6 Some Wireless LAN standards

A short gallery of the most famous Wireless LAN standard (but unfortunately not necessarily the most widespread...).

6.1 IEEE 802.11

The main problem of radio networks acceptance in the market place is that there is not one unique standard like Ethernet with a guaranteed compatibility between all devices, but many proprietary standards pushed by each independent vendor and incompatible between themselves. Because corporate customers require an established unique standard, most of the vendors have joined the IEEE in a effort to create a standard for radio LANs. This is IEEE 802.11 (like Ethernet is IEEE 802.3, Token Ring is IEEE 802.5 and 100vg is IEEE 802.12).

Of course, once in the 802.11 committee, each vendor has pushed its own technologies and specificities in the standard to try to make the standard closer to its product. The result is a standard which took far too much time to complete, which is overcomplicated and bloated with features, and might be obsoleted before products come to market by newer technologies. But it is a standard based on experience, versatile and well designed and including all of the optimisations and clever techniques developed by the different vendors.

The 802.11 standard specifies one MAC protocol and 3 physical layers : Frequency Hopping 1 Mb/s (only), Direct Sequence 1 and 2 Mb/s and diffuse infrared (can we really call it a "standard" when in includes 3 incompatible physical layers ?). Since then, it has been extended to support 2 Mb/s for Frequency Hopping and 5.5 and 11 Mb/s for Direct Sequence (802.11b). The MAC has two main standards of operation, a distributed mode (CSMA/CA), and a coordinated mode (polling mode - not much used in practice). 802.11 of course uses MAC level retransmissions, and also RTS/CTS and fragmentation.

The optional power management features are quite complex. The 802.11 MAC protocol also includes optional authentication and encryption (using the WEP, Wired Equivalent Privacy, which is RC4 40 bits - some vendors do offer 128 bits RC4 as well). On the other hand, 802.11 lacks to defines some area (multirate, roaming, inter AP communication...), that might be covered by future developments of the standard or complementary standards. Some 802.11 products also implement proprietary extensions (bit-rate adaptation, additional modulation schemes, stronger encryption...), those extensions may or may not be added to the standard over time.

When 802.11 was finalised (september 97), most vendors were slow to implement 802.11 products because of the complexity of the standard and the number of mandatory features (and in some cases they also need to provide backward compatibility with their own previous line of products). Some of the optional features (encryption and power saving) did only appear months after the initial release of the product. But things seem to be sorted out and we now have fully featured products on the market. The complexity of the specification, the tightness of the requirements and the level of investment required made 802.11 products expensive compared to the previous generation of wireless LANs, but because of the higher standardisation and higher volumes, prices are now dropping.

Even if vendors eventually have launched 802.11 products, the standard doesn't fully guarantee inter-operability : the products have to use at least the same physical layer, the same bit rate and the same mode of operation (and there is so many other little important details...). The most cooperative vendors have been busy lately sorting out interoperability issues with independent testing labs, but it is still a touchy subject...

6.2 802.11-b and 802.11-a (802.11 at 5 GHz)

After 7 years of arguing in sub-committees making 802.11, you would think that most people would had enough of it. In fact no, the 802.11 committee is now busy pushing a new standard at 5 GHz, and also higher speed at 2.4 GHz (by tweaking the Direct Sequence physical layer). Both standard makes changes only to the physical layer, so that the 802.11 MAC can be reused totally unmodified, saving costs.

802.11-a (802.11 at 5 GHz) was standardised first (spring 99), based on OFDM (see chapter 4.7.4), and using the UNII band (see chapter 4.2 - so it won't be available in Europe and Japan). The OFDM physical layer is a very close copy of the one used in HiperLan II (so they might be some sort of compatibility - see chapter 6.4), using 52 subcarriers in a 20 MHz channel, offering 6, 12 and 24 Mb/s and optional 9, 18, 36, 48 and 54 Mb/s bit-rates. No products are yet on the market.

Very soon after, 802.11 did standardise 802.11-b (802.11 HR), based on a modified DS physical layer (see chapter 4.7.3). The goal was to extend the life of the 2.4 GHz band by overcoming the major drawback : low speed. On top of the original 802.11-DS standard, 802.11-b offer additional 5.5 Mb/s and 11 Mb/s bit rates. It was approved by the FCC and they are now products on the market (which are quite popular).

6.3 HiperLan

HiperLan is the total opposite of 802.11. This standard has been designed by a committee of researcher within the ETSI, without strong vendors influence, and is quite different from existing products. The standard is quite simple, uses some advanced features, and has already been ratified a while ago (summer 96 - we are now only waiting for the products).

The first main advantage of Hiperlan is that it works in a dedicated bandwidth (5.1 to 5.3 GHz, allocated only in Europe), and so doesn't have to include spread spectrum. The signalling rate is 23.5 Mb/s, and 5 fixed channels are defined. The protocol uses a variant of CSMA/CA based on packet time to live and priority, and MAC level retransmissions. The protocol includes optional encryption (no algorythm mandated) and power saving.

The nicest feature of Hiperlan (apart from the high speed) is the ad-hoc routing : if your destination is out of reach, intermediate nodes will automatically forward it through the optimal route within the Hiperlan network (the routes are regularly automatically recalculated). Hiperlan is also totally ad-hoc, requiring no configuration and no central controller.

The main deficiency of Hiperlan standard is that it doesn't provide real isochronous services (but comes quite close with time to live and priority), doesn't fully specify the access point mechanisms and hasn't really been proved to work on a large scale in the real world. Overhead tends also to be quite large (really big packet headers).

HiperLan suffers from the same disease as 802.11 : the requirements are tight and the protocol complex, making it very expensive.

6.4 HiperLan II

HiperLan II is the total opposite of HiperLan (see above ;-). The first HiperLan was designed to build ad-hoc networks, the second HiperLan was designed for managed infrastructure and wireless distribution systems. The only similarities is the HiperLan II is being specified by the ETSI (Broadband Radio Access Network group), operate at 5 GHz (5.4 to 5.7 GHz) and the band is dedicated in europe.

HiperLan II was the first standard to be based on OFDM modulation (see chapter 4.7.4). Each sub-carrier may be modulated by different modulations (and use different convolutional code, a sort of FEC), which allow to offer multiple bit-rates (6, 9, 12, 18, 27 and 36 Mb/s, with optional 54 Mb/s), with likely performance around 25 Mb/s bit-rate. The channel width is 20 MHz and includes 48 OFDM carriers used to carry data and 4 additional are used as references (pilot carriers - total is 52 carriers, 312.5 kHz spacing).

HiperLan II is a Wireless ATM system (see chapter 5.1.4), and the MAC protocol is a TDMA scheme centrally coordinated with reservation slots. Each slot has a 54 B payload, and the MAC provide SAR (segmentation and reassembly - fragment large packets into 54 B cells, see chapter 5.2.2) and ARQ (Automatic Request - MAC retransmissions, see chapter 5.2.1). The scheduler (in the central coordinator) is flexible and adaptive, with a call admission control, and the content of the TDMA frame change on a frame basis to accommodate traffic needs. HiperLan II also defines power saving and security features.

HiperLan II is designed to carry ATM cells, but also IP packets, Firewire packets (IEEE 1394) and digital voice (from cellular phones). The main advantage of HiperLan II is that it can offer better quality of service (low latency) and differentiated quality of service (guarantee of bandwidth), which is what people deploying wireless distribution system want. On the other hand, I'm worried about the protocol overhead, especially for IP traffic.

6.5 OpenAir

OpenAir is the proprietary protocol from Proxim. As Proxim is one of the largest Wireless LAN manufacturer (if not the largest, but it depends which numbers you are looking at), they are trying to push OpenAir as an alternative to 802.11 through the WLIF (Wireless LAN Interoperability Forum). Proxim is the only one having all the detailed informations on OpenAir, and strangely enough all the OpenAir products are based on Proxim's module.

OpenAir is a pre-802.11 protocol, using Frequency Hopping and 0.8 and 1.6 Mb/s bit rate (2FSK and 4FSK). The radio turnaround (size of contention slots and between packets) is much larger than in 802.11, which allow a cheaper implementation but reduces performance.

The OpenAir MAC protocol is CSMA/CA with MAC retransmissions, and heavily based on RTS/CTS, each contention slot contains a full RTS/CTS exchange, which offer good robustness but some overhead. A nice feature of the protocol is that the access point can send all its traffic contention free at the beginning of each dwell and then switch the channel back to contention access mode.

OpenAir doesn't implement any encryption at the MAC layer, but generates Network ID based on a password (Security ID). This provide some security only because Proxim controls the way all the implementation behave (they don't provide a way to synchronise to any network as 802.11 manufacturers do). OpenAir also provide coarse power saving.

6.6 HomeRF & SWAP

NOTE : this chapter was written when I was finishing writing the SWAP 1.0 specification in December 98. After I left the HomeRF, a lot of big political game did happen, which triggered some critical changes to the specification (SWAP 1.1). I don't really know how much of it is still accurate, but I believe that the standard is no longer as open and vendor neutral as it was and that performance has been dramatically reduced.

The HomeRF is a group of big companies from different background formed to push the usage of Wireless LAN in the home and the small office. This group is developing and promoting a new Radio Lan standard : SWAP.

The Home is a good market for Wireless LAN because very few houses are nowadays cabled with Ethernet wire between the different rooms, and because mobility in the home is desired (browse the web on the sofa). The use of the 2.4 GHz band allows a free worldwide deployment of the system.

The HomeRF has decided to tackle the main obstacle preventing the deployment of Wireless LAN : the cost. Most users just can't afford to spend the money required to buy a couple of Radio LAN cards to connect their PCs (without talking of the access point).

The main cost of a radio LAN is the modem. As this is analog and high power electronics, it doesn't follows Moore's law (the market trend that allow you to buy a Cray at the price of a calculator after a few years) and modems tend to be fairly stable in price. Frequency Hopping modems tend to be less expensive, but the 802.11 specification impose tight constraints on the modem (timing and filtering), making it high cost. The SWAP specification, by releasing slightly those constraints, allows for a much cheaper implementation, but still keeps a good performance.

The MAC protocol is implemented in software and digital, so doesn't contribute that much to the final cost of the product (except in term of development cost). Releasing some hardware constraints prevented the use of the 802.11, which anyway was much too complex and including too many features not necessary for the task.

The main killer application that the HomeRF group envisages is the integration of digital cordless telephony and the computing word, allowing the PC to reroute the phone calls in the home or to offer voice services to the users.

A new MAC protocol has been designed, much simpler, combining the best feature of DECT (an ETSI digital cordless phone standard) and IEEE 802.11 : a digital cordless phone and ad-hoc data network, integrated together.

The voice service is carried over a classical TDMA protocol (with interference protection, as the band is unlicensed) and reuse the standard DECT architecture and voice codec. The data part use a CSMA/CA access mechanism similar to 802.11 (with MAC level retransmission, fragmentation...) to offer a service very similar to Ethernet.

The 1 Mb/s Frequency Hopping physical layer (with optional 2 Mb/s using 4FSK) allows 6 voice connections and enough data throughput for most users in the Home. The voice quality should be equivalent to DECT in Europe and much better than any current digital phone in the US. Data performance should be slightly lower than 802.11. The MAC protocol has also been designed in a very flexible way, allowing to develop very cheap handset or data terminals and high performance multimedia cards for PCs...

The SWAP specification is an open standard (in fact, more open than 802.11, because there should be no royalty or patent issues), quite simple and straightforward. In fact, the combination of voice and data gets already most marketing people drooling ! The only drawback is that you will have to wait a bit before seeing SWAP products in your favourite supermarket...

6.7 BlueTooth

BlueTooth should not even be mentioned in this document, but people keep thinking that BlueTooth is a Wireless LAN. BlueTooth is a cable replacement technology mostly developed and promoted by Ericsson with the help of Intel, offering point to point links and no native support for IP (need to use PPP). It may be good for some applications, but not for Wireless LANs.

I personally read the BlueTooth specification, and I was not impressed, expect by the size of the thing (more than 1500 pages !). My take is that BlueTooth offers the functionality of a Wireless USB, and in fact looking into the huge specification we can see some similarities in the design.

BlueTooth offers the possibility to create a set of point to point wireless serial pipes (RfComm) between a master and up to 6 slaves, with a protocol (SDP) to bind those pipes to a specific application or driver. The BlueTooth mindset is very vertical, with various profiles defining every details from bit level to application level. TCP/IP is only one profile, implemented through PPP in a specific pipe. There are other pipes for audio, Obex... With BlueTooth, nodes need to be explicitely connected, but they remember bindings from one time to another.

This is miles away from the current wireless LAN approach (connectionless broadcast interface, native IP support, cellular deployement, horizontal play), so BlueTooth doesn't fit TCP/IP and wireless LAN applications too well. On the other hand, as a wireless USB, it fulfil a role that regular wireless LANs can't, because TCP/IP discovery and binding protocols are more heavyweight.

Currently, BlueTooth is moving very slow (my first reading of the spec was autumn 97 - then called MC-Link) due to its complexity and the inherent limits due to the protocol design (people are learning how to workaround "features"), but eventually some products should reach the market and later on software support should come...

In summary, if all you want is to run TCP/IP, you may find it cheaper and more effective to NOT wait for BlueTooth and live without the hype.

up


Linux Wireless LAN Howto - jt@hpl.hp.com
Converted to html from Frame Maker - 25 August 98
Updated 3 August 00
Copyright © 1996-2004 Jean Tourrilhes
    Project hosted and sponsored by :
HP home page