How to test a PSU?

I have a PC Power and Cooling 470W unit that's a few years old but
hasn't had a lot of mileage on it and the rig it was in didn't come
near to stressing its limits. It was modestly pricey when I got it so
I figure it's worth seeing if it's still working the way it's supposed
to. If I want to ensure it's not doing anything that might fry a h/d -
voltage spikes perhaps...or? - is there a way to test it, or someplace
you trust to send it for testing?

Thanks.

Doc wrote:
I have a PC Power and Cooling 470W unit that's a few years old but
hasn't had a lot of mileage on it and the rig it was in didn't come
near to stressing its limits. It was modestly pricey when I got it so
I figure it's worth seeing if it's still working the way it's supposed
to. If I want to ensure it's not doing anything that might fry a h/d -
voltage spikes perhaps...or? - is there a way to test it, or someplace
you trust to send it for testing?

Thanks.

There are all sorts of test facilities out there, but the price
they'll charge, you could easily buy half a dozen to a dozen supplies,
just for talking to them.

*******

At the power supply factory, they use various flavors of "Chroma Tester".
The tester might be able to do things, like dynamic tests, such as
a load step, and observe the results. Chroma is the brand name.
Other (cheaper) brands likely exist.

When you see such things advertised, they're shown mounted in a
rack mount form factor. A thing like this would sit next to the
test station perhaps. With computer programming, a simple "pass/fail"
can be shown on a screen somewhere, so staff can sort the good from
bad supplies. The boxes would be cabled together, and a computer
would remotely program them with the various tests to run.

http://img.directindustry.com/images...85-2703793.jpg

There are tests of things that "don't matter" to end users. One
test is the "hi pot" test, which checks for isolation between
the line and the DC outputs. It's that isolation that
prevents you from getting a shock. And it's tested to perhaps
1100 volts or so. I don't know if a Chroma tester does that, or
some other setup is used. It requires applying a potential across the
thing, to test whether it can withstand the high voltage. While that
test can just be applied to the transformer, it's also helpful
to know there are no violations on the finished assembly (PCB)
either.

*******

For a home tester, the cheapest test to do, is to build a "load box".
That puts a steady DC load, and does not create any transient conditions.
I made my own, using power resistors from my "good" electronics store.
A switch connects between PS_ON# and COM, to control the supply when
it sits on the bench. I also power a 12V fan from the supply (80mm)
and it blows over the power resistors, to remove some of the heat.

If a power supply was weak, and you connected a "representative load",
such as 200W worth of resistors, then you get some idea whether it could
power a "real" computer.

So the ingredients, include a 20 pin or 24 pin shell similar to the
motherboard connector. A bag of crimp pins, for wiring things up.
In my case, I soldered crimp pins to the power resistors, and just
plug them into the main connector. My load box does not present
a heavy load. It's not even "representative".

For each rail, you have to decide on what load to use.
Perhaps it would be 5 amps from 3.3V, 5 amps from 5V, and 10 amps from 12V.
That would be a total of 162W say. You then do the math, to figure out
what resistors to use. 3.3V/5A = 0.66 ohms. I was at the electronics
store today actually, and bought a package of 0.33 ohm resistors for $2.00.
By placing two of those resistors in series, and putting them across 3.3V and COM,
I have my load for that rail. On the 5V rail, it would be 5V/5A = 1 ohm,
and there are 1.0 ohm resistors at the store too. Actually, the stock
at my store is pretty thin, and there's only about one package of
resistors on each hook. So if I needed a whole bunch of the same
value of resistor, I'm screwed.

Those resistors are rated at 10W. I know that 3.3V*5A is 16.5 watts,
which is more than the 10W resistor is rated for. But, my plan called
for two resistors in series, the resistors have equal values, and without
any additional math, I see 8.25 watts being dissipated in each one. The
resistors will get boiling hot, and some level of moving air over them
is recommended. (The resistors have a ceramic body.) This is one reason
I didn't built a purely representative load box, because I didn't want
to have to build a good cooling solution for it. In a way, a good sized
load box, is like a "hair dryer", in terms of the heating.

In school, the favored apparatus, was a panel full of 120V light bulbs.
I've never seen the thing used, but the device was pointed to as being
a load that was used for experiments. And preferred, because of the
cheapness of light bulbs. The problem with a light bulb, is the "cold"
resistance is a lot lower than the "hot" resistance, and for any electronic
supplies, they "tip over" when the light bulb is connected. To illustrate
that, I tried connecting a 12V automotive bulb, a marker light, to a 12V @ 2A
wall wart, and the "cold" load would trip the overcurrent on the supply. So
even though the bulb draws well less than 2 amps when running, the cold
resistance draws enough current, that the overcurrent stops it immediately.
The power resistors I buy at the store, are a bit better than that, and
are rated for 5% initial tolerance, plus some degree of temperature coefficient.
The only resistance that has next to no temperature related effect is
manganin wire, and its resistance is pretty close to a constant. But
many other materials have a significant effect from cold to hot.

You can build yourself a load box, then measure the voltage with
a multimeter, but it's not much of a test. I use my load box for
testing brand new supplies. I leave them running for a couple
hours, just to weed out any supplies that "explode immediately"
when you use them. After the couple hours are up, I again measure
the voltages with the multimeter, and then conclude the supply
is safe to use with my new motherboard.

I have one supply here, with an extremely weak 12V output. Drawing
more than 0.1 amp, flattens the 12V output. My load box could be
used to evaluate such an ancient supply, and conclude it was broken.
So if your supply is weak, the load box does have a use. But for
more subtle problems, the load box is useless. A multimeter can't
see "glitches" very well. And more expensive instruments are
needed to do anything fancy.

*******

Back when hard drive controller boards, had the component side
facing outwards, you could see two transient suppressor parts, near
where the DC comes into the PCB. One on the 5V rail, one on
the 12V rail. If you were to rip the Molex out of the drive
while the drive was spinning, a small inductive arc might appear
at the power terminals. The transient suppressor, is intended to
"clip" the transient. So perhaps the 12V rail begins to clip,
when a voltage of 15V or more appears there.

Those two protection devices will burn, if a power supply overvolts.
The reason is, the transient protection devices are only intended to
handle short, high energy events. A long slow burn, will fry them.
And a power supply that has gone nuts, and put 8V on 5V or 15V on 12V,
will burn those parts. I only found out about this, when a poster
did the leg work, and traced down the part numbers, and from
that, we figured out what it was there for. At those voltages,
there's a good chance something on the controller board got
fried too. Once burned, those things would no longer
be protecting anything.

Since current controller boards, have the components facing inwards,
it's no longer possible to visually inspect for burned transient
suppressors, or burned motor controller chips. If the drive is
dead, you can always take the thing apart, to satisfy your curiosity.

Paul

On Friday, December 28, 2012 3:34:08 PM UTC-7, Doc wrote:

I have a PC Power and Cooling 470W unit that's a few years old but
hasn't had a lot of mileage on it and the rig it was in didn't come
near to stressing its limits. It was modestly pricey when I got it so
I figure it's worth seeing if it's still working the way it's supposed
to. If I want to ensure it's not doing anything that might fry a h/d -
voltage spikes perhaps...or? - is there a way to test it, or someplace
you trust to send it for testing?

I think anybody who can test it properly will charge at least $100.

Here's how XbitLabs.com does it, using their homemade load tester
consisting of several MOSFETs controlled by a computer that also
reads Hall effect sensors to measure amps:

http://www.xbitlabs.com/articles/cas...thodology.html

A simple load tester made of resistors needs to be rated for at least
twice the maximum power it will handle or else the resistors will get
hot enough to melt plastic or burn wood or skin.

I avoid frying my computer by hooking up the PSU to someone else's
computer because it's the morally right thing to do. I doubt a PC
Power & Cooling product will damage a computer, even if the PSU is
faulty, but you can check it before hooking it up by shorting the
green wire on the 20-24 pin connector to either black ground wire
next to it and measuring all the voltages with a meter, including
the +5Vstandby on the purple wire (the latter is active even without
shorting the green wire).

If you load down the PSU with just a computer, it's unlikely the
load will be more than 100W - 300W, but that should be good enough
for verifying your PSU:

http://www.xbitlabs.com/articles/cas...m-wattage.html

OTOH I think the only things in a PSU that become spoiled are the
fans (oil or grease gums up) and electrolytic capacitors, and the
latter vary greatly in how well they hold up over time. But PC
Power & Cooling isn't known for having many caps fail, even when
they use not so good brands, like Teapo (not all PSU models). So
take a look inside for any caps that are bulging on top or
corroding. However old glue used to hold caps in place can look
like corrosion, and some caps rot without bulging or leaking. Do
NOT operate the PSU with its cover removed because there's a lot
of exposed high voltage in there, sometimes even on the big heatsinks.
Replace the cover and all its screws before plugging in the AC.

On 12/28/2012 04:34 PM, Doc wrote:
I have a PC Power and Cooling 470W unit that's a few years old but
hasn't had a lot of mileage on it and the rig it was in didn't come
near to stressing its limits. It was modestly pricey when I got it so
I figure it's worth seeing if it's still working the way it's supposed
to. If I want to ensure it's not doing anything that might fry a h/d -
voltage spikes perhaps...or? - is there a way to test it, or someplace
you trust to send it for testing?

Thanks.




Generally it's as simple as looking in the bios
for voltage readout to see if the supply is in spec.

More extensive testing would require special equipment.

--
https://www.createspace.com/3707686

"Doc" wrote in message
...
I have a PC Power and Cooling 470W unit that's a few years old but
hasn't had a lot of mileage on it and the rig it was in didn't come
near to stressing its limits. It was modestly pricey when I got it so
I figure it's worth seeing if it's still working the way it's supposed
to. If I want to ensure it's not doing anything that might fry a h/d -
voltage spikes perhaps...or? - is there a way to test it, or someplace
you trust to send it for testing?

Thanks.

If you wanted to test it out of the machine, you can use something like
this:
http://www.amazon.com/Manhattan-Products-Digital-Supply-Tester/dp/B0016MHQA8/ref=sr_1_7?ie=UTF8&qid=1356785873&sr=8-7&keywords=psu+tester

If you're looking to monitor the voltages while the PC is running, you can
use a software program like HWMonitor http://www.cpuid.com/hwmonitor.php
or any number of other monitoring programs.
--
SC Tom

SC Tom wrote:


"Doc" wrote in message
...
I have a PC Power and Cooling 470W unit that's a few years old but
hasn't had a lot of mileage on it and the rig it was in didn't come
near to stressing its limits. It was modestly pricey when I got it so
I figure it's worth seeing if it's still working the way it's supposed
to. If I want to ensure it's not doing anything that might fry a h/d -
voltage spikes perhaps...or? - is there a way to test it, or someplace
you trust to send it for testing?

Thanks.

If you wanted to test it out of the machine, you can use something like
this:
http://www.amazon.com/Manhattan-Products-Digital-Supply-Tester/dp/B0016MHQA8/ref=sr_1_7?ie=UTF8&qid=1356785873&sr=8-7&keywords=psu+tester


If you're looking to monitor the voltages while the PC is running, you
can use a software program like HWMonitor
http://www.cpuid.com/hwmonitor.php or any number of other monitoring
programs.

The Manhattan one is a "little lacking in the digits department".

http://www.manhattan-products.com/en...-supply-tester

You'd be better off, in terms of general investments, in getting yourself
a multimeter. As it can be used for more purposes than a dedicated measurement.
(I use mine all the time, for checking the state of charge of my car battery.)

Devices like the Manhattan, may load one rail, and draw a tiny bit of
power, in order to keep the supply happy. But it is not a replacement
for a load box.

If the Manhattan device used an ADC that was the equivalent of what a
multimeter might offer, it might be more worthwhile. As it is, it's pretty
hard to check how close to the +/- 5% limits on the supply things might be.
And the product spec, offers no error bars on the measurement itself.

*******

This meter is 5x the price of mine, and about 100x better at doing it.
It's too bad these have to be priced out of range for hobbyists. At least
with this, you'd be pretty confident on whether each supply rail is
within the +/- 5% spec.

http://www.rigolna.com/products/digi...m3000/dm3058E/

Paul

Paul wrote:

Purely for it's comedy value, here is my load box.

http://img827.imageshack.us/img827/6915/loadbox.jpg

Ghetto gold. The 80mm fan with Molex, is missing from the photo,
but is recommended for cooling of the resistors (so you don't
burn yourself). The critical part, is the proper ATX connector
pins, which are soldered to each resistor. The pins, being
square in shape, resist rotation and tend to hold things in
place while testing. My job then, is to orient the resistors
so nothing shorts out. Some of the shrink wrap tubing is a bit
too long, and makes it hard to make the necessary voltage
measurements at the end of the test.

For someone without the odds and ends in the junk box,
starting with an ATX power supply extender cable, will provide
the necessary bits of wire, and also give a nice connector.
You can chop one end off this, and connect your load box
resistors. And the 20 pin connector, is still suitable
for testing 24 pin power supplies - just make sure to align
pin 1 to pin 1. The four pins on the 24 pin supply are
redundant, and add nothing new to the picture.

http://www.vesalia.de/pic/atxatxext.jpg

The last time I bought one locally, that cable was $8.00.

Paul

On Sat, 29 Dec 2012 09:49:57 -0500, Paul put
finger to keyboard and composed:

Paul wrote:

Purely for it's comedy value, here is my load box.

http://img827.imageshack.us/img827/6915/loadbox.jpg

Ghetto gold. The 80mm fan with Molex, is missing from the photo,
but is recommended for cooling of the resistors (so you don't
burn yourself).

I built something like that using aluminium clad power resistors.
These had their own heatsink and were rated at 25W or 50W. I screwed
them onto an aluminium plate.

http://www.rapidonline.com/catalogue...-8170P01WL.jpg

You could mount the resistors inside a square metal tube and install a
fan over the end of the tube.

- Franc Zabkar
--
Please remove one 'i' from my address when replying by email.

Franc Zabkar wrote:
On Sat, 29 Dec 2012 09:49:57 -0500, Paul put
finger to keyboard and composed:

Paul wrote:

Purely for it's comedy value, here is my load box.

http://img827.imageshack.us/img827/6915/loadbox.jpg

Ghetto gold. The 80mm fan with Molex, is missing from the photo,
but is recommended for cooling of the resistors (so you don't
burn yourself).

I built something like that using aluminium clad power resistors.
These had their own heatsink and were rated at 25W or 50W. I screwed
them onto an aluminium plate.

http://www.rapidonline.com/catalogue...-8170P01WL.jpg

You could mount the resistors inside a square metal tube and install a
fan over the end of the tube.

- Franc Zabkar

My local store doesn't carry those. If I want those aluminum ones
I have to buy them from Digikey or similar. I do have a few of the
aluminum ones, from some "surplus kits" I bought years ago. But I now
have a much larger collection of random ceramic ones. The thing
about the aluminum ones, is they can be mounted to other stuff if
you want. The ceramic ones, there's nothing to fasten to. And you
can't put a clamp around the ceramic ones and crank down on them,
because they'll crack. Whereas the aluminum ones, you have more
options.

The aluminum ones are nice when you can get them.

Paul

On Sat, 29 Dec 2012 16:08:42 -0500, Paul put
finger to keyboard and composed:

The aluminum ones are nice when you can get them.

I, too, try to make do with what I have at hand. One other possibility
is a water cooled 12 ohm resistor. That should dissipate 12W on a 12V
supply. You can find them in 1200W 120V electric kettles. They're
called heating elements. :-)

In fact you don't even need to dismantle your kettle. Just make a
suitable AC-DC adapter cable.

- Franc Zabkar
--
Please remove one 'i' from my address when replying by email.