The roadmap towards mobile WiMAX

Rupert Baines, VP Marketing, picoChip

WiMAX has established itself with the promise of providing a standard for wireless broadband access – in the same way that WiFi did for wireless LAN – that will bring costs down with the introduction of volume off-the-shelf components and subsystems. This article describes the options that must be addressed along the road to mobile WiMAX, and analyses whether mobile WiMAX will be complementary to 3G or will broadly compete with it.

Although there has been an undeniable element of hype, the importance of WiMAX is hard to deny. Most analysts agree that standardisation under the WiMAX banner will help drive the fixed wireless market forward, after many false starts. It provides equipment vendors with off-the-shelf silicon and reduces costs to levels where volume deployment is possible. Just as many of the two billion cellular users are in areas with no fixed network, WiMAX has the huge potential of bringing broadband access to the hundreds of millions of people around the world who do not have access to cable or DSL.

Even though it has been designed to maximise interoperability by taking only a subset of the much wider IEEE 802.16 standards on which it is based, WiMAX encompasses a wide range of different options, each with slightly different technology or requirements, of which arguably the most significant is mobile WiMAX, or IEEE 802.16e. Many see this as the most exciting area in wireless today, with the potential to leapfrog 3G..

One of the most fascinating things on this technology is how it encompasses a wide range of applications: more than address a market it serves a complex portfolio of different markets, each with somewhat distinct characteristics. Another unique aspect is the wide variety of radio options supported within the one standard.

Applying WiMAX for fixed applications, as a broadband technology in places with little infrastructure, is very attractive. But the opportunity that is getting everyone more excited is for mobility.

(Fig.1)

(Fig.2)

Deployment Choices

There are several ways in which WiMAX can be deployed. One is high-bandwidth, point-to-point backhaul, for example from 2G/3G sites or WiFi hotspots. A second market is “metro Ethernet” where bandwidths of 10Mbit/s and above are provided on a point-to-multipoint basis, competing with fibre. Another is ‘Access’ where competitive operators use WiMAX in the 1Mbit/s to 10Mbit/s range as an alternative to DSL or cable modem, potentially with longer range and hence better economics. While this can be a competitive offering, the opportunities are most powerful in territories without much installed copper plant, using WiMAX to obtain access to the Internet– a potential market of billions of users worldwide. Finally, of course there is mobility: ranging from nomadic use (“super hot spots”) through portable to true high-speed mobile data services, adding a further range of options.

Uniquely for wireless standards, WiMAX does not specify the radio types, and allows for both FDD and TDD with channels of 1.75MHz-28MHz, and almost any carrier. The WiMAX Forum has defined several standard profiles to suit the most common requirements around the world, the most popular channel bandwidths being 3.5MHz FDD for licensed bands, and 10MHz TDD, which can be used for both licensed and unlicensed applications. Other standard WiMAX profiles include 1.75MHz, 3MHz, 5MHz, 7MHz and 10MHz. Other options that can potentially be employed are to re-use either the 1.25MHz CDMA profile or the 6MHz TV profile.

Although 10MHz TDD is the primary profile that has been adopted by Forum members, many operators, such as Sprint, own spectrum that requires 5MHz channels, while Clearwire in the US is using a band with 6MHz channels.

(Fig.3)

In addition any of these profiles can be superimposed on a variety of carrier frequencies ranging from 450MHz up to 5.8GHz and above. For example, there are operators using WiMAX at 450MHz, 700MHz, 1.9GHz, 2.3GHz, 3.5GHz, 4.9GHz and 5.8GHz – and probably a number of others as well. This variety of frequencies and modes is a big difference compared to other wireless standards – especially WiFi, which uses spectrum in bands that are unlicensed in most of the world’s regions. While the WiFi installed basis is both large and homogeneous, the situation for WiMAX is much more complex due both to the higher transmit power levels and to the fragmented radio spectrum, which spans both licensed and unlicensed bands that also differ from country to country.

(Fig.4)

This diversity of frequency and channel options places enormous demands on the silicon that is used in WiMAX systems. Software flexibility is the key to allowing a system to adapt to all of these variants. The picoChip PC8520, for example, is designed to support 1.25MHz, 1.75MHz, 3MHz, 3.5MHz, 5MHz, 5.5MHz, 6MHz, 7MHz, 8.75MHz and 10MHz channel bandwidths with either FDD or TDD. It is also being used in systems at a variety of carrier frequencies in the range 450MHz to 5.8GHz.

Certification of WiMAX products will begin this summer at Spanish test house CETECOM, and is focused on the most popular profiles - 3.5MHz FDD at 3.5GHz, and 10MHz TDD at 5.8GHz. However in Korea the licence has been awarded at 2.3GHz for service next year, while in the USA the 2.5 - 2.6GHz band is owned by carriers planning on using it. The 5.8GHz frequency allocation is unlicensed or “lightly licensed” in many countries. In addition several operators in both Japan and the USA are working at 4.9GHz.

This ad hoc attitude is in stark contrast to the cellular philosophy, which invents a new description for each frequency shift that is applied to an established standard, for example cdma450 or PCS as distinct from GSM.

Licensed bands (for example, at 3.5GHz in many countries) allow operators to manage frequency planning, for unlicensed bands (e.g. 5.8GHz) different techniques are needed. While carrier sense multiple access is sufficient for WiFi, a much more rigorous radio access control mechanism is required for WiMAX, leading to increased complexity in the physical and MAC layers.

However, while the technology can support a wide variety, it is important to stress that the laws of physics are stricter. It is very hard to get mobile coverage to work at high frequencies, and obviously cells get smaller. As such, while 3.5GHz is fine for fixed coverage most mobile deployments are around 2GHz (2.3GHz in Korea, 2.5GHz in US).

Another distinction to mention is simply the environment. WiFi is intended for local coverage with relatively short distances and simple radio environments. In contrast, WiMAX can be used in a huge variety of ways, many with extremely long range and corresponding variety in channel conditions. Mobility, with its fast fades, further complicates things. This is why the PHY signal processing is so much more sophisticated in .16 than in its “little brother” .11, and the DSP algorithms are dramatically harder.

WiMAX profiles

The first version of the IEEE 802.16 standard released addressed line-of-sight environments using comparatively high frequency bands in the 10GHz to 66GHz range. The most recently published standard, 802.16-2004, describes 2GHz to 11GHz, allowing it to support non line-of-sight environments. Three completely new physical layers were added together with a number of modifications to the MAC, with knock-on effects on the digital processing needed. Further changes have been proposed to allow more efficient use of the radio spectrum at lower frequencies, for example 450MHz.

Although WiMAX was created to promote 802.16, it has deliberately defined a small subset of options and pre-defined profiles in order to simplify implementation. For example, WiMAX only supports one of three possible options in 16d: 256OFDM (which specifies Orthogonal Frequency Division Multiplexing with 256 tones) which is highly suited to non-line-of-sight environments, and excludes the single carrier and OFDMA2048 modes.

With 802.16-2004 published, attention has shifted to developing the 802.16e mobile standard, which opens up competition with 3G cellular networks (see sidebar). 16e extends this with a scalable OFDMA system, delivering further improvements at the expense of complexity, with a scalable FFT size proportional to channel. This standard will add further complex PHY-layer processing together with handoff signals to allow users in vehicles to switch from basestation to basestation seamlessly. Forward error correction in 16d used convolutional coding; in 16e that is optionally extended with a very powerful but very complex convolutional turbo code (Double-Binary Circular Recursive Systematic Convolutional Turbo (CTC). In WiBRO that is a mandatory option.

Uplink subchannelisation is an optional feature in 16d, and is generating a lot of interest from operators. This allows a subscriber station to concentrate its transmit power on a subset of the total OFDM subcarriers, leading to link budget improvements in the uplink which translate into coverage and capacity benefits. Multiple subscriber stations can be scheduled to transmit simultaneously on different subchannels. In essence it is a form of OFDMA, although not “the OFDMA” of the 16e standard.

WiMAX was designed from the start to support smart-antenna systems, including RX antenna diversity, TX antenna diversity, beam forming, Space Time Coding (STC) and ‘multiple input, multiple output’ (MIMO). These systems are becoming more affordable and their ability to suppress interference and increase system gain will see them introduced to WiMAX implementations in the near term. A WiMAX Forum White Paper [1] suggests that for the same circumstances (3.5MHz FDD, 3.5GHz band) adding subchannelisation, diversity and STC to a basestation could increase coverage from 2km range to 9km range – that is a twenty-fold increase in coverage, and potentially in subscribers. Consequently, most practical systems will choose to support these options, despite the complication they bring.

Mobile WiMAX does compete with several other mobile broadband technologies, including WCDMA, cdma2000 EVDO, TD-SCDMA, Flash-OFDM and others. Among the advantages that mobile WiMAX claims over these are as follows:

Superior airlink technology: Scalable OFDMA is a very modern, sophisticated modulation method that can reliably delivery high performance (bps/Hz) even in challenging multi-path environments.

Network efficiency: WiMAX is an inherently IP-based system, to create an open architecture for mobile data networks, significantly reducing complexity and cost.

Full QoS: WiMAX includes a sophisticated and versatile MAC layer, with extremely good support for management of QoS. This is especially important for multi-media and voice (VoIP) services.

Applications: Superior transparency to applications in WiMAX will encourage faster adoption of the service by enabling performance equivalent to and, in some cases, better than wireline access technologies. Mobility data with this level of performance promises to open up new applications as well.

While there is some truth in these, there is an element of caution. In reality, the laws of physics do impose some hard limits; better air-interfaces are an improvement but Shannon’s Law still holds.

But an area where Laws of Physics don’t have so much sway is financial, and here there are very good reasons why WiMAX has advantages. Fewer royalties, less expensive system architecture, increased competition from open markets all deliver lower costs. The possibility of higher performance yields higher average revenues for significantly better economics.

Mobility

The present WiMAX standard, 802.16-2004 (also known as 802.16d), is for fixed applications only: 16e adds support for mobility (although, it is also expected to be adopted for some fixed installations since it offers a better link budget than 16d).

Despite the name, there is no backwards compatibility between 16e and 16d. Another variant is the Korean standard WiBRO, which is essentially the first version of 16e, in much the same way as FOMA was for WCDMA. Licences for WiBRO have already been issued and the service is due to launch next Spring.

Developing the standard for 802.16e is the responsibility of 802.16 Task Group e within the IEEE, and is centred on amending the PHY and MAC layer specifications laid down in 802.16-2004 to provide for both fixed and mobile access. Things get a little complicated as there are various profiles within each version, but it is important to say that WiMAX based on 16e will not be backwards compatible with WiMAX based on 802.16-2004 (or 16d). The PHY profile in 16-2004 uses OFDM256, while that of 16e is based on Scalable Orthogonal Frequency Division Multiple Access (OFDMA) and there are also planned enhancements to the MAC to make it more suitable for mobility. WiBRO, the Korean standard is the only type of mobile WiMAX that is currently defined, offers a theoretical peak data rate of 18.4Mb/s in the downlink and 6.1Mb/s in the uplink from a 10MHz channel bandwidth, making it the fastest ‘mobile’ data standard apart from WLAN, which is essentially portable rather than mobile. 16e also offers significant benefits (improved link budget) in fixed applications, and while it is expected that the two incompatible WiMAX standards will exist side-by-side, eventually 16e will take over completely for both fixed and mobile use.

Like 802.16d, the 16e standard uses QPSK, 16QAM and 64QAM modulation, but with an OFDMA scheme with 128, 512 1024 or 2048 sub-carriers rather than the 256 sub-carrier OFDM used by 16d.The selectable channel bandwidths are the same, but the expected cell radius of 1.6 – 3.2km is only around half that of a 16d cell (to allow for mobility). 16e is also specified only for systems operating at below 6GHz, compared with up to 11GHz for 16d.

OFDMA is scalable, and allows the sharing of the allocated frequency band using sub-carriers, and improves the performance with varying channel bandwidths. Thus the power of the mobile device can be centred on the carriers that provide the best propagation characteristics in a given environment, and can adapt these as the environment changes for a mobile application. In a fixed scenario the sub-carrier allocations can be altered to a tighter distribution, which will improve coverage (although at the expense of tolerance to mobility that is not required).
When mobility is imposed upon a system, there will be many channel changes and distortion effects that will become evident, such as deep fades, sudden frequency shifts and Doppler effects. Some of these effects, for example Doppler, do impact on the PHY, while some affect the PHY-MAC interaction, for example more frequent updates to compensate for frequency changes. Adaptive modulation will improve performance significantly by increasing the data rate when propagation conditions are good and dropping the rate to improve reliability when conditions are poor.

These requirements mean that a new PHY design is required to ensure that there are enough pilots in the OFDM tones, and sufficient preambles and mid-ambles to guarantee that signalling is robust and that the timings are synchronised. OFDM256, as used for 16d, does not have sufficient design flexibility to allow this. Although it is possible to improve the underlying algorithms, mainly by updating more often to allow for Doppler, it is essentially lacking in the required degrees of freedom to allow this. Various expansions are being discussed - under the banner of ‘16d+’ - but these are essentially proprietary modifications. 16e adds more features by providing a more robust PHY properly designed to handle mobility.

Handoff

A major consideration is seamless handoff. Handoff has always been a challenge for mobile systems, but with the complexity of the OFDMA modulation scheme it becomes even more difficult. Mobile IP, with “slow” handoff, will be fine for web-browsing but not enough for decent voice capability. Many services require the appearance of seamless connections (VoIP, VPNs etc) – much of the complexity (and latency) in the cellular network is from maintaining these connections across cell boundaries.

Such flexibility will require a sophisticated management infrastructure. Cellular systems have been designed from the beginning to facilitate handover, but it is not easy to introduce it to an existing standard.

When the signal quality deteriorates close to the edge of the cell it is necessary to switch seamlessly to a new basestation. But this poses enormous problems, since the terminal needs to communicate with two or more basestations simultaneously, which means that each of them must be synchronised and with full communication in place so the mobile terminal can make measurements and decide the appropriate point at which to switch over primary control. The handover itself entails a complicated dialogue in order to disconnect one flow, start another, transition contexts, switch circuits, and so on. Although complex, this procedure needs to appear seamless. Just as in a voice call the human ear is sensitive enough to detect even small discontinuities, a data network such as a VPN should not be affected by the handover, even though the whole circuit has effectively changed, and could equally change again or change back to the original basestation.

In GSM handover is managed by the basestation controller (BSC), and in 3G by the RNC. In WiBRO the equivalent is known as the Access Control Router (ACR). This requires various gateway nodes that ensure IP addresses look the same to the outside world while managing the internal identity.

WiBRO will launch with mobility (i.e. support for fast moving users in a train) but with “simple” handover, where the session is passed but not totally transparently. True mobility will follow, perhaps in the 2007/08 timeframe.

This is being standardized by the WiMAX Forum (as it is “above” the air-interface and therefore outside the scope of 802.16). There is a strong debate still going on, with various architectures being proposed. Some have an all-new, all-IP flavour, such as WiBRO’s ACR. Others hope to leverage existing architectures, such as 3GPP or 3GPP2. One common thread seems to be to align around IMS for core network.

HSDPA and WiMAX: competitive or complementary

The debate between the merits of HSDPA and WIMAX promises to be this decade’s equivalent of the competition between GSM and CDMA in the 1990s.

HSDPA is already a real and demonstrable technology, and to a large extent delivers what WCDMA was initially intended to provide: a technology optimised for high speed, reliable, cost efficient data services.

WIMAX, in contrast, is not an incremental development, but a whole new family of standards, in fact a brand for the IEEE 802.16 family (as Wi-Fi is the brand for the different varieties of IEEE 802.11). The IEEE 802 community has a very different approach to their 3G colleagues: they deliberately re-use protocols very aggressively, and they emphasise openness and real interoperability. As a result matters have progressed more quickly than in the 3G community, with far greater competition and a pace of innovation.

Many in the industry, including picoChip, believe that the potential for fixed wireless broadband is greatest in developing countries, because of the lower availability of fixed phone lines. Without a wired connection, people would not get access to the Internet or other data services, broadband or otherwise. Using wireless instead, two billion people get voice services using GSM – and they will likely get broadband with WiMAX (and then voice using VoIP over it).
There are many examples of how the two technologies – HSDPA and WiMAX – are entwined. For instance, WiMAX is an ideal technology for backhaul applications because it eliminates expensive leased line or fibre alternatives. An HSDPA picocell with wireless back-connection would be very cheap and very easy to deploy, and could offer voice as well as high-speed, high-quality data, inside a corporate office or at a super hot spot. A second example is the provision of seamless service, always using whichever is the better connected technology. For example, a passenger on a train waiting in a station would connect using WiMAX, and when the train starts moving into the countryside, there would be a smooth transition between networks to HSDPA for full high-speed mobility with handoff using the cellular network.

Nevertheless it would be foolish to pretend that HSDPA and WiMAX will not also be direct competitors. The fact is that if a customer sends a megabyte of data on WiMAX, that megabyte will not be sent on HSDPA, and if a WiMAX service offers cheap flat-rate voice over IP (VoIP) services as well there will be a huge impact on both revenues and margins for the 3G operators.
UMTS does have one huge advantage in that it is already deployed. There are operators, manufacturers, regulators and most importantly there is a customer base, with a huge pool of 2G users who will migrate in the future. Most importantly of all, there are major operators with service, with cell sites already installed and operational, while WiMAX still has the difficult process of site acquisition, planning and deployment ahead of it.

But the harsh view is that if 3G had delivered what it promised, there would be no need for WiMAX. Two years ago, the competition offered by WiFi was hypothetical, but today many business data users routinely use hot spots in preference to UMTS. WiMAX is also likely to launch astonishingly quickly, on datacom timescales rather than traditional telecom timescales. This is a race that UMTS could lose unless something changes, because it is failing to move sufficiently quickly, and it is failing to implement genuine open standards, rather than quasi-proprietary systems like Iub.

It should be stressed that this challenge applies primarily to manufacturers. Operators have a huge vested presence in customer base, brand, cell-sites and backhaul, and are well-positioned regardless of which acronym succeeds. Although there may be new competitors, the barriers to entry remain dauntingly high. But for OEMs WiMAX may be a disruptive change that allows new suppliers into the carriers, potentially displacing the traditional partners.

The conclusion is that, although the two technologies can support each other, the WCDMA development community needs to seriously its approach and its priorities, and has much to learn from other communities whether competitive or complementary.

Conclusion

Because there are so many options that all fall under the WiMAX umbrella, it will clearly be a much more complex standard to support than WiFi. Flexibility will be the key to successful adoption, as deployment challenges in the field will lead to changes being demanded. This will be a prime consideration when it comes to selecting an architecture that supports the planned, and possibly as yet unplanned, evolution of WiMAX from its position today, when standardisation is not yet complete,.