Second Generation, Multi-Mode Wireless LAN Clients

White Paper

 

By Yoram Solomon
Director, Strategic Business Development
Wireless Networking Business Unit
Broadband Communications Group
Texas Instruments

 

Abstract

In this article, I’m addressing the future of wireless LAN connectivity, referring to it as the Second-Generation Wireless LAN.  The four major issues that need to be addressed by wireless LAN product manufacturers are: (1) coexistence of physical layer standards such as 802.11a, 802.11b, and 802.11g, (2) security, quality of service and other “maturing” factors, (3) interference immunity and (4) the user experience. 

For every item, I am describing the current “state of the industry,” as well as what will need to take place in order to have mature products ready for mass deployment.  This article starts and concludes with a vision of the proliferation and mass-adoption of wireless LAN through the deployment of the second-generation wireless LAN products beyond today’s “early adopter” market.

Foreword

In the first quarter of 2001, the PC market continued its slowdown, as did the entire economy; however, the revenue from wireless LAN products was reported to grow 15 percent quarter over quarter.  Has this market matured?

Today’s wireless LAN market is in its early adopter stage.  Today’s products are not as simple to install, and not as seamless to operate.  The users of wireless LAN products are typically very technical and can overcome the hurdles associated with the installation and use of wireless LAN.  However, the benefits of using wireless LAN are tremendous.  Personally, I use only wireless LAN at the office, roaming from office to conference room. At home, I get connected to the Internet through my cable modem and wireless residential gateway, using my computer anywhere in the house.  When I travel, I use my MobileStar subscription to connect to the Internet at booths located near almost every American Airlines’ gate, and make the most out of the downtime before my flight.  Oh, and by the way, I also use it in the Starbucks near my house while I sip my “grande coffee Frappuccino®.”

There is no doubt that wireless LAN is going to be adopted.  However, there are several trends that must be observed and acted upon in order to deliver the right products for mass adoption.  This article covers the four most important trends and issues around wireless networking.  Addressing those correctly will promote wireless LAN adoption beyond our wildest dreams (did anyone ever believe that there would be 400 million cellular telephones sold in one year?), introducing the second-generation wireless LAN.

Frequency bands, Modulations, 802.11g, and the PHY Layer

Wireless LAN products can use several frequency bands allocated by the Federal Communications Commission (FCC) and other regulatory agencies worldwide.  The most common frequency bands are the Industrial-Scientific-Medical (ISM) frequencies of 902-928MHz and 2.4-2.4835GHz, and the Unlicensed National Information structure (U-NII) frequencies of 5.15-5.25GHz, 5.25-5.35GHz, and 5.725-5.825GHz. 

Due to the lower bandwidth available in the 900MHz band (only 26MHz), the achievable data rates in that band are relatively lower, and with the increased use of cordless telephones in that band, it has never been widely used for high-data rate wireless networking. 

The 2.4GHz band was the first one adopted widely for wireless networking, mainly by the 802.11b standard (also known as Wi-Fi), by the wireless home networking standard (HomeRF), and by the wireless personal area networking standard (Bluetooth), known as PAN.  One of the leading reasons for such a wide proliferation of products and standards in this band is the availability and feasibility of silicon operating components.  The resulting cost reductions, mostly through moving to low-cost CMOS processes, increased the penetration and deployment of such products. 

The 5.4GHz band is not yet used for such products, as the availability and feasibility of silicon components in these frequencies are still being evaluated.  The IEEE created a standard known as 802.11a that uses the 5.4GHz band for high-speed networking, and several companies are developing components to support this standard and offer data rates higher than Wi-Fi in this band.  It is expected that the first products will hit the market toward the end of 2001, followed soon by price reductions, causing this standard to then be widely adopted. 

While the 802.11b standard offers a data rate of 11 Mbps at the 2.4GHz frequency band, the 802.11a standard offers a data rate of 54 Mbps at the 5.4GHz band.  Much like the migration from 10Mbps to 100Mbps (and then to Gigabit Ethernet and beyond) for wired LAN, there is a perceived need for higher data rates in the wireless LAN arena, beyond the 11 Mbps offered through Wi-Fi products.  Texas Instruments (by acquiring Alantro) has offered a new modulation scheme called PBCC (Packet Binary Convolutional Coding), capable of delivering data rates of up to 22Mbps, still using the 2.4GHz band.  This technology was proposed to the IEEE for a working group called 802.11g.  Shortly thereafter, Intersil offered a counter-proposal using OFDM (Orthogonal Frequency Division Multiplexing), capable of delivering 36 Mbps in the same frequency band.  Initially, both proposals violated the FCC rules (15.247) for the use of that frequency band.  However, the new FCC proposed ruling (Docket 99-231, discussed later in this paper) suggests modifying the rules such that neither proposal will be in violation.  Meanwhile, in the IEEE 802.11g working group, the TI proposal was voted off, but the OFDM proposal still needs to get the support of 75 percent of the members to become the 802.11g standard. 

What will the impact of 802.11g have on the use of 802.11b and 802.11a?  The effect on the current 802.11b (Wi-Fi) standard is pretty predictable.  Once a higher-data rate standard gets approved, the existing 802.11b will simply become a fallback data rate to the higher data rate standard.  This is similar to the original 802.11 (supporting 1 and 2 Mbps) that became the fallback data rate to the 802.11b (driving 11 Mbps or 5.5 Mbps).  The main question is, what will the impact of a higher data rate 2.4GHz standard (802.11g) be on the higher frequency, higher data rate 802.11a standard?  I will assume two scenarios, depending on the first standard to hit the market with products.

Scenario 1 – 802.11g Products Available before 802.11a Products

With the OFDM modulation for 802.11g being able to deliver as high as 36 Mbps, the main question is whether 802.11a will still be needed.  I believe that there will still be a need for 802.11a for several reasons.  One reason is the allocation of additional bandwidth that provides a higher aggregate data rate usable for wireless LAN.  If 802.11g is used within a certain physical area, then only 36 Mbps becomes available in this area.  If, however, both 802.11g and 802.11a services (or coverage) are offered in an area, then the aggregate data throughput offered at this area increases to 90 Mbps, and therefore can support more users.  Furthermore, the 2.4GHz band sees more interference today from non-wireless LAN products, such as Bluetooth devices and cordless telephones.  Adding the 5.4GHz band for wireless LAN applications will help in avoiding some of that interference.  The conclusion from this scenario is that the introduction of 802.11g will not delay or prevent the later deployment of the 802.11a-based products.

Scenario 2 – 802.11g Products Available after 802.11a Products

As in scenario 1, the introduction of 802.11a products before 802.11g products will not prevent 802.11g products from being deployed.  While the 5.4GHz band is limited to 54 Mbps, having 802.11g deployed will increase the aggregate data throughput by 36 Mbps to the total of 90 Mbps.  Furthermore, with the 802.11a products suffering from a shorter range, there may be somewhat different applications for both standards.  In any event, the conclusion remains the same: the deployment of 802.11a products before that of 802.11g products will not delay or prevent the later.

Given these conclusions, it seems that there is opportunity for the 5.4GHz and 2.4GHz standards to coexist.

Multi-Mode Access Points

The next question that comes to mind is, what will the migration path be for deploying the high data rate standards? 

One possibility is that all existing (installed) access points will be upgraded to either the high data rate standard, or become multi-mode.  Most of the installed access points are not scalable or upgradeable to the new standards or do not support multiple standards.  Therefore, the scope of this proposition is that the access points will be replaced in their entirety.  This is a very expensive proposition, and one that will probably never happen, much like the old analog cellular infrastructure was not replaced once PCS and CDMA were introduced.  I believe that we will continue to see multiple 802.11b access points, along with new multi-mode access points. 

Another question is whether there will be new access points that only support the higher data rate standards, but not the existing 802.11b standard.  With the current rate of 802.11b client product deployment, it is estimated that by mid-2002 (when we expect to see some 802.11a or 802.11g products available), there will be some 20 million 802.11b-only client adapters installed in computers and other devices.  Installing access points that do not support this 802.11b standard means that 20 million users will not be able to access them, and that is unimaginable.

Multi-Mode Client Devices

Now comes the question of whether the new clients are going to support only a high-data rate standard.  Assuming that not all of the currently installed (802.11b) access points will be replaced by multi-standard or high-data rate standards, due to the required investment in infrastructure, there will still be many 802.11b-only access points installed.    Therefore, installing wireless network cards in computers that only support the high-data rate standards would disable these computers from accessing the existing 802.11b access points.  Wireless Internet Service Providers (W-ISP), such as MobileStar, WayPort, AirWave and others, have already installed access points in public places such as airports, hotels and even Starbucks’ locations.  They probably will not replace the existing 802.11b-based infrastructure, and users wanting to access the Internet or their corporate Intranets in these locations will have to have equipment supporting the 802.11b standards.  MobileStar, for example, offers its users both 802.11b cards and FH cards to support the older infrastructure for the lower data rate FH standard.  However, the problem with that is much less acute than it might be if they had to provide all of their users with both a high-data rate card (supporting 802.11a and/or 802.11g) and an 802.11b card. 

As new computers and other non-PC devices, such as PDAs are becoming smaller, the amount of space allocated for extension accessories (such as wireless LAN cards) becomes more limited.  Installing two cards (one to support 802.11b and one to support 802.11a) will be nearly impossible. 

Moreover, wireless LAN is becoming an integral part of new PCs, as PC manufacturers such as Dell, Compaq, Toshiba, IBM and others are shipping new computers with bundled wireless LAN cards in a mini-PCI form factor. While PC card (PCMCIA) accessories can be replaced during operation, the embedded mini-PCI cards can’t. 

Therefore, the best solution for multi-standard wireless LAN client support is multi-standard wireless LAN cards, supporting more than one standard on the same card.  These cards should support 802.11b, 802.11g, and 802.11a due to their coexistence in the marketplace.

Service Discovery, and 802.11b as the Negotiation Standard

With the coexisting environment described above, products will need to implement “intelligent” service discovery procedures, allowing them to identify the available networks, and to select the best one to use. 

We must assume that beginning in 2002, there will be a mix of clients, some of which will have an 802.11b-only access card, and some of which will have high-end multi-standard cards capable of higher data rates.  There must be a common way to communicate with them.  I propose that this will be accomplished through the use of 802.11b standard at its basic form.  Since all client devices will be able to communicate through 802.11b, it can be used as the initial link between the client and establish the access point.  Then, still using the 802.11b standard, the client and/or the access point will attempt to negotiate a higher level of communication, either using a higher data rate standard (such as 802.11a or 802.11g) or enabling higher level of security or quality of service, or other enhanced features.  If both ends (the access point and the client) can support a higher data rate or enhanced functionality, they will negotiate their link up to the highest common ground.

Security, Quality of Service, and the Upgradeable MAC Layer

Security

In February 2001, the Internet, Security, Applications, Authentication and Cryptography (ISAAC) research group at the University of California, Berkeley, published an article attacking the level of security provided by the Wired Equivalence Privacy (WEP) protocol supported by the Wi-Fi standard.  Following this publication, the public perception of the security offered by wireless networking was that these networks are not to be trusted.  A not-very-well-known fact is that most equipment vendors and wireless ISPs recommended that WEP be disabled altogether “to simplify the installation and operation” of such equipment. 

 

The IEEE formed a working group named 802.11i to address the security needs of the wireless LAN user community.  Two future developments are certain: some action will be taken, and it will render the existing wireless LAN cards obsolete.  These changes will most likely affect the Medium Access Control layer (MAC), which is implemented within the chipset that constitutes the core of the wireless LAN card.  When a new security standard for wireless networking emerges, new chipsets will be developed, installed on new cards and sold to end users.

Quality of Service

The basic 802.11b standard has one disadvantage when compared to HomeRF and Bluetooth -- its inability to guarantee quality of service.  This disadvantage translates into an inability to deliver non-data services (such as voice and video) in a reliable, high-quality manner.  Once again, the IEEE addressed the issue by forming a working group called 802.11e that proposed a QoS standard be implemented in the MAC layer.  If approved, the next generation wireless LAN products will support multi-media applications with a high level of quality.  Once again, when QoS is implemented, it will render the existing wireless LAN products obsolete.  They will be incapable of supporting the new standards and will force the end users to replace their hardware.

The Upgradeable MAC

The introduction of new standards that address quality of service, security and other issues will automatically make existing products obsolete.  End users are typically reluctant to purchase products that might become obsolete and prefer to wait until the next generation is available.  The way to reduce that risk is to offer a field-upgradeable MAC layer that is fully programmable in real time.  When a new standard is released, the MAC can be upgraded by simply downloading a new driver. 

This might appear to be a simple task, but the key to its successful execution is to correctly forecast the future standards that will be released, and to incorporate a powerful enough programmable platform that will be capable of supporting these standards.  The tradeoff can be in terms of flexibility versus cost, and flexibility versus power consumption.  A higher level of flexibility can be achieved when functions are implemented in software rather than in hardware; however, this typically involves a higher-cost platform (multi-processor), and consumes more power due to the high-speed processors deployed.  Moore’s Law and the overall progress in semiconductor development will compensate for these tradeoffs in the long run.  The correct trend still needs to be a shift from hardware implementation to software implementation, offering the user the programmability and upgradeability that will reduce the risk of purchasing products before they fully mature.

FCC, IEEE 802.15.2 and Interference

On May 10, 2001, the FCC proposed changing the rules for the unlicensed 2.4GHz ISM band.  The commission proposed “to reduce the amount of spectrum that must be used for frequency hopping spread spectrum systems operating in the 2.4 GHz band (2400-2483.5 MHz), and to eliminate the processing gain requirement for direct sequence spread spectrum systems.”  This new proposed rule is known as “Docket 99-231.”  This ruling will allow both techniques proposed to the IEEE 802.11g committee (Texas Instruments’ PBCC modulation and Intersil’s OFDM modulation) to comply with the new FCC rules.  This ruling will also allow Frequency Hopping systems (such as Bluetooth and HomeRF) to avoid frequencies used by a Direct Sequence 802.11b (or 802.11g) system, not having to hop throughout the entire frequency band.

While this ruling would seem to promote the adoption and deployment of wireless networking in this frequency band (through higher data rates and better mutual avoidance of the existing standards), it also allows for greater interference in this frequency band. 

We must remember that this unlicensed band is open to use by other types of devices, such as cordless telephones.  The FCC previously regulated all the products using this band so that they had less interference with one another by spreading the signal over a wider band (whether through higher Direct Sequence processing gain or through a higher number of frequency hops in a Frequency Hopping system).  With the proposed ruling, a device now operating in this band might generate a more “concentrated” interference with another device.  Even before the ruling, devices operating in the 2.4GHz band included cordless telephones, microwave ovens, and others.  Now, with the new ruling, the expected interference from cordless telephones will increase. 

The IEEE has addressed the mutual interference between Wi-Fi (802.11b) devices and Bluetooth devices through a newly formed working group, named 802.15.2.  This working group, however, is proposing to solve only the mutual interference created by Bluetooth and Wi-Fi products and does not address any interference to or from other devices using the same band.

Today, the deployment of products in the 2.4GHz band is relatively low.  However, we are expecting wireless LAN (mainly Wi-Fi), Personal Area Networking (mainly Bluetooth), and even cordless telephones operating in this band (and thus offering longer range and multiple handsets) to proliferate.  Once that happens, the level of man-made interference in this band will increase significantly.  Devices incapable of avoiding such interference will suffer degradation of quality through lower throughput (or completely loose connection), range decrease and noise.  On the other hand, devices with intelligent interference-avoidance capabilities will outperform their competitors, offering superior range, throughput and overall quality.

The User Experience

Microsoft has announced that the Windows® XP operating system, to be launched in October 2001, will support wireless networking in a way that no other previous operating system has done before.  This support addresses zero configuration, security, authentication, roaming and service discovery.  At the Wireless Ethernet Compatibility Alliance (WECA, the promoter of the Wi-Fi log and certification) annual member meeting in Helsinki in June 2001, Microsoft even promoted the idea of multi-mode wireless LAN client radios to simplify the user experience.

Where are we today?  The installation of an 802.11b product is somewhat complicated.  A one-hour installation can easily turn into a 6-hour installation.  Roaming between access points is not trivial, and extending this roaming to cover residential gateways and public “hot spots” (such as hotels, airports, and even Starbucks’ locations) requires some “unnatural acts.”  Having to change configuration, release the current IP address, renew the IP address, and other tasks that need to be performed, reduce the pleasure of experiencing a wireless LAN.  To compare this experience to that of a cellular telephone, imagine what it would be like if, when you roam between cell sites (every 1-2 miles), you had to change your cell-phone configuration, release its telephone number, renew its telephone number, and do all of that so you could still accept phone calls.  Unimaginable?

From this comparison, you can understand that in order for wireless LAN to be as widely adopted as cellular telephones, their operation must be as seamless as that of cellular telephones.  While Microsoft® XP might offer a solution, the solution should be supported with existing operating systems that are still bundled with new computers.

Conclusion

In this article I described second-generation wireless LAN clients.  The coexistence of the different frequency bands used for wireless LAN will lead to the adoption of multi-standard (or multi-mode) products.  The proliferation of protocols and functions (such as QoS and Security) in an immature standard will lead to the development of software-programmable products that can be upgraded as the standard matures.  The ability to avoid the growing in-band interference will differentiate good products from bad ones and give the manufacturers of the better products a competitive advantage.  Last, but not least – the user experience will have to be as close as possible to that of a cellular telephone, with seamless operation that can leverage on the pervasiveness of wireless LAN into our lives.