Measuring (AC) load represented by a PC

[Don Y]

Hi,

I've got to put a number on the sort of load a particular
model PC represents in order to determine the appropriate
ampacity of the branch circuit feeding them.

How reliable would the reading from a current clamp be
in quantifying the load as seen by the breaker feeding the
circuit?

I suspect staggering the application of those loads will
go a long way towards making the breaker happy (i.e.,
putting them all on a single outlet strip with a "master
power switch" would be asking for trouble?)

Any other advice/suggestions? Looking at the PSU's in
the machines is pretty useless as that only determines an
upper limit about what they *might* eventually draw over
their lifetime (accommodating additional boards, disks, etc.).
And, of course, measuring one particular machine isn't going
to tell me much about some *other* machine...

Thx,
--don

 

[tm]

Hi,

I've got to put a number on the sort of load a particular
model PC represents in order to determine the appropriate
ampacity of the branch circuit feeding them.

How reliable would the reading from a current clamp be
in quantifying the load as seen by the breaker feeding the
circuit?

I suspect staggering the application of those loads will
go a long way towards making the breaker happy (i.e.,
putting them all on a single outlet strip with a "master
power switch" would be asking for trouble?)

Any other advice/suggestions? Looking at the PSU's in
the machines is pretty useless as that only determines an
upper limit about what they *might* eventually draw over
their lifetime (accommodating additional boards, disks, etc.).
And, of course, measuring one particular machine isn't going
to tell me much about some *other* machine...

Thx,
--don

Why don't you grab yourself one of those Kill-o-watt devices and then you
will know.

http://www.ebay.com/itm/P3-INTERNAT...ultDomain_0&hash=item1e7fea34c9#ht_1362wt_911

 

[Don Y]

Why don't you grab yourself one of those Kill-o-watt devices and then
you will know.

Can you attest to their accuracy regardless of load characteristics?
And, how their "measurements" relate to the rating on a typical
circuit breaker? (i.e., have you just traded one problem for
another?)

 

[tm]

Can you attest to their accuracy regardless of load characteristics?
And, how their "measurements" relate to the rating on a typical
circuit breaker? (i.e., have you just traded one problem for
another?)

They cost ~~ $30. Get one and test it.

I find them surprisingly accurate considering the price. Just how accurate
is your circuit breaker?

They measure power (watts), VA, PF, volts, frequency, and current.

If you need more, it's only a mater of emptying your wallet. For that, you
get to choose.

 

[Martin Riddle]

Hi,

I've got to put a number on the sort of load a particular
model PC represents in order to determine the appropriate
ampacity of the branch circuit feeding them.

How reliable would the reading from a current clamp be
in quantifying the load as seen by the breaker feeding the
circuit?

I suspect staggering the application of those loads will
go a long way towards making the breaker happy (i.e.,
putting them all on a single outlet strip with a "master
power switch" would be asking for trouble?)

Any other advice/suggestions? Looking at the PSU's in
the machines is pretty useless as that only determines an
upper limit about what they *might* eventually draw over
their lifetime (accommodating additional boards, disks, etc.).
And, of course, measuring one particular machine isn't going
to tell me much about some *other* machine...

Thx,
--don

If you like to play with calculators...
<http://www.power-on.com/calculator.html>

Most modern PC's have a PFC front end, so throwing on a current clamp
will give accurate results.


Cheers

 

[Martin Riddle]

They cost ~~ $30. Get one and test it.

I find them surprisingly accurate considering the price. Just how accurate
is your circuit breaker?

They measure power (watts), VA, PF, volts, frequency, and current.

If you need more, it's only a mater of emptying your wallet. For that, you
get to choose.

There are two versions, one has a battery backup and one does not.
I think neither has a backlit display.
Still usefull either way.

Cheers

 

[mike]

The devil is in the details.
Based on the Kill-O-Watt recommendation below
Here's what my 2.8GHz. quad-core intel does.

State Watts Volt-amps
off 3 7
booting 120 174
Idle 88 132
Sleep 10 20

A newer system would more likely have a power factor
corrected power supply. I don't have anything that new.

That doesn't include either of the two 24" monitors.
93W 95VA each.
Or the desk lamp.
Or the electric heater I turn on to keep my feet warm
when it's cold.
Or the laser printer when it fires up.A current clamp can give you accurate results. Just be sure
it's one with a spec and it's measuring what you want to measure.
They don't all measure RMS. If you care about peaks, RMS may
not be what you want.

I've compared my Kill-O-Watt to my Valhalla 2101 Digital Power
analyzer and found them to read almost exactly the same
for the power factors encountered here.How much averaging are we talking about?
Averaging over 300 devices is more likely to work than
trying to average over 5.

I don't think you have much opportunity to average the peak
load that the breaker cares about. There are events that make
everything happen at once. 4AM virus updates. 1PM when everybody
comes back from lunch. 8AM when they arrive.
Power glitch that reboots everything. Heat in the wire won't
be as much a problem, but the NEC won't let you put 15amp wire
on a 20amp breaker anyway.

Lost work for 15 people due to tripping a breaker is VERY expensive.
Your job won't survive many of those.
Back when I was trying to figger out the electrical code,
I ran across some anecdotes. Don't trust me, look it up...
but as I recall, 3-phase building power is designed around
resistive loads.
With low power factor loads, like computers, there's a dramatic
increase in the current in the neutral. Fires are not uncommon
when old buildings get updated with computers. Someone with the
actual facts can probably weigh in on this.
Looking at the PSU's inMy experience is that most computers needed by an office worker
draw about 150 watts or so..just the computer box, not the display
and other peripherals. And it's easy to sample what you have now.
And there are green forces trying to drive that down.
Yes, you can easily build a system that takes 10x that much, but
it should be obvious if you have any serious computing going on.I think your best bet for general advice is a computer vendor.
Dell can probably give you case studies covering a wide range of
scenarios.

 

[Don Y]

They cost ~~ $30. Get one and test it.

AGAINST WHAT??? Even if I test against a set of 0.1% resistive loads
covering a 0-20A range, that doesn't tell me how this correlates to
what the *breaker* thinks "trip current" should be!

I find them surprisingly accurate considering the price. Just how
accurate is your circuit breaker?

Circuit breaker is *the* standard. Doesn't matter if your
meter claims 19.032A are being drawn and the breaker *shouldn't*
have tripped.

 

[Don Y]

Hi Mike,

The devil is in the details.
Based on the Kill-O-Watt recommendation below
Here's what my 2.8GHz. quad-core intel does.

State Watts Volt-amps
off 3 7
booting 120 174
Idle 88 132
Sleep 10 20

A newer system would more likely have a power factor
corrected power supply. I don't have anything that new.

I don't have anything that *efficient*! :> (gobs of
fans, redundant power supplies, 4-16 spindles, etc.)

That doesn't include either of the two 24" monitors.
93W 95VA each.
Or the desk lamp.
Or the electric heater I turn on to keep my feet warm
when it's cold.
Or the laser printer when it fires up.

I first have to determine what *else* is on the circuit(s)
in question. I'll pay them a visit tomorrow and start
poking around. Of course, difficult to do when folks are
actually *using* the office ("Gee, I wonder what's on
*this* circuit..." "Hey! Who turned off the coffee pot?!")

A current clamp can give you accurate results. Just be sure
it's one with a spec and it's measuring what you want to measure.
They don't all measure RMS. If you care about peaks, RMS may
not be what you want.

What I care about is whatever the breaker will ultimately
base "its decision"! I don't want to start plugging in PC's
to see when/if the breaker trips -- anymore than I'd plug
in electric frying pans to determine how many I can support
on a given counter circuit! :>

I've compared my Kill-O-Watt to my Valhalla 2101 Digital Power
analyzer and found them to read almost exactly the same
for the power factors encountered here.
How much averaging are we talking about?
Averaging over 300 devices is more likely to work than
trying to average over 5.

Staggering, not averaging. I.e., don't put all machines on the
same "outlet strip". Don't let machines automatically restore
power after an outage. Etc.

E.g., when I spin up the drives in my arrays, they don't all
try to apply power to the spindle motors simultaneously. You
can tolerate a few extra seconds in the power up sequence
for the array. And, you can tolerate a few extra seconds in
the powering up of multiple computers!

I don't think you have much opportunity to average the peak
load that the breaker cares about. There are events that make
everything happen at once. 4AM virus updates. 1PM when everybody
comes back from lunch. 8AM when they arrive.

These are used by students. Unlikely that they'll all decide
to "drop in" at the same *instant*. Also, I suspect someone
else will turn things on in the morning and shut them down at
night.

Power glitch that reboots everything. Heat in the wire won't
be as much a problem, but the NEC won't let you put 15amp wire
on a 20amp breaker anyway.

Don't let machines restart unattended.

Lost work for 15 people due to tripping a breaker is VERY expensive.
Your job won't survive many of those.

Not my job :>

But, that doesn't mean I don't want to consider factors that
would make them unhappy with the results ("pro bono" doesn't
imply "incompetent")

Back when I was trying to figger out the electrical code,
I ran across some anecdotes. Don't trust me, look it up...
but as I recall, 3-phase building power is designed around
resistive loads.
With low power factor loads, like computers, there's a dramatic
increase in the current in the neutral. Fires are not uncommon
when old buildings get updated with computers. Someone with the
actual facts can probably weigh in on this.

I *believe* they occupy an older, remodeled building. But, I
haven't yet looked to see how "current" the wiring is. On my
last visit, I had other primary concerns (hadn't even been
inside the facility). I will be more observant, tomorrow.

(And, better equipped to get into more pertinent details)

Looking at the PSU's in
My experience is that most computers needed by an office worker
draw about 150 watts or so..just the computer box, not the display
and other peripherals.

Yes. The problem lies in the fact that people take power
for granted. I.e., replacing a 17" monitor with a 21".
Adding another printer so B&W output can use cheaper supplies.
Tacking on USB accessories. Adding a second disk drive.

When you are dealing with *one* machine, its no big deal
(usually). But, adding a second disk to a dozen machines
is now an extra amp on the supply...

And it's easy to sample what you have now.
And there are green forces trying to drive that down.

If you "upgrade" and have a say in that "upgrade"...

 

[Phil Allison]

"Don Y"

I've got to put a number on the sort of load a particular
model PC represents in order to determine the appropriate
ampacity of the branch circuit feeding them.

How reliable would the reading from a current clamp be
in quantifying the load as seen by the breaker feeding the
circuit?

** Providing it is a " true rms " clamp meter AND you open up a three core
lead so the clamp goes around only one conductor - it will give you the
long term current that operates the thermal part of the breaker.

I suspect staggering the application of those loads will
go a long way towards making the breaker happy (i.e.,
putting them all on a single outlet strip with a "master
power switch" would be asking for trouble?)

** The inrush surges produced by most SMPS are a real hazard for breakers,
quantifying them and relating that to the published curves is a nonsense
idea.

IME, if the RMS current of the total load is close to the breaker's rating,
then the inrush surge will trip the same breaker at least occasionally.

The inrush surge even with a given SMPS varies over a wide range, depending
on the part of the AC cycle existing at switch on AND the recent history of
use.



..... Phil

 

[Don Y]

Hi Martin,

If you like to play with calculators...
<http://www.power-on.com/calculator.html>

Most modern PC's have a PFC front end, so throwing on a current clamp
will give accurate results.

I'll see what sort of numbers I get tomorrow. I can always make
an educated guess seeing what's inside the box and how well that
correlates to "measurements".

Most important thing is to evaluate alternatives so they have
*a* solution -- even if it is not the solution they originally
expected.

 

[Phil Allison]

"Don Y"

It also depends on the characteristics of the "overload".

** Meaningless pedantry.

The RMS current value of the load is the only thing that trips a thermal
breaker.

Unstated but obvious assumptions are that the RMS current level is steady
and there is no unusual heating of the breaker itself.



.... Phil

 

[[email protected]]

Can you attest to their accuracy regardless of load characteristics?
And, how their "measurements" relate to the rating on a typical
circuit breaker? (i.e., have you just traded one problem for
another?)

If you don't believe one, test it with a 100W light bulb. They're
fairly accurate (couple of percent).

 

[tm]

If you don't believe one, test it with a 100W light bulb. They're
fairly accurate (couple of percent).

That plus you could get a NIST traceable HP 3458A for somewhere in the $6k
range and know what it is doing to 8 1/2 places.

 

[Jasen Betts]

Hi,

I've got to put a number on the sort of load a particular
model PC represents in order to determine the appropriate
ampacity of the branch circuit feeding them.

How reliable would the reading from a current clamp be
in quantifying the load as seen by the breaker feeding the
circuit?

depends if your meter can handle the crest factor.

 

[Robert Baer]

I second that, been using one now for a year.
I did use it to see the difference in power consumption of various
PC configurations (plus monitor).

I doubt one can do that in a cheaper simpler safer way for the same money,
it is a mass product.
The total over a year about matched what the big meter says.

Note "rating" on a circuit breaker is NOT one number - most
especially not the tag value!
Reasonable ratings of circuit breakers are like those of fuses: a set
of curves, time VS current.

 

[Phil Allison]

"Robert Baer"

Note "rating" on a circuit breaker is NOT one number - most especially
not the tag value!

** Yes it is - you illiterate, bullshitting wanker.

The nominal amp rating is the *never open* RMS value.

Reasonable ratings of circuit breakers are like those of fuses:

** Same as above.

You stupid shit.



..... Phil

 

[Neon John]

Can you attest to their accuracy regardless of load characteristics?

Yes, as a matter of fact I can. I've spent considerable time
characterizing the KAW. I have a Fluke multimeter calibration bench
and a transfer standard watt-hour meter as references.

The KAW does high speed sampling of both the voltage and the current
shunt and computes true RMS values on the fly. The algorithm starts
to break down with a PF <0.5 but then we don't see that very often.

The KAW is firmly in the 1% class. At room temperature and with a low
enough load that the shunt doesn't heat the electronics too much, it
is typically better than 0.5%.

The KAW is so accurate because each unit is individually characterized
on the production line and a set of calibration factors are loaded.

And, how their "measurements" relate to the rating on a typical
circuit breaker? (i.e., have you just traded one problem for
another?)

Circuit breakers in general are quite accurate considering their cost
and construction. I've spent quite some time characterizing the cheap
chicom breaker that we use in our induction heater product. It's the
shortest delay model they make this side of no delay. It will conduct
15.0 amps indefinitely. It will trip in a few minutes at 15.25 amps.
Close enough for me.

John

John DeArmond
http://www.neon-john.com
http://www.fluxeon.com
Tellico Plains, Occupied TN
See website for email address

 

[Robert Baer]

"Robert Baer"


** Yes it is - you illiterate, bullshitting wanker.

The nominal amp rating is the *never open* RMS value.




** Same as above.

You stupid shit.



.... Phil

Such a mouth-foaming idiot!

 

[[email protected]]

That plus you could get a NIST traceable HP 3458A for somewhere in the $6k
range and know what it is doing to 8 1/2 places.

;-)