I concerned about performance. I've seen some of the current products perform, and they usually just die (no traffic for several seconds/minutes), when subjected to heavy use. Most of the Access Point products use half-duplex, the cause of the problem.
Some of the newer Access Point products talk about supporting Time Division Duplexing (TDD),
and/or TurboCell technology from KarlNet.com, to overcome the problems of half-duplex. Which
technology is better?
Time Division Duplexing (TDD) is a technique used to transmit and receive on a single frequency by using different time intervals. 802.11b and 802.11a WLAN standards both use TDD. Compare this to Frequency Division Duplexing (FDD), where different frequencies are used to carry upstream and downstream traffic simultaneously. FDD is a traditional duplexing scheme used by protocols like GSM and CDPD operating in licensed bands. Both TDD and FDD are supported by the 802.16 wireless MAN PHY standard.
As you point out, TDD means that 802.11 devices are effectively half-duplex: when a WLAN station is receiving, it cannot transmit, and vice versa. However, duplexing is only part of the story.
To avoid simultaneous transmission (collision) with other stations on the same channel, every 802.11 station implements a CSMA/CA (Carrier Sense Multiple Access/Collision Avoidance) media access protocol. 802.11 stations transmit only when the channel is clear. Since stations cannot detect collisions, access points must return ACKs to acknowledge successful receipt of data. Because no other data can be sent while the ACK is using the channel, effective throughput is lower than the channel's actual bandwidth. Between interframe spacing, preambles, management frames, and ACKs, 802.11b overhead is at least 40%.
In a long range point-to-multipoint WLAN, you may be experiencing additional overhead caused by the "hidden node" problem. 802.11 stations that can "see" each other can sense a busy channel and use a backoff algorithm to avoid collision. However, stations that "see" the access point but not each other have an increased probability of simultaneous transmission. To combat this, whenever a station has data to send, it can first issue a Request To Send (RTS) control frame. Upon RTS receipt, the access point allocates time to the station and responds with a Clear To Send (CTS) control frame. Other stations may not "see" the RTS, but they all "see" the CTS and avoid transmitting during the allocated time slot.
RTS/CTS is not required before every 802.11 data frame. Depending upon the WLAN product, RTS/CTS may be disabled by default, with an option to enable if the collision rate is high. Even when RTS/CTS is enabled, short packets might be sent without RTS/CTS because the probability and adverse impact of collision is smaller. However, in an outdoor point-to-multipoint network where stations are far apart, the "hidden node" problem occurs more often and the delay and overhead due to RTS/CTS (including RTS collisions) can be more significant. You may be able to obtain modest performance improvement by adjusting RTS/CTS parameters.
Vendors that sell outdoor wireless LAN products have developed innovative approaches to improve
performance in this environment. For example, Wave Wireless SPEEDLAN access points use a
proprietary CampusPRC protocol which polls stations for data, trading the overhead of polling for
the (presumably greater) overhead of hidden node collisions and RTS/CTS. KarlNet's ISP software
implements a proprietary TurboCell polling protocol that can be used with certain Agere ORiNOCO and
Apple Airport products. These approaches do not make the WLAN full duplex - rather, they try to use
the available spectrum more efficiently.
This was first published in July 2002