Global Sources
EE Times-Asia
Stay in touch with EE Times Asia
?
EE Times-Asia > Networks
?
?
Networks??

WLANs are jump-starting cognitive radio

Posted: 16 Nov 2004 ?? ?Print Version ?Bookmark and Share

Keywords:cr? wlan? fcc? radio? rf?

Cognitive radio might seem to be a radical concept, but WLAN designers have already implemented many CR techniques, some of which are quite sophisticated. Because these techniques can be expanded in an evolutionary way to achieve the full promise of CR, they point the way toward the future of wireless communications.

The basic premise of CR is that radios can better use the available spectrum by detecting their environment and adapting accordingly. Regulatory agencies such as the FCC require that 802.11a radios detect radar signals and avoid interfering with them. This ability to dodge radar requires a significant amount of CR-type adaptability and it is just the beginning of WLAN CR capabilities.

WLAN radios may detect a wide variety of radio environment characteristics. These include traffic statistics and other RF events that are identifiable, such as radar, Bluetooth, microprocessor noise or microwave oven noise. They also include WLAN side effects such as collisions, failed packets, adjacent channel interference and hidden stations and unidentifiable noise sources.

By recording RF events, identifying them when possible and responding appropriately, the WLAN radio improves its ability to optimize throughput. Given the amount of interference that can exist in the unlicensed WLAN bands, the radio's CR capabilities are crucial for achieving the robust performance that users expect.

For the typical CR-related functions in a WLAN access point or client station, the entire hardware/software system can be thought of as "the radio" for CR purposes.

The hardware stores and analyzes RF events and passes significant events to software for further analysis. In this process, RF "events" usually begin with a fairly sudden increase in received signal amplitude, although the device also constantly cross-correlates low-level signals for any indication of a faint 802.11 preamble pattern.

Two sections of logic can be used to analyze RF events: an 802.11 packet-reception block and a block that detects non-802.11 signal properties. These blocks analyze events in parallel for maximum speed.

The non-802.11-detection block captures several characteristics for each RF event: signal level, length, time of occurrence and spectral content. The last named information is available because the detection logic has access to the FFT block used to decode 802.11 signals.

The events and their characteristics are passed to software unless they are so weak that they are unlikely to represent meaningful signals or interference. Programmable-threshold hardware screens out these less significant events to keep from overwhelming software. This screening is important because WLAN radios are sensitive enough to pick up an enormous number of RF events. The 802.11a/g standards require a sensitivity of -82dBm and most radios can achieve better than -90dBm. This capability must be managed carefully to avoid creating a large number of false alarms for the adaptive software.

Non-WLAN events

The WLAN radio's most powerful CR aspect is the ability to analyze non-802.11 events, determine their source and adapt accordingly. The radio's embedded software works with a fairly large database of events and searches for known signal signatures, WLAN "side effects" and any periodicity in unidentifiable noise.

Since WLANs operating in the 5GHz band must leave a channel on which a radar can be detected, the radio examines the events to see if they might be from radar pulses. Not all radars use periodic pulses, so this analysis must be fairly sophisticated. The radio even surveys channels that are not in use so it will know whether they are available should the need arise.

Identifying microwave oven noise is usually straightforward because most ovens use a half-wave rectifier that operates the magnetron during half of the 60Hz power cycle. The WLAN radio can then ensure that packets can be sent successfully in the other half of the power cycle. This adaptation requires the use of a technique known as fragmentation to shorten packets sufficiently. The radio may use this technique any time a periodic noise source is detected, so long as that source is not radar.

Varying pattern

Processor noise from the host computer should be simple to identify, but the clock for a fast processor, such as a 2GHz Pentium, is not periodic in a simple way. To comply with FCC EMI limits, the clock frequency is usually dithered enough to avoid putting all its energy in a narrow bandwidth. The WLAN radio recognizes this varying pattern and can usually minimize the effect of the interference by switching to a different antenna and adjusting signal-detection parameters.

Systems such as laptops and WLAN CardBus typically have two antennas to provide diversity; generally, one or the other of the antennas receives less host processor noise than the other. Radio spurs can also arise from the WLAN radio itself and the radio may adapt its antenna, signal-detection and signal-decoding parameters accordingly.

If software analysis detects RF events that occur at random frequencies with narrow bandwidth that have an unvarying length, Bluetooth is probably the source. Bluetooth uses a strict time-division multiplexing scheme that creates rigid time slots for packets and a frequency-hopping method that transmits the packets with a random frequency pattern.

The importance of identifying Bluetooth is to avoid trying to adapt to the interference too much. Changing channels is useless because of Bluetooth's frequency hopping, and Bluetooth signals interfere only a percentage of the time anyway. The WLAN can usually keep operating normally and simply retransmit failed packets. System designers may limit interference within the same system by taking advantage of a hardwired link that is often available between WLAN and Bluetooth chips. When one chip needs to receive, it sends a "wait" signal to the other chip to prevent it from transmitting and creating interference.

Adaptive WLANs

Even if the WLAN radio cannot identify interference sources, WLAN systems can adjust many factors to avoid the interference:

??? Change channels;

??? Change the bandwidth being used from 40MHz to 20MHz, 10MHz or even 5MHz;

??? Change data rates from 54Mbps to 1Mbps in 12 steps;

??? Use request to send/clear to send (RTS/CTS) and/or adjust clear channel assessment thresholds to minimize contentions with other WLAN traffic; and

??? Use fragmenting to minimize packet length.

Cutting back bandwidth and data rates toward their minimums is often useful to get some amount of data through noisy environments, with the ability to decrease data rates all the way to 250Kbps when necessary having already been demonstrated. More interesting is the ability to increase bandwidth to 40MHz by combining two channels. With this technique, the radios can push data rates as high as 108Mbps. With access-point chipsets supporting concurrent operation in the 2.4GHz and 5GHz bands, using a total bandwidth up to 80MHz achieves raw aggregate throughputs of 216Mbps.

The key to taking advantage of this capability is to recognize when it will not generate or receive interference to or from other systems. The radio listens across the wide bandwidth both before and during operation in this mode. The radio then falls back to narrower bandwidths whenever it becomes necessary.

- Bill McFarland

Director of Algorithms and Architecture

Atheros Communications Inc.




Article Comments - WLANs are jump-starting cognitive ra...
Comments:??
*? You can enter [0] more charecters.
*Verify code:
?
?
Webinars

Seminars

Visit Asia Webinars to learn about the latest in technology and get practical design tips.

?
?
Back to Top