Inside Google’s innovative African broadband trial

Broadband connectivity is getting more widespread by the day, but the industry is still experimenting with new techniques to spread it further. Google(s goog) and Microsoft(s msft), for example, are both using trials in Africa to test out a new wireless technology called white space broadband. Google’s trial is taking place in Cape Town and, seeing as I was on holiday there in the second half of June, I couldn’t resist popping along to see how it’s going.
I’m very glad I did so. For an experiment that’s only supposed to deliver humble speeds of 2.5Mbps to each of the 10 participating schools, it’s turned out to be a lot more useful than I initially thought it might. Although the schools generally already had ADSL broadband lines, these are of very poor quality, mostly thanks to Telkom’s monopoly in South Africa. The white space broadband trial is delivering up to 10Mbps per school, which makes a real difference — it’s perfectly normal for a Telkom line to deliver speeds that are only measurable in kilobits per second.
As Gerdus Nieuwoudt, an IT teacher at Settlers High School in the Bellville area, told me:

“We’ve come to rely on it, purely because our Telkom line is very unstable and very erratic. It gives us peace of mind – we’d had lots of packet losses, mail didn’t go through properly, and we had sat without internet for 3-4 hours or even a whole day.”

The kids and their teachers can now even use Skype to talk to each other and other schools – previously impossible due to the Telkom lines’ paltry upload speeds. “Programs on the computer weren’t able to update or run because of the connection, but since the upgrade the connection works almost perfectly,” student Conor Dirks (pictured) said, noting that the school could now more easily do basic things like updating its antivirus software.

Student Conor Dirks uses a computer hooked up to the white space broadband trial. He says it makes a "big difference".

Student Conor Dirks uses a computer hooked up to the white space broadband trial. He says it makes a “big difference”.

But what is white space broadband, and how is Google involved?

Recycling wasted airwaves

Wireless broadband usually does its thing at tightly defined radio frequencies – Wi-Fi uses the 2.4GHz band, 3G (mostly) uses 2.1GHz, and so on. White space broadband is different: it takes advantage of the buffer zones between TV channels, left there so they don’t interfere with one another. That means any broadband service exploiting these spaces also has to avoid messing with the surrounding TV signals.
And so, in order to carry out tests in a worst-case scenario, the Google trial is taking place in Cape Town. Because the city is built around mountains and hills, it has to have 3 TV transmitters in order to reach all of its residents. Each of these transmitters pumps out 5 terrestrial TV channels, meaning Cape Town’s airwaves are buzzing with a total of 15 TV channels, coming from multiple transmitters and being received at varying strengths, depending on where the viewer is.
As Arno Hart, a project manager at trial participant TENET, put it:

“Cape Town is a very active TV market. If you can prove you can do this without interference in Cape Town, you can do it anywhere.”

But, he added, there’s a total of 400MHz of fragmented white space to play with here – and that’s an enormous amount of real estate for broadband purposes.
So, with a lot of bitty little pieces of radio spectrum on offer, how do you use them for a coherent broadband service? There are theoretically two ways of doing this. One is the use of something called cognitive radio – essentially a radio that can sniff out empty frequencies and exploit them dynamically. The other is to use a database that can tell you which frequencies are available at specific geographical spots.

Spectrum database

As cognitive radio is still very much in its infancy, the Cape Town trial is taking the database approach. That’s where Google comes in – the company started trialling its spectrum database in the U.S. earlier this year and is putting it to use in South Africa, too. The data comes from ICASA, the local telecoms regulator, which tells Google where the existing primary spectrum users are based, where the TV transmitters are, which channels they’re using, the direction of the antennas, and so on.
Google then takes that data and applies its own experimental algorithms in order to create a spectrum map. That map is in turn used by the receiving equipment. “The equipment manufacturer registers with various database providers – until then it can’t operate with a database,” Hart said. “Nobody’s going to be able to manufacture equipment for TV white spaces and just turn it on and run it.”
Other participants in the trial include: TENET, the country’s national education and research network; the South African Council for Scientific and Industrial Research (CSIR); the e-Schools Network; the Wireless Access Providers’ Association (WAPA); infrastructure-builder Comsol Wireless Solutions; equipment vendor Carlson Wireless Technologies; and wireless tech firm Neul, which is very keen on using white spaces for its Weightless “internet of things” effort.

The Carlson equipment being used in the white space broadband base station.

The Carlson equipment being used in the white space broadband base station.

The transmitting antennas positioned on top of Stellenbosch University's Faculty of Medicine and Health Sciences.

The transmitting antennas positioned on top of Stellenbosch University’s Faculty of Medicine and Health Sciences.

The base station feeds off a 10Gbps connection. “We wanted plenty of backhaul so that wouldn’t be another point of failure,” Hart explained. “The furthest school is 6 kilometers away and the closest 1 kilometer… When there is no contention, depending on the day and on usage, we are averaging 6Mbps for the furthest schools and 10Mbps for the closest. When all the schools are connected and streaming at the same time, we’re averaging 2-3Mbps per school.”
The antenna on the roof of Settlers High School (the receiver is up there too, powered by an Ethernet line).

The antenna on the roof of Settlers High School (the receiver is up there too, powered by an Ethernet line).

A speed test showing the speeds the school is getting down below – 4.58Mbps on the downlink and 1.8Mbps on the uplink, with 104ms latency.

A speed test showing the speeds the school is getting down below – 4.58Mbps on the downlink and 1.8Mbps on the uplink, with 104ms latency.

Alan Norman, who leads Google’s white space research efforts, told me that the company is committed to running the trial for 6 months (it began in March) but hopes to keep it going after that. “Schools are getting connections. They like it – it’s valuable,” he said.
It’s also successful as a technology trial. Both Norman and Hart told me that there have been no reports of interference with TV channels so far. “It’s been a lot of work to get everything to the point where it’s robust, but Carlson and Neul have done a ton of work to make this work well,” Norman said.

New opportunities

So, if white space broadband is soon to be a reality, where will it be most useful?
White space broadband makes strong economic sense in rural areas, where it is very expensive to deploy wired connections to thinly distributed properties, and also in poor, densely populated urban areas, where copper gets stolen and no carrier sees enough money to justify rolling out fiber.
But, according to Norman, there are other potential uses too. “In the U.S., I’m not sure white space is only for rural areas,” he said. “With low-frequency spectrum, you can do things where you get good propagation at low power. You can do machine-to-machine [M2M] communications, and also inside the home, where the TV antenna is on the roof, there are a lot of white spaces available.”
And, if the cognitive radio piece turns out to be viable, there could be further implications too. “The TV band is just the first band we’re working on but, if you get to a point where cognitive radios are more available, we think there’s a lot more spectrum that can be freed up in other bands too,” Norman said.
These early trials may already be handy for dragging poorly-served areas into the broadband age, but the future implications of this work could be even more wide-ranging. If the so-called spectrum capacity crisis is truly a problem, we could be looking at the early stages of the solution.