More options are available to developers of 802.11 wireless-local-area-networking (WLAN) products than ever before. To support a growing number of coding schemes and data rates, industry-standards bodies continue to establish new specifications. Each implementation requires a fairly well-established set of semiconductor devices including a baseband and medium-access-control (MAC) processor. Several other integrated circuits (ICs) also are required, mostly from third-party vendors.
As the industry explores new technology standards, the architecture is steadily evolving. This trend is especially evident at the MAC level. Here, the advent of multiplexing approaches and algorithms has the potential to improve both voice and multimedia performance.
The IEEE 802.11 specification for WLAN defines both MAC and physical-layer (PHY) protocols. Its base standard and extensions specify a number of options. Only a few of them are currently implemented in consumer and enterprise WLAN products. Most of the devices found in today's market operate under the IEEE 802.11a, 802.11b, and 802.11g physical layers. Each of these PHYs has its own proprietary variants. Support mainly falls to the distributed-coordination-function (DCF) MAC option.
In the 2.4-GHz ISM band, network interface cards (NICs) operate complementary code keying (CCK) up to 11 Mbps. They operate packet binary convolutional coding (PBCC) up to 33 Mbps and orthogonal frequency division multiplexing (OFDM) up to 54 Mbps. In the 5-GHz U-NII band, only Orthogonal Frequency Division Multiplexing (OFDM) is found with data rates up to 108 Mbps (and beyond).
As illustrated in Table 1, six WLAN NIC frequency allocations currently exist worldwide. This list includes two allocations each in the United States/Canada, Europe, and Japan. For each frequency allocation, government regulations stipulate the number of channels, non-overlapping channels, and power restrictions that must be supported. All of the aforementioned data rates are maximum bit rates as opposed to the actual average data throughput. The NIC will automatically drop to a lower rate when communications cannot be achieved at a specific rate. This problem arises because of impaired channel conditions.
Essentially, the standards-based 802.11b (2.4-GHz) CCK/PBCC physical layers are highly coded forms of binary-phase shift-keying (BPSK) and quadrature-phase shift-keying (QPSK) modulation. At the minimum, the BPSK scheme is used for the preamble of all packets. PBCC and CCK techniques are utilized to achieve a statistical advantage with this modulated energy in the presence of noise or "processing gain." To accomplish this goal, more symbols are sent than are required to represent the actual bit information.
IEEE 802.11b-compliant radios support four data rates: 1, 2, 5.5, and 11 Mbps. The three higher rates are coded into an 11-MegaSymbols-per-second (MSymbols/s) QPSK waveform. The 1-Mbps rate is coded into 11 MSymbols/s BPSK. A proprietary 22-Mbps PBCC mode also is included in some devices, although it isn't supported by the 802.11b specification.
The 5-GHz 802.11a specification uses OFDM, while the 2.4-GHz 802.11g specification offers it as one of the proposed options. In the latter case, however, OFDM is expected to be the main mode of operation. This standards-based OFDM is comprised of 52 independently modulated carriers. Four of them are BPSK "pilot," or synchronization, carriers. Depending on the data rate, the remaining 48 independent carriers are modulated as BPSK, QPSK, 16-quadrature amplitude modulation (16-QAM), or 64-QAM.
The supported data rates are 6, 9, 12, 18, 24, 36, 48, and 54 Mbps. Even though they're rarely omitted, the 36-, 48-, and 54-Mbps data rates are "optional." As with CCK and PBCC, the waveform is coded. Processing gain is thereby achieved over the gain of the actual information that's being conveyed. Additional proprietary modes of up to 108 Mbps (and beyond) also are available. But they require all stations to utilize similar hardware.
Table 2 provides a summary of the different modes. Today's PHY options span a broad range of data rates, modulation schemes, preamble/header formats, and additional proprietary modes. Maximum performance is available through 802.11a in the 5-GHz frequency band. Currently, 802.11a supports proprietary modes with data rates up to 108 Mbps.
802.11 WLAN NICs are built in several different form factors. The most common forms of client adapters include Peripheral Component Interconnect (PCI); PC card (Cardbus and/or Personal Computer Memory Card International Association or PCMCIA); Mini-PCI; Universal Serial Bus (USB)/USB dongle; and Compact Flash (CF+). Although some configurations do involve direct communications between NICs (ad-hoc mode), most users communicate via an access point or wireless router. A wireless router provides an access point and additional functionality.
Table 3 compares the various form factors and their associated power requirements. Depending upon the implementation, the access point and the wireless router can each vary in size and power requirements. PC cards (PCMCIA/Cardbus) have the widest power-efficiency margins. Compact Flash devices have some of the most stringent power and size restrictions. Yet today's solutions are capable of supporting both of them.
Regardless of form factor, WLAN NICs are generally built around chip sets. These chip sets may or may not be provided by a single vendor. In the current trend, an entire direct-conversion (or zero-IF) NIC consists of only a few integrated circuits (see figure). Among the main IC components that are found in most WLAN adapters are the baseband/MAC processor, radio chip, and power amplifier (PA). Additional ICs include RF switches, a serial EEPROM, a voltage regulator, and a voltage-controlled oscillator (VCO) that's external to the radio chip.
On top of these components, 5-GHz NICs often contain a receiver low-noise amplifier (LNA). In addition, the NICs require at least one front-end filter that pre-selects the RF energy before it enters the receiver and filters transmit energy. A frequency reference is needed as well. Usually, this reference is a temperature-compensated crystal oscillator, TCXO module, or crystal. Lastly, 5-GHz NICs demand dozens of passive components. They include, for example, capacitors, inductors, and resistors.
The baseband/MAC processor is considered the nerve center of the NIC. It communicates with the host computer while controlling and adjusting the external radio circuitry. This processor combines the MAC and baseband-signal-processing functions through the combination of a microprocessor core (such as ARM) and custom logic. Contained within this type of chip are analog-to-digital converters (ADCs) and digital-to-analog converters (DACs). These converters communicate with the radio chip and power amplifiers. Here, the interface is usually analog.
Ultimately, the MAC portion of the baseband/MAC processor controls the baseband processor's functionality. Regardless of the PHY, the MAC is nearly identical between standards. It decides when the NIC should be transmitting, receiving, or idle. The MAC also controls low-level hardware connections to the access point or other NICs (depending on the mode setting).
Under the prevalent mode, which is called the distributed coordination function (DCF), each MAC is responsible for checking to see if the channel is busy before transmitting. If the channel is busy, the MAC must follow a specific procedure as to when it can try again. This scheme is referred to as Carrier Sense Multiple Access with Collision Avoidance (CSMA/CA). Alternatives to this type of contention-based MAC, such as the draft-802.11e MAC, include scheduled Time Division Multiple Access (TDMA) and the use of polling-type algorithms. Such algorithms offer improved voice and multimedia performance.
The MAC's additional functions include encryption and the possibility of power control. Encryption is implemented under the Wireless Equivalent Privacy (WEP) protocol. The key is entered either manually or via some automatic key-distribution scheme. The 802.11i draft standard addresses this issue. In contrast, the 802.11h draft standardizes power control. The intent of power control is to maintain maximum network efficiency. In Europe, the 802.11h specification is a requirement for 802.11a. It works by ensuring that each radio radiates the minimum necessary power for its particular link conditions.
The radio chip also integrates several functions. It contains a synthesizer, filters, amplifiers, and mixers. It houses very little logic. In fact, it's almost completely slaved by the baseband/MAC. The radio must tune to the intended frequency. Its internal synthesizer controls a voltage-controlled oscillator (VCO). In turn, that VCO's output relates to the desired transmit/receive frequency.
In transmit mode, the radio chip converts the complex baseband waveform from the baseband processor to modulated RF. While in receive mode, it converts the desired modulated RF signal (which has a power level that could vary by more than 70 dB) to a complex baseband waveform. The baseband processor then decodes that waveform. To guarantee a recoverable product, the baseband processor controls the gain of the receiver as well as the front-end gain (the LNA).
Usually, the modulated signals that come directly from the radio chip are insufficient to maintain the desired 802.11 link capabilities. In such a case, a power amplifier may be used to increase the signal level. Requirements mandate that this PA consume as little current as possible while minimizing distortion. At the same time, it must maximize radiated power. Often, a control algorithm that's performed by the baseband processor can be used to maximize the PA's effectiveness. It also could precisely control the NIC's output power.
Finally, radio-frequency switches select the path of the RF signals. They connect either the PA or the radio chip's receiver to one of the two antennas. These antennas are spaced so that one is likely to receive a stronger signal than the other in the presence of multipath interference (called "diversity"). This interference, known as fringing, occurs because of the alternating constructive and destructive mixing that occurs when the same signal follows multiple paths to its destination. Once again, the baseband processor is responsible for making the decisions and sending the appropriate control signals.
Together, this collection of integrated circuits can be used to create network interface cards that support a wide variety of industry-standard form factors, coding techniques, and data rates (see glossary). The advent of new MAC architectures, encryption options, and power-control specifications will heavily influence next-generation NIC design. In addition, NIC design will be simplified by the continuing trend toward higher levels of system-on-a-chip (SoC) integration and multi-chip-module packaging advances.