How Well Does 3G Broadband Work In Kiama?

The National Broadband Network (NBN) saw its second mainland site turned on last Friday, and Lifehacker was there to document the event. That trip also presented an opportunity to test just how well current 3G broadband services work in the area, which in turn lets us address the question: is mobile broadband a viable alternative to the NBN?

One of the most common issues raised in discussions of the NBN is whether investment in so much landline fibre cabling makes sense given the growing popularity and use of 3G broadband. If everyone can get online using mobile networks, why are we spending all that money for the NBN?

There are some obvious counter-examples to that argument. Mobile connections still ultimately need landline-based backhaul of some sort, so the fact that everyone is using mobile doesn't mean we don't ever need to lay new networks. The lower speeds offered even on "fast" mobile broadband aren't high enough for medical applications. And as I know from my own previous testing, mobile speeds represent a best-effort number most of the time, unlike the planned NBN model which does guarantee a minimum service speed.

However, the Kiama launch gives us a chance to actually put a real-world face on what is often an entirely theoretical argument. Right now only a handful of customers are actually using the NBN, and that will be the case until at least October. For the rest of Kiama, what's on offer if you can't even get a higher-speed ADSL2 connection, a common situation according to local mayor Sandra McCarthy? To find out, I decided to run speed tests on each of the three major mobile networks: Optus, Telstra and Vodafone.

This time around I didn't attempt to reproduce my often-used model of testing the availability of service while actually travelling down on the train. NBN connections will be to a specific home or office, so I wanted data from fixed locations on the ground.

Additionally, my previous experience on the Kiama train line back in 2009 also suggested that coverage on the line would be poor. Actually, that turned out to be a mistaken assumption: the situation has definitely improved since that time.

While I didn't run speed tests, I was working on my laptop the whole journey there and back. For the vast majority of that time, I had Next G signal on my computer and Optus signal on my BlackBerry, which is good news for south coast commuters. (I briefly tried using Vodafone on the way back from Kiama and lost all signal very quickly, though the coverage maps suggested I should have been getting OK if not spectacular results.)

In Kiama itself, I ran speed tests for Telstra and Optus on separate Android mobiles running the Speedtest.net application, and on my PC connected to a Vodafone Wi-Fi hotspot device. That meant I could run all three tests pretty much in parallel, and repeat them multiple times and average them to get better results.

I conducted those tests in two locations: at the Pavilion where the launch event was held, and at Kiama railway station while waiting for my return train. Time constraints precluded me walking further around town, but even those two locations proved instructive. Here are the results:

So what can we conclude for each carrier? Vodafone isn't in the race, quite honestly. It did have a signal to offer and surprisingly quick ping times, but its actual speeds were nothing to boast about and would be irritatingly slow for everyday use as the main source of broadband. (That might change in the future as it continues its network rebuild, but I'm concerned with what's available right now.)

Optus arguably emerges as the most consistent player, with the best results at the Pavilion and solid ping times. Its download speed of 2.25Mbps in the slower Pavilion location is better than you would get on a standard ADSL connection (though potentially not on ADSL2 and certainly not on a minimum-level NBN connection, which offers 12Mbps).

The results for Telstra fall very clearly into two camps. At the Pavilion location, the speeds on offer were fairly similar to Optus (the latter had an edge in upload speed and ping times). At Kiama Railway Station, Telstra's download performance was dramatically improved, with an average speed of 4.55Mbps (which is, for the record, faster than the ADSL2 connection I can get at home in Sydney).

However, it's now worth pointing out that the railway station is only a couple of blocks east of the Pavilion. Why were the figures so different across such a short distance? (A similar issue, albeit a little less dramatic, can be seen with Optus as well.)

I can think of two explanations, both of which remind us why mobile broadband isn't ultimately a substitute for the NBN alternative as it stands. The first is that because I was doing the first set of tests at a launch event where there would have been dozens of other mobile devices around, there was much greater strain placed on the network. You can't guarantee any kind of minimum or typical speed under those circumstances, which isn't a great argument for using it as the basis for a widely-used system.

The second explanation is that even though I was only a couple of blocks away for the second tests, I was connecting to an entirely different tower. That underscores even more heavily the inherent unpredictability of mobile networks: you simply don't know what you're going to get. It's an area filled with ifs, buts and maybes. A fibre connection to your premises offers far more certainty — something that most consumers and all businesses are likely to demand.

Ultimately, the argument shouldn't have to be an either/or scenario. I'm a big fan of mobile broadband, I made heavy use of it while I was in Kiama, and as a frequent traveller I'm grateful for it whenever I encounter it. But when I return to my office or home, I'm even more grateful to have a connection that's reliable and fast, and I'd be doubly grateful to have a wider range of options available in the future.

Lifehacker Australia editor Angus Kidman had forgotten how attractive the Kiama train journey is. His Road Worrier column, looking at technology and organising tips for travellers, appears each week on Lifehacker.


Comments

    Upload and Download titles on the graph appear to be around the wrong way..

      They were indeed! Thanks for the spot, fixed now.

    A major part of the issue with 3g is congestion. Mobile Broadband gets congested very quickly, and has quite limited bandwidth to share. In an area that doesn't have saturated 3g use, yes it can return quite usable speeds, but try any of those tests in say Sydney city, and you'll get vastly different results.

    Also important to note that the contemporary definition of broadband (as of 2010) is a service that delivers at least 4Mbps downstream and 1Mbps upstream. In this regard, only the Telstra measurements at the Railway can truly be categorised as 'broadband'.

    Every time you guys use the speediest.net application you don't mention the mirror being used. 90%of the time i have used it on mobiles it always connects to an Optus Mirror. How about trying it connected to a laptop and selecting a different mirror or a download test off aarnet mirror

    I think it's worth mentioning that wireless IS being used for part of the NBN as it is - but the technology used is LTE (currently only available on a limited basis with Telstra), it will be optimized for fixed location endpoints and there shouldn't be a congestion issue as the assumption is made that most people will be connected to it, with the landline / microwave backhaul to support that.

    DIDO!

      DIDO! read the report, it's full of if's, probably's, may be, in the future etc, etc.

    Comparing Exetel's published pricing, it becomes pretty clear that 3G is most certainly *not* a fixed line replacement.

    NBN: 25Mbit (10Mbit up)
    3G: 5-6Mbit if you're lucky

    NBN: $40/20gig
    3G: $90/12gig

    NBN: No contention, ever (or close enough)
    3G: More than 10-15 people in your suburb online and you can expect sub 2Mbit speeds.

    3G is great for on the Train, but for your home/office it's a waste of everybody's time.

    Out of interest what is the exact testing methodology used?

    How many times are the tests run? Which mirrors are used? Are the results averaged? How do you ensure time of day isn't an issue?

    i.e. if you run one provider at 11am, and then by the time you get to provider 3, it is 1230pm then you may well have entirely different results at different times.

    Also the ping results - as mentioned, if it is hitting a local ISPs server, then they have an advantage over the other providers. At least for ping results, it would be ideal to localise results so the only variable is the mobile network, and not the path taken once the packet is on the provider's network. Ideally you would get latency to one of the gateways at the provider in question.

    Same with downloads. In my experience even tests 10 minutes apart can have wild fluctuations with the servers.

    I am somewhat sceptical of these tests as the methodology to reduce variability due to external sources hasn't been provided.

      The post does point out that I ran the tests on three separate devices, so they happened at the same time. I would always do that when comparing providers.

      Speedtest automates to select the server with the fastest ping times. That means if anything the results might be overstated.

        Not necessarily, because it is offnet to the provider. So depending on who they use as an upstream provider and where the server is, a single provider could gain an advantage or disadvantage.

        I am not sure a single bandwidth test shows anything statistically relevant whatsoever. On any given single bandwidth test everything from a temporary network drop, through to a sudden network condition or temporary process on the machine in hand can affect the results.

        What I am getting at, is the isolation of the test from other factors that can affect it. If you allow the other factors, and random natural variability to come into play, then it will affect the reliability of the results. I could be wrong, but I think speedtest.net does choose on latency, which means it could well choose a server which is limited by its own load as well (which I have had multiple times on speedtest.net).

          I agree that the tests are far from perfect - in-depth broadband testing requires a lot more resources and time than I had at my disposal. Nonetheless, they do represent actual data, which is something that is often ignored in discussing broadband in general, and the NBN in particular.

    One thing that needs to be remembered is that 3G (and any wireless technology) is a shared bandwidth technology (i.e. bandwidth you get = bandwidth available /users using that service) whereas the NBN will be switched technology delivering full bandwidth (subject to engineering constraints built in).

    Wireless technology is good for its purpose, but is not the panacea to all problems.

      Absolutely - impossible to do without resources and lots of time. The results are useful, although I find more so when looking at the overall trends from your previous results.

      Unfortunately broadband services are hard enough to compare when looking at a medium with contention ratios, such as ADSL. Nevermind when you throw wireless mobile tech into the mix as well!

      What will be interesting, is when LTE has been released, and matured over the course of a year and to see what the capabilities are then.

      Geoff, at some point all fixed line infrastructure is shared just like bandwidth in a wireless network, just with a lower contention ratio. Wireless is main limitation is there is only a limited spectrum (overcome by frequency reuse, more towers, directional antennas etc), wired (fibre, ADSL etc) can lower the contention ratio by installing more cables to aggregation points.

    I use a Telstra 3G connection for work and the office is just out of the Melbourne CBD in Southbank. I have had repeated issues connecting using a plethora of devices (iPhones, HTC Desire, 3G Turbo stick, Ultimate WiFi device, 3G Ultra stick, Standard 3G [MF626] Stick... and so on). The connection is so bad that I am unable to place Skype calls using my 3G connection and am instead having to use Skype's local number service through my phone.

    So when you ask "If everyone can get online using mobile networks, why are we spending all that money for the NBN?" my immediate response is: because currently, wireless sucks.

    I have been to Telstra about this and their response was that there are known issues in the Southbank area and that upgrades are planned, but no due date is known. How is it that in Australia's second largest city we still don't have decent coverage, yet I can drive 300km's out of Melbourne (as I did yesterday) and have a blisteringly fast connection: something here is horribly wrong.

    All of Australia's Internet is about ifs and buts at the moment, you either get ADSL and pray you're not on a rim or on a congested exchange. Or you get mobile broadband and hope you're not on a congested tower...

    Internet is so great in Australia isn't it? Not to forget the whole time the goverement is wasting millions trying to filter it, just ick...

    Your results lack depth, and applications used. You should have a phone that is able to go into debug mode to see what cell you are using in the area and TX/RX levels from the cell etc. Also noting how many people are close by while testing and noting what carrier they are with. Being stuck at a train station which has 90% Telstra/Optus users all using facebook will affect the results depending on the time of the day when the number of users is at max/min for each carrier.

      It's not realistic to find out which carrier is being used by everyone at a station at the same time you're running tests. It wasn't peak period though so it definitely represents a best-case scenario.

      Facebook is a low-bandwidth application. If a couple users using Facebook has a significant impact on the results of the speed test, then that speaks louder than the test results themselves.

    My parents live in Kiama and there is practically n mobile reception in their house (or outside) for Optus and Vodafone. I can't comment on how Telstra performs since no one in the family uses them.

    I love having the internet with me on my phone, but I am so looking forward to getting my parents on the NBN and letting them use VOIp to cut out their line rental.

    They're with Exetel at present and it looks like they can get a budget plan for $34.50 a month and move their landline number over to MNF which means now more land line rental of around $22 a month. Every bit helps when ya on the pension.

    A flawed methodology to compare two different generations of technologies. How could the results be anything other than expected?

      I tend to agree, but the idea that wireless could be a complete substitute for fibre networking remains very widespread. Just ask radio shock jocks.

        Come on Angus, tell the truth! I haven't heard anyone suggesting that 3G data is a substitute for the NBN.

          You havent heard tony abbot open his mouth?

    I would suggest that quite possibly at the pavillion, a lot of other people would have been using Telstra (especially any government reps, or anyone from a larger corporation)

    That would mean that the Telstra results would be lower at the pavilion due to a saturation of users.

    Probably would explain some of the difference between the result there and at the pavillion.

    Too many variables that's for sure...

Join the discussion!

Trending Stories Right Now