PSN-L Email List Message

Subject: Re: computer power supply
From: "Charles R. Patton" crpatton@......
Date: Wed, 05 Jan 2000 10:11:08 -0800

In addition to Raul J. Alvarez=92s quite accurate comment to be concerned=

about hash from the supply due to the switcher topology used, two other
important points come to mind:
1) Many of those PC type P/S require a minimum load.  Without that
minimum load, regulation and in some cases, the P/S itself fails
catastrophically.  A typical technique we often used when trying to do
certain types of tests on the P/S itself was to load the 5V to about 2A
or more or the 12V to 1A with a fixed resistor.
2) Cross regulation is often less than satisfactory.  I.e., the +12 V
and +5 V are sometimes regulated by the ratio of the windings in the
transformer and only the +5 V is actually feedback to the switcher
control circuits, so consequently the +12 V is subject to greater
variation since the voltage drops in the P/S circuit under changing load
are not compensated for.

Anecdotal notes about P/S hash.
1) Solving EMI (ElectroMagnetic Interference problems, i.e., hash) was
one of the main aspects of my job at a disk drive manufacturing
company.  There were multiple incidents where a poorly designed switcher
was putting out incredibly difficult to remove hash on the supply=92s
output lines that found its way into the disk drive effecting the read
performance.  One way of looking at it is that a disk drive consists of
a magnetic field sensor (a read head) generating tens of microvolts of
signal in the 10 MHz to-50 MHz region inside a shielded enclosure.  If
this can be effected, then so can a seismometer.
2) I just finished putting together a new PC with all new parts.  My
wife was not happy with the fact that whenever it was on she could not
receive a certain AM station.  The cure turned out to be that I had to
add a Corcom line filter on the AC power cord entering the computer,
which completely cured my problem.  The point, though, was that the
tower case I bought was cheap (the only correct word for a $69 tower
case with P/S) and so did not have adequate line filtering on the AC
line side, and this is a brand new, ATX level P/S.  So hash problems are
not confined to old designs or just the load side.


Public Seismic Network Mailing List (PSN-L)

[ Top ] [ Back ] [ Home Page ]

Larry Cochrane <cochrane@..............>