Those who have been connected with me on Facebook would know that on Friday (21st) morning, I lost my normally-reliable naked ADSL2+ service with TPG. Coincidentally, over the night of the 20th to the 21st, Downer Engineering, on behalf of Parramatta Council, were resurfacing the road outside our strata. Coincidence? I think not.
They initially planned to have it done the night before, but they didn’t end up doing it – likely due to weather as stated in their mail-drop leaflets.
When I first started with naked ADSL2+ almost 3 years ago at this place, I had 12-13Mbit/s sync speed down, and 1Mbit/s sync speed up. Prior to the incident, it had fallen, but not by much (more likely due to a change of modem) to about 10-11Mbit/s sync speed down and still 1Mbit/s sync speed up.
Then, disaster. The modem was on the blink – syncing every few minutes, dropping out and trying over. It was achieving just 6-8Mbit/s down and 0.6Mbit/s up. A very strange thing to happen, literally overnight. Once I discovered this when I woke up, I turned off the modem to stop it from hammering TPG’s authentication servers and did some sleuthing.
Their logs, published a day later, showed the carnage:
10.20.21.192 23:47, 19 Jun 2013 23:59:51 2335.538 39.195 0.00 10.20.21.192 23:47, 20 Jun 2013 01:50:33 6.623 0.000 0.00 10.20.21.179 01:38, 21 Jun 2013 00:03:40 0.087 0.000 0.00 10.20.21.192 03:14, 21 Jun 2013 00:07:01 0.000 0.867 0.00 10.20.21.179 03:24, 21 Jun 2013 00:05:05 0.000 0.122 0.00 10.20.21.179 03:30, 21 Jun 2013 00:03:30 0.000 0.093 0.00 10.20.21.179 03:37, 21 Jun 2013 00:01:00 0.000 0.003 0.00 10.20.21.192 03:40, 21 Jun 2013 00:01:00 0.000 0.011 0.00 10.20.21.192 03:43, 21 Jun 2013 00:30:33 0.000 0.705 0.00 10.20.21.192 04:14, 21 Jun 2013 02:10:33 0.000 3.909 0.00 10.20.21.179 06:27, 21 Jun 2013 00:01:40 0.000 0.046 0.00 10.20.21.179 06:30, 21 Jun 2013 00:02:10 0.000 0.140 0.00 10.20.21.192 06:35, 21 Jun 2013 00:06:11 0.000 0.253 0.00 10.20.21.179 07:11, 21 Jun 2013 00:01:10 0.000 0.006 0.00 10.20.21.192 07:14, 21 Jun 2013 00:01:00 0.000 0.043 0.00 10.20.21.179 07:17, 21 Jun 2013 00:01:10 0.000 0.034 0.00 10.20.21.192 07:24, 21 Jun 2013 00:02:32 0.000 0.400 0.00 10.20.21.179 07:33, 21 Jun 2013 00:04:20 0.000 1.102 0.00 10.20.21.179 07:38, 21 Jun 2013 00:04:27 0.000 0.062 0.00 10.20.21.179 08:12, 21 Jun 2013 00:00:56 0.023 0.000 0.00 10.20.21.192 08:14, 21 Jun 2013 00:00:38 0.010 0.000 0.00 10.20.21.192 08:16, 21 Jun 2013 00:01:05 0.038 0.000 0.00 10.20.21.179 08:19, 21 Jun 2013 00:01:20 0.138 0.000 0.00 10.20.21.179 08:23, 21 Jun 2013 00:00:31 0.045 0.000 0.00 10.20.21.192 08:25, 21 Jun 2013 00:01:20 0.070 0.000 0.00 10.20.21.179 08:27, 21 Jun 2013 00:01:53 0.178 0.000 0.00 10.20.21.179 08:29, 21 Jun 2013 00:04:13 1.367 0.000 0.00 10.20.21.179 08:34, 21 Jun 2013 00:00:59 3.971 0.000 0.00 10.20.21.179 08:35, 21 Jun 2013 00:03:30 0.116 0.000 0.00
A careful examination of the logs shows I had no problem maintaining 24 hours of connection on the 20th. On the 21st, at 1:38am shows the first of many automated reconnection attempts – most sessions lasting no longer than a few minutes with time gaps between where there was completely no service.
Interestingly, when viewing the statistics on the modem (yes, it’s a crappy D-Link I picked up off the street), we saw a negative margin of about -6dB. This means the signal-to-noise ratio is 6dB too low to support the sync rate negotiated during the modem training sequence! Normally most modems I have met will drop the link very soon after falling below 0dB (sometimes transients to -1dB have been seen). This screenshot was taken after my little desperate attempt to reconnect by disabling all higher modulation modes other than G.DMT – still no luck on the stability.
I grabbed an old fashioned corded phone – a naked DSL service has no phone service as such, but the -48v talk battery is still supplied to the line. A decent line should have a quiet background – with the keypad of the phone operative off the loop current.
Instead, I got a loud nasty persistent crackle. In all the years of using ADSL, I never thought it would happen to me. It was the same in both sockets, with nothing else connected to the line. A quick inspection of the demarc box outside provided no clues – as the connections were punch-down (i.e. cold-welded) and had no signs of corrosion. Whatever it was, it wasn’t on our side of the fence.
I’ve contacted TPG – they’re onto it. We’re still waiting for updates though, and it’s the weekends – everybody wants internet, so what can we do?
For one, I patched up my Kogan Mobile Data as a temporary fix – with my Telstra Elite Wi-Fi modem. In order to ensure intranet access and overcome the Telstra Elite Modem’s 5-user limit, I hooked the modem up to a wireless to ethernet bridge, and connected that into the main router in place of the ADSL modem. Then I did some juggling of IP’s (modem, bridge on the 10.0.0.xxx subnet) and security settings (all ports forwarded from modem to router) and we were off.
The quota was not really enough – life in modern times means lots of background data accesses from tablets, phones etc. Even with all of mine turned off, the rest of the family’s usage was enough to blow the 400Mb/day restriction on Kogan. What happens when you exceed 400Mb/day? Nothing! It keeps working. But I anticipate that if you do it enough, that something nasty could happen.
According to Kogan’s T&Cs:
You must use the DATA 30 Plan in accordance with Kogan’s Acceptable Use Policy, which allows Kogan to suspend, terminate or refuse to renew your Service where you:
- use the service for commercial use or for use as a permanent connection;
- download or upload more than 400MB of data on a single day on three or more occasions in a 30 day period; or
- download or upload more than 1GB of data on a single day.
I hope I don’t lose my SIM over this …
So I also bought up an old Virgin Mobile 3Gb Mobile Broadband starter kit I had in reserve. All I know is that the connection quality here of Optus based services is rubbish – I’ve had anywhere from 0 to 3 bars on 2100Mhz, with no 900Mhz service. It’s definitely in the 100-200kBit/s zone most of the time, with occasional good signals up to 2Mbit/s. Not great – but better than nothing. After all, I need the internet to restore the internet! (It’s far better than to stay on the line and hold with some of the tech support staff.)
A quick examination of the outside shows the linesman had marked out the telephone bundle – and that area was not resurfaced over, but the kerb in the area had cracked, which may imply that they parked or stood heavy machinery in that area for prolonged periods. Hmm.
You can see the ‘can’ across the road as well.
Making the most of the situation – seeing as I have internet, albiet slow and flakey, is to blog about it and maybe do some testing of my own. I’m no telco engineer, and I have no telecomms test equipment of my own – but I do have bits and pieces.
I wanted to visualize the crackle – so I grabbed an old 605 to 6P2C adapter and pried it open. I splayed the two gold pins apart and grabbed a ~600 ohm resistor (in my case, 680 ohm because I had a pile of them next to me) and connected it across to simulate an off-hook phone (nominal impedance is about 600 ohm, with some reactance/capacitance as it varies for phone standards across countries). This should start the loop and we’d expect to see about 3-9V across the resistor with a clean DC signal as naked services do not have a dial tone. I connected a DSONanov2 across the resistor since I had it handy with me.
(If you’re watching closely, yes, the polarity is reversed – I forgot that the telephone lines are negative with respect to ground – no matter). Also, I should take the time to note that you should only connect Austel approved equipment to your phone line … (shh!)
The traces that came out told me the line was in trouble … the crackle was visible as momentary dips (and even a short spike) in line voltage – which appeared to hover at about 3v (the lower end of the expectation).
Short transient dip – probably more noisy but limited by the resolution of the DSONanov2 used.
Later on, I wanted to come back and test it with my Velleman PCSU1000 with a higher sampling rate, only to find the line was completely lost.
All of this makes me wish I had the NBN. Sure, a fibre snap would be just as devastating, but my assumption would be that the system would be self monitoring and would send crews out to fix it almost immediately – after all, the fibre is being laid in rings – one break takes out tens of households at a time. But at least fibre is fast – living on 3G service only for a weekend shows me how even slow broadband access (especially ones with restricvtive quotas) can make you a “second class citizen” of the internet. Unfortunately, looking up the rollout maps only tells me that across two houses, neither one has won the NBN lottery.
Lets just hope this gets fixed – fixed properly and quickly. Around my area, I’ve seen Telstra engineers sitting at the ‘can’ looking for good pairs. Apparently, we have lots of issues – and lots of industrial users in this area too. That might be the reason why a few of the neighbours have gone with Cable internet instead.
It was so reliable – I just never thought it would be me …