Subject: Re: Strange failure mode in switchmode power supplies: Followup
Date: Mon, 23 Sep 2002 16:08:42 -0400
Organization: Posted via Supernews, http://www.supernews.com
X-Mailer: Mozilla 4.75 [en] (WinNT; U)
Summarizing important points as I understand them:
1) High frequency noise in SMPS is only when powered in your
house and not in any other house; both powered by same utility
transformer (and phase).
2) Using an oscilloscope in differential mode, electrical
source of that noise cannot be observed / detected.
3) Noise source is believed so strong as to damage every power
Point 1 says the source of electrical noise is only in your
If the electrical noise exists, then it 'must' be observed on
the oscilloscope. However you only looked for differential
noise. Both differential and common mode noise can appear on
the same wire One is observed when the other is not depending
on how the scope was connected. Point 2 - IOW did you forget
to repeat all measuring for common mode noise? A scope
connected to observe differential mode signals will not see
common mode. Don't let ground fool you. Every ground -
safety, oscilloscope, breaker box, and earth is different.
Best measure for noise by using the scope in differential
mode - two probed required. If you were only using one probe,
then you were not properly searching between the many
different voltage potential points. There are many different
places that a voltage potential could exist. Don't just think
in terms of black wire to white wire. Think of every AC wire
to all other conductors such as computer chassis, DC wire,
etc. Use the scope in differential mode.
Microwaves, et al as the noise source in point 3 is not
reasonable. To damage electrical components, the noise energy
would not just damage resilient SM power supplies and not less
resilient dimmer switches, GFCI, electronic clocks, etc. To
have noise that high, another important household component
would also suffer damage - the human.
From the list of damaged equipment, I read only things that
make good common mode circuit connections. Previously stated
was that the entire power supply was destroyed. Not
possible. Some components can be damaged but not all or even
most. An autopsy must discover which components are and are
not damaged. Not difficult with a soldering iron and better
multimeter. Most good multimeters measure capacitance,
inductance, diode voltages, and transistor gain. Since the
supply is already toast, then remove and test the usual
suspects - electrolytic capacitors, schottky diodes, power
transistor, transformer. Not just for the appearance of no
damage but also to put numbers to their performance.
In murders, the body is always 'best evidence'. In your
case, the power supply is same. However we have no data from
the best source of information - the dead body.
Rightly eliminated from consideration is millihenries of
inductance in wires. John Woodgate has accurately identified
that speculation as irrelevant. But then, if wire inductance
or even a loose AL/CU junction was a problem, then it would be
obvious on the oscilloscope. If the noise exists, then it
must be observed on the oscilloscope. Either the noise does
not exist, or all possible combinations of voltage potential
differences were not examined.
If, for example, the noise was common mode, then it would
not be affected by the transformer, and would not appear on
the oscilloscope setup to observe differential mode noise
(single probe measurement).
Peter Gutmann wrote:
> Wouldn't the various isolation steps I've tried (i.e. standard line filtering,
> isolation transformer and, ultimately, the ferroresonant transformer) pretty
> much rule out any power-line related issues? Given that the latter will stop
> pretty much anything by generating its own output signal sans assorted
> possible disturbances, it would seem that I'd have to look elsewhere for the
> source of the trouble. That is, I've applied every kind of measure I can
> think of to either locate or stop anything coming in via the power lines, and
> nothing has had any effect. Whatever it is, it's also universal enough to not
> be tied to the parameters of one particular SMPS (frequency, capacitance,
> topology, etc) which would tend to rule out any fixed power-line
> characteristic, because it shouldn't affect a wide range of SMPS' the same
> Admittedly the other possible causes of this are pretty odd, but given that
> nothing designed to find or stop power-line based problems has worked, I'm
> wondering about applying the Sherlock Holmes quote about eliminating the
> impossible to find the truth. Unfortunately the price of copper sheeting to
> build a Faraday cage to run a PC in as a test is a bit expensive, but if I can
> determine that the house is (say) baking in microwaves I can at least find
> something to hunt down and kill^H^H^Hfix (this might also help explain why
> DECT phones don't work properly here). Is there anything *other than the
> power* that I can investigate?