When the National Broadband Network (NBN) launched in Kiama, we tested how well 3G broadband worked as a rough measure of how effectively mobile networks could compete with the new fibre alternative. With the first Victorian NBN site following less than a week later in Brunswick, it only seemed sensible to repeat the experiment. What did we learn? That being in the city doesn't necessarily mean you'll get better 3G performance, and that yet again you only have to move a couple of blocks to see dramatically different levels of performance.
The NBN was officially launched at Brunswick Town Hall by Prime Minister Julia Gillard last Thursday. Lifehacker showed up around 24 hours later for broadband testing. I couldn't imagine that there'd be much different in terms of announcement content, and I wanted to be able to test at the launch location without the potentially confusing effect of having hundreds of dignitaries with mobile phones all in one spot, something which might well have influenced some results in Kiama.
The NBN rollout in Brunswick is fairly constrained, covering a couple of square kilometres constrained by Lygon Street to the east, Sydney Road to the west, Stewart St to the north and Glenlyon Road to the south. That's not a large area, but it incorporates a wide range of options, ranging from businesses large and small on the eastern and western boundaries to a mixture of townhouses, standalone residences and units in the middle. According to NBN Co, the rollout passes 2,689 homes, and 1,405 have agreed to a new connection during the initial rollout. (Only 14 of those houses will actually take part in the first round of testing, but all will be able to sign up after the first phase of the trial finishes in September.)
During my visit last Friday, I used the same testing methodology I'd followed in Kiama, running Speedtest.net via Telstra on my PC and via separate Android handsets for Optus and Vodafone. Tests were run pretty much simultaneously, allowing for me having to hit three separate start buttons. I followed a roughly Z-shaped pattern from the north-east corner, stopping in six locations and recording two tests for each provider. (Side note: you get some odd looks when you sit down in a public park and pull out a laptop and three mobile phones.)
There are two points I want to make up-front. The first is that I'm not able to hang around Brunswick for a day or more, so I can't measure what happens at different times of day. With that said, visiting around 11am probably represents a best-case scenario in many respects. It's not rush hour, so people aren't returning home, and many people will be at work. However, most businesses are operating and there should be plenty of shoppers with mobile phones.
The second is that while there turned out to be notable differences in performance between each network, in many ways the key question is whether anyone can deliver a consistent experience across the suburb. If a mobile network varies widely across small distances, it's hard to persuasively sell it as a high-quality broadband option for the masses.
Here are the full test results: unlike Kiama, I've quoted both figures rather than averaging them, since there were some notable variations. Low ping times can influence quality of service for time-critical applications such as VOIP or online gaming; download speeds affect general internet content; upload speeds are a factor if you regularly share large files or upload videos. A ! indicates that the test was not able to run.
With Brunswick less than 10 kilometres from the Melbourne CBD and with a range of public transport options, you'd expect reasonable competition from each of the main mobile network providers (Telstra, Optus and Vodafone). In fact, there were some definite problem areas. Vodafone availability disappeared completely for the second test at the corner of Blyth and Lygon — absolutely no signal at all even when I waited five minutes — and Optus required a second test to work in Symons Park. Telstra had particularly low numbers near Thomas and Albert, which might well be due to the presence of a large number of Telstra repair trucks on the same street.
It would be broadly true to say that Telstra had the best performance, followed by Optus and Vodafone (with the latter performing rather worse), but this would ignore some major variations. For instance, all three networks performed best near Blyth and Lygon, even though this was where Vodafone also disappeared altogether on its second test. Optus also outstripped Telstra by a healthy margin in the south-eastern corner.
Quite aside from the implications for the NBN, this provides a solid reminder that mobile network performance can vary hugely even within the same suburb. If you're considering switching networks, a recommendation from someone two blocks away could well be entirely meaningless. You need data from your own location.
In NBN terms, there are very few parts of Brunswick where you could confidently assume a download speed of 4Mbps, and it's worth remembering that this probably represents a best-case scenario figure. In technical terms, this doesn't seem a very challenging area to roll out mobile networks: there aren't any hills to speak of and no high-rise buildings, and the population density should be high enough to require reasonable coverage. If mobile networks can't deliver reasonable speeds and suffer from frequent dropouts in this kind of environment, then it seems naive to assume that they'll be adequate Australia-wide.
Unlike Kiama, Brunswick has the advantage of having a wider range of ADSL options available for its residents, so I'm sure there will continue to be arguments about whether a fibre rollout is needed. However, the mobile data is pretty clear: it's not a realistic competitor when it comes to speed or availability.