As an Amazon Associate, Akermin earns from qualifying purchases. Product links on this page are affiliate links.
A digital multimeter works by turning every quantity it can measure into a small direct voltage, digitising that voltage with an analogue to digital converter, and then scaling the digital result for the display. Voltage ranges use a resistive divider, current ranges pass the current through a precision shunt and measure the voltage across it, resistance ranges push a known current through the unknown component and measure the resulting voltage, and alternating signals go through a rectifier or a root mean square converter first. Once you see that single chain, counts, update rate, input impedance and the two part accuracy specification all follow from it.
From probe tip to digits: the chain inside the case
Five stages sit between the probe and the display. First comes input protection: fusible resistors, varistors, spark gaps and in the current path an actual fuse, all sized to absorb a transient without passing it into the electronics. The published category rating describes the energy that stage is built to survive. Second comes range selection, a precision resistor network that attenuates a large signal or a shunt that converts a current into a voltage. Third is signal conditioning, where a high input impedance amplifier buffers the divided signal and an alternating input is converted to a direct equivalent.
Fourth is the analogue to digital converter, which is where the measurement becomes a number. Fifth is the microcontroller, which applies calibration constants held in memory, positions the decimal point according to the selected range, and writes to the display. A great deal of what feels like intelligence lives in that last stage: autoranging is the controller noticing the result is near the top or the bottom of the span and switching the divider, and hold functions are simply the controller freezing what it writes. The Fluke 177 digital multimeter states that behaviour explicitly, with autoranging that picks the range giving the best resolution for the present value, a display updating four times a second, and a 33 segment analogue bar updating forty times a second so that movement remains visible while the digits settle.
Every function is a voltage measurement underneath
The converter at the heart of the instrument can only digitise a direct voltage within a narrow window, perhaps a few hundred millivolts. Every rotary switch position is therefore a scheme for turning something else into a voltage inside that window.
| Function selected | What physically happens | What the converter sees |
|---|---|---|
| DC volts | A precision resistor divider attenuates the input | A scaled fraction of the input voltage |
| DC amps | Current passes through a low value shunt resistor | The voltage across the shunt, by Ohm law |
| Resistance | The meter sources a known current through the component | The voltage that current develops across it |
| Continuity | The same current source, with a threshold comparator | A voltage below or above the buzzer threshold |
| AC volts | Rectification and scaling, or a true RMS converter | A direct voltage representing the AC value |
| Capacitance | A known current charges the capacitor and the time is measured | A voltage ramp converted into a time interval |
| Diode test | A small fixed current is forced through the junction | The forward voltage drop of the junction |
Two details in that table explain common surprises. Because resistance is measured by sourcing current, a resistor still connected to a powered circuit gives nonsense, since the circuit own voltage adds to the one the meter is trying to create. And because alternating measurement needs a conversion step, the choice of conversion matters: an averaging design multiplies the rectified mean by the form factor of a pure sine, approximately 1.11, which is only correct when the waveform really is sinusoidal. A true root mean square design computes the heating equivalent value whatever the shape, which is why it is the specification that matters on drive outputs and switched supplies.
What the analogue to digital converter actually does
Most handheld and bench meters use a dual slope integrating converter, and its operation is worth knowing because it explains the update rate. An integrator charges a capacitor from the unknown voltage for a fixed period. Then the input is switched to a reference of opposite polarity and the capacitor is discharged until a comparator detects zero crossing, while a counter runs. The ratio of the discharge time to the fixed charge time equals the ratio of the unknown voltage to the reference, so the count is the measurement.
The clever part is the fixed integration period. Integrating over a whole number of mains supply cycles makes any interference at that frequency average to zero, which is how a meter rejects the hum picked up by a metre of test lead. That is also why a typical handheld produces only a few readings a second rather than thousands, and why bench instruments offer selectable integration times: longer integration means quieter last digits and slower updates. Meters designed to digitise quickly sit at the other end of that trade, and the Hantek HDM3065X bench multimeter states up to 30,000 readings a second alongside trend plotting and histogram display, which is a different instrument philosophy to a slow, quiet, high stability measurement.
Counts, digits, and the resolution they buy
Resolution is set by how many distinct values the display can show, and that is what a count figure or a digit figure describes. A 6000 count meter can display any value from zero to 5999 before it has to change range. Digit notation says the same thing differently: a half digit means a leading position that can only show zero or one.
| Display span | On this range | One count equals |
|---|---|---|
| 6000 counts | 6.000 V | 1 millivolt |
| 6000 counts | 600.0 V | 100 millivolts |
| 9999 counts | 9.999 V | 1 millivolt |
| 199999 counts, the usual 5.5 digit convention | 19.9999 V | 0.1 millivolt |
| 1999999 counts, the usual 6.5 digit convention | 19.99999 V | 10 microvolts |
Two cautions go with that table. The count span behind a digit figure is a convention rather than a guarantee, and the Fluke 8808A bench multimeter states 5.5 digit resolution with a dual display while the exact count span is not stated in the listing bullets. And resolution is not accuracy. The second row shows a meter resolving a volt range in steps of a tenth of a volt, which says nothing about whether the reading is correct to a tenth of a volt. Those are separate specifications and they are often separated by an order of magnitude.
Why accuracy is written as percent of reading plus digits
Two different error mechanisms exist inside the instrument, so the specification needs two terms. Gain errors scale with the signal and come from the tolerance and temperature drift of the divider resistors and the reference, so they are expressed as a percentage of the reading. Offset and quantisation errors are fixed in size regardless of the signal and come from amplifier offset, converter noise and the last digit itself, so they are expressed as a number of counts.
Work an example. That bench meter states a basic DC voltage accuracy of 0.015 percent. Reading 9.5000 V, the proportional term contributes 0.015 percent of 9.5, which is about 1.4 millivolts. The counts term adds whatever the manual specifies on top of that, and since the listing does not publish it, the honest total is 1.4 millivolts plus an unstated number of final digits. Now contrast a specification written as percent of full scale, still common on analogue instruments and panel meters. One percent of full scale on a 600 V range is 6 V of possible error whether the needle reads 580 or 30, which is why a low reading on a high range is the worst case in any measurement and why autoranging to the lowest usable range matters so much.
Reading a specification sheet as a measurement budget
Treat the chain as a list of things that can cost you accuracy, then spend your attention where the cost is largest. Select the lowest range that holds the signal, because the fixed counts term is a bigger fraction of a small reading on a large range. Prefer a longer integration time when the last digits matter and a faster update when you are watching a value move. Use a true root mean square meter on anything electronic, and remember that burden voltage on a current range changes the circuit you are measuring.
Then separate resolution from accuracy when you compare instruments, since extra digits are cheap and a stable reference is not. Four wire resistance measurement, which that bench instrument states for low value work, exists precisely because lead resistance becomes a large error below a few ohms. The selection criteria that follow from all of this are set out in choosing a digital multimeter and, for domestic wiring tasks specifically, in choosing a multimeter for home electrical work. How those compromises land on real handhelds is covered in our notes on the Veyfiy multimeter and the Elejoy multimeter.
Frequently Asked Questions
Why does a multimeter read a few millivolts with nothing connected?
Because the open leads act as an aerial and the input stage has very high impedance, so tiny induced currents produce a readable voltage. Touch the probes together and the reading collapses to near zero. It is normal behaviour on the millivolt range and not a sign of a faulty meter.
What does ten megohms of input impedance actually do?
It limits how much current the meter draws from the circuit it is measuring, so the act of measuring disturbs the circuit as little as possible. The trade off appears on high impedance nodes, where ten megohms is low enough to pull the reading down, and on long dead cables, where it is high enough to show capacitively coupled ghost voltage.
Why do more counts not mean more accuracy?
Counts describe resolution, which is the size of the smallest step the display can show. Accuracy describes how close the reading is to the true value. A meter can resolve to a millivolt and still be two hundred millivolts out, and the specification sheet treats the two as separate figures for exactly that reason.
What is burden voltage on a current range?
It is the voltage dropped across the internal shunt and fuse while current flows through the meter. That drop is subtracted from the circuit under test, so a sensitive low voltage circuit can behave differently with the meter in place. It is one of the reasons a clamp jaw is preferred for current work where accuracy of the surrounding circuit matters.
Why do bench meters give different readings to handhelds on the same point?
Mostly because of integration time. A bench instrument can integrate over several mains cycles, which rejects interference and stabilises the last digits, while a handheld updates several times a second so you can watch a value move. Faster sampling trades noise rejection for responsiveness, and both are correct for their purpose.