When I first developed an interested in radio as a hobby, the only multiple access scheme in common use was frequency division. The government would grant an organization, such as a towing company or police department, exclusive rights to use a particular frequency in a geographic area. The government could only assign a frequency to an organization well outside the range of all other users. This was inefficient for organizations with a low duty cycle, i.e. who most of the time were not transmitting.
Trunking systems appeared in the late 80’s as a way to increase spectrum use efficiency. Fifty users could share 10 frequencies in the same geographic area. A central controller assigned users a frequency each time they transmitted. This increased efficiency at the price of only allowing 10 users to transmit at any one time.
In recent years more powerful multiple access schemes have come into common use. Regulatory authorities have opened up unlicensed ISM bands. There is no licensing system to control who uses the bands so ISM-band equipment must incorporate a multiple-access scheme to share the channel. Regulatory authorities limit output power and power density, but they do not specify a scheme to share bands.
The popularity of the ISMs bands is a testament to the power of a laissez-faire approach to sharing radio spectrum. Radios operating on these bands are commonly deployed in links running several miles (despite the rules limiting output power) controlling important systems such as municipal water and sewage systems and oil pipelines. In these networks deployed in urban areas, I have seen cases where more than 50 Wi-Fi networks are visible. This does not count non-802.11 frequency hopper networks, which are visible on a spectrum analyzer but not in a Wi-Fi sniff. It is very possible when you are using a Wi-Fi hotspot at a coffee shop to watch a 1Mbps video, the coffee-shop’s access point and all the users are seemlessly sharing the same frequency with a network monitoring the city’s water system.
Wi-Fi’s most popular channel-sharing system is called enhanced distributed channel access (EDCA). Before each transmission, a Wi-Fi node listens to the channel for a period of time called an inter-frame space, typically 16us to 79us, depending on the type of transmission. If it does not receive acknowledgement from the receiving station, it listens again and re-tries the transmission after waiting a random amount of time. The amount of backoff time could be anywhere from 0us to 9.2ms, depending on the type of transmission.
This system does a remarkably good job of sharing the channel, even with non-Wi-Fi radio systems. I have done tests with frequency hopping radios transmitting continually, without checking to see if the channel is busy first, on random frequencies throughout the 2.4GHz band. A 20-MHz Wi-Fi channel takes up a third of the 2.4GHz band, and the EDCA system allows the Wi-Fi user to get 66% of the maximum throughput by transmitting only when the frequency-hopper is not using that particular 20MHz segment.
The EDCA is not mandated by government regulation. The frequency hopper transmitting without any regard to who might be using the channel is not breaking any laws. I have worked on products based on the Wi-Fi standard that intentionally listen for less time before transmitting, re-try more quickly when they do not receive an acknowledgement, and use a higher signal threshold to consider the channel busy. They want their system to work well, even if might possibly come at the expense to another system operating in the same area. The self-interested motivation in not being too aggressive about taking the channel is that it can increase the number of collisions on my product’s Wi-Fi-like network. There is no self-interested reason apart from this for exercising restraint in taking the channel.
Last year the FCC opened up new unlicensed bands in the “white space” frequencies of unused television channels. Technology companies with an interest in providing broadband Internet access called for this decision. It would be nice to create new rules that encourage efficient sharing of the spectrum in these new bands. A facile solution would be to set aside portions of the bands where governments would mandate compliance with the Wi-Fi EDCA standard. This is short-sighted, though, because EDCA is not the most efficient multiple-access scheme for all networks. Rather, we need to create rules that are liberal but that provide incentives for self-interested devices to play nice together. For example, networks might be allowed more liberal rules in some regard if they are more generous in listening for co-existing systems and avoiding interference with them. This is a complex problem because the system must work for a wide variety of applications and must not accidentally encourage greedy behavior that complies with the letter of the rule but not its spirit. Dr. K.J. Ray Liu suggests that classical game theory provides a framework to develop schemes to encourage selfless sharing of the band, even in the presence of malicious participants who want to monopolize the band or disrupt communications. Developing a good system now is important because the rules in place for devices on the new ISM bands will have a long legacy.