With the recent series of announcements by enterprise WLAN vendors about including spectrum analysis into their products, let’s dig a little deeper into spectrum analysis and see what matters and how things add up.
First off, the basics. Spectrum analyzers that are built into Wi-Fi networks are designed to detect the presence of—and identify—sources of non-Wi-Fi interference in the Wi-Fi unlicensed spectrum. Given the way that Wi-Fi works and how it uses unlicensed spectrum, there are many non-Wi-Fi devices that generate emissions that can severely cripple network throughput.
Some of these interfering devices are well known and have obvious effects, such as microwave ovens (which, after all, do own the 2.4GHz band, with Wi-Fi simply sharing it when it can). Others, however, are far more subtle, such as certain manufacturing or medical equipment, which are also allowed to send off radiation into the unlicensed bands, but usually aren’t expected to by their operators. These interfering devices do not know about Wi-Fi, and do not send signals that look anything like Wi-Fi. Detecting these devices requires using a few special techniques (usually on general-purpose spectrum analysis hardware) to narrow down the interferer’s signal and identify it.
How important is it to do a good job detecting interference? Vendors often use examples of interferers, such as non-Wi-Fi 5GHz surveillance cameras, that completely freeze the channel—resulting in zero Wi-Fi throughput. Detecting interference for this case is somewhat interesting, but is not really why spectrum analysis is important, as a complete loss of throughput isn’t likely to be missed by anyone. Where spectrum analysis really matters is when interference comes and goes, introducing network instability and threatening service quality but not rising to the level of a full stoppage. These sorts of variations are disastrous for mission-critical networks—imagine a nurse’s Wi-Fi phone cutting in and out with the interference changes—but leave the IT staff scratching their heads, being unable to figure out why the service levels keep changing. Good spectrum analysis can detect the interference that is causing the service reduction and identify it where other means can’t.
This leads to one key principle in spectrum analysis: the more the stability of the network and its service levels matter, the more spectrum analysis matters. Networks that are lightly used don’t need to worry about the impact interference causes them, because even a significant reduction in potential service isn’t all that important. But networks that require high-performance service need the sort of service assurance that comes with proper spectrum analysis.
Now, let’s compare the different approaches that are showing up in WLANs.
Existing Wi-Fi chipsets
Spectrum analysis abilities within currently available Wi-Fi chipsets are extremely coarse. At minimum, they can detect the noise floor—a slow-moving average of all signals coming in. Depending on the model of chipset, it may also be possible to divide out these noise values into buckets over the 40MHz 802.11n channel, giving somewhat better sensitivity to frequency over the basic noise measurement.
There are three problems with using this for spectrum analysis. First, the methods the chipsets use are too coarse; they can miss damaging interference and lull users into a false sense of security. Furthermore, the sheer amount of spectrum information that needs to be mined to detect and identify interferers is immense, and these chips do not provide the necessary additional processing power to offload the added burden, meaning that core access point service performance will be impacted. (Some future 802.11n chipsets with 450Mbps throughput and three spatial streams are expected to provide far better granularity, but those are not commercially available today.)
Second, the chipsets are limited to the channel the access point is on.
Third, the methods are not useful whenever the Wi-Fi radio is sending or receiving anything: the interferer’s signature tends to be lost as it proceeds through the 802.11n reception process.
The irony is that this approach requires the radio to stop serving mission-critical applications—by going silent—to get the analysis necessary to protect mission-critical applications from a stoppage of service.
Pros: no new hardware
Cons: too coarse for mission-critical use; spectrum analysis steals cycles from traffic service; must stop service to be meaningful
Grade: D
Inserting extra hardware into Wi-Fi chipsets
One way to overcome the problem of potentially-crippling coarseness is to embed a fine-grained analyzer in parallel to the 802.11n receiver in a Wi-Fi radio. Doing this lets the radio detect more interferers and more often, and can introduce enough resolution to allow the network to try to locate the interferer.
So far, so good. However, as convenient as it seems it is to place a fine-grained spectrum analyzer into the Wi-Fi radio itself, because the analyzer still depends on the Wi-Fi radio to tune to a channel, the analysis is limited to only the serving channel. If this approach were used with a single-channel network, that would be fine; however, the vendors offering this approach are microcell-based, and so the inability to detect interference on neighboring access points’ channels without switching from channel-limited to a wider band mode that cannot provide service means that the network must be disrupted for the analyzer to get enough of a view to do its job adequately.
Furthermore, if the Wi-Fi radio is also used for service, the analyzer is blinded by the intense Wi-Fi signal generated from within every time the access point transmits. That’s not good, as mission-critical networks that are most likely to be hurt by interference are transmitting a lot as it is, and when interference starts causing problems, it causes retransmissions that only increases the amount of time the analyzer gets blinded. In other words, the interference itself can lead to the analyzer missing the interference.
Pros: fine-grained; local hardware
Cons: blinded by network use; cannot detect off-channel without disruption
Grade: B-
Using dedicated spectrum monitors
Dedicated spectrum monitors look like access points, but have one Wi-Fi radio replaced with a fully-dedicated spectrum analyzer. This type of analyzer is fine-grained, with dedicated local hardware for detecting and identifying the interferer. However, in comparison with the Wi-Fi/analyzer combination mentioned previously, this analyzer has a permanent wide bandwidth that it analyzes over, thus avoiding the limitation to any given 802.11n channel.
These monitors do also have a Wi-Fi radio. However, not only does that radio not share the same RF front-end as the analyzer, but the radio does not transmit for service and is receive-only. The advantage is the elimination of the blinding of the other solutions.
With virtualized WLANs, the sensor approach is optimal: an addition of one sensor in up to six access points allow for strong coverage while still requiring up to 30% less access points than equivalent microcell networks. And the monitors can be used for rogue mitigation and wireless intrusion prevention as well.
Pros: fine-grained; local processing for scalability; no disruption in service
Cons: comparatively minimal
Grade: A
