Betting on WiMAX chips

WiMAX has the benefit of a mature and innovative chipset ecosystem. Increasing volumes is now the priority to lower prices further.

When WiMAX supporters put forward the business case for deployment, they invariably refer to a mature and highly competitive WiMAX chipset ecosystem as one of their trump cards. With over 20 WiMAX chipset manufacturers in existence-and each appearing (so far) to appreciate the importance of growing the market by keeping prices down through not making any exorbitant royalty claims on patents-it looks a forceful point to make.

But the proof of the pudding, as they say, is in the eating. Yet the WiMAX camp, with some justification, can claim they are starting to deliver on the promise of lower-cost devices directly as a result of competitively priced chipsets.

When Xohm (now Clearwire) fully launched its Mobile WiMAX service in Baltimore in October 2008, the USB dongle-manufactured by ZTE using a Beceem chipset-retailed at an impressively low $59.99. WiMAX device chipsets, says Sprint, are in the $20 range.

“This is a far cry from what 3G can do on price,” says Lars Johnsson, VP of business development at Beceem. He argues it would cost a mobile operator twice as much to put a 3G dongle into the market. “The 3G chips are much more expensive due to ecosystem deficiencies where Qualcomm calls all the shots,” he continues.

Although Sprint-which fully owned the Xohm business unit at the time of commercial WiMAX launch-may have been able to lower its WiMAX device prices to a certain extent through the weight of its purchasing power, the third-largest operator in the US says there are “virtually no subsidies” on any of its end-user Xohm devices. (The Samsung PC express card retails for $59.99, while the Xohm modem from Zyxel sells at $79.99.)

Moreover, there are no lengthy contract lock-ins attached to the Clearwire service, something that mobile operators usually insist on as a form of a device subsidy if they are to give away their USB dongles and PC cards for ‘free’.

“I thought the prices [for Xohm end-user devices] would have been higher,” says Monica Paolini, founder and president of Senza Fili Consulting. “When 3G came out, devices were much more expensive unless you signed a contract, so $60 is a remarkable price point. When wifi started, the cards cost $250 and it took some time before they came down to the $60 level.”

The WiMAX chipset ecosystem can boast a number of heavyweight players, including Intel, Motorola, Samsung and Fujitsu. Intel, as WiMAX’s principal financial backer, is making particularly strong progress in lowering the cost for WiMAX-embedded laptops, netbooks and MIDs (mobile internet devices) through developments at the chipset level.

In April 2008, the US chip giant unveiled Atom, parading it as the world’s fastest processor under three Watts. Using 45nm silicon technology, Atom is aimed at the ‘affordable’ netbook market.

And by combining the Atom processor with lntel’s EchoPeak chipsets-which integrate wifi and WiMAX on one piece of silicon-they promise to be a potent combination in driving WiMAX adoption through seeding the market with WiMAX-embedded devices at prices not much higher than if they didn’t have WiMAX capability. EchoPeak chipsets, which are now shipping to PC manufacturers, is central to Intel’s WiMAX-embedded device strategy.

“You can add WiMAX to wifi chips with almost no incremental cost to the notebook or netbook,” says Ron Resnick, president of the WiMAX Forum.” If you try and add an HSDPA modem to that device, it can cost between $60 to $100 extra.”

But there is also a slew of smaller companies and start-ups in the WiMAX chipset ecosystem, including Beceem, Comsys, Altair, Wavesat, Runcom and picoChip. Competition is helping to drive innovation, which, in turn, lowers chipset costs and increases performance.

One important chipset development is the trend towards WiMAX SoC (system-on-a-chip) solutions, which combine the RFIC and baseband components (see ‘WiMAX device chipset definitions’ sidebar). Historically, different companies manufactured the RFIC and baseband chipsets to produce two-chip solutions. By integrating the RFIC with the baseband chipset on one piece of silicon, the bill of materials (BOM) cost can be lowered by as much as 25 percent, according to Beceem, with no performance impairment.

A string of WiMAX chipset suppliers now have single-chip designs, including Beceem, Fujitsu, GCT Semiconductor, DesignArt and Sequans.

Chasing volumes
While innovation paves the way towards more cost-efficient chipsets, higher volumes are obviously key in bringing costs down. Although WiMAX is growing, it is still a nascent market and high volumes will be difficult to come by in the short term, particularly for the smaller chipset companies.

Factor in the economic downturn and the prospect of generating much lower unit shipments than previously forecast, then it seems reasonable to expect that the WiMAX ecosystem will shrink to some extent through 2009, either through consolidation or start-ups falling by the wayside simply through lack of capital.

Some baseband chipset vendors, such as Wavesat, Comsys and Altair, are adopting what they believe to be a risk mitigation strategy by using the same piece of silicon to support more than one air interface, which increases their addressable market.

The growth strategy of Wavesat, a privately-held fabless chipset supplier headquartered in Canada, revolves around its ‘multimode 4G architecture’, whose target customers are OEMs (original equipment manufacturers) and ODMs (original design manufacturers) of end-user devices (dongles and express cards) and CPE.

Dubbed the Odyssey platform, it aims to serve multiple ‘4G’. This gives it a better chance of generating higher unit volumes, which. in turn, lowers costs and should make the company more competitive.

Wavesat’s ‘standards-agnostic’ position comes from having developed its baseband chipset to be software-programmable to Mobile WiMAX or XG-PHS. (XG-PHS is the next-generation PHS (personally handyphone system) standard, which has been adopted by Willcom, a PHS operator in Japan.) Plans are also afoot to make the Odyssey platform software-programmable to LTE by April 2009.

This flexibility allows Wavesat not only to serve multiple markets (and hedge its 4G bets) but enables its OEM and ODM customers to design and manufacture end-user devices to different ‘4G’ standards without needing to change the format of the device. The Wavesat proposition to its OEM/ODM customers is that all they need do is upload software in the factory to make the device adhere to the ‘4G’ standard of their choice, which gives them flexibility in responding quickly to market demand without increasing their BOM cost. The switch between the standards can only happen, though, if the same frequency band is used.

“Cost reductions can be quite dramatic with higher unit volumes,” says Raj Singh, Wavesat CEO. “The extent of those costs reductions will obviously depend on negotiations with the foundries [which supply the silicon], but once you get between two million and three million units per year, the prices start to come down quite rapidly.”

The Odyssey platform started sampling in 2008 and Singh says enough design wins have already been secured to ensure volume ramp-up in 2009. “We’re forecasting that our volumes will be in the 6-7 million unit range by 2010, across all market segments,” he says.

End-user devices using Wavesat’s Odyssey 8500 platform will not be multi-mode in the sense of users being able to roam seamlessly between different 4G networks. Singh does envisage, however, that the same 4G device (using software-defined baseband chipsets) could still be configured to access different OFDMA-based standards if it were an operator requirement. By having different software stored inside the device’s Flash memory, a user could activate the appropriate 4G access bearer-at the press of a button, perhaps-according to the geography he or she is in. “That’s a pretty good trick,” says Singh.

For Singh, the time has not yet come when it makes sense to take the SDR concept beyond the air protocol. To optimise performance and silicon size, he argues, the core of OFDM technology still needs to be hardwired into the chipset. “We have done a lot of work in building customised DSPs that do one thing and one thing only,” says Singh.

The current crop of multimode products from Comsys, an Israel-headquartered supplier of integrated baseband chipset solutions, is focused on combining Mobile WiMAX with established cellular standards; it is already shipping its multimode WiMAX/EDGE CM1125 chipset. “It should be noted that as GSM/EDGE is not OFDM/A based, there is a requirement for a minimal amount of extra hardware for support of this standard,” says Ehud Reshef, WiMAX product marketing director at Comsys. “But the CM1125 is true multimode in the sense that a single firmware can support GSM/EDGE and Mobile WiMAX, and service continuity between the two, given the right upper layer protocols support.”

Comsys is now working on its next-generation ComMAX solution, which is focused on LTE, but it is not currently looking at a 4G multimode architecture along the lines of Wavesat. “Due to the different operators deploying WiMAX and LTE, we currently do not see the need for a WiMAX/LTE multimode solution,” says Reshef.

Beceem’s Johnsson is not convinced, however, that software-programmable air-interfaces is either cost-efficient or even necessary, particularly as he believes LTE is still some way off. “It is an understood phenomena that software modems, which have high-performance DSPs, do not have the power performance efficiency of the more hardwired silicon that we have,” he argues.

Beceem, which counts Intel and Samsung among its investors, is a fabless semiconductor company exclusively focused on producing chipsets for Mobile WiMAX devices. And to win orders, and so increase chipset volumes, Johnsson says the key battlegrounds are performance, multiple radio capability (to address a global market) and having a large IOT (interoperability testing) footprint.

In late 2006, Beceem launched what it claimed to be the world’s first Wave 2 compliant Mobile WiMAX chipset in the market. “We subsequently set the MIMO application layer speed record at 33Mbps and nobody has been able to do that as of today,” says Johnsson. “Our nearest competitor does 25Mbps.”

Beceem produces 65nm WiMAX chipsets with in-built triband radio covering all the standardised WiMAX frequency bands: 2.3GHz, 2.5GHz and 3.5GHz. Johnsson adds that Beceem has done ‘very deep IOT” with ten base station vendors, including Alcatel-Lucent, Alvarion, Motorola, NEC, Samsung and ZTE. “These factors combined,” says Johnsson, “gives us a lot of leverage with the ODMs in Taiwan, which can address a global market with Beceem chips.”

Despite the gloom surrounding the chipset market with the prospect of lower demand for laptops and netbooks in the wake of the credit crunch, Johnsson is confident that 2009 can still be a good year for Beceem. “Clearwire will bode well for us in 2009.” he says. “Last year the market for mobile WiMAX chipsets was around one million units across all suppliers, but this year we expect that to go up to four or maybe five million, and we’re looking to get a 50 percent share of that. We stand to profit from that very nicely.”

WiMAX device chipset definitions
WiMAX device chipset: A chipset that is based on or evolving to IEEE 802.16 standards-specifically 802.16-2004 and/or 802.16e-2005-and which is embedded in a device that a WiMAX subscriber uses to access a WiMAX service. Furthermore, the chipset typically consists of both RF/IF and baseband components.

RF/IF: A radio frequency integrated circuit and associated intermediate frequency components used in a WiMAX chipset, including the RF transceiver, power amplifier and power management. The RF/IF component is built to, or evolving to, IEEE 802.16 standards-specifically 802.16-2004 and/or 802.16e-2005-and operates in WiMAX spectrum, such as bands in and around 2.3GHz, 2.5GHz, 3.5GHz and/or 5.8GHz.

Baseband: A baseband integrated circuit used in a WiMAX chipset, including the associated memory and other components supporting key baseband functions such as MAX (media access control). The baseband component is built to or evolving to IEEE 802.16 standards, specifically 802.16-2004 and/or 802.16e-2005.

Source: Informa Telecoms & Media

CE device market potential
Another potential way to increase WiMAX chipset volumes could be through the embedded consumer electronic (CE) device market. If CE devices-such as digital cameras, MP3 players, gaming consoles and the like-could be WiMAX-enabled without adding significantly to the cost of those devices, then a new and extremely large market could be opened up to the chipset suppliers.

Wavesat is looking to develop a ‘pared down’ WiMAX chipset for the embedded CE device market and achieve price parity with 802.11n chipsets (the higher performance version of wifi). Singh says this might be achieved in 2010 at a unit price of around $8. “If the cost of WiMAX chipsets are the same as wifi chipsets, the ODM perception of the [CE] embedded market completely changes,” he says.

But if the claims of Coresonic are to be believed, the added cost of WiMAX connectivity to CE devices will be much lower than the $8 level envisaged by Singh. “Because our solution is so small and low power, the incremental cost to someone else’s chip is less than $1 in terms of additional silicon area,” claims Rick Clucas, Coresonic CEO. In fact, he argues the incremental manufacturing cost of using its solution could drop as low as $0.20-if the volumes are high enough-when added to an SoC and using an existing CPU (in the SoC) to run the MAC layer.

Coresonic, headquartered in Sweden, is not a chipset maker in its own right. It is an IP (intellectual property) semiconductor company that licences its programmable architecture to companies further up the eco-system-the chipset suppliers-so they can manufacture SoC silicon that has an extremely efficient and cost-effective baseband component for processing applications.

Dubbed LeoCore, Coresonic’s baseband architecture can support a range of wireless standards-including 3G/HSPA and DVB-T/H-plus protocol-specific firmware, but it is in the WiMAX-enabled CE device market where Clucas sees one of the biggest opportunities. One reason for that, he says, is because baseband competition for CE chipsets is far less fierce than in the cellular chipset market where the ecosystem relationships-from mobile operators through to OEMs and ODMs-are much more entrenched. Clucas sees chipset makers for CE devices, rather than the WiMAX chipset makers, as his main addressable market.

The Swedish company makes a number of bold claims in its ‘LeoCore WiMAX Personality Pack’ literature. One is that the company’s complete Mobile WiMAX solution (from the RF interface through to the central processor interface) can be implemented in a 65nm design using less than 1.5sqmm of additional silicon. Moreover, says Coresonic, Mobile WiMAX (using MIMO) will only require a maximum clock speed of 250MHz to run on LeoCore, which, along with a ‘very small gate count’, translates to a power consumption of less than 50mW in a 65nm process. And if you factor in power management and real-life usage patterns, continues the Coresonic marketing pitch, then power consumption would be a fraction of this.

“The solution we are offering is so radically smaller and much more efficient than anything else available on the market that we have had potential customers coming up to us saying they didn’t believe it,” says Clucas. “That’s why we put a platform together to demonstrate it [at WiMAX World 2008, held in Chicago, September 2008].”

Coresonic still has its detractors and sceptics. Beceem’s Johnsson finds Clucas’ incremental cost claim of $1 and below as “simply not credible”.

But Coresonic would have the last laugh if it could make good on its promises, and Clucas expects the market to eventually vindicate LeoCore. “You might see chipsets [using LeoCore] by the end of 2009 with commercial products available in the 2010 timeframe,” he says. “This fits in with the time when WiMAX rollouts will be mature.”

More specialisation
How will the WiMAX chipset market likely develop? Wavesat’s Singh says that chipset suppliers will try and differentiate more through greater specialisation as the WiMAX market grows. “As with all nascent markets, there is a bunch of start-ups but there isn’t enough granularity of definition to segment it,” he says. “There isn’t a big enough market to do multiple chips, so everyone has one chipset, occasionally two.”

As the WiMAX market develops and grows, a single chip for all uses becomes less practical. “The requirements of a CPE in terms of memory and processing power are significantly different from the requirements of a handset,” says Singh. “You are going to get a scenario where companies will have to bifurcate their product lines and do multiple things; one chipset design for handsets, one for more powerful CPEs, and a more integrated chipset for laptops.

Singh says the CE market is promising, as are home base stations. “The femtocell market may well be the killer app that makes this thing really take off,'” he says. “If you are inside the home, and your operator can give you five bars of signals on your cell phone, you actually have no reason to have a home line.”

Wifi is unmanaged spectrum, and for penetrating walls it is not ideal, so WiMAX should have an advantage argues Singh. “We’re shipping for femtocell usage today using existing silicon,” he says. “It’s not completely optimised but it is a solution. In two or three years’ time, we could have two or three product lines but you need to have big enough markets that can support each of them.”

One strain that the smaller chipset companies might find difficult to bear is to scale up adequately to develop ever more sophisticated software that will be required to interoperate with more and more players. While Wavesat falls into the 70-100 employee bracket, Singh ruefully points out that Broadcom has a 300-stong team working on software development alone for its wifi chipsets.

But while WiMAX supporters typically talk about having a much more developed chipset eco-system than LTE, that advantage may well be short lived, says Paolini. “A lot of WiMAX chipset vendors are getting ready for LTE as they too want to keep their options open,” she says. “It’s compelling for them to do that, because a lot of the work that they’ve done on WiMAX is going to be done for LTE as well. Yes, the uplink is different, but there are a lot of commonalities. Once you have a WiMAX chipset, adding LTE doesn’t cost huge amounts. And it’s a way to address a market that’s going to be bigger. I think all of them, eventually, will have a WiMAX/LTE product.”