A Field Guide to Wireless LANs for Administrators and Power Users

Clearly, to make a WLAN, one needs at least two devices to make an ad hoc network, and at least three devices to make an infrastructure network (an infrastructure network with one STA and one AP would be possible, but the STA would not have any other STAs with which to communicate…). As a practical requirement, one also needs a wired network (the Distribution System in IEEE 802.11 parlance) to which the AP may be attached.

APs are available at a variety of price points. Entry-level IEEE 802.11b APs for home use might be available (refurbished) for less than $40. On the opposite end of the price spectrum, APs designed for corporate deployment will cost up to $750 or more. The latter type of AP distinguishes itself by supporting features like manageability, upgradeability (e.g., the ability to add IEEE 802.11a and/or IEEE 802.11g support to an IEEE 802.11b AP), as well as robust support for the full range of features in the standard, even those that may not be required for the necessary testing to acquire Wi-Fi certification. The more expensive APs also may have higher quality components, such as their RF components (power amplifiers and such), and antennas.

In many current laptops, IEEE 802.11b is almost an afterthought. (Would you like that new laptop with or without a WLAN? Well, duh…with, of course!) In fact, the incremental retail cost of WLAN STA capabilities in a laptop is rapidly approaching $0; in other words, it is becoming a required component, like a CPU or a keyboard. Another trend is that by the end of 2004, it is likely that IEEE 802.11g-2003 support will be at least as prevalent as IEEE 802.11b is today.

The hardware necessary to implement a STA is generally divided into a chipset that implements the MAC sublayer protocols, and a chipset that implements the PHY, including the modulation schemes (in the BBP), and the RF subunit, which actually transmits the modulated digital data onto the WM. IEEE 802.11b implementations are available as both PCI^[11] cards for desktop PCs and as PC Cards (formerly known as PCMCIA^[12] cards). Wireless PC Cards can be used in certain PDAs (those that have PC Card slots) to make the device into a WLAN STA. Other nontraditional devices, such as printers, digital cameras, and so forth, are appearing with integrated WLAN capabilities.

^[11] PCI stands for "Peripheral Component Interconnect."

^[12] PCMCIA stands for "Personal Computer Memory Card International Association." This mouthful was replaced by the much shorter "PC Card," which doesn't appear to be an acronym. The joke used to be that PCMCIA stood for "People Can't Memorize Computer Industry Acronyms." The original PCMCIA standard provided an interface bus that was 16 bits wide and ran at 8 MHz (maximum throughput, 128 Mbps, or 16 megabytes per second (MBps)). The newer CardBus interface runs at 33 MHz, and is 32 bits wide, for approximately eight times the throughput (132 MBps, or 1,056 Mbps; i.e., just over a gigabit per second).

Integration of WLAN components has progressed in an orderly fashion so far, tending to follow similar integration steps that wired LAN components have already taken. Initial WLAN cards were isolated products that could be easily added to any laptop (or desktop…WLAN cards for desktops do exist). This add-in card product category is not likely to go away any time soon, especially since the rapid pace of change of WLAN standards means that whatever built-in hardware a laptop has might become obsolete.

The early days of Ethernet were dominated by the add-in card business. For cost saving reasons, and higher reliability, the second phase of WLAN integration was reminiscent of the evolution of Ethernet progressed from a NIC-oriented (i.e., add-in card) to so-called "LAN-on-motherboard" (LOM) implementations. From a customer's perspective, Ethernet had been integrated onto the motherboard, even though it was still a set of discrete components. Eventually, Ethernet controllers were further integrated into the core logic of the motherboard, although they may or may not still be logically a PCI device. The adoption of progressively faster Ethernet speeds tended to be NIC-driven, meaning that 10/100 desktops were commonplace before the network infrastructure was upgraded to support 100 Mbps to the desktop. The evolution of gigabit Ethernet is proceeding in the same way, with 10/100/1000 NICs being readily available today, well in advance of widespread deployment of gigabit-to-the-desktop.

The integration of WLAN components into PCs (i.e., laptops) has progressed in a similar fashion. In laptops, WLAN support has now reached the first level of integration, which is not to say that there is no need for the add-in cards…any laptop that was built prior to the newer WLAN-on-motherboard (to coin a term) designs is a potential customer for an add-in card. The add-in card manufacturers will be able to leverage the cost reductions in the components that the higher volume WLAN-on-motherboard chipsets will create, and the add-in card designs are likely evolve to ultimately use the same components or some closely related derivative components.

The most modern motherboard designs have the Ethernet hardware logic built in to the PCI South Bridge chip. In the early days of PCI, the South Bridge chip was simple and basically just provided an interface through which the add-in cards on the PCI bus could access the memory and CPU in the PC via the PCI North Bridge chip, which interfaces to the CPU and acts as a memory controller. As an aside, it is typically the case that the graphics controller interfaces to the North Bridge chip via the AGP^[13] interface, or it may even be integrated into the North Bridge chip in some designs.

^[13] AGP stands for "Accelerated Graphics Port."

One trend in current PC design is to have the South Bridge chip be a package that includes Ethernet MAC logic, USB^[14] and FireWire interfaces, and any other type of logic that is involved in talking to a peripheral device (even storage interface protocols like SCSI^[15] and IDE^[16] could be integrated into the South Bridge itself, thereby eliminating the need for standalone disk controllers). Figure 8-1 shows the evolution of the PC motherboard, in terms of block diagrams that identify the major components.^[17]

^[14] USB stands for "Universal Serial Bus."

^[15] SCSI stands for "Small Computer System Interface."

^[16] IDE stands for "Integrated Drive Electronics."

^[17] In the diagram, the following abbreviations are used: Cntlr (Controller), CPU (Central Processing Unit), DC (Disk Controller), E (Ethernet), FSB (Front-Side Bus), LOM (LAN on Motherboard), NB (North Bridge), NIC (Network Interface Card), SB (South Bridge), SC (Sound Card), and W (Wireless).

The evolution of WLAN designs can be expected to follow the same course as Ethernet did, for much the same reasons. Newer PCs (especially laptops) will probably have integrated WLAN capabilities embedded in their core logic (i.e., in their PCI South Bridge chip, which already houses integrated Ethernet and other interfaces; e.g., USB, FireWire, etc.). This level of integration is desirable for PC motherboard OEMs,^[18] since their design and production costs are lower than if they have to accumulate all of those functions via separate PCI cards. Ultimately, these cost savings translate into some combination of increased profits and lower prices at retail. There are really two major metrics that are used to compare laptops…battery life and weight. In laptops, it is conceivable that the integrated solution might be thriftier when it comes to using the battery, but that's not necessarily true. The weight shouldn't be that different either way (integrated or not).

^[18] OEM stands for "Original Equipment Manufacturer."

Besides the degree to which WLAN hardware could be integrated with PC motherboards, the actual implementation of the WLAN product has evolved considerably. First- and second-generation WLAN products tended to implement the MAC sublayer protocol in a small CPU, with the BBP and RF subunits being external to that CPU, connected by proprietary interfaces (these interfaces did not need to be exposed to third parties because, generally, these chipsets were sold as a package from the beginning, the MAC and BBP were trending toward integration with each other so there would be some ability for product designers to choose their own radio, much as they can choose their own Ethernet PHY today).

Newer designs are integrating the MAC (which is sometimes still implemented as firmware running on a CPU core, with that CPU core embedded in the South Bridge package, or perhaps implemented as purpose-built logic circuitry in the South Bridge) with the BBP inside the same package (the BBP has always been purpose-built logic). Thus, the MAC alone or MAC and BBP can be physically integrated in the South Bridge chip.

Figure 8-2 shows some ways in which the various WLAN components could be combined with each other. The fully integrated solution is likely to be useful for add-in cards (where minimizing the cost is especially critical), but "complete integration" may not be desirable for motherboard applications. The RF subunit is likely to remain separate for motherboard applications, since it is desirable to keep the high-frequency analog signals as close as possible to the antenna.^[19] The trend for the MAC and BBP components is clearly toward integration within the South Bridge chip.

^[19] These signals are close to the clock frequencies of modern PCs, so the issue of interference is real. The analog signals probably wouldn't interfere too badly with the CPU clock, but the CPU clock's square wave harmonics could make a real mess of the WLAN RF signal.

Over the next several years, it is highly likely that most laptops will ship with WLAN interface logic, in much the same way that Ethernet has become ubiquitous in laptops (and everywhere else).