5 GHz or Bust

George Ou has some interesting points about 5 GHz, but there's more to the story: Ou hates the 2.4 GHz band with something greater than a passion--it's crowded, there aren't enough non-overlapping channels, and it's just so out of fashion. He's right about all that. The 5 GHz band has lots of possibilities, including 23 channels open for use. Before we get too excited, though, let me point out a few things that Ou didn't cover--primarily, the signal power output restriction for 5 GHz. 5 GHz 802.11 standards can't send signals as far as 2.4 GHz for a good hunk of the band. (23 channels are available for 802.11 specs; there are technically 24 possible in a different configuration.)

One of N's possible advantages of double-wide channels--instead of 22 MHz, they can use 40 MHz channels, which effectively doubles throughput. When you combine a newly efficient design for encoding, two or more radios, and double-wide channels, that's when you get the high symbol rate of 300 Mbps, with effective throughputs that could go well over 100 Mbps. The 100 Mbps throughput factors in--as I understand it--the expectation that N devices will have brief periods in which they can bond two channels.

Here's my executive summary so you don't have to read my entire analysis:

While 5 GHz is uncrowded and has more clear, non-overlapping channels that can be combined for the highest speeds in 802.11n, the rules governing 5 GHz for indoor use with omnidirectional antennas mean that only two double-wide channels in 5 GHz are available at anywhere near the maximum power (and thus potential range) of interior Wi-Fi. A number of 5 GHz channels offer power levels that are comparable to 2.4 GHz--and that assumes manufacturers will allow their 5 GHz radios to output enough juice to produce ranges similar to 2.4 GHz. But the lack of competing networks in 5 GHz could mean that 5 GHz N networks will almost certainly work much better in crowded RF environments--apartment buildings, for instance--than 2.4 GHz N networks which have very little chance to use double-wide channels. (Proviso: Some countries don't allow double-wide channels in 5 GHz.)

Ou says that "802.11b was supposed to have given way to its sibling standard 802.11a which operated in the 5 GHz range." That's not my recollection back at the time this was happening. 802.11b and a were seen as having somewhat different purposes: 802.11a had the potential for high-speed, short-range indoor connections, and long-range point-to-multipoint outdoor hookups. This was largely due to the higher cost of A, as Ou points out, but also due to the shorter range possible in indoor applications due to lower signal strength allowed. (Also, there were doubts about A's ability to be produced using CMOS chip processes, which Atheros put to rest, even as it moved into G for competitive reasons.)

He notes that while 802.11g boosted speed, "the problem was that in order to maintain backward compatibility, 802.11g also had to operate in the limited 2.4 GHz space and worse it had to switch to 802.11b if even one legacy 802.11b device joins the party." This isn't strictly correct as described, although it's widely stated this way. An 802.11b device forces a G network to drop down to B speeds only when the B device is transmitting or receiving data. There's a bit of extra network overhead, as well, if you choose for your G device to support B that does produce a single digit reduction in throughput even on G devices. The same will be true with N: older devices will occupy disproportionate amounts of time while transmitting or receiving, but if they aren't heavy network users, the N network should still work well.

I would argue that dual-band gateways didn't succeed in the consumer marketplace because of an absence of a compelling reason to use them. Throughput wasn't a big motivator. There were initially few adapters, none affordable, for 802.11a when 802.11g hit its stride. Remember Steve Jobs telling us 802.11a was dead? (Hey, their new device supports A, so I guess it got better.)

If you had mostly 802.11b/g devices and wanted to use A, you'd have to switch all your adapters over because no consumer devices supported simultaneous dual-band operation in 2.4 GHz and 5 GHz. Some enterprise hardware did (and still does) allow both bands either through two baseband chips or through two separate radios.

Ou complains that MIMO's leap into the market caused the lack of 5 GHz expansion. "Since it already required multiple radios for single band operation, adding an additional set of radio on the access point would have increased the already-high prices even higher." But there are few except very high end devices that would have duplicated radios; typically, as I just noted, it's one radio with two frequency ranges supported for either 2.4 GHz or 5 GHz operation.

Now here we get to the crux: Ou writes, "802.11n in my opinion should have NEVER permitted 2.4 GHz operation in the first place and should have only used the 5 GHz band." However, that idea has two flaws, one of which Ou admits and addresses. First, it breaks compatibility, which means that it's not a solution for people with B and G that want to gradually move to N. Ou suggests that a cheap single-radio could have been inserted into access points to handle B/G clients and the N system could have worked just in 5 GHz. Let's leave out all the radio engineering issues involved in that--like antenna coordination, cost of manufacture, firmware support, and so forth. The second flaw is more critical: 5 GHz has too many limitations in range.

The four spectrum hunks of legal 5 GHz channels each carry restrictions that don't dog 2.4 GHz, and that's why 5 GHz hasn't caught on. Where 802.11b/g/n devices can transmit as much as 1 watt of power at the antenna and use any channel indoors or outdoors, rules for 5 GHz prescribe in the U.S. and some other markets which channels may be used indoors only, and has much lower power levels for omnidirectional indoor use than 2.4 GHz allows.

There are now four bands in 5 GHz channelized for 802.11 in the US, although they're numbered somewhat strangely. In brief, there is total of 555 MHz across 23 channels in 802.11a/n. The lower four are indoor only; the higher 19 are indoor/outdoor. The lowest four (5.15 to 5.25 GHz) can have 50 mW of output power, the next four (5.25 GHz to 5.35 GHz), 250 mW; the next 11 (5.47 to 5.725 GHz), 250 mW; and the top four (5.725 to 5.825 GHz) up to 1 W. (There are further restrictions on 5.25 GHz to 5.725 GHz in terms of detecting and avoiding stepping on military radar transmissions, which share those bands. And the 802.11a spec specifies 40 mW/200 mW/800 mW instead of 50, 250, and 1,000, just to make it even more complicated.)

There are an enormous number of details about effective output, antenna gain, and so forth, but most of that affects the use of directional antennas and point-to-multipoint outdoor connections, not the use of interior omnidirectional antennas.

Because 5 GHz signals have shorter wavelengths than 2.4 GHz signals, at the same amount of power, they propagate shorter distances. They're also worse at penetrating solid objects. This means that even if you use the top four channels for 802.11a or single-wide channels for 802.11n in 5 GHz, you will only be able to send data less than half as far if that. There are only two double-wide channels possible in that top band.

In the 250 mW restricted range of 5 GHz, you could achieve the same range by using higher power in 5 GHz than in 2.4 GHz. But many of the devices that offer 2.4 GHz and 5 GHz radios don't compensate in 5 GHz by having higher-powered signal output. Thus a device that gives you 100 interior feet in any direction in 2.4 GHz could span less than 50 feet for this reason in 5 GHz. The lack of interference from competing networks could compensate for the shorter distance, however.

(Another issue: Some 802.11n device makers may not let you use double-wide channels in 2.4 GHz. Apple's new AirPort Extreme with 802.11n says in its advanced configuration manual--online already long before the product ships--that what it dubs the Use Wide Channels options is only available in 5 GHz. Conversely, Apple is promoting its AirPort Extreme with N in some European as only offering 20 MHz channels in 5 GHz because of regulatory limits. Thanks to Iljitsch van Beijnum for pointing to the manual and the European issue.)

Still, 5 GHz does offer some hope, and while Ou thinks the boat was missed, I see Apple's support in clients and adapters for 5 GHz in N, and Intel's support in its Centrino client for 5 GHz as a sign that that band will pick up steam. Note that Intel is certifying access points and routers with a Connect with Centrino label--and those devices will likely have to support 5 GHz, like this Intel-co-branded Buffalo router.

Let me reiterate one point here at the end: Manufacturers often limit their devices to 100 mW or even much less of output power for 2.4 GHz for reasons of cost. The problem in using 5 GHz will come entirely from whether those manufacturers decide to use the same power output limits for 2.4 GHz and 5 GHz even though they don't have to, or whether they'll actually take advantage of 5 GHz by boosting its power to put its range into parity.