The 1990s have seen the very rapid growth of wireless networking, analogous in many respects to the early growth in copper LANs. The most significant event in this process was the ratification, in mid 1997, of the new IEEE 802.11 industry standard for wireless networks.

Initiated at the beginning of this decade, often controversial and definitely protracted in its definition, the 802.11 standard finally sets a basic measure for interoperability of wireless LAN equipment. While many still argue that holes exist in the standard, setting restrictions on full interoperability, the standard in its existing form addresses nearly all areas of interest.

In terms of basic technology, 802.11 is essentially "Wireless Ethernet", with the caveat that the much trickier radio/infrared transmission environment results in a more complex and in many respects more refined standard than the established copper LAN 802.3 10/100-Base-2/5/T/F standards.

The practical implications of 802.11 are that direct head-to-head competition between vendors will result in the "commodification" of wireless LAN hardware, and the option of mixing and matching hardware to implement multivendor networks to achieve the desired match in capability and cost.

Systems Magazine considered a detailed review of the standard in 1997, but consensus between the editor and author was to wait for the first compliant chipsets to emerge, to provide readers with a more comprehensive picture of the technology base. The technological ideas underlying the spread spectrum techniques used in 802.11, and the basic functional issues in wireless networking were addressed in earlier issues.

The IEEE 802.11 Standard - A Rationale

The 802.11 standard was devised primarily to provide interoperability in a multivendor environment, while also providing the capability to exploit Direct Sequence Spread Spectrum, and Frequency Hopping Spread Spectrum modulation techniques in the microwave radio frequency (RF), and infrared (IR) transmission bands. An important issue was the provision of a model which would allow simple and robust inter-operability with devices based upon the established 802.3 standards.

Wireless networking introduces a number of rather nasty problems which have hitherto not been seen in copper/glass networks:

The high complexity of the resulting 802.11 standard, in comparison with 802.3, directly reflects the much more demanding transmission environment. Specific features in the standard are designed to accommodate these potential problems.

IEEE 802.11 Standard - Topological Models

The conventional copper/glass LAN is built around the model of a logical peer-peer network, using a linear segmented, star/tree, or combination thereof for its physical topology.

The 802.11 standard, by the nature of its transmission environment, departs from this long established model, since it must accommodate a common logical/physical topology.

Two models are therefore supported in the 802.11 standard. The first of these is the "ad hoc" model, which is a peer-to-peer arrangement of 802.11 devices, which all exchange traffic. In the standard, this model is referred to as the Independent Basic Service Set (IBSS). An IBSS network is typically not connected to a larger network, although this does not preclude one of the nodes in the network providing IP routing access via a different channel.

The second model is commonly referred to as an "infrastructure" model, in which wireless nodes connect through a shared Access Point. In the standard, this model is referred to as a Basic Service Set (BSS) or Extended Service Set (ESS) scheme. A BBS network typically employs a single AP, whereas an ESS network ties in a number of APs to cover a much larger footprint. The APs are tied together via the established copper/glass cable LAN infrastructure.

These models are in turn reflected in the Medium Access Control (MAC) layer of the standard. The layers above the MAC comply with the existing 802 series standards.

IEEE 802.11 Standard - The Physical (PHS) Layer

The PHS layer in the 802.11 standard is defined for direct spreading (DSSS), frequency hopping (FHSS) and IR environments. Each has unique idiosyncrasies, designed to specifically exploit each environment.

The DSSS PHS model is based upon direct spreading using an 11-bit pseudorandom Barker code, a technique widely used in modern radar pulse compression designs. Each bit in the 802.11 packet is thus mapped into a sequence of 11 PN chips. The 11-bit code is pretty much minimal, in comparison with other spread spectrum schemes, and provides only a very modest 10.4 dB of interference rejection, and corresponding approximately tenfold spreading of the baseband bandwidth. It does however provide good performance in rejecting multipath reflections.

Two forms of carrier wave modulation are supported, Differential Binary Phase Shift Keying (DBPSK), and Quadrature Binary Phase Shift Keying (DQPSK), with an 11 MHz chip rate and 1 Mbit/s and 2 Mbit/s data rates, respectively. In comparison with common BPSK and QPSK, the state of the data bit/dibit in these differential variants depends on the state of the preceding bit/dibit received. This is different in concept to the use of Manchester encoding in 802.3, where the direction of transition determines the value of transmitted cell.

The modulation scheme is in principle similar to that used in CDMA mobile phones, but the shorter code length precludes code division muxing. Therefore, 802.11 DSSS PHY operates in 12 discrete channels, with 5 MHz spacing, in which the sidebands of adjacent channels overlap. Because the Barker code is unbalanced, and does not yield a very "white" pseudo-noise spectrum, the code is fed into a scrambler to "whiten" its pseudo-noise properties, and DC balance the code for reception.

The FHSS PHS model is based upon frequency hopping between 79 discrete channels, spaced at 1 MHz, using a PN code to select the instantaneous channel frequency. The data symbol to be transmitted at 1 Mbit/s is first given a Gaussian shape, and then frequency modulated on to the carrier wave. A 1 Mbit/s data rate is achieved with 2 level Gaussian Frequency Shift Keying (GFSK), a 2 Mbit/s rate with 4 level GFSK.

The intent of the FHSS PHY scheme used is to allow multiple channels to coexist in the same bandwidth. 78 PN hopping patterns are divided into three sets of 26 channels. As a result, 26 colocated networks can coexist within the same slice of spectrum.

It was intended by the designers of the standard that FHSS 802.11 devices have provisions for multiple antenna channels, to allow a receiver to select the antenna which provides the best possible multipath interference level, for a given transmitter.

The Infrared (IR) PHS model is based upon Pulse Position Modulation (PPS) techniques, whereby the relative timing of the transmitted IR pulse contains the modulation envelope. PPM techniques have the advantage, against other IR modulation schemes, of allowing a constant energy IR pulse to be used by a transmitter. As a result, a smaller LED or laser diode can be employed, in comparison with schemes such as Pulse Frequency Modulation (PFM), or Amplitude Shift Keying (ASK).

All of the 802.11 PHS schemes differ fundamentally from 802.3 in the use of Carrier Sense Multiple Access / Collision Avoidance (CSMA/CA). In this arrangement, the backoff scheme is arranged so that contending receivers avoid collisions. Each receiver first senses the presence or absence of a carrier wave. If a carrier is present, the receiver backs off, exponentially in time, and then retries again, sensing for a busy channel. Only when the channel is free will it attempt an access.

The scheme implements a Request To Send / Clear to Send (RTS/CTS) arrangement to facilitate fair bandwidth use.

This scheme is supported through the use of a Physical Layer Convergence Protocol (PLCP) which is prepended to each transmitted packet. In the DSSS PHS model, the PCLP comprises a 144 bit preamble, followed by a 48 bit header. The preamble contains 128 synchronisation bits, enabling a receiver to lock on to the modulation a set receive gain, select the antenna channel if applicable, and compensate any frequency drift, and a 12 bit Start Frame Delimiter. One the receiver identifies the SFD, it clocks in subsequent bits in the header.

The header contains an 8 bit Signal field indicating 1 or 2 Mbit/s data rate, an 8 bit Service field indicating 802.11 compliance, a 16 bit Length field for the payload, and a 16 bit CRC for the header.

The PCLP format for the FHSS model differs in a number of respects form the DSSS scheme. The preamble is 80 bits long, a 16 bit frame start sequence is used, and the 32 bit header is structured differently. A 12 bit field is used for defining the payload length, a 4 bit field is used to identify the data rate, and a 16 bit CRC is employed.

The header is always transmitted at the lower 1 Mbit/s data rate, and the rate field in the header is used by the receiver to set the demodulation for the following MAC level payload. The payload is scrambled.

The complexity of the PHS layer in 802.11 is striking in comparison with copper/glass 802.3 variants, but the complexity is entirely justified given the pathological properties of the transmission environment. Importantly, the PCLP frame structures have provisions for further evolution of the standard downstream.

IEEE 802.11 Standard - The Medium Access Control (MAC) Layer

The MAC layer for all 802.11 variants is common, with PHS variant specific differences all pushed down into the PHS PCLP header. This clever strategy was devised to provide simpler MAC level hardware implementation. The MAC layer is designed to provide seamless inter-operation with 802.3 LANs.

Unlike 802.3, 802.11 provides for MAC level acknowledgment of and if necessary retransmission of messages. This is a major improvement over the wired LAN standard, which must rely on higher level protocols for managing the retransmission of trashed or lost packets. This is however as much a virtue as a necessity in an interference rich environment.

Another very important and new feature used in the 802.11 MAC layer is fragmentation, whereby a transmitter can break a message into a series of much smaller messages. In an interference rich environment, smaller packet sizes are advantageous, since less data needs to be resent if a packet is trashed in transmission. In this fashion, noisy and clean environments can be used and throughput balanced accordingly. The standard is defined to allow a vendor to use a proprietary adaptive fragmentation management scheme, yet still be interoperable with other implementations.

Roaming is supported in some detail, for BSS/ESS environments. An 802.11 device can scan for whatever APs are within range, and through the use of signal power level detection, and protocol provisions, change its association with an AP transparently to the user. There is some criticism of the handoff mechanism between APs when roaming, and this area of the standard is expected to further evolve downstream.

Power management features are included in the MAC layer, to maximise battery life in portable devices, such as laptops, palmtops and "wearable computers". Two schemes are used, to allow devices to "sleep" in battery saving mode.

Where an AP is employed, it queues up messages to sleeping devices, which are required to periodically wake up and poll the AP for queued traffic from other devices in the network.

Supporting a sleep mode in an ad hoc IBSS network is more complex, and the scheme is designed so that all nodes wake up on the receipt of a beacon signal, the transmission of which is randomly alternated between the nodes in the IBSS network.

Encryption, termed in 802.11 Wired Equivalent Privacy (WEP), is another optional, and very important feature which is new to the 802.11 standard. The MAC payload can be encrypted using a 40 bit secret key RC4 algorithm. Encryption is provided only for node-to-node links and is not supported for through connections.The limitation of this scheme lies in key management, since this must be implemented either manually, or via a separate higher level protocol, not defined in the standard. The default operating mode is unencrypted operation.

In my opinion, the WEP scheme will not be adequate in the longer term, since it is locked into a fixed 40 bit key size and does not provide an embedded mechanism for key distribution, such as is common in digital envelope schemes. However, it is also an area where the standard could later evolve.

Since very little research has been published in the literature on the embedding of MAC level encryption, and supporting key management schemes, the 802.11 committee had little choice than to specify a minimalist solution to the problem.

The framing structure for the MAC layer is unique but similar for the three defined frame types, Control Frames, Management Frames and Data Frames. The basic frame format contains a 30 byte MAC header, a 0-2312 byte variable length data payload, and a 4 byte CRC field.

The MAC header contains a 2 byte Frame Control field, a 2 byte Duration ID field, a 2 byte Sequence Control field used for managing acknowledgments, and four 6 byte address fields. The Frame Control field is broken down further, into a 2 bit protocol identifier, a 2 bit Type field, a 4 bit Subtype field, a pair of To/From bits, a bit to flag more fragments, a bit to flag a retry, a bit for power management state, a bit to flag more data, and a bit to flag the use of WEP.

The complexity of 802.11 MAC layer reflects, in comparison with 802.3, the much more demanding requirements of the transmission environment. Importantly, the standard makes many of these features optional. Therefore, an end user shopping for 802.11 compliant equipment must be extremely careful to ensure that the equipment to be acquired does indeed fully support all of the options which are intended to be used in the organisation's LAN. Vendors should therefore provide proof of specific features being supported, or the offered equipment should be tested by the end user against in service 802.11 gear.

Areas of Future Growth

The 802.11 standard is expected to grow further as it matures. A number of existing holes in the protocol area, especially roaming and encryption, will clearly have to evolve as the standard proliferates and the installed base grows.

The area where most pressure is being exerted at this time is that of providing higher data transmission rates than the very modest 1-2 Mbit/s provided by the first release of the standard. While 1-2 Mbit/s is easily adequate for lightly loaded laptops and palmtops, accessing files on a central server, it is barely adequate for applications such as multimedia, where the 802.11 channel will be saturated by a single MPEG application.

At this time proposals exist for 10 Mbit/s variants of the 2.45 GHz standard, and a 20 Mbit/s variant is under discussion for the 5 GHz band. Existing 802.11 development tasks are to provide for backward compatible variants running at 3 Mbit/s (FH) and 8 Mbit/s (DS).

Bell Labs (Lucent) are claimed by US industry sources to have developed a DS scheme which embeds additional * Mbit/s throughput via PPM, and substituting DQPSK with QAM. A device would then select which mode it uses depending on the capabilities of other devices it communicates with.

The most recent 802.11 publications indicate that at least three schemes are under serious consideration. The Alantro proposal uses QPSK and provides for 1, 2, 2.75, 5.5, 11, 14.333, 16.5, 17.6, 18.333, 19.25 Mbit/s operation. The Lucent/Harris proposal under evaluation uses Complementary Code Keying techniques and provides for 1, 2, 5.5 and 11 Mbit/s operation. The Micrilor proposal uses 16 level DBOK modulation, to provide 1, 2, 8.7, 10 or 18 Mbit/s throughput.

In terms of basic design, the standard is flexible enough to absorb extensions in a number of areas, while retaining a high level of backward compatibility. At this time a user's focus in purchasing should be concentrated upon whether the devices on offer are indeed genuinely complaint with the full standard, or relevant portions thereof.

Implementation

The complexity of the 802.11 standard makes it a difficult one to implement well. At this stage few vendors are offering genuinely compliant equipment, and most of this equipment would appear to be built around the chipset jointly developed by Lucent and Philips. The Lucent-Philips chipset is the basis of the WaveLAN OEM 802.11 modem module, which is being marketed to OEM equipment manufacturers intending to build 802.11 based I/O cards and APs.

Odds are that in the near term most equipment you may source will be built around the same chipset and therefore basically interoperable. Difficulties may arise downstream as other chipsets begin to proliferate.

The Philips-Lucent chipset comprises three devices, and supporting components. The SA 1630 is an IF quadrature transceiver, designed for DSSS, operating at an IF frequency between 70 and 400 MHz. It provides 70 dB of digital gain control in 2 dB steps, and includes quadrature signal generation and a pair of mixers.

The SA 2410 is a 2.45 GHz band integrated power amplifier, which includes a Transmit/Receive switch and can output up to 18.5 dBm (260 mW) of RF power, to drive an antenna.

The SA 2420 is an integrated 2.45 GHz RF transceiver which contains a preamplifier, receive and transmit mixers, and is designed to provide the carrier frequency portion of the receiver.

An 802.11 interface would be built up using this chipset, and additional logic using either CMOS ASICs or EPLD style devices.

Summary

Size limitations preclude a more detailed discussion of the 802.11 standard, implementation alternatives, and emerging products. What is clear at this point in time is that the standard is ready for use in the marketplace, and that we can expect a decent number of products to emerge in the next 18 months.

The caveat with 802.11 is to choose cautiously, since we are in the transitional period between a marketplace filled with "almost compliant" products built around the draft standard, and early products built around the first release of the standard. Providing that this is observed, the risks in implementing an 802.11 network are easily manageable.

The essential conclusion is that 802.11 will revolutionise networking in a manner no different to 802.3 in its day.