reducing the number of bands in a phone to support one sectionof a country or one service provider, starts looking like the early ’NokiaModel’ where at the peak of Nokia’s phone presence, they manufactured andsupported over 200 different phone products.

CM: Wouldyou like to add some final words for our readers?

WM: The transition from 3G to LTEbrought significant demand for high performance acoustic BAW filtering insmartphones, but only after 1 to 2 years of network densification. FBARfiltering now addresses these needs, and in very high volume. We see aparallel scenario in 5G. As filters proliferate in the MIMO-rich NRenvironment, and as multiplexing needs increase to support CA of new bands aswell as DC, we expect our FBAR technology to continue to offer some of thehighest performance solutions available. Some OEMs and operators willeven require high performance BAW filtering in their first smartphoneimplementations. With expanded 8” capacity, continual effort on solution size reduction,BAW performance improvements and capability to meet requirements up to andbeyond 6GHz, Broadcom iswell placed to meet the demands of the 5G New Radio era.

RR: Imagine a world withoutpiezoelectric resonators for RF filters. Phones would be using theearlier ceramic filter technology (ceramic duplexers were 5 X 5 X30 mm3 in volume). In contrastBroadcom FBAR filter volumes are typically 3000 to 4000 times smaller. Next, imagine a world where the only piezoelectric filters were SAWs. SAWs work very well around 1 GHz or lower frequencies, but struggle to meetpower, insertion loss, isolation and linearity specs for frequencies at 2 GHzor higher. While SAWs are used at higher frequencies for Rx-only applications,such as in diversity modules, the power limitation is fundamental: SAWs are notgood for transmit at higher frequencies. And Tx is needed to maintain thedownload link, e.g. for synchronization. However, there is simply notenough bandwidth below 1 GHz to give 100 MHz bandwidth to a user and meet thegoal of Gigabit per sec download speeds. With only SAWs, downloadinglarge data content (movies, streaming, etc.…) would simply not be part of ourvocabulary. FBARstarted around 1993 at HP Labs. It was the first high volume BAW technology for cellularapplications; we introduced the first stand-alone FBAR duplexer in 2001. By 2013 we were in every smartphone sold. Today, all of thehigh-performance smartphones use either FBAR or other BAW technology. Inshort, you could say FBAR was a key enabler of the modern smartphone.

Interviewees

WilliamMueller has a BSEfrom Harvey Mudd College, and an MSEE from UC Berkeley.

Williamhas over 45 years of experience in RF, with emphasis on component level designand RF front end architectures. The last 34 years have been withAvantek-HP-Agilent-Avago-Broadcom (the same work group, different names) wherehe is presently Principal Technology Strategist for Broadcom Inc.’s WirelessSemiconductor Division.

Williamis familiar with RF power amplifier and low noise amplifier design, as well asFBAR filter design. He is active in standards bodies (3GPP, MIPI), industryforums (IWPC, GTI, IEEE), and with regulatory (FCC, Ofcom).

Williamhas presented numerous papers at symposia, and holds three patents relating toRF front end components.

Rich Ruby, PhD (U.C. Berkeley) is Director ofTechnology, IEEE Fellow.

Richjoined HP Labs in 1984 working on superconductivity, E-beam lithography, X-Raylithography and packaging. In 1993, he started work on Free Standing BulkAcoustic Wave Resonator devices (FBAR) and has stayed with that technologysince.

He hasmade many contributions to the acoustic properties, manufacturability and thepackaging of FBAR filters and duplexers. Rich commercialized the first FBARduplexers HPMD7901 and the 7904 back in 2001 to 2003. The first all-silicon,chip-scale packaged FBAR duplexer was introduced in 2004. Today, Avago/Broadcomsells over 2 Billion FBAR filters per year into the mobile market.

Over theyears, Rich was awarded the Samuel Silver Award, the Barney Oliver Prize, theBill Hewlett Award, the C.B. Sawyer Memorial Award, the Institute of AmericanPhysics Prize for Industrial Application of Physics and recently DistinguishedAlumni of U.C. Berkeley. Rich has over 90 patents in the area of FBAR devices.

Interviewer

As aTechnology & Market Analyst, specialized in RF devices & technologieswithin the Power & Wireless division at Yole Développement (Yole), Cédric Malaquin is involved in the development oftechnology & market reports as well as the production of custom consultingprojects. Prior his mission at Yole, Cédric first served Soitec as a process integrationengineer during 9 years, then as an electrical characterization engineer during6 years. He deeply contributed to FDSOI and RFSOI products characterization. Hehas also authored or co-authored three patents and five internationalpublications in the semiconductor field. Cédric graduated from Polytech Lille in France with an engineering degree inmicroelectronics and material sciences.

本文为全文引用i-micronews的访谈文章。

仅仅标题做了修改。蓝色的字体为编者特意作为警示突出。

感谢原作者的付出。