Physical Layer Selection and Design
Selection of the 802.11 physical layer is typically driven by user requirements rather than the physical design. In most cases, it is also driven by the need to build the fastest network possible, at least when next to an access point. Choosing a physical layer is a matter of engineering. None of them are inherently superior to the others. In choosing a physical layer, you are trading off many different factors. In a nutshell, the 2.4 GHz ISM band is less affected by obstacles, so 802.11b/g signals will travel farther. However, backwards compatibility can limit throughput, and it difficult to lay out a network with only three channels. Furthermore, many devices use the 2.4 GHz ISM band, and it is likely that there will be some sort of interference from a Bluetooth device, 2.4 GHz cordless phone, X10 video camera, or some similar widget. If not, the use of only three channels will be limit throughput as the channels overlap with each other. 802.11a is ideally suited for high-density, high-capacity networks because it does not have backwards compatibility limitations and the radio spectrum is much larger. Table 23-2 shows a comparison between 2.4 GHz and 5 GHz 802.11 networks.
802.11b/g (2.4 GHz) |
802.11a (5 GHz) |
|
---|---|---|
Performance (throughput) per AP |
Low for 802.11b; 802.11g may vary from medium to high |
Highest throughput per channel |
Potential performance per unit area |
Lowwith only 3 or 4 channels, channel overlap and interference is practically guaranteed |
Highgreater number of channels means less self-interference between network elements |
Range |
Lower frequency has longer range |
Worsefree-space loss of higher frequencies is higher |
Interference |
Many other uses of frequency band Very limited channel selection leads to lots of co-channel interference |
Frequency used by many fewer devices More channels make layout easier, especially in three dimensions |
Backwards compatibility with older hardware |
Compatible with 802.11 direct sequence and 802.11b hardware |
None |
One of the ways to ease the pain of large-scale wireless LAN deployments is to automate as much as possible. Several products offer the ability to automatically lay out channels. One major channel layout approach is based on physical measurements. Network administrators place access points based on expected user density, mounting convenience, and environmental constraints. When the network is powered on, the APs communicate with each other through a wired network to select the optimum channel assignment. Some products continuously monitor the radio space to adjust the channel settings dynamically in response to environmental changes. The major alternative to physical measurements is based on virtual modeling of a building. When creating a virtual model of the building, channels can be assigned based on calculations with the mathematical model to minimize interference. As with may other aspects of wireless LAN design, many products combine both techniques.
2.4 GHz (802.11b/g) Channel Layout
802.11b and 802.11g share the same frequency band. They are subject to the same regulatory requirements, and use an identical channel map (Table 23-3). Although there are 14 channels, each channel is only 5 MHz wide. A direct sequence transmission is spread across a much wider band than its assigned channel. (See Figure 12-5.) For best effect, you will want to keep both the first and second lobes free of interference. Ideally, that leads to a 33 MHz separation between channel assignments.
There is not quite enough radio spectrum assigned to have three fully nonoverlapping channels in most jurisdictions. Rather than have two interference-free channels, most users allow a slight degree of overlap in the second lobe and run on three channels with at least 25 MHz separation; the resulting channel set is 1, 6, and 11. This causes a small amount of interference to each channel, but it is worthwhile to accept a small reduction in throughput per channel to get the third channel.
In some situations, it may make sense to perform a four-channel assignment (1, 4, 8, and 11).[*] The trade-off is that the signal overlap will be more crowded, further reducing peak throughput. Trading peak speed for a higher total area speed may be worthwhile in some circumstances, but I have generally found this to be a poor trade-off.
[*] European regulations allow the use of wider-spacing, such as 1, 5, 9, and 13 in a four-channel layout.
Channel number |
Channel frequency (GHz) |
US/Canadaa |
ETSIb |
---|---|---|---|
1 |
2.412 |
|
|
2 |
2.417 |
|
|
3 |
2.422 |
|
|
4 |
2.427 |
|
|
5 |
2.432 |
|
|
6 |
2.437 |
|
|
7 |
2.442 |
|
|
8 |
2.447 |
|
|
9 |
2.452 |
|
|
10c |
2.457 |
|
|
11 |
2.462 |
|
|
12 |
2.467 |
|
|
13 |
2.472 |
|
|
a 802.11 allows different rules regarding the use of radio spectrum in the U.S. and Canada, but the U.S. Federal Communications Commission and Industry Canada have adopted identical rules. |
|||
b Not all of Europe has adopted the recommendations of the European Telecommunications Standards Institute (ETSI). Spain, which does not appear in the table, allows the use of only channels 10 and 11. |
|||
c Channel 10 is allowed by all regulatory authorities and is the default channel for most access points when they are initially powered on. |
Part of the site survey is to lay out coverage areas in a way that minimizes channel overlap. Antennas can be a valuable tool in doing this because they can tailor the coverage area to fit the shape of the building or room. No matter what type of antennas you select, there is a general pattern that can be used. The cellular-telephone industry uses a hex pattern as the background for a channel layout. Figure 23-4 shows the problems in planning out a large coverage area. The center channel, in bold, is set for one of the three nonoverlapping channels. In this case, I chose channel 1 arbitrarily. To avoid overlap, the next set of channels around the center needs to alternate between the two remaining nonoverlapping channels in a circle around the channel. After assigning channels to the ring, successive center channels can be laid out. Two such focal points are shown in Figure 23-4 with circles.
Naturally, Figure 23-4 presents a channel layout under ideal circumstances. Within a building, radio propagation is subject to obstacles, and it is usually impossible to avoid overlapping channels. For example, the first "ring" of channels 6 and 11 around the outside of the center channel 1 may very well interfere with each other, especially if the hexagons represent target operational rates that are much larger than the minimum.
Figure 23-4. Frequency planning
Limitations of the 2.4 GHz channel layout
The inability to perform a channel layout with three channels is not surprising. In mathematics, the general result is known as the Four-Color Map Theorem. Map makers discovered in the mid-19th century that an arbitrary two-dimensional map can be filled in with four colors.[*] Unfortunately, 802.11b/g networks have only three non-overlapping channels. Maps where adjacent regions share a color are unsightly. Adjacent AP coverage areas sharing the same channel are not unsightly, unless you can see into the microwave spectrum, but they do suffer from throughput-sapping co-channel interference.
[*] The Four-Color Theorem was finally proved in 1976 with the aid of 1,200 hours of time on a Cray, and was one of the first theorems to be proved with extensive computer assistance.
It is often possible to minimize channel overlap by careful AP placement or the use of external antennas, but you will have to devote time and energy to minimizing channel overlap. In three dimensions, it gets even harder. Radio signals may bleed through the floor or the ceiling, requiring you to lay out channels in three dimensions.
Even in large, open, two-dimensional spaces where you would not expect an interference problem, there can be one. In 2002, the Associated Press ran a story about the problems of intereference, and cited a neighborhood in Florida that had formed its own ad-hoc frequency allocation committee to ensure that neighbors were using nonadjacent channels for their home networks!
5 GHz (802.11a) Channel Layout
802.11a has two major advantages when laying out your network. First, there are at least 12 channels, so channel layout is a snap. Table 23-4 shows the channel frequencies. Be aware that some of the first 802.11a cards to hit the market did not support the highest frequency band, so channels 149 and up may not be usable with all cards. Either replace the old cards, or restrict the network to the eight remaining channels. In either case, with 8 or 12 channels, there are more than enough channels for 2-dimensional layouts, and 3-dimensional layouts can be taken care of in most cases.
Channel number |
Frequency (GHz) |
Notes |
---|---|---|
36 |
5.180 |
Lowest maximum power |
40 |
5.200 |
Lowest maximum power |
44 |
5.220 |
Lowest maximum power |
48 |
5.240 |
Lowest maximum power |
52 |
5.260 |
Slightly higher maximum power |
56 |
5.280 |
Slightly higher maximum power |
60 |
5.300 |
Slightly higher maximum power |
64 |
5.320 |
Slightly higher maximum power |
149 |
5.745 |
Not supported by all cards |
153 |
5.765 |
Not supported by all cards |
157 |
5.785 |
Not supported by all cards |
161 |
5.805 |
Not supported by all cards |
Mixed Channel Layouts (802.11a+b/g Networks)
Most networks are being built with a combination of 802.11b/g radios for compatibility with past hardware, while building in 802.11a for future capacity. Dual-radio APs from many vendors are only marginally more expensive than single-radio models. In many cases, the cost of a tri-mode/dual-band network is only a few hundred or thousand dollars more than the cost of an 802.11g-only network. Unless your budget is extremely tight, it is worth paying for more than double the capacity.
For the foreseeable future, most 802.11 devices are likely to operate in the 2.4 GHz band. Most chipset vendors have concentrated on building 802.11g devices. There are several tri-mode chipsets capable of supporting 802.11a and 802.11b/g simultaneously, but they are not yet widely purchased as a built-in laptop option. When I worked on the Supercomputing wireless network, we had 1,300 users at the peak, and only about 100 were using 802.11a. As the price of chipsets and cards continues to drop, and 802.11a becomes more widely appreciated, I expect the balance to shift drastically towards 802.11a for indoor usage. In outdoor environments, transmission range remains the prime concern. Lower frequencies travel much farther, which will enable 802.11b/g to retain its current dominant position.