Minimizing GFCI vampire / phantom load loss and heat? Which model is most efficient?
Woah! There’s hot spots on my walls on the thermal cameras. It’s clear that each GFCI outlet is using power 24/7.
Can anyone provide a pointer to detailed measurements of GFCI (or AFCI) models and their standby, parasitic or “vampire” power draw? I see numbers all over the map from .1 Watt to 7 Watts.
The big vendors do not specify standby power consumption, and I have not yet found any EnergyStar or certification guidance. AFCI’s are of course controversial for a number of reasons: add to the list the standby power.
Any check your kitchens: a lot of them have redundant GFCI’s (only one is required, the rest benefit from the downstream protection if wired correctly).
GBA Detail Library
A collection of one thousand construction details organized by climate and house part
Replies
Most GFCIs use a dedicated integrated circuit to handle the ground fault sensing function. I looked up a datasheet for one such chip:
https://www.onsemi.com/pub/Collateral/RV4145A-D.pdf
This datasheet states the chip uses a supply current of 18mA, and that it’s usually powered using a shunt regulator. A shunt regulator is a type of voltage regulator that regulates voltage by managing the current through a resistor, which means the regulator’s output voltage is controlled by varying the current. Shunt regulators act as a constant-current load, which means they always use the same power (watts), and divert what isn’t used into that resistor which converts the excess power into heat. 18mA at 120VAC gives about 2.2 watts which is the power consumed by the GFCI.
Note that there are more efficient ways to handle the power supply, but all require more parts which means both more cost and more parts that can fail. Remember that these are devices that are intended to be very reliable which means that energy efficiency is generally not the highest priority design consideration.
If I were to design such a device, I’d probably use a so-called “transformerless” power supply which uses a capacitor as part of the shunt regulator to reduce the amount of wasted power. I don’t know if there are any regulatory requirements that prohibit the use of such a power supply for a GFCI device.
The only way to reduce the total energy consumption of these devices in your home is to minimize how many you have. I prefer to use only GFCI circuit breakers (they tend to be FAR more reliable over time), which means you only use one for each circuit requiring GFCI protection and they are installed in the circuit breaker panel. It’s also possible to daisy chain additional receptacles off of a single receptacle-type GFCI, but you MUST be careful how you do this to preserve the protection function (the “load” terminals feed downstream outlets, both hot AND neutral wires must loop through the GFCI load terminals).
Bill
Other models are about 1/10 of that. Sounds like it is worth it for someone to test a variety of current models and report the power draw.
https://www.digikey.com/products/en/integrated-circuits-ics/pmic-power-management-specialized/761?k=GFCI
I don’t know about 1/10 of that, but at least some models are apparently less. I did a quick lab this morning and tested one of the 20A single pole (120V) GFCI breakers in my panel at home. This is a Siemens type QPF2 GFCI breaker and is the current production version (it’s about a 2 year old breaker). I measured about 9.3mA of 120V power running the breaker, and that’s using a lab-quality meter. That’s about 1.1 watts, so about half of the power the datasheet I posted lists for the other chip.
I don’t have any other GFCI breakers here to test (except for a 15A version of the same Siemens breaker, but I’d expect it to use the same amount of power as the 20A breaker), so I can’t do any comparisons. My Leviton receptacle-style GFCIs in the bathrooms are more difficult to isolate for testing.
My test methodology is as follows:
One breaker was swung out from the busbar and a folded piece of 12 gauge solid copper wire was inserted into the breaker’s slot to make a terminal for the test lead to attach to. The breaker’s neutral wire was left connected to the neutral busbar in the panel. This particular breaker runs one receptacle in the kitchen through approximately 60 feet of new 12-2 NM cable and has no loads connected. The meter is connected in series with the improvised breaker terminal and a wire connected to a spare 15A non-GFCI breaker as a power source. The meter is configured to measure AC current and is capable of reading up to 3A with a resolution of 0.01mA. I’ve attached some pics. I did not attach a pic showing about 0.04mA of noise, so I’m not using the last digit of the meter reading.
Note that my usual portable meter was not able to accurately read such a low level of AC current which is why I used my fancier lab meter. Because of this, I’m not sure if others would be able to make simple home measurements of these breakers without special equipment — normal meters might not have enough low-range sensitivity to give reliable measurements with the low current levels involved.
Bill
Bill,
What is your usual portable meter? Even a relatively cheap hand-held can usually handle down to a few mA without any problem, plus or minus a couple of percent.
My portable meter is a tektronix DMM916. Not a cheapie. It’s the blue thing on top of the HP meter in the pic. It would show about 4mA all the time this morning, so I didn’t trust it for this test. I should probably calibrate it but I haven’t bothered in a while, even though I have a full cal setup so I don’t need to take it to a cal lab. Too many other things to do and I don’t rely on the portable meter for critical stuff.
DC current is easier. AC current is more difficult to accurately measure. It was on the AC range that my little meter showed the 4mA all the time, and didn’t reliably register a change when the breaker was connected. Since I had the much better HP meter I just used that. That HP meter is my mid-level meter and is still relatively easy to carry around. My best meter is mounted in a rack of test gear.
One last thing too: the “few percent” error is usually specified relative to full scale. That means if your meater is accurate to 1% of full scale, but you’re on the 10A range, you are accurate to +/- 1% of 10A which is +/- 100mA. That’s a 200mA error band. On a 1A scale you’d have a 20mA error band. Even the HP34401A meter I used is specified as +/- 0.1% of reading + 0.04% of the 1A range. That means my measurement is accurate to about +/- 0.4 mA, and this meter is much better than even the best handheld meters.
Bill
I ran a calibration lab for many years, so I'm familiar with the DMM916. While it was definitely not a cheap meter, in terms of dollars, it was not a good meter. I remember when that line came out, and they pushed it as a competitor to the Fluke 80 series. I was very disappointed when I got my hands on one, both in terms of internal build quality and functionality. It was clearly a cheap, Chinese meter that Tektronix just slapped their name on. Aside from the quality problem, it took a shortcut with the current ranges, skipping the 40mA and 4A ranges you would normally expect to see. That effectively increases the measurement floor of readings in the 4-40mA range and 400mA-4A by a factor of 10. So taking a reading at 10mA is going to be about +/-8%, which is pretty abysmal. All that aside, assuming you were using the 400mA range, it sounds like the meter is broken. There is no calibration adjustment on that unit for offset in the current range (as there isn't in most hand-held meters). If you were on the 10A range, then it's just way outside the capability of that range.
But back to the original point, if anyone with a handheld meter wanted to measure the current draw on a GFCI, the chances are very good that their meter will be capable of providing reasonable results. The DMM916 is an unusual case, rather than the typical case. Even a moderately competent handheld meter would actually exceed the 34401A performance at that particular measurement, which is hampered by the lack of any ranges below 1A, limiting the resolution to 1uA. It actually looks like you used the 3A range in the photo, in which case the uncertainty for that measurement was about +/-1.8mA (using 1 year specs). Even on the 1A range, it would be +/-0.4mA. Using a Fluke 187, the uncertainty would be +/-0.09mA, and with a really cheap Klein MM400, it would be +/-0.22mA.
I know what you mean about the DMM316 being a bit odd. Years ago I was using it to measure current for a DC motor. The motor drew way more than I was expecting and opened the fuse in the meter. At least that’s what I thought — but it turned out the fuse was OK, but a trace on the PCB fused open (!!) instead of the 10A fuse. I was not real happy to see that, but I fixed it with some solder. I mostly use that meter for simple voltage and current measurements in the field where I don’t need really high precision. Looks like we know why Tek stopped making these meters...
The 34401A was set for auto ranging and should have been on the 1A range. Admittedly I didn’t bother to really set things up for a maximum precision measurement (I didn’t even warm things up more than a minute or two). I got a reasonable reading that held stable and then put things in the panel back together. Close enough for my purposes.
I’ll have to look into portable meters again. I haven’t needed a really good one for a while. Last I checked, Agilent’s handheld was the best for accuracy. I think it’s carried over to keysight and still available. Most of my home lab has been setup for RF design work for years so I haven’t been as concerned with precision AC/DC measurements since the 90s. My DMM316 is mostly used to check battery string float charge voltages (48 and 480v strings) and UPS output voltages (480 and 120). Neither needs to be better than a few percent.
Bill
So what are we talking about on the average well designed circuits of a house? Is this something worth addressing? It seems to me we are right on the edge of it being inconsequential.
1-7 watts depending on model, times millions of houses, is a lot. Add to that new requirements for AFCI breakers basically everywhere, and we're talking a lot of energy.
I'm just not sure chasing these small things is that productive.
It's like worrying about the thermal bridging at door handles, or not having foam under one interior footing pad. The loses are eclipsed by occupant behaviour - like leaving some small device on its charger, chatting with a neighbour at the front door on a cold day, or taking a less efficient route to the market on the weekend.
If we are talking about the aggregate amount of energy wasted, you could easily find sources the same size with no utilitarian purpose. Cancel one college football game, or a couple of vacation flights a weekend (and no, I'm not suggesting those are good ideas).
Malcolm, I think people look at these losses because they see a number and think “AHA! Waste!”, but engineers look at overall system efficiency. Keeping your cars tires properly inflated will make a bigger difference in overall energy use than a 10% improvement in GFCI power consumption.
I actually worked out for a previous post, but didn’t post it, that if we assume 9 GFCI devices per house like I have, and 100 million homes like that, it works out to about 1.3% of the generating capacity of my electric utility here in southeast Michigan. It sounds like a lot, but it’s not when compared to the entire system. Also, chasing those tiny percentages is also going after the most expensive gains. I like to tell my customers that every gain of 1% costs ten times as much as the previous 1% gain (I’m usually talking about reliability though).
Replace a CFL security light with an LED and you’ve just gained as much as all your GFCIs put together. Replace the weatherstripping on your front door, probably also the same as a few watts of savings. Replacing everyone’s front door weatherstripping will probably save more energy than a 10% improvement in everyone’s GFCI devices.
To put it in perspective another way, it’s common to design power lines for 95% efficiency. Notice how the “100 million homes” GFCI power consumption isn’t even 1/4 of just the power line loss in terms of percentage of the total system. There are much bigger energy losers than GFCIs, and there isn't much that can be done about many of them.
Bill
Bryce, this shouldn't surprise you, after all the gov is watching out for our safety, while they benefit off their supposed control of power.
I’m not sure how you think the govt benefits off of anyone’s use of electric power. Most electric utilities are private companies. I can tell you that as someone who works in a quasi-utility, the govt does more to get in the way than they do to help utility customers. I have been involved in numerous legal disputes with muncipalities over right of ways and easements, for example, and I’ve won every one. The municipality just wastes everyone’s time and money.
Bill
Basically I’d say the typical GFCI is using about 1 to 2 watts 24x7 to run the sensing chip. I have, in my house, 5 GFCI breakers and 4 GFCI receptacles (in all the bathrooms where I can’t isolate the circuits). If I assume they all use 1.1 watts like the one I measured, that’s 9.9 watts 24x7. That’s a little more than a pair of 4w incandescent night lights. A little over 7kWh per month, about 85kWh per year. At our electric rates, that’s about $9/year.
I don’t have any AFCI breakers to test. My understanding is that the requirements for AFCIs are going to be relaxed in the next revision of the code due to all the problems with them and their limited benefit. I’m thinking about borrowing an AFCI breaker from one of my electricians so that I can test it and share the results here.
Over millions of homes, yeah, it’s a decent amount of power. But there is a legitimate safety benefit to using GFCIs in the areas where they’re required. Could the energy use of these devices be reduced? Possibly. I actually have a contact at Eaton so I might ask them what they think. I also have lutron caseta smart switches in my house, they all use a little power too I’m sure. I would try to minimize the number of GFCIs used. Don’t use one in every receptacle in a kitchen, for example — use one for every CIRCUIT in the kitchen instead. It’s both cheaper (fewer devices), and less energy use so it’s a win-win with zero downside.
My guess is the losses in the transformer powering your house are probably an order of magnitude higher than the total GFCI power consumption for your entire house, and probably the houses of your neighbors too. Those transformers are usually pretty efficient these days too so there isn’t a lot of room for improvement.
Bill
The true power used is likely to be less than the 1.1 watts of apparent power as calculated. We have not considered the phase angle of the current relative to the voltage this could change the answer by 30%. I am sure this is not news to Bill or Trevor. I think a better test would be to connect the breaker to a Kill-A-Watt device and let it run for 24 hours.
Walta
I measured a Leviton GFCI on a Kill-A-Watt EZ
03 Watts
.15 Amps
.20 Power Factor
Ran for 24 hours, showing $7.20/year at $0.274/kwh
Leviton 07599 INSP 1K20I 97481-01F 7
As I expected, the draw is the same if the device is tripped or in operation mode.
The device has a permanent magnet latching relay, so only uses peak current to engage or disengage.
--------------------------------------------
That's terrible, in a world were wall wart supplies now have to be 0.100W or less. That's proof a device like this could use 30 times less power.
The Department of Energy Level III, IV, V and now VI standards are way stricter.
If nobody's measuring this, there's no incentive for Leviton and the other vendors to step up and improve.
See
https://www.cui.com/catalog/resource/efficiency-standards-for-external-power-supplies.pdf
That’s super crappy power factor (0.8 is usually considered a minimum, 0.9+ is good), but it’s also an itty bitty load so it doesn’t matter too much in the big picture.
It’s important to remember that GFCIs are built for reliability first, energy efficiency is NOT a priority. To significantly improve the power consumption would likely require a transformer power supply, and to make it fit in the tiny package would require a switching-type (high frequency) power supply. That’s a lot more parts to potentially fail, and a much higher risk of the GFCI failing in a way that won’t be obvious. That is potentially dangerous. Remember that you’re supposed to test GFCIs monthly, but almost no one actually does.
It may be possible to slightly reduce power consumption with a capacitor type “transformerless” power supply, but if no one does that then I suspect there is a regulatory reason.
Try to minimize the number of GFCI devices by using one per circuit instead of one per receptacle. In a new build, you could probably put all the bathroom receptacles in a house on one circuit so that they could all share a single GFCI in the panel. That would mean only one hair dryer running at a time though! I’d also do a quick review of residential codes before doing this just to be sure no one will fail you for not enough circuits.
This is a tiny amount of power, and not easily addressed. You could probably save the same amount of power, and probably even more, by using 12 gauge wire instead of 14 gauge for all circuits to reduce voltage drop, and the associated power losses, in your house.
Bill
I'm a product design engineer. I can say with confidence that the GFCI package is not tiny at all by modern standards, and the efficient supplies are highly integrated and built in massive volumes: they're cheap and reliable.
The shunt regulator is probably sized for the monitoring electronics PLUS the relay activation. One could design with two: one for the standby current, the other normally OFF but used to activate the relay currents.
The pictured GFCI above failed in use, and it failed "on". There are black marks on the PCB and one of the transistors exploded. The enabling relay is latching: good for reducing power consumption compared to an active relay, but not fail-safe. It's fail-dangerous.
I'm willing to spend $7.20/year on GFCI protection per outlet. True. But if I had a choice of paying $20 for a model that uses $1/year, I would. Multiplied by the millions of homes AND the new AFCI requirements, it's a lot of energy. Again if nobody cares, no vendor will do better.
It's a lot of trouble to share GFCI's, and impossible if the wires in the walls don't support it. One GFCI relay per bathroom is fine, it's still going to be 15 or 20 total amps, so individual GFCI's don't affect the number of hair driers usable at once.
I’m an EE. Much of my work is with power system designs in critical facilities. Remember that the GFCI package has a lot of things in it (receptacle contacts, mounting structure, etc), and clearance requirements between things. If you look at the datasheet I posted earlier, you can see that the shunt power supply doesn’t really power the relay coil, the coil is essentially triggered off the line with a triac.
I’ve seen many GFCIs fail, usually those that are installed in outdoor locations (which is why I recommend always using GFCI breakers for outdoor circuits). Additional components are going to increase the chance of failure. There may be regulatory requirements the restrict the use of other types of power supplies.
In power systems terms, a watt or two isn’t much. Resistive losses in the wiring FAR exceeds this amount. The telecom facilities I design often have many thousands of watts of losses just in branch circuit wiring, for example. It’s a different world from consumer electronic devices. This is not to say it’s not worth trying to optimize a GFCI for energy efficiency, it’s just that there may be other concerns that are more important for that specific type of device.
BTW, I’ve been told that AFCI requirements will be relaxed in the next code revision. AFCIs were a bad idea to begin with, and cause a lot of problems in practice.
Bill
Michigan has toyed with dropping AFCI's https://forums.mikeholt.com/showthread.php?t=173784
Outdoor GFCI/AFCI is a problem waiting to happen, from all sorts of reasons, including humidity and temperature swings. The addition of an efficient supply for the electronics won't change that basic issue. Breaker based is the way to go.
What's not measured won't be optimized. That's the point of the entire thread. So who else has measured specific brands and models of AFCI/GFCI for power draw?
While substituting 12AWG for 14AWG wire would certainly save a measurable amount of power, intuitively I would guess it's going to be less than the power usage of all the GFCI's in the house. Using a voltage drop calculator, I compared the losses of a 50' line of 14 and 12 gauge at 12A. The difference is 12W. That sounds like a lot, but what fraction of the time is there a full load on the line? Very small. If we're talking about wall outlets, the portion of the time there's ANY load on the line is small.
If you did this for the whole house, the cost would be rather immense. In my house, I used about 5 rolls of 14/2. If all of that was replaced with 12/2, the incremental cost would have been about $650. If you're not doing the wiring yourself, you can also expect an upcharge in labour.
What I think does make sense, and what I did was to target the highest load lines and upsize them. I went one gauge bigger on the circuits for the microwave, washing machine, dishwasher, cove heaters, cooktop, and two long runs to outlets before branching off.
I tested two different GFCIs. Both read 1W with a Kill-A-Watt. An ammeter doesn't work (because of power factor).
Which models did you test?
One is a Carol Plug-It (external type) and the other is a in-wall type (but I don't see any external labels).
So here's a practical approach. A 7W LED lamp has a rather large heatsink area to shed heat from and still gets so hot I can't hold onto it. That's 7W.
The GFCI outlet in my kitchen may be warmer than its surroundings, but barely, and has less surface area to shed heat from than the LED bulb just mentioned. Based on that, I'd be surprised if it took more than an order of magnitude less power to run it.
That’s a bit oversimplified. A 7w LED bulb isn’t giving off 7w or heat. 7w of heat would mean a 0% efficient lamp with no light given off.
A GFCI consuming 1 watt is giving off ALL of that watt as heat, there no portion of that watt coming off as a useful form of energy like the light that an LED lamp gives off. Also, a large amount of slightly warm air may well contain more energy than a small amount of very hot air.
The basic goal is to reduce the power consumption as much as possible in the case of a GFCI. For something like an LED, the goal is to minimize the percentage of the input power that goes into heat since light is the useful output, heat is waste.
Bill
True, but if the LED lamp in my example is only 50% efficient that makes the probable power consumption of the GFCI outlet in my kitchen even less since that 7W LED is only rejecting about 3.5W to be that hot.
I just checked the GFCI outlet in my kitchen again, and if it was any cooler it would be hard to tell by touch that it was any warmer than its surroundings. It's hard to imagine it's consuming any more than a tiny fraction of a watt.
OK, I grabbed my FLIR and checked, the GFCI in the kitchen is running at around 29C (84F) and the AFCI in my main panel is running at 23C (73F). The AFCI is a Siemens QAF, not sure what brand the GFCI in the kitchen is.
Either way, they are not anything to be concerned about.
Here are the pics:
Just curious, but which of FLIRs cameras are you using for those pics?
Bill
It's a 2nd generation FLIR One, the dongle-style that attaches to a smartphone. I don't put much stock in the temperatures being all that accurate given the emissivity of various surfaces, but it's an excellent tool for visuals and relative temperatures between similar surfaces. If more accurate temps are required a piece of black electrical tape on the target works well.
You can also see the shift between the visible light camera information and the IR camera data. You can adjust that in the app to clean things up if you want, but at arm's length or farther it lines up pretty well. I originally bought it to look for thermal bridging and air leakage, something it seems to work pretty well for.
Bill / Zephyr,
Any electrical device that consumes 7 watts is producing 7 watts of heat. This rule applies to everything -- refrigerators, televisions, radios, and, yes, LEDs.
Martin, you are correct, but in the case of a bulb a certain percentage of that energy leaves as light to be absorbed and converted into heat by its surroundings. Only the waste heat coming off the back of the LED die is warming the bulb itself.
Perhaps splitting hairs here, but the portion of light that goes out the windows does not end up as heat inside the house.
...or by a dog that is then taken out for a walk on a cold day.
This is a funny thread.
Putting your dog outside on a cool summer night and bringing it inside in the morning is a good source of free air conditioning.
Yes and no. It depends on the definitions. I like to joke at work that all the energy that we put into a datacenter as electricity comes out as heat through the cooling system, except the very tiny amount of energy that leaves the facility as pulses of light in the fiber optic cables.
For an LED lamp, the energy leaving as useful putput (light) isn’t leaving as heat in a heatsink. That light will eventually become heat when it hits the walls, etc, except any energy that’s leaves out windows as Trevor mentions.
Efficiency is defined as how much of the total input energy goes into the useful output product. In the case of an electric resistance heater, the losses are heat which is also the useful output product, so such a heater can be called 100% efficient. For a motor, the heat is the energy loss, since it didn’t going into torque making the shaft turn.
Since a GFCI doesn’t convert the electric energy to any useful output product, all th energy is essentially wasted as heat. In this regard, a GFCI is different from other devices that convert electrical energy to some other form like lights, motors, etc. While all the energy does ultimately end up as low-grade heat, it’s what the energy does on this way there that makes things interesting.
Bill
A typical LED lamp is about 20% efficient.