Desktop PC power supply

B

Bob Engelhardt

Guest
I have an old PC that recently lost its PS. An identical one was hard
to find, but I didn't know the PC's power requirements, so I couldn't
use a generic PS. I should say that I thought I couldn't use a generic one.

I measured the PC's power use so next time I'll be ready. It was quite
surprising - the PS is "400W", but the PC only uses 54W. And there's
40A of 5V available, but only 5A are used. Etc.

The biggest surprise was the PS's efficiency, or lack thereof. A
Kill-a-watt on the input showed 200w being used, and 54w output. 25%
more or less. I thought that a SMPS would be way more efficient than that.

Bob
 
In article <o99c4q015kg@news6.newsguy.com>, BobEngelhardt@comcast.net
says...
I have an old PC that recently lost its PS. An identical one was hard
to find, but I didn't know the PC's power requirements, so I couldn't
use a generic PS. I should say that I thought I couldn't use a generic one.

I measured the PC's power use so next time I'll be ready. It was quite
surprising - the PS is "400W", but the PC only uses 54W. And there's
40A of 5V available, but only 5A are used. Etc.

The biggest surprise was the PS's efficiency, or lack thereof. A
Kill-a-watt on the input showed 200w being used, and 54w output. 25%
more or less. I thought that a SMPS would be way more efficient than that.

Bob
Click to expand...

The energy has to go somewhere. If your measurement is to be trusted
(real RMS?) then the PSU fan must be working like a hairdryer!

And what do you mean "lost its PS"? It vanished in a puff of smoke,
leaving not even a specification plate behind?

Mike.
 
In article <o99c4q015kg@news6.newsguy.com>, BobEngelhardt@comcast.net
says...
I have an old PC that recently lost its PS. An identical one was hard
to find, but I didn't know the PC's power requirements, so I couldn't
use a generic PS. I should say that I thought I couldn't use a generic one.

I measured the PC's power use so next time I'll be ready. It was quite
surprising - the PS is "400W", but the PC only uses 54W. And there's
40A of 5V available, but only 5A are used. Etc.

The biggest surprise was the PS's efficiency, or lack thereof. A
Kill-a-watt on the input showed 200w being used, and 54w output. 25%
more or less. I thought that a SMPS would be way more efficient than that.

Bob
Click to expand...

I don't have a Kill-O-Watt but do have a meter similar to it. That one
reads out Volts, Amps,and Watts. I have noticed that if I hookit to a
resistive load like a heat gun the volts and amps are very close to the
watts. If I hook it to a reactive load there is a big difference in the
Watts vers the volts times amps. That I understand.

I have not paid any attention to a switching supply that way. Is it
possiable the the watts and the V times I are not the same for your
meter ?
 
On Thu, 02 Mar 2017 09:58:26 -0500, Bob Engelhardt
<BobEngelhardt@comcast.net> wrote:

The biggest surprise was the PS's efficiency, or lack thereof. A
Kill-a-watt on the input showed 200w being used, and 54w output. 25%
more or less. I thought that a SMPS would be way more efficient than that.
Click to expand...

I'll assume it's an ATX style power supply.
<http://www.corsair.com/en-us/blog/2012/august/80-plus-platinum-what-does-it-mean-and-what-is-the-benefit-to-me>
"Intel’s ATX specification only requires that a power supply
is 60% efficient at 50% load. Most decent quality power
supplies made in the last decade are around 70% efficient
at 50% load."

If you're getting 25%, either something is wrong, but I can't tell if
it's your measurements, a defective design, a failed power supply, or
how you're measuring the efficiency. I've made similar measurements
of an ATX power supply using a dummy load on the +5v and +12V lines,
and a Kill-a-watt meter. Typically, I obtained 75% efficiency at
about 50% of full load on a cheap ATX PS.

As MJC mentioned, the heat has to go somewhere. If your efficiency at
at 50% of maximum load (200 watts) is only 25%, then you should have
150 watts of hot air and flames coming out the back end of the PS with
the temp controlled fan spinning furiously. Something is wrong.

Photo of one of my loads:
<http://802.11junk.com/jeffl/pics/drivel/slides/dummy-load-inductance.html>
--
Jeff Liebermann jeffl@cruzio.com
150 Felker St #D http://www.LearnByDestroying.com
Santa Cruz CA 95060 http://802.11junk.com
Skype: JeffLiebermann AE6KS 831-336-2558
 
On Thu, 02 Mar 2017 08:13:54 -0800, Jeff Liebermann <jeffl@cruzio.com>
wrote:

On Thu, 02 Mar 2017 09:58:26 -0500, Bob Engelhardt
BobEngelhardt@comcast.net> wrote:

The biggest surprise was the PS's efficiency, or lack thereof. A
Kill-a-watt on the input showed 200w being used, and 54w output. 25%
more or less. I thought that a SMPS would be way more efficient than that.

I'll assume it's an ATX style power supply.
http://www.corsair.com/en-us/blog/2012/august/80-plus-platinum-what-does-it-mean-and-what-is-the-benefit-to-me
"Intel’s ATX specification only requires that a power supply
is 60% efficient at 50% load. Most decent quality power
supplies made in the last decade are around 70% efficient
at 50% load."

If you're getting 25%, either something is wrong, but I can't tell if
it's your measurements, a defective design, a failed power supply, or
how you're measuring the efficiency. I've made similar measurements
of an ATX power supply using a dummy load on the +5v and +12V lines,
and a Kill-a-watt meter. Typically, I obtained 75% efficiency at
about 50% of full load on a cheap ATX PS.

As MJC mentioned, the heat has to go somewhere. If your efficiency at
at 50% of maximum load (200 watts) is only 25%, then you should have
150 watts of hot air and flames coming out the back end of the PS with
the temp controlled fan spinning furiously. Something is wrong.

Photo of one of my loads:
http://802.11junk.com/jeffl/pics/drivel/slides/dummy-load-inductance.html
Click to expand...

I was going to respond to the OP, but Jeff beat me to it. Switching
supplies are notorious for having very poor power factors under light
load. I looked up the specs on the kill-a-watt and it is supposed to
read power factor correctly when seeing a pure sine wave. Of course,
there is no guaranty the current waveform is a pure sine wave. It is
more likely to have spikes all over the place. (That's one of the
reasons switching supplies can cause RF noise.) My guess (and that's
all it is) is your supply is using under well under 100 watts, but
your kill-a-watt is confused by the strange current waveform and says
it is using 200 watts.
 
On Thursday, March 2, 2017 at 6:58:36 AM UTC-8, Bob Engelhardt wrote:
I have an old PC that recently lost its PS. An identical one was hard
to find,
... the PS is "400W", but the PC only uses 54W.

The biggest surprise was the PS's efficiency, or lack thereof. A
Kill-a-watt on the input showed 200w being used, and 54w output. 25%
more or less. I thought that a SMPS would be way more efficient than that.
Click to expand...

Unless your machine is a loud-fans-and-multiple-XEONs type, 54W is a likely
power draw, and 200W is unlikely.

It is possible (on old Apple computers, it's common) to repair a dead power
supply. They're cramped and sometimes have unlabeled tiny chips, but
a check for burst capacitors and shorted rectifiers, and of course fuses, will often
bring 'em back to life.

If you find shorted power FETs, be aware that there's a gate drive which might
have failed, too; those repairs get complicated.
 
Bob Engelhardt wrote:

The biggest surprise was the PS's efficiency, or lack thereof. A
Kill-a-watt on the input showed 200w being used, and 54w output.
Click to expand...

** A Kill-A-Watt reads "watts", power factor and "VA".

What does yours give for PF & VA ?


...... Phil
 
In sci.electronics.repair, on Thu, 02 Mar 2017 08:13:54 -0800, Jeff
Liebermann <jeffl@cruzio.com> wrote:

On Thu, 02 Mar 2017 09:58:26 -0500, Bob Engelhardt
BobEngelhardt@comcast.net> wrote:

The biggest surprise was the PS's efficiency, or lack thereof. A
Kill-a-watt on the input showed 200w being used, and 54w output. 25%
more or less. I thought that a SMPS would be way more efficient than that.

I'll assume it's an ATX style power supply.
http://www.corsair.com/en-us/blog/2012/august/80-plus-platinum-what-does-it-mean-and-what-is-the-benefit-to-me
"Intel’s ATX specification only requires that a power supply
is 60% efficient at 50% load. Most decent quality power
supplies made in the last decade are around 70% efficient
at 50% load."

If you're getting 25%, either something is wrong, but I can't tell if
it's your measurements, a defective design, a failed power supply, or
how you're measuring the efficiency. I've made similar measurements
of an ATX power supply using a dummy load on the +5v and +12V lines,
and a Kill-a-watt meter. Typically, I obtained 75% efficiency at
about 50% of full load on a cheap ATX PS.

As MJC mentioned, the heat has to go somewhere. If your efficiency at
at 50% of maximum load (200 watts) is only 25%, then you should have
150 watts of hot air and flames coming out the back end of the PS with
the temp controlled fan spinning furiously. Something is wrong.

Photo of one of my loads:
http://802.11junk.com/jeffl/pics/drivel/slides/dummy-load-inductance.html
Click to expand...

Wow. I want one of those.

The meter I mean.

It's a shame I have no use for it. In all these years I've never
wanted to measure inductance and I've never needed to measure
capacitance.

But what is the diff between these two things:
https://www.amazon.com/AideTek-multimeter-capacitance-Inductance-self-discharge/dp/B0187LWDBM/ref=sr_1_1?ie=UTF8&qid=1488498176&sr=8-1&keywords=dm4070+lcr+meter
and
https://www.amazon.com/AideTek-multimeter-capacitance-Inductance-self-discharge/dp/B01N2KDB18/ref=sr_1_2?ie=UTF8&qid=1488498176&sr=8-2&keywords=dm4070+lcr+meter

Same brand, same model, but one is 2 1/2 times the price!

Interestingly, the more expensive one takes much longer to ship, has no
ratings, and only one picture, compared to 7. Different vendor. How
do they hope to sell it. (I didnt' mean these differences. I thought
maybe they're not actually the same.)

They both save Vici in the upper right but afaict, yours doesn't.
 
On 3/2/2017 10:13 AM, MJC wrote:

The energy has to go somewhere. If your measurement is to be trusted
(real RMS?) then the PSU fan must be working like a hairdryer!
Click to expand...

It would take a very long time to dry your hair with it. 150w vs 1500w
in hair dryer. Yeah, the PSU exhaust is warm, but 150w is pretty easy
to blow away.

And what do you mean "lost its PS"? It vanished in a puff of smoke,
leaving not even a specification plate behind?
Click to expand...

It died. The problem was matching ALL the name plate specs. There are
a lot of combinations.
 
On 3/2/2017 10:50 AM, Ralph Mowery wrote:

I don't have a Kill-O-Watt but do have a meter similar to it. That one
reads out Volts, Amps,and Watts. I have noticed that if I hookit to a
resistive load like a heat gun the volts and amps are very close to the
watts. If I hook it to a reactive load there is a big difference in the
Watts vers the volts times amps. That I understand.

I have not paid any attention to a switching supply that way. Is it
possiable the the watts and the V times I are not the same for your
meter ?
Click to expand...

I assumed that the Kill-a-watt does true rms, but possibly not.
 
On 3/2/2017 11:13 AM, Jeff Liebermann wrote:

I'll assume it's an ATX style power supply.
http://www.corsair.com/en-us/blog/2012/august/80-plus-platinum-what-does-it-mean-and-what-is-the-benefit-to-me
"Intel’s ATX specification only requires that a power supply
is 60% efficient at 50% load. Most decent quality power
supplies made in the last decade are around 70% efficient
at 50% load."

If you're getting 25%, either something is wrong, but I can't tell if
it's your measurements, a defective design, a failed power supply, or
how you're measuring the efficiency. I've made similar measurements
of an ATX power supply using a dummy load on the +5v and +12V lines,
and a Kill-a-watt meter. Typically, I obtained 75% efficiency at
about 50% of full load on a cheap ATX PS.

As MJC mentioned, the heat has to go somewhere. If your efficiency at
at 50% of maximum load (200 watts) is only 25%, then you should have
150 watts of hot air and flames coming out the back end of the PS with
the temp controlled fan spinning furiously. Something is wrong.

Photo of one of my loads:
http://802.11junk.com/jeffl/pics/drivel/slides/dummy-load-inductance.html
Click to expand...

I measured the current (inline meter) on all the MB power lines and the
hard drives power lines. As I posted, the input was from a Kill-a-watt.
One divided by the other for efficiency.

Although the PSU is using 200w, it's rated for 400w output, so my 54W is
only 13% of rated power. (The name plate input is 10A at 115v). Maybe
efficiency goes to hell for really small loads.

My intuition is that 2 75W light bulbs in a box could be cooled by a PSU
fan. Without flames.

It's kinda wonky, but not crazily so.

Bob
 
On 3/2/2017 11:45 AM, Pat wrote:
I was going to respond to the OP, but Jeff beat me to it. Switching
supplies are notorious for having very poor power factors under light
load. I looked up the specs on the kill-a-watt and it is supposed to
read power factor correctly when seeing a pure sine wave. Of course,
there is no guaranty the current waveform is a pure sine wave. It is
more likely to have spikes all over the place. (That's one of the
reasons switching supplies can cause RF noise.) My guess (and that's
all it is) is your supply is using under well under 100 watts, but
your kill-a-watt is confused by the strange current waveform and says
it is using 200 watts.
Click to expand...

So, "maybe, maybe not" on the Kill-a-watt number. My scope is packed
away waiting for its once-a-year use or I'd look at the PSU input current.
 
On 3/2/2017 7:34 PM, whit3rd wrote:
Unless your machine is a loud-fans-and-multiple-XEONs type, 54W is a likely
power draw, and 200W is unlikely.
Click to expand...

Very modest machine. And I'm much more confident about the output
measurements.

It is possible (on old Apple computers, it's common) to repair a dead power
supply. They're cramped and sometimes have unlabeled tiny chips, but
a check for burst capacitors and shorted rectifiers, and of course fuses, will often
bring 'em back to life.

If you find shorted power FETs, be aware that there's a gate drive which might
have failed, too; those repairs get complicated.
Click to expand...

I did actually try to repair the old one. I don't remember many
specifics except that when I got a big spark and a burned trace I gave up.
 
In article <o9abh111jk1@news6.newsguy.com>, BobEngelhardt@comcast.net
says...
On 3/2/2017 10:50 AM, Ralph Mowery wrote:

I don't have a Kill-O-Watt but do have a meter similar to it. That one
reads out Volts, Amps,and Watts. I have noticed that if I hookit to a
resistive load like a heat gun the volts and amps are very close to the
watts. If I hook it to a reactive load there is a big difference in the
Watts vers the volts times amps. That I understand.

I have not paid any attention to a switching supply that way. Is it
possiable the the watts and the V times I are not the same for your
meter ?

I assumed that the Kill-a-watt does true rms, but possibly not.
Click to expand...

I did hook up the meter I have and the volts times amps was showing 120
watts but the power in watts was showing about 75 watts when hooked to a
switching power supply. Did not measure the output to the load.
 
On Thu, 02 Mar 2017 19:12:38 -0500, Bob Engelhardt
<BobEngelhardt@comcast.net> wrote:

I measured the current (inline meter) on all the MB power lines and the
hard drives power lines. As I posted, the input was from a Kill-a-watt.
One divided by the other for efficiency.

Although the PSU is using 200w, it's rated for 400w output, so my 54W is
only 13% of rated power. (The name plate input is 10A at 115v). Maybe
efficiency goes to hell for really small loads.
Click to expand...

I didn't have time for a proper test today. I just stuffed the
kill-a-watt meter into the bench power strip and used it to power an
HP 8200 Elite SFF desktop. Something like this with an i3 CPU:
<http://h20564.www2.hp.com/hpsc/doc/public/display?docId=emr_na-c03412787>
<http://h20564.www2.hp.com/hpsc/doc/public/display?docLocale=en_US&docId=emr_na-c02781555>
I think it has power factor correction for:
320 W active PFC 87%/90%/87% efficient at 20%/50%/100% load.
The input power consumption varied from 30 watts to 70 watts depending
on what the machine was doing. Eyeball average was about 50 watts.
Playing a video burned 70 watts. The power factor varied from 0.90 to
0.96. However, I was unable to obtain VA and watt reading pairs to
determine if the PF was for real.

However, I had no way to measure the current at the motherboard and
drives. I also couldn't find my ATX dummy load. I'll throw something
together this weekend if I have time.

My intuition is that 2 75W light bulbs in a box could be cooled by a PSU
fan. Without flames.
Click to expand...

I don't use light bulbs as loads (except for UPS testing). Mostly, I
use power resistors and head sinks. My load is intended to simulate a
game machine, which typically burns about 350 watts with a high end
video card installed. At that power level, I need either a fan or a
bucket of water to cool the heat sink.

>It's kinda wonky, but not crazily so.

Well, like I suggested... something is wrong. It's unlikely that your
PS has PF correction or is highly efficient. However, it shouldn't be
as low as 13%. Something is wrong, but I can't tell where from here.

If you have the time and inclination, you might try testing a known
good ATX power supply and see if the numbers magically improve. If
the efficiency is fairly high, then the original power supply is
broken somewhere. However, if it too shows 13% efficiency, then
there's something wrong with either your instruments, or your
methodology.

--
Jeff Liebermann jeffl@cruzio.com
150 Felker St #D http://www.LearnByDestroying.com
Santa Cruz CA 95060 http://802.11junk.com
Skype: JeffLiebermann AE6KS 831-336-2558
 
On Thu, 02 Mar 2017 18:51:08 -0500, micky <NONONOmisc07@bigfoot.com>
wrote:

Wow. I want one of those.
The meter I mean.
Click to expand...

I think I bought it from this vendor probably 3 years ago:
<http://www.ebay.com/itm/322363697016>
It's a tolerable LCR meter. I bought it for a specific project
(inductive vehicle detector) and then bought 2 more (for my office and
vehicle).

It's a shame I have no use for it. In all these years I've never
wanted to measure inductance and I've never needed to measure
capacitance.
Click to expand...

You must live a sheltered life. Haven't you ever made your own
inductors or capacitors and needed to measure the inductance or
capacitance? In RF, that's common practice. Even simple things like
estimating the length of coax cable or CAT5 by measuring the
capacitance and dividing by capacitance per foot, is useful. Then
there are loading coils, filters, drifting caps, etc. Also, if you
trust the values printed on the parts, you might be in for a rude
surprise when you test them.

Incidentally, I also use an earlier version of one of these M328
meters:
<http://www.ebay.com/itm/252290685197>
<http://www.banggood.com/M328-LCD-12864-Transistor-Tester-DIY-Kit-Diode-Triode-Capacitance-LCR-ESR-Meter-p-1041588.html>
I prefer a curve tracer for testing and matching active devices, but
this is good enough for a quick test. It also does LCR. There are
many variations. Google for "M328":
<https://www.google.com/search?q=m328&tbm=isch>

--
Jeff Liebermann jeffl@cruzio.com
150 Felker St #D http://www.LearnByDestroying.com
Santa Cruz CA 95060 http://802.11junk.com
Skype: JeffLiebermann AE6KS 831-336-2558
 
On 3/2/2017 10:51 PM, Phil Allison wrote:
** A Kill-A-Watt reads "watts", power factor and "VA".

What does yours give for PF & VA ?
Click to expand...

It's consistent: VA = 310+-, PF=.69, W=210+-. So, it knows that there's
difference & it knows how to divide, but it still might not know how to
do true RMS power.
 
On 3/3/2017 12:48 AM, Jeff Liebermann wrote:
On Thu, 02 Mar 2017 19:12:38 -0500, Bob Engelhardt wrote:

... HP 8200 Elite SFF desktop. ...
The input power consumption varied from 30 watts to 70 watts depending
on what the machine was doing. Eyeball average was about 50 watts.
...
Click to expand...

It's highly improbable that your machine is weaker than mine, so your
50W in versus my 200W means that mine is way too high. Unless my
Kill-a-watt is wonky.

My intuition is that 2 75W light bulbs in a box could be cooled by a PSU
fan. Without flames.

I don't use light bulbs as loads (except for UPS testing). ...
Click to expand...

I just used the light bulbs as a common example of power. I have an
intuitive feel for the power of a light bulb & if it could be kept in
limit with a fan. Not so much a ceramic power resistor.

It's kinda wonky, but not crazily so.

Well, like I suggested... something is wrong. It's unlikely that your
PS has PF correction or is highly efficient. However, it shouldn't be
as low as 13%. Something is wrong, but I can't tell where from here.
...
Click to expand...

My 13% is my 54W draw versus the 400W rating of the PSU.

The PSU name plate ratings: 1200W in (10A), 400W out. I'm using 200W
in, 54W out (25% efficient). 54 / 400 = 13% rated load. But even at
12% rated load, the efficiency shouldn't be 25%.

BTW - I don't really care, I'm just curious.
 
In article <2u0ibc52ot4skkdnoudsvf372tivkcb8sc@4ax.com>,
jeffl@cruzio.com says...
Incidentally, I also use an earlier version of one of these M328
meters:
http://www.ebay.com/itm/252290685197
http://www.banggood.com/M328-LCD-12864-Transistor-Tester-DIY-Kit-Diode-Triode-Capacitance-LCR-ESR-Meter-p-1041588.html
I prefer a curve tracer for testing and matching active devices, but
this is good enough for a quick test. It also does LCR. There are
many variations. Google for "M328":
https://www.google.com/search?q=m328&tbm=isch
Click to expand...

There are many versions of that device and similar ones. I have one
that was already built for about the same price I bouth several years
ago. Too bad they did not have things like that around 20 or 30 years
ago when I was working on equipment all the time.
Retired now and just do electronics for a hobby, but it is nice to be
able to get some of the equipment to play with. Especially in the $ 10
to $ 50 range.
 
On Fri, 03 Mar 2017 01:46:08 -0500, Bob Engelhardt
<BobEngelhardt@comcast.net> wrote:

On 3/3/2017 12:48 AM, Jeff Liebermann wrote:
On Thu, 02 Mar 2017 19:12:38 -0500, Bob Engelhardt wrote:

... HP 8200 Elite SFF desktop. ...
The input power consumption varied from 30 watts to 70 watts depending
on what the machine was doing. Eyeball average was about 50 watts.
...

It's highly improbable that your machine is weaker than mine, so your
50W in versus my 200W means that mine is way too high. Unless my
Kill-a-watt is wonky.
Click to expand...

That's a possibility. I have several kill-a-watt meters. When I
upgraded to the latest greatest, I took them apart to determine the
difference.
<http://802.11junk.com/jeffl/pics/drivel/slides/kill-a-watt.html>
Much the same except for the clock crystal, 5 instead of 4 buttons,
different uP, and some beefier components.

However, I was having problems with erratic readings with the old
model (top photo) which I traced to marginal soldering on the AC power
connector pins and current sense "resistance wire". I resoldered
everything that looked suspicious and the problems disappeared. You
might want to disassemble yours and check for similar soldering
problems.

Incidentally, the marginal soldering in the later unit (bottom photo)
is stock. I later went through it and resoldered all the rough
conections. That's when I found that the ground pin on the AC power
plug had a cold solder connection. That should have any effect on the
readings, but it does suggest that the soldering problem might be
common.

--
Jeff Liebermann jeffl@cruzio.com
150 Felker St #D http://www.LearnByDestroying.com
Santa Cruz CA 95060 http://802.11junk.com
Skype: JeffLiebermann AE6KS 831-336-2558
 
You must log in or register to reply here.

Welcome to EDABoard.com

Sponsor

Back
Top