Why a Multimeter Reads AC Volts on a DC Supply

This page contains affiliate links. If you buy through one we may earn a commission, at no cost to you.

Before anything else: this page is arithmetic about a measurement, not a safety assessment. Nothing below establishes that any particular reading on any particular circuit is safe to touch, and the whole reason the category ratings discussed later exist is that a voltage alone never tells you that.

With that said, the reading in the question is usually not the supply. Three different things produce alternating volts on a meter pointed at a direct voltage, and only one of them belongs to the equipment. The other two belong to the meter, and one of them scales directly with the length of the leads you bought.

The first mechanism: the leads are an antenna and the meter is the other half of a divider

A handheld meter's voltage input is about 10 megohms. That is a huge resistance by the standards of a circuit and a small one by the standards of empty space, and the gap between those two facts is where the reading comes from.

Lay a test lead down on a bench. It is a conductor, and every conductor in a building has a capacitance to the mains wiring in the walls, the flex of the lamp beside it, and the bench's own metalwork. That capacitance and the meter's 10 megohms are in series across the mains, so the meter reads its own share of a divider, with nothing connected to the probe tip at all.

How big is the capacitance? For a wire at some height above a return path it is two pi times the permittivity of free space, times the length, divided by the natural logarithm of twice the height over the conductor radius. The constant out front is 5.5632 times ten to the minus eleven farads per metre. At a half metre height and a 0.8 millimetre conductor radius the logarithm is 7.1309, and the whole length term sits on top:

A set of 60 inch silicone multimeter test leads publishes exactly the figure this needs. Sixty inches is 1.524 metres, which comes to 11.89 picofarads. A 12 inch lead of the kind that ships in the box with a meter is 0.3048 metres and comes to 2.378 picofarads. Capacitance is strictly proportional to length, so the ratio is exactly five, the ratio of the lengths, with nothing else in it.

Now the divider, at 120 volts and 60 hertz. The reactance of a capacitor is one over two pi times the frequency times the capacitance, which is 223.1 megohms for the long lead and 1,115.6 megohms for the short one. The meter reads 120 volts times its own 10 megohms over the magnitude of the series impedance:

The 60 inch lead reads 5.37 volts alternating. The 12 inch lead reads 1.08. On a 230 volt 50 hertz supply the long lead reads 8.59. These are not errors in the meter. The meter is correctly reporting a real potential difference across its own terminals, and it appeared because you bought longer leads.

Why that reading is both real and worthless, and how to prove it in one move

Work out the current. 120 volts across 223.1 megohms is 0.537 microamps. Half a millionth of an amp. The same leads are rated for 10 amps on a meter's current input, which is 18.6 million times the current behind the number on the display.

That ratio is the whole character of the thing. A capacitive source has voltage and effectively no current, so anything you connect across it wins. Put a 10 kilohm resistor across the meter's terminals and recompute: 120 volts times 10 kilohms over 223.1 megohms is 5.38 millivolts. The reading falls by a factor of 999.

So the diagnostic is one resistor. A capacitively coupled reading divides by roughly a thousand when you shunt the input. A real source with a low output impedance barely moves. If the alternating volts on your supply survive a modest shunt, they are the supply's. If they collapse, they were the room's, arriving through your own leads.

And there is a reason to prefer the longer leads anyway, which that record states well: in a consumer unit, an engine bay or an air handler, the point you need to probe and the place you can safely set a meter down are rarely the same place, and five feet lets you separate them so you read the display instead of guessing at it. Full record in our review of the 60 inch silicone test leads. The trade is simply that the number which buys you that reach is the same number that multiplies the pickup by five.

The second mechanism: ripple is real, and your meter only sees part of it

Sometimes the alternating volts genuinely are on the output. A supply that rectifies mains and smooths it with a reservoir capacitor leaves a residue, because the capacitor discharges between charging peaks. Drawing a current I from a capacitance C at a ripple frequency f leaves a peak to peak residue of I over f times C, and a sawtooth's root mean square value is its peak to peak figure over two root three.

One amp from a 2,200 microfarad reservoir at a 120 hertz ripple frequency is 3.79 volts peak to peak, which is 1.09 volts on an alternating volts range. The same amp from 10,000 microfarads is 0.833 peak to peak and 0.241 volts. Those are substantial readings and they are entirely honest.

Here is the part that inverts everyone's intuition. A handheld meter's alternating volts path is specified over roughly 40 hertz to 1 kilohertz. It is a band limited instrument, deliberately, because mains frequencies are what it is for.

A modern switching charger does not ripple at 120 hertz. It ripples at its switching frequency, which for a compact USB-C adapter is commonly somewhere near 100 kilohertz, a hundred times above the top of the meter's specified band. So the meter's reading on a switching supply is not a measurement of its ripple at all. A switching supply with far more ripple energy can read far less than a linear one with less, and the reading on it tells you about the 40 hertz to 1 kilohertz slice and nothing else.

Which means the sentence worth keeping is this: the meter does not measure how much alternating voltage is sitting on your direct voltage. It measures how much is sitting on it between 40 hertz and 1 kilohertz.

The supplies in this catalogue, and the figures they do not publish

The objects in the question are exactly the sort of thing we sell. A 50 watt five port USB-C charger block, sold as a two pack publishes five ports, two USB-C and three USB-A, a 20 watt single USB-C fast charge port, IEC 62368-1 and FCC certification, multiple protections and a fireproof housing. A 140 watt smart auto-stop USB-C charger publishes up to 140 watts and a full charge detection that physically cuts the connection when charging completes.

Neither publishes a switching frequency, an output capacitance, or a ripple figure. Nor does any other supply in this catalogue, and there is no linear bench supply here at all. So the second mechanism above cannot be evaluated against our own products, which is why the 2,200 microfarad and one amp in the arithmetic are representative numbers rather than published ones. The one genuinely useful number in this area comes from the five port block's own drawbacks field, and it is about division rather than ripple: "50W is a shared total, not per port." That is the honest version of a specification and we have a page on exactly that habit, USB-C charger wattage explained.

Two more record notes. The 140 watt unit's auto-stop is a genuine mechanical disconnect rather than a current taper, which is a different claim from the usual one and a better one. And both are covered in our review of the five port charger block and review of the 140 watt auto-stop charger.

A voltage rating and a test voltage are different numbers, and so is a category

While we are separating numbers that look alike, two more are worth doing, because both appear on things in this catalogue and neither means what the larger figure suggests.

A 1000 volt insulated electrician screwdriver set, eleven pieces publishes VDE and GS certification, S2 steel shafts, magnetic black tips, flame resistance, a thin Slimfix insulation layer, a tether hole, sizes from Pozidriv PZ1 and PZ2 through Phillips PH1 and PH2, slotted 3.5 to 6.5 millimetres and Torx T15, T20 and T25 in a foam tray, and two voltages: tested to 10,000 volts and rated to 1,000.

That is a factor of ten, and it is not a safety margin you get to spend. It is a proof test: the insulation is subjected to ten times the working figure briefly, as evidence that it will hold the working figure indefinitely. The number you use is the smaller one. Our page on whether insulated screwdrivers prevent electric shock covers which hazard that insulation addresses and which it leaves bare, and how to choose an insulated screwdriver set has the selection side. More in our review of the FESA insulated set.

And the test leads above carry the pair that explains itself best. They publish CAT III 1000V and CAT IV 600V, and their own record says why those differ: "The rating is about the energy available at the point of measurement, not just the voltage, which is why the two categories carry different voltage figures." That is exactly right, and it is the ideal foil for everything above. The entire category system exists because available energy, not voltage, is what decides what a measurement can do to you. And the reading in this question has 0.537 microamps behind it. A number can be a real voltage, correctly measured, and carry no energy at all.

Why this is not the hum on your stereo, although it looks like it

Our page on why your speakers hum when everything is plugged in describes an unwanted mains frequency voltage appearing where nobody put one, which sounds like the same thing and is a different coupling with a different cure.

That one is magnetic. A closed loop of wire in a changing magnetic field has a voltage induced in it equal to two pi times the frequency, times the field strength, times the area the loop encloses, and the fix is geometric: make the loop smaller.

This one is electric. There is no loop and no area term. There is a capacitance proportional to length, and a cure that is equally simple and completely different: present a low impedance, or use shorter leads, or keep them off the mains flex. Two mechanisms, two different geometric terms, two different remedies. Worth knowing which you have before changing anything.

And a practical note on holding the probes while you do any of this. An AORAEM helping hands magnifier with dual alligator clips publishes a cast iron base, a 63 millimetre 2.5x glass lens on a 360 degree rotation, and two clips on four way swivels. Clipping both probes means the reading is taken with your hands nowhere near the leads, which removes you from the capacitance calculation entirely, because a person is a far larger coupling plate than any wire. Its own limit is scope: "built for medium to small items only." The clip springs themselves are the subject of why a clip that stays shut loses its grip first, and the record is in our review of the AORAEM helping hands.

The honest limit, and three things not computed

What is exact. The ratio of the two lead capacitances is exactly the ratio of their lengths, 5.00, and that does not depend on the height, the radius or the frequency, because those sit in a term both leads share. The divider algebra, the 0.537 microamps, the 999 fold fall under a 10 kilohm shunt, and both ripple figures are arithmetic.

What is representative rather than published, and the list matters. The 10 megohm input impedance is near universal on handheld meters and is published by no record we carry. The 0.8 millimetre conductor radius and 0.5 metre height are estimates, and here the mathematics is kind: they enter only inside a logarithm, so being wrong about the height by a factor of two changes the capacitance by about 9.7 percent, while the length, which is published, scales it one for one. The expression itself is for a wire above a conducting plane and a workbench is not one, so treat the absolute picofarads as an order of magnitude and the ratio as exact. The 2,200 microfarad reservoir, the one amp load, the 40 hertz to 1 kilohertz band and the 100 kilohertz switching frequency are all typical values.

Three things this page deliberately does not compute. It does not compute the actual ripple of either charger we carry, because neither publishes a switching frequency, an output capacitance or a ripple specification, and inventing those three would be inventing the whole answer rather than deriving it. It does not compute what a meter's own accuracy specification contributes, the "plus so many digits" term that dominates every reading taken well below full scale, because no meter in this catalogue publishes one and that deserves its own page. And it does not say whether any reading is dangerous, which is a question about available fault energy at a particular point in a particular installation, and is precisely what the category ratings exist to answer and what arithmetic on a published lead length cannot.

Three numbers to carry. A loose lead and a 10 megohm input make a divider with the room, and because capacitance is linear in length a 60 inch lead reads about five times what a 12 inch one reads. The current behind that reading is around half a microamp, so shunting the terminals with 10 kilohms collapses it by a factor of a thousand while a real source does not move. And a meter's alternating volts range is a band, roughly 40 hertz to 1 kilohertz, which is why a switching supply full of 100 kilohertz ripple can read lower than a linear one with a tenth as much.