Measuring standby power

[rr100]

Measuring the standby power on AC mains devices is not as easy as it sounds and many consumer devices (Kill-A-Watt types but also just using an ammeter in series with the load) give a value that is much higher than in reality, sometimes with more than an order of magnitude.

What would be a proper way to measure this without a dedicated equipment? I don't need ultimate precision, even 30% off would be ok.

Probably the ultimate would be to just measure the current (over a series shunt) with a scope while triggering from the mains but I can't say I particularly like this idea (both because you need to probably write a program to get you the result from the resulting waveform and also because of the safety issues, although you probably can get away by using a transformer to isolate the scope from the shunt you try to measure).

Any other ideas?

[TriodeTiger]

I know nearly nothing, but..

A current transformer? I'd assume the average power could be found through the average current, and the average current remains a constant while in standby. Find the peak through a peak detector, and multiply by the RMS voltage of course.

[tszaboo]

There are analog front ends to measure mains power consumption. I was using something like the ADE7768 to do this once upon a time. It is really easy, you just put a shunt on the mains, and also divide it so the IC can measure it. The output is pulsed, so you only need an optocoupler and an universal counter on the "safe side" to do accurate measurements. Of course if you dont have a counter, you can put together any digital logic counter, Arduino or whatever, you only have one LED which will blink every now and then depending on the consumption.
I guess I dont have to mention to be careful if you design something like this, use proper layout, protect your inputs and your life etc...

[rr100]

TriodeTiger: NO, this is precisely the pitfall I try to avoid. You can't assume much with these black-box power supplies and the peak method can give you EASILY 200W for something that takes basically nothing. A "true-RMS" ammeter would do better but still can show considerably more than real.

That ADE sounds certainly promising.

[NiHaoMike]

Use an isolation transformer, a DSO, and a relatively high value shunt resistor.

[djacobow]

You might want to check out this thread:

https://www.eevblog.com/forum/testgear/yokogawa-cw-10/

I originally posted about wanting to measure standby power on a budget. It is not easy.

There are a handful of < $500 devices that can do the job, but really, anything quality is north  of $2k.

One of the more intriguing suggestions, which I have yet to try is to modify a Kill-A-Watt by removing the 2mOhm (or so) current shunt and replacing it with a 200 mOhm (or so) current shunt. Device should read approximiately 100x more sensitive than it is. Obviously, you don't want to plug something that draws 15 A into it after modification, but it should be good for generating low power measurements. Probably very inaccurate, but better than nothing.

[rr100]

As I said I want something that can be even 30% off, but I don't want to be order(s) of magnitude off (well, up to a point, I have no problem if 0.0001W is shown as 0 or 0.1W).

Any caveats with the Kill-A-Watt mod? Not safety-wise but I'm thinking these switched power supplies really don't behave as any theoretical nice inductive (for example) load so what is the Kill-A-Watt doing, is it actually sampling the current many times while minding where the voltage is in the cycle or something much simpler just assuming some specific waveform for the current?

[tszaboo]

is it actually sampling the current many times while minding where the voltage is in the cycle or something much simpler just assuming some specific waveform for the current?

If you check the IC mentioned by me, that ie samples at 450Khz. After the inductive filtering of a cable you can pretty much see any spikes caused by any switching supply.

[NiHaoMike]

One of the more intriguing suggestions, which I have yet to try is to modify a Kill-A-Watt by removing the 2mOhm (or so) current shunt and replacing it with a 200 mOhm (or so) current shunt. Device should read approximiately 100x more sensitive than it is. Obviously, you don't want to plug something that draws 15 A into it after modification, but it should be good for generating low power measurements. Probably very inaccurate, but better than nothing.

You can add a pair of inverse parallel diodes for overload protection. Then it might be good for checking standby draw of a device that initially draws a lot of current.

[The Electrician]

You might want to check out this thread:

https://www.eevblog.com/forum/testgear/yokogawa-cw-10/

I originally posted about wanting to measure standby power on a budget. It is not easy.

There are a handful of < $500 devices that can do the job, but really, anything quality is north  of $2k.

One of the more intriguing suggestions, which I have yet to try is to modify a Kill-A-Watt by removing the 2mOhm (or so) current shunt and replacing it with a 200 mOhm (or so) current shunt. Device should read approximiately 100x more sensitive than it is. Obviously, you don't want to plug something that draws 15 A into it after modification, but it should be good for generating low power measurements. Probably very inaccurate, but better than nothing.

I did some more measurements on the modified Kill-a-Watt mentioned in the referenced thread.

First, I measured an LED light bulb with the modified Kill-a-Watt, and also with the power measurement application on my Tek scope.  Here's the scope capture of the measurement.  The green trace is line voltage, purple is bulb current, and the red trace is instantaneous power.  The modified Kill-a-Watt measured 5.97 watts:



Next I measured the no-load power consumption of a small switcher wall wart (it's the charger for my electric razor) without a load.  The power consumption is very small and as can be seen from the scope capture, the current waveform has a very high crest factor.  The modified Kill-a-Watt measured .067 watts, 1.5 mA current draw, and .178 VA:



The modified Kill-a-Watt is doing a good job of measuring the true power with these quite non-sinusoidal current waveforms.  It reads a little low but I'm not going to bother trying to trim the .2 ohm shunt; I just mentally increase the reading by about 2%.

[rr100]

ADE7768 sounds like the way to go here; also cheap and probably they'll even send free samples but the pitch seems to be outside my comfort zone for a "quick one". However I would consider it for any dirt cheap integrated solution (-> serial, bluetooth, etc). Any idea where to find the evaluation board (or some other similar product, because they mention it in the datasheet but is nowhere to be found)?

Other than that I REALLY appreciate the Kill-A-Watt idea AND the research with the scope, many thanks! I only have the Rigol DS1052E that can do Math AxB but I can't seem to find a way to average that in the Rigol (to get the power). So I'll have to just guesstimate or go through the computer somehow.

Now let's see where to find a cheap "kill-a-watt" clone (I'm in Europe so I'll have to settle for something else and see for myself how it works).

[djacobow]

Yup. I haven´t cracked mine open yet. When I do I´ll see if diodes like that are already there. I mean, they need to protect from spikes anyway.

You can add a pair of inverse parallel diodes for overload protection. Then it might be good for checking standby draw of a device that initially draws a lot of current.

[NiHaoMike]

There aren't any since the original shunt can take a very large amount of current without overheating (a PCB trace would probably fail first!), but a 0R2 would not be as robust.

I just modified one of mine with a 0R2, but also put in a switch to change between 100x and 1x (original) mode. In 1x mode, the 0R2 is bypassed and the input routed to the original shunt. In 100x mode, the 0R2 is enabled (with a pair of Schottky diodes in inverse parallel to handle overloads) and the input routed to that shunt.

[mjkuwp]

What level of standby power are you talking?  What makes you say that meters are an order of magnitude off?   how do you know the error exists?

You can use the eevblog uCurrent ! and a DMM if the load is really low and you take the necessary precautions. - that will give you the current and depending how you do the grounding it seems you could get that signal to a scope.

[NiHaoMike]

The uCurrent is designed for low voltage use (low insertion resistance) and would be easily damaged by current spikes (e.g. charging of the capacitors in a switching supply). In addition, the insertion resistance of a common multimeter is already insignificant for low current mains, but that won't help as AC power measurement requires reading both voltage and current and vector multiplying the values.

[rr100]

What level of standby power are you talking?  What makes you say that meters are an order of magnitude off?   how do you know the error exists?

You can use the eevblog uCurrent ! and a DMM if the load is really low and you take the necessary precautions. - that will give you the current and depending how you do the grounding it seems you could get that signal to a scope.

I am talking about the standby for totally normal devices from media players to TVs and maybe even wall-warts, something like 0.1-5W maybe.

Please take a look at the second scope capture/picture. There is no way to read a good power value based on measuring the current without sampling the voltage the very same instant (for the very simple reason that the voltage isn't constant). The usual ways to still do it (as in measure the power with the ampermeter) are:

- take the peak value and assume that's good for the average, divide it (I think) by sqrt(2) and multiply by Vrms. This would give in this case a huge value of power compared with the actual (because the peak is short)
- take the RMS value of the current and multiply by Vrms. This would give here a somehow lower value compared to actual because the peak current (and most current and power consumption in fact) is at peak voltage, not at average. However, if for some reason the peak current is when the voltage is really low there is no limit at how much we would OVERestimate the actual power

The uCurrent is absolutely not needed and is extremely poorly suited for this application. The uCurrent isn't mainly solving the problem of measuring low currents (that's trivial, just use a big enough resistor as a shunt), instead is solving the problem of actually having some voltage to measure if you want enough resolution. Having a 0.5V drop across a shunt might be a problem when analyze a device powered by a 1.2V NiMh but is absolutely of no importance for something powered by mains. In this case the uCurrent can be fully replaced by a simple resistor. In fact the resistor is much safer, simpler and has almost infinite bandwidth.

[mjkuwp]

What level of standby power are you talking?  What makes you say that meters are an order of magnitude off?   how do you know the error exists?

You can use the eevblog uCurrent ! and a DMM if the load is really low and you take the necessary precautions. - that will give you the current and depending how you do the grounding it seems you could get that signal to a scope.

I am talking about the standby for totally normal devices from media players to TVs and maybe even wall-warts, something like 0.1-5W maybe.

Please take a look at the second scope capture/picture. There is no way to read a good power value based on measuring the current without sampling the voltage the very same instant (for the very simple reason that the voltage isn't constant). The usual ways to still do it (as in measure the power with the ampermeter) are:

- take the peak value and assume that's good for the average, divide it (I think) by sqrt(2) and multiply by Vrms. This would give in this case a huge value of power compared with the actual (because the peak is short)
- take the RMS value of the current and multiply by Vrms. This would give here a somehow lower value compared to actual because the peak current (and most current and power consumption in fact) is at peak voltage, not at average. However, if for some reason the peak current is when the voltage is really low there is no limit at how much we would OVERestimate the actual power

The uCurrent is absolutely not needed and is extremely poorly suited for this application. The uCurrent isn't mainly solving the problem of measuring low currents (that's trivial, just use a big enough resistor as a shunt), instead is solving the problem of actually having some voltage to measure if you want enough resolution. Having a 0.5V drop across a shunt might be a problem when analyze a device powered by a 1.2V NiMh but is absolutely of no importance for something powered by mains. In this case the uCurrent can be fully replaced by a simple resistor. In fact the resistor is much safer, simpler and has almost infinite bandwidth.

afraid there is a bit of misunderstanding due to assumptions I made, sorry.  I agree with your points and understand them except I don't agree the uCurrent is poorly suited...

As you said, one must sample V and I at the same time at a high rate and multiply them together.  Digital scopes will do this so I guess the questions was more around how to make the connections.  I am building a power meter mostly for hobby and  have used the uCurrent to calibrate the I RMS portion of the meter.  I use a 4-terminal shunt resistor in my design.  The chip ADE7953 is used for all the calculations.  There are other energy metering chips that you could use. 

I think the idea of using the Kill-A-Watt with a different shunt resistor value is a really good one if you find a way to bypass the current around the meter when the device is not in standby. 

...there are many ways - it all depends whether you are trying to get one measurement, or design a product, or do this for your own education, etc..

[dannyf]

sampling the voltage the very same instant (for the very simple reason that the voltage isn't constant).

That is generally true, and need a fairly fast adc as well.

Some mcus have two independent adc modules so that can be done.

An alternative, using one adc, is to perform sequential but alternating samples (volt->current->volt->current...), relative to the zero-crossing point. Once you have enough data points accumulated, you can realign them in software.

[macboy]

TriodeTiger: NO, this is precisely the pitfall I try to avoid. You can't assume much with these black-box power supplies and the peak method can give you EASILY 200W for something that takes basically nothing. A "true-RMS" ammeter would do better but still can show considerably more than real.

That ADE sounds certainly promising.

A True RMS ammeter can help give you a quite accurate view of the VA (apparent power) but not of the Watts (real power). You pay for Watt-hours, not VA-hours. Inductive or capacitive loads use more VA than W, which means that the power factor is not 100%, i.e. means current is out of phase with voltage.  Using separate RMS ammeter and RMS voltmeter will only tell you VA, not W. You need a wattmeter that can internally multiply the instantaneous voltage and current waveforms, and average the result.

[HackedFridgeMagnet]

An mcp3901 ic can do this, obviously quite a bit of work involved though.

[rr100]

I agree with your points and understand them except I don't agree the uCurrent is poorly suited...

Can you please find at least one single reason why you would use a uCurrent and not a resistor?

[mjkuwp]

a resistor is fine. 

Just that I happened to have the uCurrent Gold on my desk and it has the banana jacks on it.  My DMM has banana jacks on it so it is an easy fit.   I have a BNC to Banana jack adapter for my scope.  I don't have a sense resistor other than the ones soldered into my circuit.

[The Electrician]

Now let's see where to find a cheap "kill-a-watt" clone (I'm in Europe so I'll have to settle for something else and see for myself how it works).

I saw this one referenced in another forum:

http://www.lowenergysupermarket.com/collections/energy-monitors/products/plug-in-energy-monitor

[Mr. Coffee]

Many Kill-a-Watt types are just an ammeter assuming a P.F. of 1, but the actual Kill-a-Watt brand model that I have truly measures wattage on inductive loads. I cannot say how accurate it is on non linear loads like switching power supplies or other high harmonic developing equipment.

[NiHaoMike]

The Kill-a-Watt is regarded as very accurate for what it does, more than good enough for checking and optimizing energy use at home.

[The Electrician]

Many Kill-a-Watt types are just an ammeter assuming a P.F. of 1, but the actual Kill-a-Watt brand model that I have truly measures wattage on inductive loads. I cannot say how accurate it is on non linear loads like switching power supplies or other high harmonic developing equipment.

You can get an idea of how accurate it is by looking at post #9 in this thread.

[NiHaoMike]

Another trick for switching supplies is to use a DC supply at the peak voltage of the mains (170V DC for 120V AC, 340V DC for 240V AC, both easily done with a bridge rectifier and isolation transformer) and some multimeters. Note that switchers with voltage doublers need to be supplied with 340V on the bus.

You can also measure the voltage drop across the current sense resistor in some topologies, along with the DC bus voltage, then add in the power used by startup or bus sense resistors if there are any.

[djacobow]

I tried the modification on my Kill-A-Watt this afternoon. I don't really have the right setup to accurately assess the results, but they look promising-ish.

However, there are some caveats I learned.

For one, not all Kill-A-Watts are the same even if they look the same on the outside. I assumed that mine would be like The Electrician's, but actually, my two KAWs are quite different from each other, and different from his.

First, one of them is obviously highly cost reduced. It has single sided boards, just cheaper construction all around. I didn't even bother to measure the shunt because I didn't have a 4-wire setup that could do it properly, so I just removed it and put in the 0.2 ohm wire resistor I had bought. Result was that the meter read approximately 2x what it used to. Deduction: the shunt was not 0.002 ohm, but 0.1 ohm. Also, in the process of removing the shunt with my underpowered iron, I used brute force, and actually delaminated the trace around one side of the shunt. Joy. Anyway, I put it together again and it is about as good as it ever was, plus a ginormous blob of solder. Measures about 5% low all the time, which is what it always did. About half of that is from the voltage error, and half from current, as near as I can tell. Oh well.

Undeterred, I decided to apply the hack to my other KAW. I removed the shunt trying to use a little less brute force this time. There was so much solder on the connection, though, and I didn't want devote my whole roll of desoldering braid to the project, so I pushed pretty hard again. Similar delamination result, but less severe this time. Anyway, got the new resistor in there, and now the meter appears to read about 55x what it used to. More deduction: That old shunt wasn't 0.002 ohm, but 3.6 mohm or thereabouts.

Which leaves you wondering if they just grab any old piece of wire, shove it in there, and then trim as necessary. Sadly, the trimming must be done by flashing some register, since there are no pots in these devices at all.

Anyway, I have limited gear to calibrate. I used a Fluke  87V, a variac, and a 43W incandescent bulb to calibrate the current at various settings. It was very linear with respect to the ammeter, so I feel pretty good about the current measurement, at least for a sinusoidal signal. The variac introduces some inductance, so I got a none unity pf, but with a more or less constant VA addition (so the variac was consuming constant VA but the bulb wasn't) but without a power meter to compare, I have no idea if the unit got that right.

I also tried driving a 47k resistor. This came up as unity power factor (good) and based on the current measurement from the 87V and a mains measurement from same, the power estimate was pretty good (relative to the 55x adjustment). It is a few percent high. Anyway, I have reasonable confidence that this thing basically works correctly for nice loads. Because it uses sampling and multiplication in a DSP I hope there is no reason to think it would not work well for less nice loads.

One more trap for young players:

 - to get a bit more precision out of my 87V, I was using the mA setting. No problem with the static current, but I zapped the fuse with the inrush current turning on the variac. :-( Goddam, it looks like an expensive fuse, too. And I knew that would happen. For some reason I thought that the fuse would not respond so fast. Silly me. Of course it will


OK, I think that's all for now. Overall, I think I'll leave the mod in. Pics attached.

Regards,
Dave J

[han]

Maybe not the best way,
I use 12V to 220V inverter and homemade P/S.
just calculate the Total power consumption before and after. and you get DUT idle power

[rr100]

I've been thinking about that, especially that I have a sinus inverter but I don't know how constant is the idle/wasted power.

I didn't think that would be too precise but I guess is time to test with some "nice" loads. In fact I'm using the inverter as isolation transformer right now, as in

bench PS (isolated&tested) -> inverter (should be isolated but did not bother to check)

I also have a low power non-sinus one in case that's somehow better (although the waveform might mess up with the standby device).

Sorry about the fuse, that might be more expensive than the .1W resolution british kill-a-watt recommended above :-(

[tszaboo]

The chip ADE7953 is used for all the calculations.  There are other energy metering chips that you could use. 

It was only suggested as an example, there are other ICs out there. This one makes it easier, that is handles the oversampling, multiplication, and the accumulator for you, so an 8 bit IC with an RTC one interrupt/counter pin and some digital communication is enough to make the system work.

[pol098]

If anybody comes to this old thread via a search: I don't know any more than what the Web page says, but I've just seen a not-too-expensive meter, AnTai ATX 9801, which claims 0.0005A - 3A , minimum resolution: 0.0005 A, 0.01 W - 660 W, Accuracy: 0.01w, input voltage: 85 V - 265 V, display resolution appears to be 1mW. It has a "universal" socket on the body, for any country, and the plug is on a cable, not fixed to the body, so it should work anywhere with AC mains. Price £42.45 including carriage (but probably plus import tax).

eBay: "0.01-660W Electric Power Energy Monitor Tester Socket Watt Meter Analyzer 9801 K"
http://www.ebay.co.uk/itm/0-01-660W-Electric-Power-Energy-Monitor-Tester-Socket-Watt-Meter-Analyzer-9801-K-/141565459690

CCI Power-Mate of Australia make meters PM10A,PM15A,PM10AHD,PM10AHDS which seem to have suitable low-power capability, at a higher price and probably with Australian connectors and for 200-240Vac.

HTH

[djacobow]

That does not look like a quality instrument, but for the price, it's a pretty low risk investment. could make an interesting teardown at least, to see just how cheap they went inside. I'm still looking for a nice power meter, that can handle high crest factors and has good accuracy at low levels current, and which is in the mid 3-figure range, with commensurate construction quality.

[rr100]

Certainly looks tempting even if it is impossible to tell what would that thing actually measure :-)

[quarks]

Now let's see where to find a cheap "kill-a-watt" clone (I'm in Europe so I'll have to settle for something else and see for myself how it works).

Unfortunately I have not seen this thread until now. Otherwise I would not have opened this

https://www.eevblog.com/forum/projects/standby-power-measurement/msg594715/#msg594715

I have quite a few Power/Energy meters (from low cost to reference).
If you are in Europe using 230VAC I recommend you should have a look at ELV Energy Master and Conrad/Voltcraft Energy Monitor 3000, if resolution 0.1 Watt is enough. If you need more resolution, have a look at ELV EA8000. I have all of them and they are probably the best "cheap" meters you can buy. The ELV meters are about 1% accurate. That is a bit better than the Voltcraft.  I personally prefer the tripple display of the Monitor 3000 compared to the ELV displays.

https://www.eevblog.com/forum/testgear/yokogawa-cw-10/msg397795/#msg397795

[djacobow]

Another cheap power meter. Perhaps too cheap.

https://www.eevblog.com/forum/testgear/teardown-antai-atx-9801-handheld-plug-power-meter/