This week, my internet connected was knocked out for about an hour after achieving a record 60+ days continuous uptime. The cause was quickly traced down to a failed power supply which runs my ADSL2+ modem resulting in a boot-loop. This would make it the second failure of a power supply connected to a modem that I have experienced as being the cause of an internet outage.
The modem I’m using is a fairly old D-Link model which was abandoned on the side of the road, sans power supply. This might have been because the supply it was provided with had failed. Regardless, it was found that the wireless modem router had permanently buggy wireless and incompetent routing, so was thus augmented by having a wireless router placed behind it, with wireless and routing disabled on the modem router.
Because it didn’t come with a power supply, I needed a suitable one. At around the same time, I had purchased a second hand Dreambox 500 satellite TV set-top box which sadly had a blown LNB power output stage (could be fairly easily repaired but I didn’t bother). Ultimately, its lack of HD output and blind scanning made it redundant with my PC based satellite cards, which liberated the PAG024F power supply to run the modem.
It’s a pretty generic looking power brick, of the sort that takes a figure 8 lead, so will not consume valuable plug space as a “wall wart” would. This particular supply has a 2A rating, and the modem requires 1.25A, meaning that the supply is being run underneath its maximum rating, thus should last longer.
The supply had served probably three or four years in the modem-powering role, and maybe another four before it running the STB. It was well worn and due for replacement – but as with many pieces of equipment that “just keep working”, we tend to forget about it until it actually collapses.
I knew something was wrong when I couldn’t get into the admin page for the modem. A power cycle drew my attention to the flicker pattern of the LEDs which implied a boot loop. This immediately drew my attention to the power supply unit as a potential culprit.
The PSU’s LED continued to shine steadily despite the boot-looping router. Some failing switchmode PSUs like to make squealing noises and have the LED flicker just as the boot loop happens. This one was quiet. I wasn’t entirely sure, but out of caution and a desire to “get the family off my back”, I quickly substituted a wall-wart (12V 2A Asian Power Devices PSU from an unused external hard drive which I had shucked) and that bought the modem back to life. The modem stayed running over 48 hours. My hunch was correct.
As this unit had no screws in its construction, taking it apart by prying at the seams wasn’t an option mainly because of how well the seams met and how hard it was to actually force the case apart. Ultimately, I resorted to hot-knifing the unit apart, although it did make my room smell quite badly of melted plastic. My curiosity needed to be solved.
A look inside shows the cause of failure immediately – a failed capacitor on the secondary side which vented. It also smelt pretty much like capacitor fluid in there. Everything else was in order with no signs of distress.
A look at the underside showed that the heat concentrated on where the capacitor was – in the output stage due to the rectifying diode (quite likely). The higher ripple currents and noisy output might not have been very good for the downstream device either, although it seems no immediate major damage was sustained.
The unit could be repaired, sans having a nice case, by changing out the capacitor. After all, most things still look okay. However, ultimately, because of a glut of 12v PSUs, I decided against it. This was chiefly because of the fact there were other questionable brand capacitors in the supply – a Yalecon primary side supply capacitor, a no brand one just in front of it, a Seacon and a United (vented) on the output which would all need to be replaced to have a chance of keeping it running reliably for a while. I suspect the newer supplies would be more efficient even if they were more frustrating due to their wall-wart form factor.
Even after all this time, we still see products fail with questionable brand caps at the helm. That being said, this one had served probably around 8 years from manufacture, a whopping 70,000 hours, which given the temperatures it was likely to operate at, means the capacitors still reached or surpassed their nominal lifetimes (generally 2,400-8,000 hours at full temperature for these sorts of electrolytics).
For equipment that just “works”, most of the time, these things are out of sight and out of mind. However, replacing older equipment and upgrading periodically can serve to improve reliability by taking such “worn out” items out of production service. If you are like me, and don’t mind having to put out a few fires or have things go down from time to time, then you can keep them going and see just how long they can run.