03.04.09

Wi Not WiFi on Steroids?

Posted in business, white space at 11:00 am by JamesNeel

Though the term was first applied to WiMAX (link), “WiFi on Steroids” now commonly refers to a use-case for the white spaces (link) for WiFi-like access points in the UHF bands, but better, stronger, faster and so on due to the ongoing advances in technology.

As an initial business model, it’s a good place to start as there’s ALOT of demand for high-speed access at the home as evidenced by WiFi and if someone  can build a better WiFi access point, the world should beat a path to their door.

However, I don’t think we’ll be able to successfully market white space devices as WiFi on Steroids.

Here’s why.

All else being equal, you get higher data rates (what most people think of for “WiFi on Steroids”) by either grabbing more spectrum or using spectrum more efficently. Sure, as a nation, allowing the opportunistic use of unused spectrum improves nation-wide spectral efficiency, but it doesn’t improve spectral efficiency for a given link. Which means that on its on, you’re not going to be squeezing higher peak bps/Hz over your WiFi link just because it’s now running DSA. In other words, simply downbanding WiFi won’t get you massively increased throughputs that would be required to encourage people to abandon their ISM-band WiFi. Augment, maybe in a dual-mode cross-band channel bonded mode to increase throughput, but the extra front ends and necessary isolation seems expensive and wouldn’t be anything that couldn’t theoretically be done with ISM and UNII bands and dual RF ports, which isn’t exactly driving the WiFi market now.

That means that in order to increase data rates up, we’ll have to use wider bandwidths or improve spectral efficiencies the old fashioned way, i.e., higher-order modulations and / or antenna arrays (ala .11n). Since I don’t think that in practice we’ll be using higher-order modulations, that leaves antenna arrays. Typically, that means MIMO (which in its own way is grabbing more “bandwidth” via spatial channels). But for MIMO you need low spatial correlation, which in practice means antenna spacing that is a significant fraction to multiples of a wavelength. I’ve seen studies showing gains for as low as quarter wavelengths, but a full wavelength or more is the rule of thumb.

But what’s a  wavelength in the UHF bands?

To simplify the math, let’s use 300 MHz (right in the middle of the band). The wavelength is then 1 meter (a little better than half your height) which is not a good form factor when compared to the 0.125 m (3e8/2.4e9) spacing for the ISM band.

So perhaps we just grab more spectrum to offset the MIMO loss (roughly, call it a factor of 2 loss that from not being able to use MIMO), which is reasonable because the spectrum is free.

That may not be much of an option in the big cities where the customers are. The studies I’ve seen have shown 14 available channels in a city such as LA once channel adjacency requirements are met. That means there’s 6×14 = 84 MHz, which is not much more than what is used in the the ISM band (e.g., 79 1 MHz channels in Bluetooth). Sure there’s a lot of interference in the ISM bands now, but with longer propagation in the UHF bands, interference will quickly become a factor there too.

Now there are creative ways to get around this, particularly with cooperative / synthetic MIMO techniques (e.g., co-deploying UHF-WiFi with your home stereo system to use the speakers as widely separated antennas or just add on antennas with wires for your WiFi AP), but they’re kinda klugey and not with the plug-and-play factor that gave WiFi it’s mass appeal.

So in short, I don’t think the WiFi on steroids as it relates to higher data rates makes much sense as a business model / use case. And since it’s been marketed as “WiFi on Steroids” (much higher throughput which I think is not possible in the most valuable markets) and not just a different band for WiFi (ala the differing bands for cordless phones), I think the market will be really slow to adopt these devices.

What I think may make more sense are applications that don’t need as much throughput (so exotic channel bonding techniques aren’t required) but would benefit from better coverage (due to the downbanding) and free spectrum. For example controls sorts of applications ala zigbee (cognitive zigbee if you will) makes sense for home automation applications and factory applications and meter reader sorts of things (which for maximum buzzword density makes white space devices a good candidate for supporting the smart grid). These sorts of coverage (range) intensive, but low throughput applications are what I think will win out in the coming white-space deployment race.

Comments are closed.