I run a number of computers at home, and was wondering how much
electricity each one consumes. Is there some kind of device that can be
attached between the PC and the wall outlet to measure power
It's likely to be significantly (double the real reading) inaccurate,
the one I got from maplin is.
It works fine on simple loads like motors/heaters.
In practice, you can look up all the figures on websites.
2-300W is unlikely.
I measured my server (12 drives spinning, 1.3G duron) at around 150W.
That's discouraging. So according to your post, there's an error of a
FACTOR OF 2? How could one know what the real power consumption is...
Maybe I can calibrate it somehow, using devices with known power
consumptions. But I can't think of any such devices...
The meters (the cheap ones, the maplin and Lidl 6.99 ones) work out the
current by measuring the voltage/current several times per cycle.
Unfortunately, the lowest cost is got by reducing 'several' as much as
For things like heaters, or motors, you can almost trivially get a pretty
good reading of power and power-factor (amount current is leading or
lagging voltage) with just 3 or 4 samples per cycle.
This is because the current they draw is a nice sine wave.
Many PC (and other switched-mode PSUs) unfortunately have significant
very fast variations in current.
So, the meter misses these changes, and assumes that the power supply is
drawing a higher current than it should.
It's basically a similar problem to trying to join up the dots.
If you've got lots of dots, then you can make a nice smooth curve, without
worrying about the shape of the curve you make being different from the
As you reduce the number of dots, eventually you get to a stage when
More expensive meters should do better, unfortunately, determining how much
goes into the electronics is hard.
I bought one of the current Maplin ones, and I think that might
not be too bad. I'm having trouble getting my proper true power
meter back from someone I've lent it to (my boss;-), so I haven't
been able to do a real check on the Maplin one.
I bought one of the older ones Maplin stocked a few years ago,
and that's hopeless on non-resistive loads (out by a factor of
3 on SMPSU's -- too high, which is a puzzling way for it to be
My proper true power meter actually uses an analogue multiplier
and integrator, so it has more than enough dots so there's no need
to guess how the line goes between them ;-)
Yes, that's the crap one which is way off -- I think the model was
PM230. The one Maplin stock now is a different make and I can't
remember what it is (nothing well known though, but it is Asian
Yes, I guess if the spike width at the voltage peaks is small
compared with the sampling rate (hello Mr Nyquist;-) and the
sampling is somehow always catching it (prehaps due to synching
with the voltage waveform), then that would account for it.
Isn't it as much to with low power factor as anything. The
cheap/simple power meters simply measure average (and I mean average,
not RMS) voltage and current, then multiply the two together and apply
the right factor to convert to RMS and say the product is power. It
matters not one whit how fast your samples are if you do this, the
answer will always be wrong.
Even the cheap ones do power factor.
I'm fairly confident that most of these devices have a single channel
8 bit A/D (minimum current is typically 40mA = 13000mA/256), sampling at
400Hz or so. (or maybe 50Hz, with undersampling, I'm not sure.)
Add a 4 bit micro, and an LCD, and you've got a power meter.
A 0 power factor device can have a very spiky current draw.
My PC power supply draw looks like a sharks fin, with a very sharp rise, and
a slow decay to a fast fall.
Typical older SMPSs with a bridge rectifier feeding a capacitor (which
read correct) have a current draw that looks like the top of a sine wave.
But *how* do you measure "total energy consumption over a long
period", that's the original problem! If the meter's view of
instantaneous power consumption says 1kW thenit's going to consume
(according to the meter) 9999kWh in 9999 hours (unless the
instantaneous reading changes of course).
If the draw looks like a sharks fin, then the instantaneous reading
*will* vary won't it? Unless the sample times are synchronised to the
Never mind. I'm quite happy to believe that the readings are unreliable
for PC power supplies etc.
The example was a SMPSU, not a capacitive or inductive load.
Averaging a SMPSU current pulse or failing to take into account
it only happens when mains voltage is around 340V would give
an under reading. I can't imagine what mistaken assumption
they've made which results in such a load 3 times over reading.
To quote from an earlier post I made on this.
A quick patch involving a 1.8 ohm resistor in the neutral lead and
dividers protected by zeners indicates that the current has a really
The couple of switchers I checked exhibited the expected short
looking pulses at maximum voltage.
The power indicated by the meter was within 20% of my eyeballing the
On trying it on my PC, I found a rather odder shape sort of sharks-fin
The current rises to a peak of 2.2A in .4ms, then decays over the next
of milliseconds more or less linearly to 0.6A, then drops in 0.5ms to
That's a total of 2.25Ams per half-cycle, at 320V, that's .721Ws per
half-cycle, or 72W when idle.
The meter however reads 138W.
And in suspend, it's around 30W, around 10W 'off'.
It may well be assuming that the current will be either a 'normal' sine
wave, with a power factor of whatever or a sharp chopped-sine pulse near the
This worked fine, until various regulations about harmonic content on power
supplies came in.
Then the PSU makers started working out the cheapest way to meet these
This ended up as very odd waveshapes.
As to why.
Well, let's assume that it samples at the peak of this pulse, 2.2A.
Let's try 2.2A * 2ms = 4.4mAs.
4.4mAs * 320V = 1.4J
* 100 cycles per second = 140J/s, or 140W.
Surprisingly this works quite well.
However, I suspect the real answer is that it's trying to make a nice smooth
wave out of it.
To get an accurate number, you're going to need to sample at better than
Alternatively, if it's sampling at 500Hz or so, and trying to assume that
the waveshapes may be symmetrical (be it intentionally, or through filters)
then it may get vastly misleading answers.
Especially as we're not talking about a 2000 quid scope that's had hundreds
of man-years spent on the design, and verifying it's correct, but maybe a few
hundred hours and checking on a few loads around the office.
Anyway, hypothesising on how it fails is kind of pointless, in the face of
the fact that it can't be repaired (practically)
HomeOwnersHub.com is a website for homeowners and building and maintenance pros. It is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.