Dynamic Frequency SelectionBy CWNP On 01/21/2010 - 22 Comments
In the beginning stages of my Wi-Fi journey, 802.11h was confusing. Just like everyone else, I was drowning in the world of faceless acronyms. Over time, the meaning of Dynamic Frequency Selection (DFS) started to take on a concrete form and I assimilated the term into my armory. If you’re feeling that way, I wanted to share some thoughts on DFS that should help to secure it in the memory as an important topic.
As a reintroduction, Dynamic Frequency Selection was introduced along with Transmit Power Control by 802.11h back in 2003. DFS came into play as regulatory domains were freeing up more channels, namely the 5.25-5.35 GHz (UNII-2) and 5.47-5.725 (UNII-2e [extended]) GHz frequencies, for WLAN usage. European countries were the initial proponents for DFS requirements, but almost all regulatory domains joined the movement, including the FCC in the US. As with other unlicensed bands, there are other RF occupants in these bands. Military, weather, and aviation radars are all used in this spectrum, and as you might guess, Wi-Fi plays a second fiddle in comparison.
So, the compromise for sharing these frequencies with Wi-Fi devices was to mandate some type of radar detection and avoidance. Hence, we have DFS. To meet the requirements of regulatory domains, WLAN devices operating in these frequencies must perform periodic scans to detect pulses from radars. When a radar signature is detected, the WLAN must avoid interfering on this channel by switching the entire BSS to a new channel. After switching to a new channel, the WLAN must cease transmissions on the previous channel for at least 30 minutes. This requirement applies to client devices as well as APs. In fact, clients that are scanning for an AP in DFS frequencies must wait to hear beacons (passive) before they can transmit a probe request (active).
This may lead you to believe that DFS channels should be avoided altogether, but this is not the intended takeaway. DFS channels can still be used with a significant amount of confidence, but due to the potential impact of radar on these channels, they should be used with caution, and only after testing has been performed. How do you test? Here are a few ideas.
Since radar pulses are only visible at the RF layer, the initial instinct may be to test for radar with a spectrum analyzer. While this is a good second layer of protection and validation, I would encourage that you test for radar by performing pilot phases with the intended infrastructure equipment. If a radar pulse is present, APs operating on that channel will use Channel Switch Announcement (CSA) frames to switch channels. While RF folks probably read that advice and think it’s crazy, I do have some sensible reasoning.
Since there are several types of radar that could be detected, RF signature detection by WLAN infrastructure vendors is not foolproof. In an effort to assure compliance with regulatory domains, AP vendors must be extra cautious when testing for DFS. In other words, if the radar detection is imperfect, it will tend to err on the side of caution. So, it will detect false positives and possibly trigger DFS events when there is no radar. Each vendor develops its DFS recognition method independently, so one vendor may trigger a DFS event while another vendor keeps operating normally—unless there is an actual radar, in which case both vendors should trigger a DFS event. So, you may test with a spectrum analyzer and determine that no radar pulses are present, but then deploy your solution and find that the system is experiencing a large number of false positive DFS events. The implication of this poor design approach is that some amount of redesign is necessary, which may require new channel plans, adjustments in transmit power, and possibly moving, adding, or removing APs. This is a headache, extra cost, and downtime…all bad. By all means, use a spectrum analyzer to detect radars as a part of your design process. But, don’t trust a spectrum analyzer exclusively.
After a careful testing phase, deployment in the DFS channel range could be done with confidence. Of course, if you are deploying in an area near an airport, military facility, or other location where radars are somewhat likely, it would be advisable to simply avoid these channels. Unfortunately, DFS channels make up more than 60% of the available bandwidth of unlicensed 5 GHz for Wi-Fi. Without DFS channels, WLANs are left with only four 40 MHz channels. This may sound sufficient when compared with the three 20 MHz channels in 2.4 GHz, but it’s an unfortunate limitation when you consider the possibility of twenty-four 20 MHz channels or eleven 40 MHz channels.
GT Hill also has a nice video discussing DFS if you’d like a helpful visual treatment of the topic. http://www.youtube.com/gthillwifi