26.6 Troubleshooting Power Supplies
Suspect a power supply problem if you
experience any of the following symptoms, particularly in
combination:
Parity check errors. Such errors may be caused by defective or poorly
seated memory or by overheating, but insufficient or poorly regulated
+3.3VDC or +5VDC (depending on memory type) from a failing or
inadequate power supply is a likely cause.
Sporadic or regular boot failures. Obviously, such errors may instead
be caused by hard disk, cable, or disk controller problems, but
inadequate or poorly regulated +12VDC (less commonly, +5VDC) is also
a common cause of this problem.
Spontaneous reboots or system lockups during routine operations, not
attributable to running a particular program. Numerous other factors
can cause this problem, but one common cause is insufficient or
poorly regulated +3.3VDC and/or +5VDC being provided to the memory
and/or processor.
Lockups after installing a new processor, memory, disk drive, or
expansion card. Driver issues and resource conflicts aside, this
problem commonly occurs when new components overload a marginal power
supply. This is particularly likely to occur if you make dramatic
changes to the system, such as replacing a slow CPU with a fast,
high-current CPU; if you expand memory significantly, e.g., from 32
MB to 128 MB; if you add a high-current expansion card such as a fast
AGP video card or internal modem; or if you add a high-current drive
such as a high-performance SCSI hard disk or a CD burner to the
system. Note that the power supplies provided with commercial
systems, particularly inexpensive ones, often have very little
reserve.
Failure to function with a Wake-on-LAN (WOL) motherboard. The
motherboard and power supply may both be operating properly but be
incompatible. Many early ATX power supplies (and some current models)
provide 100 mA or less of +5VSB. Although that
output met the ATX 2.01 requirements, WOL motherboards require
+5VSB of at least 720 mA.
Slow disk performance. Although this may seem an odd symptom to be
related to power supply problems, inadequate voltage and current can
cause disk retries on both reads and writes. The error correction
circuitry built into hard disks and controllers means that this
problem often (usually) goes undiagnosed. People often say to us
something like, "I replaced the power supply as you
suggested, and now my hard disk seems a lot faster. Is that
possible?" Yes, it is.
|
A very common source of problems is using a noncompliant ATX-like
power supply. We say "ATX-like"
because many power supplies that fit ATX cases are not ATX compliant.
Motherboards vary in their tolerance for voltages that are slightly
out of spec, and a marginal power supply that works fine with one
motherboard may not work with another, even of the same model.
|
|
Troubleshooting power supplies is difficult for several
reasons:
Other than an outright
power supply failure, problems caused by an inadequate or failing
power supply are likely to be subtle—occasional memory errors,
lockups, slow disk performance, and so on—and easily
attributable to something other than the power supply.
It is difficult to
"bench test" a suspect power supply
because PCs use switching power supplies rather
than linear power
supplies. Unlike linear power supplies,
switching power supplies by design do not operate unless minimum
loads exist on specific voltage rails. The minimum load required
varies from model to model, but in practical terms you must connect
at least a motherboard, processor, and hard drive to most power
supplies before they function at all.
Systems draw varying amounts of current at different voltages
during routine operations. For example, a drive spinning up draws
down +12VDC, which in turn may cause another voltage rail to
fluctuate, causing problems that are not clearly linked to the +12VDC
load. Even processors use varying amounts of current, depending on
what they happen to be doing at the moment. This constant variation
in draw and the interdependency of currents on different voltage
rails make troubleshooting very difficult.
Working inside a power supply is
dangerous because high voltages are present, but testing only
external connectors makes it difficult to troubleshoot effectively.
Despite this, we do not recommend removing the cover from a
power supply for any purpose. If you do so and electrocute
yourself, don't blame us.
With the high cost of labor, it is
usually more cost-effective in a business environment simply to swap
out a suspect power supply for a new or known-good unit, particularly
if the suspect unit is old and/or was inexpensive to start with.
Paying for an hour or two of technician time makes little sense when
the alternative is installing a new $50 power supply.
If you're working on your own system, however,
and if you have a DMM, you can do a few quick tests that may isolate
the problem to the power supply. These steps involve testing voltages
on specific wires of the main power supply connector while it is
connected to the motherboard. Some connectors have built-in probe
contact points that provide easy access to each signal. If yours does
not, slide the probe down inside the body of the connector until it
contacts the crimp-on connector to which the wire is secured.
Ideally, use a DMM that permits logging maximum and minimum voltages
over a period of time while you use the system. If
you're using an inexpensive DMM,
you'll have to settle for instantaneous readings,
but those often suffice. To test your
power
supply, have a list of pinouts and signals for your power supply type
(AT or ATX) handy, and take the following steps:
With the black probe touching the power
supply case, touch the red probe in turn to each Ground/Common
(black) wire on the main power supply connector, on any subsidiary
connectors, and on the Peripheral Connectors and Floppy Drive
Connectors. The DMM should show 0.00V. Significant voltage present on
any ground wire indicates a serious problem in the power supply.
If the system is completely dead
when it is plugged in and turned on, the power supply may not be
asserting Power Good. Even if the system runs, check the Power Good
voltage, because voltage variations on Power Good commonly cause
subtle system problems. With the black probe of your DMM touching the
power supply case or other grounding point, touch the red probe to
the Power Good line. Power Good is nominally +5VDC. The DMM should
indicate between +4.0V and +6.0V. Most motherboards trigger at from
+2.0V to +2.5V, so a reading below +4.0V may allow the motherboard to
boot, but indicates a possible power supply problem. If the DMM
indicates less than +3.0V or more than +6.0V, replace the power
supply. If no voltage is present, the power supply is not asserting
Power Good, and is likely defective.
Test each voltage rail against ground to verify that they are
within specifications. If a particular voltage appears on multiple
pins, test that voltage at each pin.
For each Peripheral Connector and Floppy Drive Connector, test
each of the two voltages present against ground. That is, touch the
red probe to +12V (yellow wire), and then touch the black probe to
the adjacent ground pin (black wire). The DMM should read +12V within
tolerance. Then touch the red probe to +5V (red wire) and the black
probe to the adjacent ground pin. The DMM should read +5V within
tolerance. Finally, touch each probe to one of the ground pins. The
DMM should read 0.00V or something very near it.
If any of these tests fails, a
defective or overloaded power supply is the most likely cause. In
that event, replacing the power supply is usually the best choice. We
have never attempted to repair a power supply ourselves, and do not
recommend doing so. If the power supply is under warranty—good
units often have three to five year warranties—call the vendor
for an RMA number and ask if they are willing to cross-ship a
replacement unit. If the power supply is not under warranty but is an
expensive (high-wattage or redundant) and relatively new unit,
contact the vendor about having it repaired. Some vendors quote a
fixed price, while others charge time and materials. Be wary of the
latter.
|