White Space Networking with Wi-Fi like Connectivity

Summary
This paper tackles the problem of harnessing the newly opened up UHF spectrum for data transfer. This presents a number of problems not present in the 2.4Gz band:

• The transmission range is bigger, so there is more spacial variation wrt a single AP
• Exiting users of the spectrum fragment the space making the problem of picking a channel for the AP more difficult.
  
• Because the FCC states that devices must give up a channel to incumbents (like wireless mics), and because these devices and come and go, systems must be able to quickly switch channels and do so automatically and without interfering with the incumbent device.

The paper presents two main ideas, WhiteFi, which is a WiFi like system that runs in UHF white spaces which uses SIFT, which is a fast way of locating WhiteFi signals in the entire UHF band.

The idea behind sift is to do analysis on the signal itself, and not to do a FFT first which is expensive. By looking for blips in the signal that are temporally spaced at the right time to look like WhiteFi packets and ack they can quickly detect where APs are likely to be.

WhiteFi uses SIFT for AP discovery but also establishes a secondary backup channel. If an incumbent is detected on the main channel a client or AP will immediately switch channels and 'chirp' on the backup channel to inform peers of the change. If the chirp is lost SIFT can be used to re-acquire the AP.

Comments
This paper was certainly interesting, and the prospect of using UHF white spaces for networking is quite attractive. A few issues I had:

• The authors discover that even a single packet can cause interference for wireless mics. If nodes are streaming data there is no way the detection of an incumbent plus the chirping to switch time is going to prevent a few packets still being transmitted on the old channel. This seems to violate the FCC requirement of no interference.
• The analysis only tested UDP flows. As we've seen before, TCP is much more sensitive to unusual network environments and it seems important to evaluate it's performance in WhiteFi.
• Given the long range in the UHF band it would have been nice to see numbers about how much more area can be covered with how many fewer APs compared to traditional WiFi as a motivator for this work.

This was still a cool paper, and gets into some of the signal processing problems we haven't talked about yet, so I would keep it on the reading list.