Wireless networks depend on radio waves, the result of the weird and wonderous magic of electromagnetism. Radio waves are magnetic fields caused by electrical currents; every electrical current creates a magnetic field of some fashion. If you've ever seen a monitor or TV flicker when a fan or vacuum cleaner was moved nearby, you've witnessed electromagnetic interference at work. Radio is the result of harnessed electromagnetic fields.
The flow of the current
An electrical current is passed through a specially crafted set of wires in the antenna at particular frequencies to generate a signal. An antenna's size and shape is mandated by the wavelength of the signal to be transmitted. Antennas are typically even multiples of the wavelength. There are numerous antenna designs, but the two most common are the omnidirectional "whip" antennas and the flat directional antennas.
Regardless of antenna design, radio waves emit from the antenna based on the direction of current flow, and they create large loops. The loops emanate at a right angle to the direction of flow, and the signal gets weaker the farther it gets from the antenna, as shown in Figure A.
|Top-down view of antennas|
Line weight and color indicate the strength of the signal. The closer you are to an antenna, the more powerful the signal. At a certain point, you can be too close to the antenna to get a good signal. The basic principle remains the same for both antennas: Get as much cross-section of the receiving antenna facing the most cross-section of the broadcast antenna.
There are a variety of resources on the Internet for creating a stock dipole antenna. Methods range from trimming potato chip cans to using discarded mini-dish reflectors. While in most cases these are harmless modifications, be aware that if you're creating a hot spot for public use, some antennas on base stations can push the signal strength beyond the Federal Communications Commission's accepted maximum. The most recent record for a non-amplified terrestrial Wi-Fi connection is 55.1 miles, set by a couple of teens from Ohio.
Omnidirectional antenna placements
To capture as much signal as possible, you should orient the receiving antenna parallel to the transmitter for an omnidirectional antenna. However, it should not be in line with the transmitter in order to maximize the shared facing between antennas. In Figure B, you'll find a simplified example of three receivers and one transmitter located in various parts of a three-story building.
|Three receivers, one transmitter|
The gray antenna is the broadcaster. The white antenna is pointed end-on at the end of the antenna. It is basically receiving a minimal amount of radio frequency (RF) energy and is virtually useless. This would be like having a client one floor above the transmitter with both antennas oriented vertically. The pale-blue antenna is perpendicular to the broadcaster, as if the downstairs antenna were horizontal. It has a better signal than the upstairs client, but it's still not optimal. Finally, we have the dark-blue client where the receiving antenna is parallel to the broadcast antenna and at the same elevation. This receiver is intercepting a significant amount of RF energy and has a strong signal.
Directional antenna placements
Now let's look at directional antenna placements, as shown in Figure C. The white antenna is oriented parallel to the directional antenna, but is not facing the antenna. Unlike the dipole antenna, it's not receiving any RF energy and is completely useless. The pale-blue antenna is perpendicular to the broadcaster and just barely in the broadcast cone. It will function, but not well. With the dark-blue antenna, we finally have a client where the receiving antenna is in the broadcast cone and oriented parallel to the face of the transmitter. This receiver is intercepting a significant amount of RF energy and has a strong signal.
|Directional antenna placements|
How far can you go?
With the theory out of the way, it's time to look at something a bit more concrete. In Table A, we have typical ranges you'd expect to see from an 802.11b/g device. The slight spike in transfer rates at 18 Mbps is because the 11 Mbps assumes a direct-sequence spread spectrum/complementary code keying (DSSS/CCK) broadcast, while the 18 Mbps is based on orthogonal frequency division multiplexing (OFDM). Since OFDM has better signal management, it's able to eek out a bit more range.
As you can see in the table, with a "typical" setting, the indoor environment cuts the range down to 50 percent (20 percent if outdoors). The reason is that material objects absorb the radiation. The amount of absorption varies with the material, but generally the more mass in the object, the more the absorption. Metal provides copious amounts of shielding due to how it interacts with electromagnetic fields. Furthermore, the angle that the signal passes through the wall affects the amount of interference. In Figure D, you can see the difference between a signal that passes directly through a wall and one that crosses at an oblique angle. Figure E shows the length of common angles as they pass through a hypothetical 1-foot-thick wall.
|Difference between signals|
A signal passing through a 1-foot wall at a 60-degree angle is two walls for all intents and purposes.
Table B shows how much signal remains after passing through common wall types at different angles.
Unfortunately, there's no simple guide of "subtract 1 percent per foot" rule when deciding upon the impact of distance on a signal. Distance is a funny business because energy dissipates in a sphere, which means the relationship is nonlinear. The most generic rule of thumb you can use is that the signal is one-fourth as strong when twice as far away. Thus, if you think of a hypothetical square flashlight, it can create a 1-inch by 1-inch pattern on a wall that's 1 foot away. At 2 feet away, it creates a square that's 2-foot by 2-foot, which is four square feet. Since there's the same amount of light coming out of the flashlight, the light at 2 feet is a quarter of the light produced at the 1-foot distance, as shown in Figure F.
The wireless distance calculation is further complicated because different receivers will have different sensitivities. If the signal doesn't exceed the background noise, you won't get a connection. Fortunately, you don't have to perform any difficult computations to figure out signal strength. Instead, you can use the effective range to indicate your signal strength. If you have the equipment, take a laptop, an extension cord, and the access point to the parking lot. Set the access point up and start walking, checking the connection. Barring that, you can refer to the typical range chart provided earlier.
For instance, we take a warehouse with an access point 50 feet from a plaster wall. The access point has an open-air range of 250 feet at 54 Mbps. It can get a 54-Mbps signal to a PC located across the bare warehouse 250 feet away without problems. The signal in the other direction travels 50 feet to pass through a plaster wall at a right angle. It has 200 feet of signal remaining (250 feet - 50 feet), which the plaster wall reduces to 100 feet (200 feet x 50 percent), giving a total range of 150 feet (Figure G).
|Range of 150 feet|
You'll notice that this range is much lower than that in typical manufacturers' documentation, but this is far more in line with my personal experience. I suspect the "typical" interior used for their testing is a cubicle environment with half-height walls, the Wi-Fi antennas located on top of the desks, the access point mounted to the ceiling, and all wiring and lights properly shielded. I've seen plenty of open-air connections meet manufacturer specifications, but I'm still waiting to see indoor 165-foot ranges at 11 Mbps, let alone 54 Mbps.
Figure H estimates how far a signal will stretch in a hypothetical office. The interior walls are typical plaster with a cinder-block exterior wall. This process starts with the outdoor range and then reduces the remaining range by the obstacles it encounters.
A signal is just noise without data. A simple wireless network consists of a pair of transceivers—two devices capable of transmitting and receiving data. The 802.11b/g system requires devices to associate with each other using a service set identifier (SSID), so each network knows which signals to listen to. In most cases, you have an access point, which is a wireless hub that talks to multiple devices, and one or more clients that talk only to the access point. This configuration is known as an infrastructure network because it involves centralized configuration.
There is also an ad hoc network that consists of multiple clients without an access point. Ad hoc networks are convenient for temporary connections, but the additional overhead makes them significantly slower (typically about half the speed of an infrastructure network). Because of the extremely low performance and the low expense of access points and wireless routers, ad hoc configurations should be avoided.
Every communication medium has to contend with collision, but wireless networks face special challenges. Collision happens when there are two transmissions at the same time. Electronic devices do not have the processing power to sort things out even if the medium has the bandwidth to support two conversations, so we need a way to manage the network. In the most simplistic management schema, the clients listen to see if any other devices are transmitting and wait until it gets quiet to transmit. Each client is supposed to pause occasionally so other devices can get a word in edgewise. In technical terms, this is known as Carrier Sense Multiple Access (CSMA).
Wired networks ensure that every client can hear every other relevant client's transmissions and have few worries. On the rare times two devices simultaneously start transmitting on the same medium, they're able to recognize it, stop broadcasting, and pause a random amount of time before trying again. This pause feature is known as collision detection (CSMA/CD).
On a wireless network, we have the hidden node problem. In this case, not all transmitters can hear each other to be sure the medium is available. Looking at Figure I, the access point can hear both Client A and Client B, but unfortunately A and B cannot hear each other. Unless the access point is broadcasting, both A and B will think the medium is empty, begin transmitting away, and step all over each other.
|Clients A and B can't hear each other.|
To avoid this problem, clients must poll the access point for permission to talk. When a client wants to send a transmission, it sends a special packet indicating it's ready to send (RTS). The access point replies that it's clear to send (CTS) with another special packet. If the client does not receive the CTS, it waits a bit and tries again. For every data packet transmitted, the recipient returns an acknowledgement (ACK) to the sender. This process is known as collision avoidance (CSMA/CA).
All 802.11b/g devices use the CCK modulation to send RTS and CTS messages according to the manufacturer's specification. Additional bits of information in the RTS/CTS let the 802.11g device know if it can switch to OFDM mode or if it needs to stay in DSSS mode.
A typical schema might go something like this: Client A wants to talk on the network, so it sends out an RTS. Assuming Client B is quiet, the access point returns a CTS to Client A, which begins transmitting away. The access point broadcasts an ACK message for every valid packet it receives. If, during the course of events, Client B decides to get some work done, it sends an RTS, which goes unanswered by the access point, and Client B waits a while to try again. Unfortunately, that RTS interferes with the access point's ability to understand Client A's signal, and it drops a packet. Since the access point didn't send an ACK, Client A sends the dropped packet again, and the transmission continues cleanly.
The problem with the schema above is that before you begin sending data, you have to send RTS and receive CTS messages, which require bandwidth. Every RTS that doesn't receive a CTS likely results in a dropped packet for another client. Furthermore, every successful packet you do send requires a CCK-modulated ACK packet, eating up even more bandwidth. Add the CCK or OFDM preambles used to identify the signal as something other than background noise, the wireless SSID, client MAC, WEP/WPA data, and headers for the general network (TCP/UDP, IP/NetBIOS, etc.), and you end up with a serious amount of network overhead. In general, half of the available bandwidth is consumed by the network itself. My next article will deal with various tried-and-true deployment variations that you can use to maximize your bandwidth and overcome some of these limitations.