[conspire] How accurate, etc.: AC Power (Watt)meter (What's Your Watt?/...)

Wed Apr 22 09:11:24 PDT 2015

Michael,
Thank you for the very good explanation of the subtleties of AC power.
But the manual does indeed have a section on specifications:
    Specifications
        5 ft. cord connecting display and socket
        Electrical Rating: 120V~/15A/60Hz/1800W
        Maximum Watts Displayed: 1800
        Maximum Dollars Displayed: 9999
They don't state _anything_ about measurement accuracy. As someone who has had to read the many pages of detailed specs for an oscilloscopes and voltmeters, it makes me wonder what they are hiding.

The reality is that most of our homes consume a fair amount of power even when everyone is sleeping.  Obviously the refrigerator needs to cycle.  But the microwave and coffee maker consume a constant trickle just to keep their clocks on.  Then there are the assorted networking routers, etc.   On occasion, my wife has asked if I have turned off the computer, because several LEDs are always on, unless I turn off the wall outlet.  

Actually I wonder what is the difference in power between "OFF" and "SLEEP"?   The monitor screen is black.  The disk drives are off.  The only visible difference is the "power" light which is off when "OFF" and slowly blinking in "SLEEP".

      From: Michael Paoli <Michael.Paoli at cal.berkeley.edu>
 To: conspire at linuxmafia.com 
 Sent: Sunday, April 12, 2015 10:21 AM
 Subject: [conspire] How accurate, etc.: AC Power (Watt)meter (What's Your Watt?/...)

Well ... it varies.  ;-)

But yes, good points - not so much that "they" or "it" is necessarily
inaccurate or that inaccurate, but good to generally be aware of how
(in)accurate one's measurements are.

Most test instruments typically have specifications, that well spell out
their accuracy.  And yes, the specifications generally are quite
accurate - and especially from reputable manufacturing companies.  That
doesn't mean the specifications are always 100% correct, but they're
generally intended to be correct, and for the most part are.

That being said, some related bits ...

First, I was curious on the "What's Your Watt" - specifications?  How
accurate *is* it?
A trace of sleuthing and I found ...
http://www.sfenvironment.org/whatsyourwatt
http://cache-www.belkin.com/support/dl/web_f7c005_10ms041_conserveinsight.pdf
... apparently Belkin Conserve Insight
... as far as I can tell, seems they have, for USA(/Canada) power
compatible versions, 3 very slightly different versions of this device:
Part # F7C005 http://www.belkin.com/us/support-product?pid=01t80000002zwZFAAY
Part # F7C005q http://www.belkin.com/us/support-product?pid=01t80000002HrhcAAC
Part # F7C005fc  
http://www.belkin.com/us/support-product?pid=01t80000002If1gAAC
But I also notice, all 3 have the exact same User Manual:
Conserve Insight F7C005 v1 - User Manual (US) (English)
http://cache-www.belkin.com/support/dl/web_f7c005_10ms041_conserveinsight.pdf
which is exact same URL also found further above.
In all the checks, I didn't find specifications that covered the
accuracy of its power measurements! - somewhat surprising.  But it does
give wee bit about it its inaccuracy at (very) low power levels:
"
Your Conserve Insight uses sensitive
electronics to measure the amount of power
flowing to your connected device. When
the connected device is consuming a very
small amount of electricity, it becomes
very difficult to distinguish between the
power being consumed and the electrical
noise found at all power outlets. To ensure
accurate readings, the Conserve Insight
stops showing values below 0.5 watts and
displays the '0-0.5 Watt' screen.
"

Anyway, notwithstanding the lack of more detailed specification (it does
gives specifications, but alas, doesn't state how accurate the power
measurements are), it is from a reputable instrument manufacturer, so
it's probably *fairly* accurate ... at least compared roughly to
comparable test instruments (e.g. electronic measurement instrument in
roughly the $15 to $50 USD price range ... and in probably about 2012 or
so).  So, a good guestimate might be something in the range of roughly
+-2% to +-5%, or so.  Since Belkin also quite implies it can't
accurately measure <~=0.5 Watts (W), one might guestimate accuracy of
something like the worse of:
+-2% to +-5%, or 0.5W, whichever is greater.

But more on how (in)accurate.  Measuring Watts isn't as simple as
measuring Volt-Amperes(Amps) (VA).  Watts - or true power consumption,
is the *instantaneous* product of Volts and Amps - and since that
continuously fluctuates for Alternating Current (AC), that's generally
given averaged over some reasonable interval (one or more full cycles,
e.g. for 60 Hertz ((Hz) Cycles Per seconds - CPS), one second would
cover 60 cycles).  The difference between VA and Watts, is VA is mere
product of the Root Mean Square (RMS) values for each of V and A.  Note
that VA and Watts can give *very* different numbers.  W is the true
power consumption, whereas VA is not - though knowing or also knowing
VA can be useful in certain contexts, and sometimes/commonly, VA may be
a rough, to even good, approximation of W (and knowing the VA is
generally better than having no W information).  Without explaining RMS
in detail, think of it as "equivalent".  If one has a pure resistive
load, let's say of 1 Ohm, applies a Direct Current (DC) Voltage of 1 V,
then one has current of 1A, and power consumption of 1W.  If instead of
DC, and AC voltage is applied, regardless of what waveshape applied, if
the AC RMS voltage is 1V, the RMS current will be 1A, and the power
will be 1W (at least averaged over any full cycle).  Note also here,
talking theoretical resistor (same resistance regardless of
current/voltage).  Anyway, that's a key reason why RMS values are so
useful - for their relative equivalence.  Better electrical instruments
will give true RMS values, other less accurate ones will give an
estimate, typically taking some other measurement and presuming the
waveshape is sinusoidal (most commonly it's fairly close to sinusoidal
- particularly for common power situations, but that's not always the
case), and will then give a reading based upon that presumption.

Anyway, W vs. VA ... phase matters.  The more out-of-phase they are,
the more they'll differ.  Bit of crude analogy - think of pushing
someone on a swing - regularly push as they start the downward portion
- that's in phase - adding energy to the system - to build their
swinging - or at least offset losses from various friction and such.
If instead, one pushes as they're approaching towards the top of their
swinging motion towards you, that's out-of-phase, and instead of adding
energy to the system, is taking energy out, and reducing their
swinging.  One might be doing the same amount of pushing in either
case, but not only how efficiently energy is flowing - but even which
direction - depends on the phase relationship.  Such is the case with
AC.  Measuring one's pushing, while ignoring phase, would be akin to
measuring VA (or at least one of its components), rather than properly
measuring and determining W.

Ye olde power company electric meters.  I forget the precise details of
how they do it (I read the description many decades ago), but they very
much measure W, not VA.  They very effectively and quite efficiently,
do a continuous "instantaneous" electromechanical multiplication of the
V and A - and thus measure true W.  Well, ... *almost* instantaneous.
The electrical field parts of it they use to do that are ... less some
slight bits for inductive lag and ... well, speed of light 'n
electricity 'n such, ... but also bit of mechanical inertia and such.
But interesting too on the mechanical inertia, there's also critical
damping (or nearly so) involved - and Eddy currents.  Well, there's
that aluminum disk that spins - shows power consumption.  But how to
have it stop, and not just keep sinning to only slowly slow down once a
load stops?  Eddy currents.  Some permanent magnets are placed very
close to that aluminum disk - that induces Eddy currents, which oppose
the motion.  Works rather effectively as critical damping or fair
approximation thereof - takes work to spin that disk between those
magnets - that work is effectively a "motor" ... but quite effectively
a W motor in this case - with speed proportional to the W it's
measuring ... but when the W drops or is cut off, the disk still has
those Eddy currents to slow it down - so it rather quickly slows or
stops to reflect changes in power consumption - but still not quite
instantaneous.  But, also being mechanical and all - and again,
inertia, and Eddy currents, it also doesn't spin up or spin faster
instantaneously with additional power consumption.  Those bits of not
quite instantaneous pretty well cancel each other out, so the overall
measured power consumption is quite accurate - and especially since
it's mostly used for determining kilowatt hours on an approximately
monthly basis, rather than instantaneous (kilo)watt power consumption.
However, now with "Smart Meter"s and such, though still not
instantaneous, often down to rather short intervals (e.g. 5 minute
average, or less) of average power consumption are available.

So ... Belkin Conserve Insight ... does it really measure and
accurately display W, or is it really VA?  Don't know, but I'd
guestimate it's probably W or a rather to quite good approximation
thereof.  A known significantly out-of-phase load of sufficient size
might be a way to easily determine that.  And, whether VA or W, in any
case, how accurate?  Well, I do have test instruments that can rather
accurately measure RMS A, and RMS V.  I could use that to get VA, but
measuring W isn't as easy (I can't simply multiply those for W, as the
measurements I'd have would be averaged rather than instantaneous).

Oh, also, I don't think, in general, one will improve the accuracy of
measuring power consumption of a small AC load, by starting with a
larger load, adding a smaller load, then noting the difference in total
power consumption.  Most notably, for most typical loads, the load
won't be constant - notably varying by Voltage - which does fluctuate.
E.g. that 7.5W incandescent night light one may have (before someone
makes 'em illegal or whatever), it's actual power draw will likely vary
a fair bit with voltage fluctuations, so adding a load such as that to
try and determine the power draw of a smaller load, may actually give
one less accurate information about the power draw of the smaller load.
A more accurate approach would be to apply multiples of that smaller
load (plug in many such identical devices under identical use status
and conditions) ... but alas, that may not be feasible or convenient.

Wee bit more on phase.  For larger commercial/industrial customers,
power companies will often charge them an out-of-phase penalty or
surcharge.  Why?  Because out-of-phase costs the power companies money.
The power company's meter measures kilowatt-hours (kW-h) - or
multiples thereof.  That's actual power consumed by customer.  However,
out-of-phase energy is not free for the power companies to provide.
Let's take a theoretical example.  Large commercial industrial customer
hooks up a large purely inductive theoretical load to the power
company meter.  Meter reads, and bills for 0 (zero) - no power consumed.
But there is current and VA.  And the lines the power company uses to
deliver the current - they're relatively long, for the most part, and
they're not some theoretical zero resistance lines (though some power
companies are doing some very limited trial stuff with power delivery
over superconductors - at least from what I seem to recall having read
some while back).  So ... current, resistance, volts ... power.  Power
company in such case is paying (producing power) to heat up their
lines, yet billing customer for zero power consumption.  Hence penalty
for out-of-phase - to incentivize customer to be using in-phase
current, and thus not only billable power, but more efficient for the
power company, as that reduces waste current and waste heat on
transmission lines, and optimally at fully in-phase, that waste is
minimized.

> Date: Wed, 18 Mar 2015 10:11:26 -0700
> From: Ross Bernheim <rossbernheim at gmail.com>
> To: Conspire List <conspire at linuxmafia.com>
> Cc: Nick Moffitt <nick at zork.net>
> Subject: Re: [conspire] check out AC power (Watt)meter (What's Your
>     Watt?/...) from    library
> Message-ID: <584DF350-97C9-43EF-8C9A-68CC76B69E5A at gmail.com>
>
> One thing to remember about the Kill A Watt and similar consumer  
> power/watt meters
> is that they are meant for lights and appliances and while they will  
> easily measure
> a refrigerator or 100W light bulb, they are not accurate at low power levels.
>
> The accuracy at low levels, typically below seven watts is not  
> terribly useful. As
> we move to LED light bulbs that only draw a few watts and very low power
> computers with solid state drives these meters may give inaccurate results.
>
> One suggestion is to use a multi-outlet power strip and pair the  
> device you want
> to check with another device such as a light bulb that is a constant  
> load above
> ten watts so that you can accurately measure it then add the load you want to
> test and measure the increased load accurately.
>
> All test equipment has limitations and you need to know what they are so
> you can assure the accuracy of your measurements.
>
> Ross
>
>> On Mar 18, 2015, at 3:44 AM, Nick Moffitt <nick at zork.net> wrote:
>>
>> Michael Paoli:
>>> One can check out AC power Wattmeter (What's Your Watt?) from
>>> library.
>>
>> These devices are really useful for installations where you have a lot
>> of devices running.  You can accurately measure median loads, sample and
>> get a rolling average, and take note of peak load (typically on startup
>> for devices that have spinning rust).
>>
>> It's also amazing to look at two identical pieces of hardware and notice
>> that they have dramatically different power loads. Often you can trace
>> that to software load differences, and make adjustments as necessary.
>>
>> But of course the best reason is for energy savings and capacity
>> planning.  You can work out that your mains bus can handle N servers at
>> normal load, and N/4 at peak load, or what have you.  Then you know to
>> only start up a quarter of your systems at any one time, and stagger
>> boots (or just increase capacity to cover full synchronised peak load).

_______________________________________________
conspire mailing list
conspire at linuxmafia.com
http://linuxmafia.com/mailman/listinfo/conspire

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://linuxmafia.com/pipermail/conspire/attachments/20150422/7fd7f769/attachment.html>