First thing to note is that OWL-type meters are not terribly accurate especially with 'strange'/electronic loads.

1) I can imagine that either those smoke alarms may be inefficient vampires, or that the meter is reading their consumption far too high. Up to 3-fold would not astonish me.

2) Again, those controls may not be optimised for efficiency, or they may be OK with the meter reading far too high.

3) With your office at the end of the garden: is it on an earth-leakage circuit breaker? If not, it probably should be for safety. If so, does it trip out frequently? What I'm getting at is possible genuine leakage from damp or similar.

For any of these are they plugged in such that you could temporarily plug them in via a potentially far more accurate 'Kill-a-Watt' style meter, possibly borrowed from your local library for free?

Rgds

Damon

Agree with Damon
You can, once established what it is reading when no current is drawn (except its own current), use that as an offset and minus the figure in a spreadsheet.
I use a CurrentCost device and it has a Phantom reading of 144W on one circuit when another circuit kicks in. Very annoying as it can show that there is a parasitic load of that amount when the storage heaters are off.

Good for getting an idea about what is going on though.

Thanks all!
The OWL has now stopped reading at all, so it's lost further credibility since I posted.
To answer the queries -
Damon: we'll look into what kind of circuit breaker there is, but it's never tripped. It was done about 10 years ago and seemed quite a good job, but you never know. The problem areas are all wired in, so unfortunately a Kill-a-Watt wouldn't help.
GaryB: No MVHR :( Our proposed Passivhaus is still mired in planning dispute with neighbours.
SteamyTea: It's battery powered and reads 0.00 when I put off the main fuse switch so I think it's accurate there.
I'll go back to trying to get it to work at all, shortly.

JSH: I assume that the biggest errors from OWL-et-al are more to do with assuming (a) a particular RMS voltage because it isn't actually measured and (b) assuming that it's in phase with the current that *is* measured which for much non-linear electronic gear just ain't so.

But I agree that quantisation errors will hurt too.

Rgds

Damon

Posted By: DamonHDJSH: I assumed that the biggest errors from OWL-et-al are more to do with assuming (a) a particular RMS voltage because it isn't actually measured and (b) assuming that it's in phase with the current that *is* measured.

...and (c) assuming that the consumed current is a sine wave which it's very likely not with a cheapo switching power supply which is not fully loaded (e.g., phone charger plugged into phone which is charged). Possibly the same for Claire's smoke alarms.

OWL has woken up again now, while I was doing something else (putting up gazebo for Jubilee street party)
This has gotten a bit beyond my technical competence, but I think the message is to manage my expectations of the gadget and not worry too much about what are probably false readings?
JSH - a) Am I understanding right that the quantisation errors would be random noise, so if something is consistant it probably isn't that?
b) If the device is battery powered, how would it know to auto-zero when the current it is reading is switched off? That's probably still too technical; I just need to know that it will do it.

If I were designing one of these things, and prepared to accept that there was an inherent large error from not measuring voltage or PF, then I wouldn't bother to make the resolution of the measurement system much better than the overall accuracy that results from the limitations inherent in the chosen measurement method.

The basic error budget gives the following figures:

1) Voltage error = +10% -6% (based on the allowable limits either side of the nominal 230 V UK mains supply voltage). For a domestic supply with the OWL quoted maximum current of 71 A, this gives a power error at unity PF of +1,633 W to -980 W

2) Power Factor error. Assuming that the house has an overall PF of around 0.9 (good, but perhaps not untypical - a pretty well-balanced house might get a PF of around 0.95) then measuring just current would give a further error of around +10%

The built-in error budget (as a result of the chosen measurement method) gives a total error of around +20% to +4%. If we assume that the designers of the OWL have corrected out the fixed PF offset of around +10%, assuming that the PF will be around 0.9 (which would seem the sensible thing to do) then we're left with the voltage error plus any PF error from the house PF being greater or less than 0.9.

If we assume that we need to measure to a resolution that's ten times better than the notional accuracy (which seems reasonable) then we need a measurement resolution of better than about 100 W (assuming we use the lower of the two voltage error figures).

There are large numbers of microcontrollers around with built in 10 bit analogue to digital converters, in fact there are very few around that are better than this at the affordable end; 12 bit devices tend to be a fair bit more expensive, and have more demanding power supply requirements, so aren't very common. As it happens, 10 bits is more than good enough, as it can resolve differences that are far, far smaller than the likely error.

There's no point in using a better measurement resolution with the chosen measurement method, this would only make sense if the system used voltage measurement and did true RMS power measurement, allowing for power factor errors.

My guess is that they over-sample and average (it's what I'd do). This can increase the apparent resolution, depending on how you choose to do it, and so mask the true resolution of the A-D conversion. They may even do some fancy digital filtering to get a more display-friendly apparent resolution.

I've got an OWL and find it pretty wildly inaccurate, especially below about 200-300W. It's going to be getting replaced soon, as it's also not able to detect exported power. Still been quite a handy gizmo, as it'll still detect things like the wife leaving heaters on when they shouldn't be. Pretty hard to miss an extra 2kW of juice getting slurped.

<blockquote><cite>Posted By: yclairejenkins</cite>I noticed something like that(step wise increments) with one I had before, which was also much more awkward to use as it needed mains for the display. This one seems to have 0.01kw resolution as it reads 0.02, 0.03, 0.04 at different times.</blockquote>

My guess is that what it does is over-sample (basically it'll take maybe 50 samples, then divide by 50 to get the average) and then round the result to the nearest 10 W for the display.

I do something similar with the temperature/humidity logger I've built to convert temperature readings with a true resolution of 0.0625 deg C to one with a resolution of 0.1 deg C, by rounding up or down as required (i.e. if the true reading is 21.125 deg C the logger rounds down to 21.1 deg C, if the true reading is 21.1875 deg C the logger rounds up to 21.2 deg C.

Worth remembering that resolution (the apparent accuracy displayed in terms of minimum step size) is often significantly different to true accuracy.

The building at the end of the garden doesn't seem to have any RCD on its fusebox. If it continues to give a reading with everything switched off, perhaps we should get an electrician to look at it.
I guess the smoke alarms and boiler controls, etc, we will just live with.

Posted By: SteamyTeaI think Skywright worked out out that the Current Cost device over estimates about 15%.

Yes, that's what I find, with our particular local combination of typical daily loads and mains voltages and looking at daily readings.

The overnight usage suggested by the CurrentCost is around 150% of (my best assessment of) actual. I spent quite a lot of time trying to find that extra 50% vampire load before coming to the conclusion that it was a phantom not a vampire!

N.B. The above applies to "normal" mixed domestic loads. When monitoring pure restive loads (i.e. our storage heaters) I find that the CurrentCost under reports.

Really I think they are more so that people can see what a bad idea leaving 10 halogen lights on 24x7 is, ie the first 90% of the waste, not the last few Watts.

I think they're fine for that purpose.

Rgds

Damon

The over-reading of the small loads did encourage me to track them down using a plug-in meter. I found quite a number of 1W to 7W 'standby' loads which were subsequently dealt with.

Perhaps if they had been accurately displayed I wouldn't have gone looking for them and therefore would not have found them...

The worst offender BTW was the Sky box which draws 13W - 15W when on 'auto standby' which in my opinion is a disgrace for a piece of modern equipment. Changing from Sky Plus to Sky HD+ made little difference.

I eventually found out that if you hold down the 'off' button on the remote control the box can go into proper standby <1W and the light turns red. If the light on the box is amber it's wasting energy. Manufacturers please wake up!

I have mine on a timer which shuts it off overnight just in case.

<blockquote><cite>Posted By: DamonHD</cite>...though of course what we *should* possibly be charged for is yet a third quantity, taking account of power factor as larger users have to (or pay for), since it does otherwise represent a loss in distribution...

Rgds

Damon</blockquote>

I agree, but as I understand it new houses may be penalised if the have a PF of less than 0.95. Quite how this will be enforced through life remains to be seen.

The assumption that's been made up till now is that the domestic sector PF would average out over the local phase loop (the houses connected to a single phase from the sub-station), on the basis that the current returned to the network by some houses will be utilised by other houses, and the resulting lagging/leading PFs will cancel. I suspect that with the widespread adoption of switched mode power supplies for everything from washing machines to computers and TVs this assumption is now deeply flawed. The widespread use of CFLs (and maybe LEDs) with switched mode supplies adds to the problem.