Earlier this year, at the 3GSM summit in February, femtocells really grabbed the headlines for the first time, with the main focus on their potential to extend UMTS/HSPA networks and business cases. With even the mighty Ericsson appearing to endorse the concept (though in reality, very half-heartedly), the 3G femtocell was firmly on the operators’ radar, and some significant RFPs have followed, including Vodafone’s (see Wireless Watch July 9 2007). The femtocell for WiMAX (or other emerging OFDM networks like LTE) remained a more shadowy development, until Sprint Nextel issued its ground breaking RFP for the technology in April and set its seal on the approach. In a WiMAX World show last week almost over-dominated by Sprint’s Xohm WiMAX-based initiative, the WiMAX femtocell stepped into the limelight.
Sprint Nextel’s ‘president of 4G mobile broadband’ Barry West ‘ eclipsing even Intel’s Sean Maloney as the ubiquitous tub thumper for WiMAX ‘ waxed lyrical about ‘low cost internet base stations’, as the operator refers to femtocells. The terminology deliberately highlights Sprint’s insistence that its WiMAX deployment is designed to support a new set of services and business cases, geared around the open internet model, rather than merely to supplement its 3G activities and the classic cellular approach. In this context, the WiMAX femtocell will play a different role, at least in the early stages, than its 3G counterpart. Interest in UMTS/HSPA femtocells has been driven mainly by the urgent need for operators to extend their macro networks cost effectively into the home, in order to support new tariff options such as homezones (in a more controlled way than using dual-mode Wi-Fi/3G), and better indoor penetration, to enable future converged services and high bandwidth, home-oriented wireless applications such as broadcast. In WiMAX, of course, there are no existing macro networks, and so the role of the femtocell is less obvious, but potentially even more significant to the operator case, particularly given that, for now at least, WiMAX is working in higher frequencies than most 3G, and so faces even greater indoor penetration challenges.
For operators like Sprint Nextel, the miniaturized base station, typically backhauled by the customer’s own broadband line, offers three opportunities, as outlined by West ‘ one, to offer very advanced services to a highly targeted user base in the first instance, rather than being forced to blanket large areas containing few early adopters; two, to support Sprint’s bid to incorporate WiMAX into the whole range of consumer devices, not just handsets and modems, with West particularly interested in the trend to combine femtocells with home gateways, as NEC and Netgear have done; and three, to maximize the ability to support a new breed of internet oriented, high bandwidth services without a non-viable level of investment in the macro network. In particular, Sprint is likely to look for wholesale opportunities, where femtocells could conceivably replace Wi-Fi access points with a branded partner (poaching Starbucks from T-Mobile Hotspot would be an obvious target). Another benefit of WIMAX femtocells, according to Sprint Nextel, will be to support enterprise services, which it expects to be a strong early market with a quicker ramp-up than consumer offerings. Atish Gude, senior VP of mobile broadband at Sprint Xohm, believes enterprises will be the first investors in femtocells.
Companies like Sprint and Intel know that, for WiMAX really to become a dominant force in global wireless, it needs to hijack the apparently unstoppable trend towards open access and the mobile internet, and ensure that it becomes the primary network to support that (hence the prospect of Google or another internet giant acquiring spectrum such as US 700MHz licenses and building out in WiMAX). To succeed commercially, such ventures need a far lower cost of deployment than UMTS/HSPA has offered, because of the huge levels of data traffic that will need to be supported, much of it non-revenue generating for the carrier ‘ hence the potential significance of a ‘bottom-up’ approach that uses commoditized products to a large extent and can be flexible scaled, and where the consumer will often fund the backhaul and even the device (another of Sprint’s most critical objectives).
Sprint has gained the highest profile, but many of the larger operators that are serious about WiMAX are also demonstrating high levels of interest in femtocells, and the suppliers are starting to respond to the potential. Korea’s KT has been working both with its key vendor Samsung and with picoChip (the UK-based pioneer of femtocell chips) on WiMAX femtocell concepts for a couple of years. And Japan’s NTT DoCoMo, which recently formed a $620m joint venture with smaller operator ACCA to bid for 2.5GHz WiMAX spectrum in its home country, was last week talking up inhouse developments of femtocells for 3G, WiMAX and LTE.
To date, most of the vendors promising commercial femtocells have been specialists ‘ such as ip.access and Ubiquisys ‘ and 3G-oriented. In 3G, most of the equipment majors are now at least tipping their hats to the technology, and some, like Alcatel-Lucent, are taking it very seriously, while other smaller players have joined the fray, including Airvana from the CDMA space and RadioFrame from the iDEN market. In WiMAX, the early moves by major operators will stimulate action from their would-be suppliers, so all three Sprint Nextel vendors ‘ Samsung, Motorola and Nokia Siemens ‘ are on the move, while the femtocell opportunity could also be a chance for a so-far neglected vendor like Nortel or Alcatel-Lucent to get a piece of the largest WiMAX build-out plan to date. ZyXel, the world’s largest maker of home gateways, also threw its hat in the ring at WiMAX World with its first femtocell product plan.
Many challenges remain in implementing femtocells so that they live up to their promises, whether in 3G or WiMAX. The greatest may be the timescales involved ‘ large operators are talking about going commercial in mid-2008, and Vodafone is already setting price expectations at ‘substantially less than $100 per unit’ as CEO Arun Sarin said in July. Given that Wi-Fi access points can be half that, the target needs to be realistic for the mass market, but will be almost impossible to achieve within a year, at least without significant discounting or bundling by the major suppliers. WiMAX femtocells do not have the early stage volume potential and so pricing is expected to be higher (from a theoretical $200 now to about $150 by mid-2008), but as in 3G, achieving even that level of pricing while ironing out real world problems in the field, all in the aggressive timescales set by operators like Sprint, will be a major challenge.
It will also be interesting to see which chip companies get serious about the market. picoChip has been the frontrunner and for a long time the only real contender, especially in WiMAX, but is now being joined by specialist and tier one challengers. In WiMAX, Sequans has reaffirmed its commitment to making both base station and CPE silicon ‘ having been, for some while, apparently veering towards CPE-only ‘ and with that, has announced a femtocell strategy. Another WiMAX specialist promising femtocell silicon is the maverick NextWave Wireless, although an actual product is at least a year away.
Sequans has already shown off its 802.16e wave 2-ready Mobile Station product, and at WiMAX World was focusing on its newer base station chip and its move into providing its own RFIC. The French start-up concentrates heavily for differentiation on its decoder technology and consequent low power consumption claims; and on its ability to supply both base station and CPE elements, with resulting guaranteed interoperability. The former is likely to become increasingly significant as WiMAX moves into smaller devices, while the latter should fade in importance as confidence in multivendor interoperability grows and mobile WiMAX operators look for free choice of subscriber unit. The vendor also makes capital out of the cost efficiency of combining the PHY and MAC on an ASIC (though with resulting loss of flexibility), and from being optimized specifically for WiMAX.
In base stations, Sequans has tended to play in smaller form factors, and is now pushing down into the femtocell arena, claiming that its ASIC-based approach will achieve commodity pricing levels ahead of DSP alternatives. Though this gives the company a differentiator from picoChip, TI and others, the limited flexibility of the ASIC does carry risk at the early stage of the market, where 802.16e profiles and femtocell designs themselves remain immature and fluid ‘ a fact Bernard Aboussouan, VP of marketing, acknowledges, admitting to the compromises inherent in the low cost/simplicity versus flexibility decision Sequans has taken.
In femtocells, Sequans is working with PMC-Sierra to create a reference design using its SQN2130 base station ASIC and its partner’s PM8800 RFIC. Although Aboussouan claims this RFIC is the ‘only one with the features needed for femtocells ‘higher modulation, higher output power, 2Tx/2Rx configuration, and multiple frequency bands’, Sequans has also decided to offer its own RF product, though this is ‘only really for vendors who insist on both baseband and RF from us’, suggesting that, at this stage, Sequans believes an optimized solution would use third party RF.
It is similarly candid about the limitations of its femtocell product in its first iteration, presenting this largely as a statement of direction and placeholder rather than a finished solution. ‘Our current product is not really optimized for femtos, it needs to be cost optimized further,’ said Aboussouan, promising a next generation product that would be tailored specifically for this market. But like TI (see below), the company wants to see volume commitments from operators first before it commits to the level of investment required for that development. The same goes for LTE, where it believes, like its rivals, that the crossover with WiMAX will make a move to the new standard relatively easy. ‘Our chip has some flexibility for other standards ‘ not loads, but we are definitely looking at 16m and LTE,’ commented Aboussouan. ‘We could make an early LTE move, though the actual 2130 chip would probably not be adaptable for LTE. It can do FDD and 20MHz now, and has enough programmability to adapt to 16e changes, but it would probably only be a 75% solution for LTE and 16m.’
In 3G or multi-platform developments the field will be more crowded, and as in the base station equipment field, a great deal of medium term success will rely on the vendor being able to apply its technologies to multiple platforms, a task that should theoretically get easier as wireless networks become more similar. Thus companies like picoChip will look at leveraging these similarities to make an early move on emerging markets like LTE and the Chinese pre-4G technologies. For now though, despite the growing interest in femtocells ‘ now approaching the hype stage ‘ many chip majors remain ambivalent, at least in public. Qualcomm and Broadcom are not revealing their hands, though the latter will be a natural candidate to enter the market if the femtocell does become a consumer commodity product like Wi-Fi access points, where Broadcom has enjoyed stellar success. Sources within the once enthusiastic STMicro now say the French company is pulling back from the market (potentially an odd move given its new closeness to Nokia in the 3G space); while ADI’s solution remains expensive and appears to have no clear roadmap on the baseband side (it is more likely to play heavily on the RF front end, where it is promising a single-chip HSPA product for under $12 and the AD9352/53 RFCMOS direct conversion WiMAX transceiver for femtocells.
Texas Instruments has the potential to influence the market significantly but so far has been uncommitted, at least in public, though it says it can adapt existing products for femtocells where customers require this. It admits this is ‘too expensive and not optimal’ but says it has not made a firm decision yet as to whether (or more likely, when) to create an optimized platform specifically for femtocells. Currently, according to Dr Arnon Friedmann, head of software at TI’s infrastructure division, there are several business case issues to solve before TI would be willing to commit the $30m he estimates would be needed to create a full system on chip (SoC). ‘Nobody is yet putting their money where their mouth is to encourage us to make that investment,’ he commented.
Challenges that TI perceives include the increase in customer support required once the subscriber has his or her own base station; the issues of co-channel interference with the macro network and of base stations being moved around frequently; the fact that performance and QoS will rely on the quality of the user’s DSL connection; and whether operators, despite the pronouncements of Sprint Nextel, will have to end up subsidizing femtocells as well as devices, at least in the early stages (as Sprint is doing with its CDMA femtocell, Airave). There will also be the complexity of agreements between mobile and DSL operators, where these are not the same company. Until these business and technical issues are fully addressed, TI officially remains cautious.
As Alan Gatherer, CTO of the wireless infrastructure business, put it in a recent interview: ‘The problem with femtocells at the moment is that they could be a mass market but they are not today. Reuse is tremendously important in a market that has no volume yet so our general philosophy is to get as much reuse out of the base stations that exist today. We already have devices that can do femtocells and because of reuse, customers do not have to start from scratch for a femtocell. But my finance director will not let me build these parts until there is a clear market.’
In next week’s Wireless Watch we will analyze the developments in subscriber unit chip technology showcased at WiMAX World, and look into the current position and strategy of Intel.
MobileAccess:
MobileAccess has developed an in-building WiMAX module that works with its
Universal Wireless Network, a product geared to improving wireless coverage in malls, office buildings and other large indoor areas. The product is a blade that slides into the telecoms closet.
The company’s Universal Wireless Network uses fiber/coax-based in-building capabilities for cellular voice, public safety radio, wireless Lans, and wireless medical telemetry over the same shared infrastructure.
CEO Cathy Zatloukal said: “Our new solution will be instrumental in bringing the services inside buildings because it eliminates the need for customers to install a cabled infrastructure dedicated to supporting WiMAX.”
Nokia Siemens:
Airvana has announced that its femtocells will interoperate with Nokia Siemens’ 3G infrastructure, a sign of a wave of interoperability alliances that will start to take place to reassure operators that they can mix and match femtocells with their existing macro networks.
“Delivering operators a complete, standards-based femtocell solution will enable them to offer new services and expand relationships with their customers,” said Timo Hyppola, from Nokia’s Indoor Radio Solutions unit.
Sprint and Airave:
As planned, Sprint Nextel has rolled out its first commercial femtocells, in an initial launch in Denver, Colorado and Indianapolis, Indiana. Unlike the grand objectives of its WiMAX femtocell roadmap, Sprint has rather more conservative aims for its CDMA devices, primarily to boost indoor coverage and support homezone tariffs on offers such as unlimited voice calls within the home.
All Sprint handsets except iDEN models are compatible with the femtocell, which is supplied by Samsung and costs $49.99 to the user (a price tag heavily subsidized by the operator). The service then costs an additional $15 per month ($30 for a family plan) on top of the existing fees. The Airave launch is a direct response to T-Mobile USA’s roll-out of its HotSpot@Home service, which uses a Wi-Fi access point, dual-mode handsets and UMA (Unlicensed Mobile Access) instead of a cellular femtocell, but otherwise supports a similar offering. The Airave ‘ based on Samsung’s Ubicell product – will plug into a home broadband connection to create a home cell of up to 5,000 square feet. It can handle three calls at a time, or CDMA 1xRTT data, and the user can leave it open or give Sprint a list of up to 50 authorized phones that can use the cell. It does not yet support EV-DO.
picoChip and Samsung:
Samsung has made a strategic investment of undisclosed size in picoChip, reflecting the Korean vendor’s intensifying interest in femtocells, where the UK chip designer is a pioneer. Samsung’s injection is part of a $27m D round announced in July and led by Highland Capital Partners, which brings total funding to date to $70.5m. CEO Guillaume d’Eyssautier says he expects this to be the last funding requirement before picoChip positions for IPO. In a statement, d’Eyssautier said: ‘The new round of funding reaffirms our leadership in multicore DSP for advanced wireless markets with our WiMAX, HSPA and femtocell products.’
DSP benchmark results endorse multicore:
In an independent benchtest conducted by Berkeley Design Technology (BDTI), the picoChip PC102 multicore DSP has emerged with a fortyfold price/performance advantage over traditional DSPs from companies like Texas Instruments.
The BDTI Communications Benchmark compared the PC102 ‘ the first volume shipment multicore product to be tested – with conventional products from TI, Freescale, Altera and others (see table), concluding a cost per channel of $6.78 from implementing 14 channels in one device at 160MHz. This was 40 times less than the TI TMS320C6455 at 1GHz, though TI also offers a cost optimized variant which emerged at $48 per channel. The Freescale MPC8144 at 1GHz delivered 3.5 channels for $57 each.
picoChip CEO Guillaume d’Eyssautier said the company expects a further threefold price/performance improvement from the newer PC20x family.
The results justify the company’s commitment to multicore architectures, he commented. These architectures are pushing DSP technology further towards FPGAs in terms of capacity and cost per channel, said BDTI. Processors like TI’s can handle only one channel, and so multiple DSPs are needed, or a combination of DSP and an FPGA, both expensive options for a multichannel application. The FPGA can handle more channels than any DSP, but remains very expensive. BDTI said: ‘The Xilinx FX140 ‘ is much more expensive than the PC102, but has a lower cost per channel.’
High Capacity Results
TI TMS320C6455 1GHz 1.09 channels $268.50/channel
TI TMS320C6455 1GHz <1.8 channels >$163/channel
Freescale MSC8144 3.5 channels $57/channel
Altera 1S80-6 60 channels $10/channel
picoChip PC102 @
160MHz 14 channels $6.78/channel
Xilinx Virtex-4 FX140
? 10 speed grade 432 channels $2.97/channel
Cost-Optimized
TI TMS320C6410 400MHz 0.31 channels $48.23/channel
Freescale MSC8144 3.5 channels $66/channel
Altera 1S80-6 60 channels $10/channel
picoChip PC102 14 channels $6.78/channel
Xilinx Virtex-4 SX25
Slowest speed grade 64 channels $1.39/channel
Altera Stratix II 2S15
Slowest speed grade 20 channels $2.75/channel
Results for the BDTI Communications Benchmark (OFDM)TM ‘ 2007 BDTI. For more info and results see www.BDTI.com