 |
 |
 |
 |
Channels in 802.11
Overlapping Channels in the 2.4 GHz band
One of the findings of my honours research topic was that two wireless nodes can send and receive messages even if they are not on the same channel. The reason for this is because 802.11 uses 20 MHz channels for transmission. However, the channels in the 2.4 GHz band are only separated by 5 MHz. This means that some channels heavily overlap. Prior to my honours research, I had never considered that messages could actually be received and interpreted on overlapping channels. Most people who have done courses on wireless networks have seen diagrams similar to the one shown in Fig 1, which shows the extent of this overlap in the 2.4 GHz spectrum. Fig 2, which was captured using a Wi-Spy 2.4i, shows the channel usage of a contunuously transmitting 802.11b device.
Fig 1: Overlapping channels in the 2.4 GHz band
Fig 2: DSSS Spectrum use in the 2.4 GHz band
My claim is that the transmissions on overlapping channels don't just interfere, but can also be interpreted by a receiver on a different frequency. The video below shows my laptop which is connected to a wireless AP one floor above on channel 6. Note that the initial signal strength in the video is imperfect. In the video I disconnect from the AP on channel 6 (2.437 GHz) and switch the wireless mode to monitor mode. With the radio tuned to the same frequency, I collect some beacons from the AP with Wireshark. I then switch the radio onto channel 5 (2.432 GHz). When I restart the packet capture note that I can still collect beacons from the original AP. Although fewer than the normal 10 beacons per second are captured, this shows that frames can be interpreted on overlapping channels. Despite being 1 floor beneath the AP, a substantial number of beacons are still captured. Our exerimentation shows that being close to the AP will enable greater numbers of beacons to be captured on overlapping channels. Once you have seen the demo, have a go yourself. Be mindful that you will need a wireless card capable of entering monitor mode.
This artifact has relatively little impact on the majority of 802.11 functions, however, it does affect scanning. The affects are covered in further detail in my honours thesis and papers. I show how it can significantly lengthen the amount of time spent scanning. I am unaware of other impacts but I would be interested if others have found/thought of any.
Inter-Radio Interference
The majority of my current research involves multi radio mesh networks. I have noticed that at very close range, even non-overlapping channels interfere. Numerous other research efforts [3, 4, 1, 2] encountered similar problems from two closely located radios. This effect was highlighted in an experiment by Raniwala et al [3] where, despite using completely orthogonal channels (channels 1 and 11 were used), the results show substantial drops in performance when both channels were sending and/or receiving simultaneously. The performance loss was so severe that the benefits of a multi radio system are completely negated. The reason for this loss of performance is the physical proximity of the radios combined with the closeness of the operating frequencies. Fuxjager et al [2] refers to this as the near-far problem.
One solution is to mount external antennas. A problem with this solution is that long pigtails might reduce radio sensitivity and external antennas may raise costs of multi radio systems. Another solution to this problem in multi radio ad hoc networks is to utilize frequency effectively by separating co-located radios by as much spectrum as possible. By using one radio in each of the ISM, UNI-1 and UNI-2 bands, inter-radio interference can be eliminated. Large channel separation enables nearby radios to transmit or receive concurrently. To utilise more radios within a node additional frequency such as the ~3600 MHz freqencies ratified with 802.11y must be used.
To further illustrate this point, fig 3 and 4 shows the spectral use of an an 802.11b (DSSS) and 802.11g(OFDM) device. The spectrum analyser, a Wi-Spy captured this channel usage at the distance of 5 metres. Fig 5 and 6 show the spectral usage of the same transmissions at 10cm and Fig 7 and 8 show the same transmissions when the Wi-Spy device is located ontop of the transmitter. What we are attempting to show here is that differenent distances will affect the spectral usage of a 802.11 device. When two WiFi devices are locate within the same device, they will be unable to simultaeously send or recaive if they are both trying to utilse the 2.4 GHz band. The power emitted when devices are located so closely is enough to trigger the carrier sensing mechanism on a nearby radio. Another interesting artifact of these results is that it appears that 802.11a/g, which use OFDM, have more defined channel use.
|
|
 |
 |
 |
 |
 |
References
[1] Richard Draves, Jitendra Padhye, and Brian Zill. Routing in multi-radio, multi-hop wireless mesh networks. In MobiCom '04: Proceedings of the 10th 1 annual international conference on Mobile computing and networking 114-128, New York, NY, USA, 2004. ACM Press. , pages
[2] Paul Fuxjager, Danilo Valerio, and Fabio Ricciato. The myth of non- overlapping channels: Interference measurements in ieee 802.11. In WONS: Wireless on Demand Network Systems and Services, 2007.
[3] Ashish Raniwala, Kartik Gopalan, and Tzicker Chiueh. Centralized channel assignment and routing algorithms for multi-channel wireless mesh networks. SIGMOBILE Mob. Comput. Commun. Rev., 8(2):50-65, 2004.
[4] Ashish Raniwala, Rupa Krishnan, and Tzicker Chiueh. Wireless Mesh Networking: Architectures, Protocols and Standards, chapter IEEE 802.11- Based Wireless Mesh Networks, pages 79-109. Auerbach, 2007.
Wi-Spy in Linux
In case you were wondering, getting wi-spy to work in linux is really easy.
sudo apt-get install libusb-dev libncurses5-dev libgtk2.0-dev
Download Spectrum tools from http://www.kismetwireless.net/spectools/
unzip and untar
./configure
make
sudo make install
sudo ./spectool_gtk
