“White Space Networking with Wi-Fi like Connectivity”

White Space Networking with Wi-Fi like Connectivity” discusses how to achieve wireless networking using so-called “white space”: unused portions of the UHF spectrum (approximately 512-698 Mhz). This is attractive, because using the UHF spectrum can allow wireless networking over a relatively long range (1 mile or greater, compared to 300-600 feet for 802.11n). UHF signals are also better able to penetrate walls and buildings in urban environments.

The difficulty with using UHF spectrum is that the spectrum has already been assigned for use by two “incumbents”: analog TV and wireless microphones (although the use of the spectrum by analog TV should be reduced since this paper was written, due to the recent conversion of analog TV to digital). Per FCC rules, non-incumbent use of the UHF spectrum must avoid interfering with incumbent use. While TV signals are relatively stable, wireless microphone transmissions can begin without warning, which poses some challenges to providing wireless networking using the UHF spectrum. To gain improved throughput, the authors also suggest that multiple contiguous free channels should be aggregated together and used for networking.

To enable networking over UHF, several problems must be solved:

  • Spectrum Assignment: What portion of the 180 Mhz UHF band should be used for networking in a given locale? The UHF spectrum is typically “fragmented”: some portions of the spectrum are in use, while others are free. Furthermore, to increase throughput, multiple contiguous free channels should be used by a single network. The right spectrum to use will also change over time, e.g. as hosts move and wireless microphones are activated. The channel used must be free for both the AP and all clients, which is made challenging by the larger spacial extent of UHF networks.
  • AP Discovery: WiFi APs emit beacons every 100ms on the WiFi channel they are using. To find APs, clients can easily scan each channel, looking for AP beacons. This is more challenging for UHF, because there are more possibilities to scan (30 UHF channels, and 3 possible channel widths, yielding 84 combinations).
  • Handling Disconnections: If incumbent use of a channel is detected, networking transmissions must immediately stop using the channel (the authors found that even transmitting control packets can result in audible interference for wireless microphones using the channel). If clients and APs are prone to abruptly disconnect from a channel, a method is needed to choose another channel and continue communication.

Design

To solve the spectrum assignment problem, clients and APs exchange information about which channels are free near them. APs probe for new channels when they detect incumbent use on their current channel, or if they observe a performance drop. Channel probing is done by considering which channels are free for all clients and the AP, and by then estimating the available bandwidth on each channel.

To enable AP discovery, the authors propose some signal processing magic. They propose a technique called “SIFT” which essentially allows them to scan the spectrum range and cheaply determine the presence of an AP on a channel; they can then tune the radio transceiver to the identified channel, and decode the beacon packet as usual.

To handle disconnections due to incumbent use of the spectrum, they propose using a separate 5Mhz backup channel, which is advertised as part of AP beacon packets. If a client senses that a disconnection has occurred (e.g. because no data packets have been received recently), it switches to the backup channel, and both listens for and emits chirps. Chirps contain information about the available white spaces near the chirping node, which is used to pick a new channel for communication. APs periodically scan for chirps.

It’s possible that the backup channel is already in use by another incumbent. In this case, the client picks an arbitrary available channel as the secondary backup channel, and emits chirps on it. The AP periodically scans all channels to attempt to find such chirps. The same signal processing magic (SIFT) that is used to enable cheap AP discovery is used to make scanning all channels periodically feasible.

Related Reading

Matt Welsh discusses this paper on his blog. The slides for the SIGCOMM talk on this paper are also available online.

Advertisements

Leave a comment

Filed under Paper Summaries

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s