From: firstname.lastname@example.org (Peter Gutmann)
Subject: Strange failure mode in switchmode power supplies: Followup
Date: Fri, 20 Sep 2002 05:29:36 -0000
Organization: Posted via Supernews, http://www.supernews.com
Sender: email@example.com (Peter Gutmann)
A few months back I posted a message about strange failures in assorted
switchmode power supplies in computers, printers, routers, etc etc. Various
people offered suggestions, and I've been trying those out (along with
anything else I can think of) to try and find the cause of the problem. Just
to recap (from the original post, page down to get to the new stuff):
The immediately obvious symptoms are that any SMPS plugged into the power in
my house starts producing large amounts of high-frequency (ultrasonic)
switching noise, both audibly and in its DC output. At 50W it's mostly
inaudible (but enough to cause the usual unpleasant effects of ultrasonics
if you sit next to it for awhile), at 100W you can sense it easily (although
not necessarily directly hear it), and at 200W it's so bad you can't stand
being in the room for more than a few minutes without feeling ill.
What's bad for humans is even worse for any elecronics sitting downstream,
so far this problem has caused about $2K of damage in assorted computers,
routers, etc etc (the damage is pretty spectacular, one device which failed
was an online UPS which I'd been using to protect the PC downstream, every
high-speed switching component in it had been fried).
The things I'd tried originally:
- Check neighbour's houses (nothing, only this one).
- Check times (it's present 24/7)
- Check for wiring problems (eg ground & neutral reversed, nothing)
- Check the power quality with a line analyser (nothing unusual apart from the
expected irregular fluctuations, spikes, etc)
- Check the (input) power with a scope (nothing, with the exception of about 5V
p-p ripple at maybe 1kHz on the sinewave peaks which appears for about 5s and
vanishes for 30-60s, it's not continuous so I doubt it's the problem)
- Get the power company to check the power into the house. "OK, it's about
240V, everything looks fine, connections look OK, sorry, that's all we can
do" (well, it was worth a try).
- Assorted line filters, power conditioners, and anything else I could get my
hands on (no effect).
- Run a direct line from the power board where the power enters the house to
the affected gear (no effect).
- Physically disconnect the power to everything else in the house except the
single line going straight to a sacrificial PC (no effect).
- Get the same power company (after several months of arguing and a written
complaint) to switch the house over to a different phase in case there's
something like a large reactive load on the same phase causing problems (no
- Get a new earth fitted in case the existing one (~25 years old) had corroded
through about 10cm below the surface (no effect).
- Connect a large linear load in parallel with the SMPS (no effect).
- Took readings at various locations around the house and in neighbouring
houses with a scope (much newer than the clunker I used earlier, and a
portable LCD one you can hold in one hand... drool): A generally clean
sinewave with some (very slight) odd-order harmonic distortion in the
150-300Hz range, but nothing serious, and not much different from either the
neighbouring houses or the power in the house of the person I borrowed the
scope from (which is miles from here). In other words, nothing obviously
different from locations where there was no problem.
- (I wanted to see what I could find on the line with a spectrum analyser, but
the only one I could get my hands on starts at 50kHz and doesn't have a HV
probe, and the owner wasn't at all enthusiastic about letting me put one
- After much searching tracked down a ferroresonant transformer which is
probably about the most you can do to filter (or regenerate the signal from)
a power line (I know there's some folklore out there about these upsetting
some SMPSes, but I never had any problems with several sacrificial systems
I tried) (no effect).
This means that either it's something incredibly insidious in the power which
doesn't appear on a scope and can walk through a ferroresonant circuit, or it
isn't anything related to the power at all. Unfortunately I can't even blame
whatever's causing this on power problems any more now, and there's nothing
left... at best I can handwave some sort of EMI issue, but (a) I'm nowhere
near any source of this and (b) I'd have thought an earthed metal box (SMPS
case) inside another earthed metal box (PC/router cases) would offer at least
some amount of shielding (modulo the power cord and other impromptu aerials
provide a means of access, obviously). I really don't want to sell the house
to fix this... does anyone have any other ideas? Anything I missed? Exorcism?