The number of wireless-LAN users is growing exponentially. But with the majority of this equipment operating in the 2.4-GHz band with only three clear channels, as per 802.11b (11 Mbits/second) and 802.11g (54 Mbits/s), it will not be long before we face a wireless-bandwidth crunch. And things will get worse if the unlicensed 2.4-GHz band itself gets crowded and becomes susceptible to interference from other technologies like Bluetooth, cordless phones and microwave ovens. Finally, like all Internet access media, WLANs are facing increasing bandwidth demands by rich-media applications that require large and deterministic data pipes between clients and the Internet.
The solution to these challenges comes in the form of another WLAN variant called 802.11a, which, like 802.11g, provides a raw data rate of 54 Mbits/s. But unlike the 802.11g standard, 802.11a operates in the 5-GHz band where eight discrete, nonoverlapping channels are available and there is little interference from other technologies.
But 802.11a is not backward compatible with its more prevalently installed 2.4-GHz cousins, and Wireless Fidelity (Wi-Fi) hotspot merchants are loath to invest in access-point equipment that only supports a relatively uncommon standard. But given the skyrocketing need for wireless data bandwidth, adoption of 802.11a is only a matter of time and timing. So how do we get from today to that inevitable point?
The obvious as well as popular solution for accelerating adoption of 802.11a is the development of dual-band, multistandard access-point equipment. This enables compatibility with both 2.45- and 5-GHz WLAN carrier frequencies and all three 802.11 standards, theoretically making everyone happy and enabling rapid growth of 802.11a, the standard variant that solves the bandwidth crunch.
The only problem is that most common solutions offered for dual-band equipment inelegantly cobble together radio-card hardware that supports each band, respectively. This side-by-side approach results in a bill of materials (BOM) that costs about as much as a 2.4-GHz system and a 5-GHz system together. Furthermore, equipment built around this kind of dual-band implementation can indeed support either 802.11a or 802.11b/g connections, but not both concurrently. To be truly effective, dual-band infrastructure or access points must support connections with clients in both bands at the same time.
A technique called band interleaving works much better because the access point transparently switches between the bands to provide service to all three prevailing (and growing) 802.11 user communities concurrently. The ultimate benefit of this approach is financial, since the band-interleaving technique enables an access point that supports both 2.4 and 5 GHz with a BOM equivalent to a single-band 2.4- or 5-GHz product alone. This means that dual-band networks can be deployed for pretty much the same cost as a single-band 2.4-GHz network. Solving both the BOM redundancies and concurrent band-connectivity issues, however, means tackling some nontrivial engineering challenges, both in the baseband and RF domains.
Band interleaving rapidly switches between the 2.4- and 5-GHz 802.11 bands, allowing an access point to share the available bandwidth between 802.11a and 802.11b/g devices, using a single radio. Switching between the bands occurs on the order of a few microseconds, a rate that is invisible to Wi-Fi clients, and to higher-layer protocols such as TCP.
To make the switching transparent, the access point cannot send acknowledgements for frames sent to it on the one band while it is servicing the other band. Thus, a key requirement for interleaving is that the clients connected on one band must be instructed not to send frames during the time when the access point is "away" servicing the other band.
The basic method for achieving this uses a randomized-contention mechanism that allows statistically fair sharing without a central controller. When a device wishes to transmit, it first waits until the medium is clear, and then waits a further random back-off period. At the end of the back-off period, if no other device has started to transmit first, the client station (e.g., a notebook PC with a wireless network interface card) will start its transmission.
This mechanism is acceptable for general data transfer, but there are circumstances where it is desirable to provide guaranteed access to the wireless medium for one or more stations. The 802.11 protocol specification includes a mechanism by which this can be achieved called the Network Allocation Vector. When a client has a nonzero value in its NAV counter, it will not transmit any frames. Instead, it will queue frames for transmission, while regularly counting down at a specified rate until the NAV counter again reaches zero. At this point the station will attempt to transmit the queued frames using the normal access mechanisms. The maximum delay value is equivalent to a delay of about 32 milliseconds, allowing the NAV scheme to easily implement band interleaving.
Overcoming RF interference
Whenever three or more wireless stations attempt to transmit data simultaneously, RF interference can become a problem. The Clear to Send (CTS), now incorporated within the 802.11 standards as part of the RTS/CTS (Request to Send/Clear to Send) exchange, enables a device to reserve a channel for long enough to send a frame, thus overcoming interference.
Station A, for example, may wish to transmit a frame to station B. But station C will not be able to detect that station A is transmitting, and so may start transmitting at the same time. The interference between the two transmissions may mean that station B is unable to correctly receive either of the two transmissions.
This can be overcome by station A transmitting a special short frame to station B called an RTS. Station B immediately replies with a CTS. Both the RTS and CTS frames include a duration value that will cause the NAV counters of receiving stations (in this case C) to be set for long enough for station A to send a frame to station B.
See related chart |
Band interleaving uses the NAV mechanism to schedule when a device is permitted to transmit, rapidly switching between bands before further packets or acknowledgments are sent.
If a receiving station receives either the RTS or CTS, then it will set its NAV counter appropriately, ensuring that all possible interfering stations will defer transmission for the requested period.
In the case of band interleaving, only the CTS part of this exchange is used. There is no need to use a complete RTS/CTS exchange, as by definition all stations associated with an access point must be in range to receive its transmissions. The stations on the channel will receive the CTS and stations on other channels (or bands) will not stop transmitting.
In determining when to send the CTS, the band-interleaving access point will leave transmission until as late as possible, giving itself time to switch over to the other band before the stations on the first band restart transmission. The speed with which the hardware can switch bands is a key factor in determining performance of a band-interleaving access point in this mode.
Another important aspect of this approach is that the CTS frame dynamically sets the real value of the NAV counter. This means that actual periods for which transmission in any channel is suspended can be dynamically allocated. This allows the band-interleaving access point to adapt to varying throughput demands on the two channels and to ensure that throughput is never compromised.
Ground-up RF design
In designing a dual-band system that shares as many BOM components as possible between 2.4 GHz and 5 GHz, the best place to start is in the baseband. By using a "zero-IF" approach, the main point of departure between the bands is in the final upconverting mixer, so the design challenge is in executing clever frequency selection and frequency planning that allows sharing of the phase-locked loop frequency synthesizer. Having achieved this, the sharing of everything except the upconversion mixer allows the implementation of a concurrent, dual band which in practice adds circuitry amounting to less than 1 square mm of silicon area.
In this way, a shared radio that implements concurrent dual-band support through band interleaving allows the introduction of an 802.11a access point that fully supports legacy 802.11b and 802.11g client stations, but costs the same as a single-band-only access point.
John Prince is product-marketing manager at Synad Technologies (Reading, England).
See related chart