Phone companies love to boast about how broad their coverage is. We put those claims to the test by measuring mobile broadband performance for Optus, Telstra and Vodafone on a cross-country train trip between Melbourne and Adelaide.
Long-time Lifehacker readers will know that we’re keen on testing mobile broadband while travelling on trains. We’ve looked at everything from performance in Sydney’s train tunnels to what happens between Melbourne and Sydney to how well-connected you can stay on the Eurostar and how well Perth’s suburban network is covered.
I’ve even done limited testing on the Overland, the train service that takes a leisurely 10 hours to wind its way between Melbourne and Adelaide. Back in 2010, one of the surprising findings was that there were remote locations where Telstra had no signal but Optus was fine, which goes against the conventional wisdom that Telstra is always the better choice in rural locations.
While I’ve taken the Overland several times (once in order to edit a novel), I’ve never done a test that compares how the three main networks perform, so it seemed like a good idea to run a comparison. Plus, it’s a very relaxing and enjoyable trip, especially if, like me, you enjoy train travel. (It’s worth springing for Premium class for extra leg room and at-seat meal service.)
This is the approach I adopted, drawing on my previous experience in this field:
- Speed tests were conducted at each station stop on the way to avoid variation in where the tests stopped and started for different networks. The Overland makes 10 total stops (seven in Victoria and three in South Australia), so that still gives a reasonable number of data points.
- I ran tests using the Speedtest.net app on three separate devices. I had a Nexus 7 tablet connected to a Telstra 4G hotspot, an LG G3 with a Vodafone SIM, and a Samsung S5 with an Amaysim SIM. Amaysim uses the 3G Optus network rather than 4G, so that wouldn’t necessarily represent the maximum possible speed. That said, Optus’ rural 4G coverage is still fairly limited.
- Each test was run a minimum of three times and averaged — with extra tests if there was an obviously strange or outlier result.
Here are the results. In the table, V stands for Vodafone, O for Optus and T for Telstra. P is ping speed in milliseconds (the lower the better), D is download speed in Mb/s (the higher the better), and U is upload speeds in Mb/s (the higher the better).
Location | V P | V D | V U | T P | T D | T U | O P | O D | O U |
---|---|---|---|---|---|---|---|---|---|
Melbourne | 29 | 14.35 | 1.26 | 30 | 9.63 | 10.65 | 53 | 6.39 | 2.25 |
North Shore Geelong | 26 | 48.89 | 12.07 | 30 | 36.83 | 30.65 | 28 | 8.63 | 3.57 |
Ararat | 133 | 0.42 | 0.32 | 52 | 10.54 | 0.28 | 355 | 2.12 | 1.83 |
Stawell | 40 | 1.4 | 1.06 | 68 | 6.97 | 1.23 | 60 | 11.95 | 1.91 |
Horsham | 45 | 11.55 | 2.76 | 39 | 28.67 | 26.72 | 63 | 5.72 | 2.04 |
Dimboola | 107 | 6.5 | 1.77 | 64 | 13.38 | 0.83 | 188 | 1.94 | 1.88 |
Nhill | 273 | 0.36 | 0.01 | 2246 | 1.87 | 0.28 | 65 | 1.26 | 1.13 |
Bordertown | 264 | 5.23 | 2.62 | 73 | 11.51 | 0.65 | 50 | 1.52 | 2.24 |
Murray Bridge | 150 | 5.41 | 3.09 | 52 | 36.89 | 13.89 | 126 | 8.68 | 1.49 |
Adelaide | 44 | 2.99 | 0.47 | 29 | 13.97 | 7.94 | 74 | 2.92 | 0.81 |
And some notes on those findings:
- The location with the worst overall performance was Nhill — we only managed to complete one test on Vodafone and two on Telstra.
- Our Adelaide test was before Vodafone’s recent activation of 4G in Adelaide, so I’d expect those numbers to be higher if we did the test again.
- Telstra generally produced the fastest performance numbers, but not invariably. Vodafone produced blisteringly good numbers in Geelong, and Optus 3G was the best performer in Stawell.
Overall, I was impressed with the coverage on this trip — all three networks have lifted their game compared to previous journeys, and the speeds offered are usable even with the worst results. There’s still a clear divide between metropolitan and rural coverage, but the long outages I’ve experienced on previous trips seem to be a thing of the past. As ever, the right network for you will depend on what’s available where you live and what your needs are, but all three major carriers produced decent results.
Lifehacker Australia editor Angus Kidman still hasn’t completed the set of cross-country trains in Australia. His Road Worrier column, looking at technology and organising tips for travellers, appears regularly on Lifehacker.
Comments
10 responses to “Train Torture Test: Melbourne To Adelaide”
@anguskidman Wouldn’t the different antennas in each device skew the results?
Don’t be silly.
Hmm seems to be testing at stations where you’d expect reasonable coverage but does’t say what its like between stations for example I get frequent drop outs between Melbourne and Ballarat with my Telstra conections
In general it was really good — I wasn’t testing all three every moment of the journey but didn’t notice any lengthy drop-outs.
I’m surprised to hear that: whenever I train between Melb and Canberra, the Optus cell signal is gone pretty much the entire way (both 3G and GSM) except at stops. It really bums me out, because the train is soooo much better than a Greyhound.
Lol’d at describing a train trip from Melbourne to Adelaide as “cross-country”. Now try Melbourne to Perth.
Almost Nil speed in Nhill
Whilst the test slightly flawed with the use of 3 different phones. It does show the big T still dominates rural access speeds. Also I will mention that I’ve noticed when riding trains in metro areas different models of train block 4G frequencies differently compared to some others like a Faraday cage. So some trains are worse than others. The phone tends to switch to the lower 3G frequencies which seems to penetrate better. Once you step outside the train then you’re back to 4G.
Probably should have done this test on 3 phones of the same type to give the results validity.
Nice thought though to try and give us the info – but sampling method was flawed, so can’t really read into the results accurately.
While I agree three identical phones would be ideal, I very much doubt the differences in speed would have been dramatically different had that been the case. If all the results were close, it would matter rather more.