Saturday, April 21, 2012

Is 2.4GHz Wi-Fi the next GPS/LightSquared?

No, unlicensed devices in the 2.4 GHz band (2400 – 2483.5 MHz, operation under Part 15.247) probably won’t be the next GPS/LightSquared, where a large installed base of unlicensed devices with significant susceptibility to out-of-band interference was threatened by the deployment of a cellular service in an adjacent band. However, some similar characteristics raise concerns: tens of millions of devices, poor adjacent channel rejection, and a quiet band next door. What would happen if there were a large cellular deployment next door to 2.4 GHz?

There are significant differences to temper concerns: Wi-Fi devices don’t depend on such exquisitely low signal levels as GPS receivers; we’re not talking about safety of life applications; the RF front-ends of Wi-Fi devices are not open many tens of MHz away from the allocated band; and there already is some cellular operation nearby, at least in the US (Clearwire/Sprint’s 4G service in the 2.5 GHz band).

Still, as I’ll argue, the fact that interference has been observed between 4G service in 2.5 GHz and unlicensed devices in 2.4 GHz even with at least 10 MHz of guard band between them suggests that we’ll see interference problems to and/or from unlicensed devices if a cellular service were allocated in the fallow 2360 – 2400 band. That in turn suggests that it could make sense for the FCC to start encouraging or mandating better filtering for unlicensed devices over the 2.4 GHz band now, well before the 2.3 GHz band starts being populated with a potentially interfering service.

Update 4/23/2012: Monisha Ghosh kindly let me know that  2360-2400 MHz has been requested by healthcare device manufacturers (GE, Philips etc.) for Medical Body Area Networks on a secondary basis to Aeronautical Telemetry (OET proceeding 08-59). The June 2009 Notice of Proposed Rulemaking (PDF) provides good background information on the current uses of the 2360-2400 band.

Susceptibility to adjacent channel interference

The receiver performance specifications for generic 802.11 devices are not very stringent. The most demanding adjacent channel rejection requirements is 16 dB attenuation (for the most robust, low throughput modulation, BPSK with 1/2 coding), and only -1 dB for the high throughput modulation 64-QAM ¾ (IEEE Std. 802.11-2007, Section 17.3.10.1, Table 17-13)

The transmit mask for 802.11 is not very stringent, and it’s probably a good assumption (made e.g. by Villegas et al., who examine adjacent-channel interference effects in 802.11 WLANs in http://dx.doi.org/10.1109/CROWNCOM.2007.4549783) that devices use an identical filter for  transmission and reception. There is therefore both significant leakage from Wi-Fi devices operating in channels 1 and 11 (in the US; in some other countries, channel 12 and 13 are allowed) outside their allowed band, and susceptibility to signals in the adjacent bands.

Here’s the mask in the IEEE spec:

801.11 transmit mask (IEEE Std. 802.11-2007 Fig I.1, click for full size image)

and here’s what it looks like in the band context:

2.4 GHz Wi-Fi channels (wikipedia, click for full size image)

Interference with neighbors in 2.5 GHz

The FCC Spectrum Dashboard indicates quite a few services above the 2.4 GHz ISM band. In addition to one satellite earth station licensee with an estimated 1,620,000 satellite earth stations in 2483.5 – 2495, Globalstar operating a LEO constellation for satellite phone and low-speed data communications in 2495 – 2500, some Broadcast Auxiliary Service (Part 74) used to relay TV broadcast programming material in 2483.5 – 2500 (limited to grandfathered stations that were authorized prior to July 25, 1985; 502 licensees), Federal mobile-satellite service limited to earth stations operating with non-Federal space stations, and Federal radiolocation service, the big player is Broadband Radio Service and Educational Broadband Service licenses (Part 27) being transitioned to Clearwire in 2495 – 2655.

According to wikipedia, Sprint resells Clearwire's 4G network service as Sprint 4G in over 71 markets across the United States (Clearwire coverage, Sprint coverage).

I found a couple of reports on the web of interference between 2.5 GHz WiMAX service and Wi-Fi. It’s worth noting that Clearwire’s licenses only start at 2495 MHz, so that there is already a guard band of at least 11.5 MHz between Wi-Fi and WiMAX.

Conversations on EVDOforums.com in 2008 and 4Gforums.com in 2010 suggest that the Wi-Fi signal broadcast by a router will interfere with a WiMAX signal if they are too close to each other. Apparently the Cradlepoint router’s documentation warns users to move their Wi-Fi to channel 1 or 2, i.e. the far side of the ISM band, about 60 MHz away from the WiMAX service. One commenter (apparently a 3GStore employee) noted that several customers continued to experience significant interference even with Wi-Fi set to Channel 1 or 2. The only solution was to disable Wi-Fi all together or use a shielded 10ft USB extension cable.

There was also a report in 2008 that 802.11 radios in conjunction with the Bluetooth standard will interfere with nearby WiMAX cellular networks. However, supposedly the side effects can be prevented if they are located approximately 8 meters apart for 2.6 GHz (WiMAX?), and 16m apart at 2.3GHz (WCS?).

Still, it looks like coexistence between WiMAX and Wi-Fi/Bluetooth is well understood. A 2007 article in EETimes discusses the blocking of a Bluetooth link (e.g. to a headset) when a handset’s WiMAX transmitter is operating; the authors propose solving the problem at the MAC layer by synchronizing the protocol time bases so that Bluetooth and WiMAX frames can be interleaved. Sync obviously won’t work with Wi-Fi since it’s an asynchronous protocol. In this case, the authors propose forcing the Wi-Fi node into sleep node while the WiMAX link is in use; the access point buffers transmissions that are released when the handset exits sleep mode.

All this raises some obvious questions: What’s the interference mode? Is it leakage into a service from the adjacent band (i.e. poor transmitter filtering in the adjacent), or overload due to strong signals in the adjacent (i.e. poor receiver filtering)? Does Wi-Fi suffer from WiMAX interference the same way Bluetooth does? How good is the out-of-band filtering of actual Wi-Fi and Bluetooth devices?

There's also an intriguing regulatory question: would Wi-Fi operation (say) be held liable for causing harmful interference per Part 15.5 (b) ("Operation of an intentional, unintentional, or incidental radiator is subject to the conditions that no harmful interference is caused") to a service in an adjacent band even if there is no spill-over from 2.4 GHz into the adjacent band - that is, the "victim" receiver suffers interference because it isn't doing a good enough job filtering out signals transmitted in the unlicensed band?

The moral seems to be that one can expect engineering solutions to coexistence problems when potentially interfering services in adjacent bands are all handled in the same device. This is the case with Wi-Fi/4G, but was not the case with LightSquared/GPS (in general; coexistence solutions could’ve emerged for consumer handsets combining LightSquared and GPS service).

The quiet neighbor downstairs

I was stunned when I realized how empty the 2360 – 2400 band actually was.

The Dashboard indicates a handful of licenses, mostly around aircraft plants in Southern California, Dallas/Fort Worth, Wichita and Seattle, for aeronautical radar and FAA's GPS-based Wide Area Augmentation System (WAAS) used in aircraft landing and safety systems in 2360 – 2395, and 736,811 ham radio licenses (697,887 licensees) 2390 – 2400.

Below this there’s Wireless Communications Service (WCS) in 2305 – 2320, some more lightly used aircraft landing and safety systems in 2310 – 2320, SiriusXM satellite radio in 2320 – 2345, and WCS and aviation again in 2345 – 2360.

The observations of the IIT Spectrum Observatory in Chicago (http://dx.doi.org/10.1109/DYSPAN.2011.5936195) show, not surprisingly, that the 2.2 – 2.4 GHz band is very quiet: 5% average occupancy vs. 17.5% for 2.4 GHz ISM. (Average occupancy of 2.5 – 2.7 GHz is similar to that of 2.4 GHz ISM.)

I’m sure there are very good reasons why at least 2360 – 2400 hasn’t been fast tracked for re-allocation to mobile broadband data allocation. Both the ham radio community and the FAA have a reputation for, um, vigorously defending their allocations, but we’re talking 40 MHz of “fallow spectrum” here…

The bottom line

Given that there’s already some interference between 4G service in 2.5 GHz and unlicensed devices in 2.4 GHz given at least 10 MHz of guard band, there’s a prima facie case that one would see interference problems to and/or from unlicensed devices if a cellular service were allocated in the fallow 2360 – 2400 band. That in turn suggests that it would make sense for the FCC to start encouraging or mandating better filtering for unlicensed devices over the 2.4 GHz band.

No comments: