| Episodes | 183 |
| Mentions | 356 |
| Cited here | 50 |
| First — last | #2 — #720 |
| Top guests | Martin Lorton, Fran Blanche, Paul Thompson |
| Related | oscilloscope · power supply · soldering iron · analog to digital converter · kickstarter |
A multimeter is an electronic test instrument that measures voltage, current and resistance, with secondary functions that commonly include continuity testing, diode testing and short-circuit detection.[481][72] It is the foundational instrument of electronics work: continuity testing alone can recover the complete wiring of a machine stripped of its harnesses, and a step-by-step diagnostic procedure built on four basic meter operations resolves roughly eighty percent of field faults in some repair programmes.[463][481] The instrument exists in handheld and bench forms across a requirement space wide enough to support on the order of a thousand distinct models, with no single optimal design.[554][148]
Specifications
Resolution and accuracy
Resolution and accuracy are independent specifications that are routinely conflated. Resolution is the number of digits displayed; accuracy is a percentage figure with many contributing factors, and it rarely matches the resolution, so describing a meter as having four and a half digits of accuracy is a category error, since digits are a resolution figure.[72] Extra digits remain useful even where the underlying accuracy is poor, because resolution supports relative measurement: a meter accurate to ten percent but resolving six digits will reliably distinguish one component from another, which is why high-resolution instruments carry accuracy figures specified several digits above their least significant digit.[24]
A meter cannot be specified from a single prototype. The design is built to a target specification, and the achievable specification is then established by testing a batch to characterise the production spread, since what matters is the distribution across units rather than the performance of any one of them.[491] Published specifications shift slightly between revisions of the same instrument, in both directions and by amounts on the order of 0.03 to 0.05 percent; the underlying change is commercial, an older approach having set the specification first and spent whatever the components cost to reach it, whereas a price-conscious product lets the achievable figure follow from the component budget.[491]
What an expensive meter buys is measurement confidence over time rather than initial accuracy, since calibrating any meter to read correctly on the day is straightforward; the question the price answers is how the instrument drifts with time, temperature, ageing and vibration, which is bound up with the reputation of the manufacturer and of the specific model.[353] Consistent with this, older precision instruments hold their calibration well enough that a decades-old meter bought second-hand, provided it still functions, will generally read accurately, making the used market a legitimate route to high-quality measurement.[716]
Range and conversion behaviour
Bench instruments set their integration time to a whole number of mains cycles so that line-frequency interference averages to zero, with further averaging applied on top; a meter configured for one line frequency and used on another loses that rejection, which is a real consideration for imported equipment.[643] True-RMS converters perform poorly on low-level signals, which is why the alternating-voltage millivolt ranges carry much looser accuracy figures than the direct-voltage ranges, often relegated to a footnote in the manual; the limit is component quality in the signal path rather than the conversion principle.[464]
Failure modes and limitations
Range switching blanks the measurement while the analog path settles, for milliseconds or longer depending on the instrument, and the display gives no indication that data is missing; the behaviour is general to test equipment and makes range-switching instruments unsuitable for measurements that must be continuous.[607] Putting a meter in line to measure current is a reasonable starting point and far better than not measuring, but it fails where dynamic range is required, because the burden voltage needed to read small currents will brown out the device being measured.[607]
Susceptibility to electromagnetic fields is specified against an international standard, with manuals for high-end meters claiming compliance at a field strength on the order of three volts per metre, and instruments nonetheless fail in ordinary bench conditions: one meter’s current reading swung from one amp to ten amps as a hand approached it, and simply placing a meter beside coaxial cable carrying a ten-volt square wave from a function generator moved the reading out of specification with no electrical connection at all.[319] The susceptibility is frequency-dependent and can peak in a band that ordinary equipment occupies; one meter was reported worst around 13.56 MHz, the frequency used for near-field identification, and a separate case involved a mobile phone placed beside a meter locking the instrument up and in some cases rendering it unusable, a defect that took about a year to resolve in a new revision.[319]
A high-impedance input mode, above the usual ten megohms, accumulates charge on the input capacitance when the probes are left open, so the reading ramps upward slowly instead of settling, which can be mistaken for a fault in the instrument.[293] The warning that sounds when probes are left in the current jack is implemented with a split current jack and a resistor of around five megohms feeding a high-impedance threshold detector; because that sensing node is deliberately high impedance, dirt, grime or moisture inside the jack produces the warning with nothing plugged in at all, and cleaning the jacks with alcohol usually restores correct behaviour.[688] A meter that drains its battery while switched off points to a component connected across the battery ahead of the rotary switch, the classic case being the reverse-polarity protection diode.[574]
Measurement technique carries its own limits. A continuity reading of zero ohms establishes only that two nodes are joined at direct current, because the meter measures by passing a small current slowly; two grounds that read as identical can still behave as separate nets at speed, so the measurement does not settle questions about high-frequency behaviour.[410] Measuring resistance in circuit is complicated by the meter’s own test voltage being high enough to forward-bias semiconductor junctions, which changes what is being measured.[140] User-facing controls can also betray the operator: an instrument whose reading was wrong by volts under a poorly explained sensitivity control demonstrates that a setting can silently change the digits of precision in the wrong direction, a specification problem presented as a feature.[455]
Design
Nine-volt batteries and three-cell supplies persist in meters for functional reasons rather than inertia. Two cells give about three volts, which caps the diode-test voltage below what is needed to turn on an LED, and raising it would require a boost converter; three cells at four and a half volts is the common compromise, preserving the diode range while allowing each cell to discharge to around a volt so that most of its capacity is used.[555] Component choices inside the instrument reflect manufacturing economics: the reverse-polarity protection diode is often rated far above the requirement—a thousand volts across a nine-volt battery—because the same part is already used in the input protection elsewhere in the circuit, and carrying one reel rather than two is worth more than the component saving.[574]
Some meters provide a low-ohms mode specifically to hold the test voltage below the threshold at which semiconductor junctions turn on, preserving in-circuit resistance measurements.[140] A latched continuity tester pulse-stretches a momentary short into an audible tone long enough to hear, so that a contact lasting a microsecond registers; without that stretching, lightly touching probes together produces a broken, scratchy indication as surface contamination makes and breaks the connection.[690] Sealing on industrial meters is not for operation under water but so that a fully sealed instrument admits no dust or chemicals and survives being dropped into a drum of oil, after which it can be wiped off and remain within calibration; the cost is that a sealed instrument is harder to hold.[65] Probe leads are a design decision in their own right: coiled leads were supplied as standard on the reasoning that long leads clutter a desk, and in practice they tangle and catch on everything, which is why they did not persist.[696]
Firmware
Most meters on the market use one-time-programmable microcontrollers, so a defect found after purchase cannot be corrected in the field; manufacturers will occasionally supply a replacement chip to be desoldered and fitted for a technically capable customer, but not as a general remedy, and moving to a flash-programmable part is what makes bug fixes possible at all.[619] Regulation is closing off the one-time-programmable option: rules requiring that anything electronic bearing a conformity mark be updatable would exclude meters built on non-updatable memory from that market entirely, turning a component decision into a market-access decision.[720] Manufacturers separately resist publishing firmware for meters on a safety argument: an instrument’s reading has direct safety consequences, so a modified firmware that reports zero volts on a live conductor creates a hazard the manufacturer would carry, which is why very few meters permit public firmware update.[680]
Continuous measurement architectures
Continuous current measurement over long periods requires a different architecture from a range-switching meter. Measuring on a separate channel that is always set to the highest range, in parallel with the switched ranges, allows the instrument to remain valid through a range change, and the design must also address charge injection from the switching devices themselves.[607]
Use and practice
A single meter is insufficient for ordinary bench work because voltage and current frequently have to be observed at the same instant; characterising a converter’s input and output power simultaneously requires four, and can be done with two only at considerable inconvenience.[18] Handheld instruments are preferred over bench instruments by some practitioners on grounds of mobility rather than performance: a bench meter is tied to one bench while the work moves between benches, offices and homes, and the alternative of carrying the project to the instrument means probing across whatever else already occupies that bench.[148] Three and a half digits covers the great majority of practical work, on the order of eighty to eighty-five percent, so a six-and-a-half-digit instrument sits unused for most measurements even when it is available.[313]
An inexpensive meter is adequate provided its limitations are understood, with the attached caution being not to measure mains with it; the advice against cheap instruments is about knowing where they stop being trustworthy rather than about avoiding them.[97] Starting on modest equipment is defended on the grounds that capability is only appreciated against a remembered limitation, and that a beginner’s difficulties are misattributed to the instrument when they start at the top.[470]
Diagnostic procedures
Continuity testing is the fallback when optical tracing of a board fails: on a four-layer board a via that disappears into an inner layer cannot be followed from photographs, and the remaining method is to buzz out every pin systematically with a meter.[221] Measuring the voltage across every resistor while a device is in its sleep state is a quick way to locate unexpected current paths, and requires nothing beyond a meter before any component is removed from the board.[527] Where a design has separate power domains, fitting two-pin headers rather than soldered links makes measurement practical, since a meter’s leads can be inserted directly into the header to isolate and measure a rail instead of lifting a surface-mount component and tacking wires to it.[132]
In his reverse-engineering work, Trammell Hudson recovered the wiring of industrial robot arms cut free of their cabling through hours of point-to-point continuity checking, which was enough to write a control library for them.[463] Sam Zeloof’s small-scale semiconductor fabrication uses a meter as process control: a test wafer grown with an oxide of the target thickness reads as an open circuit on the resistance range, and is etched in timed steps until it reads as conductive, which establishes the etch time for that layer.[390] Paul Thompson’s field repair programme authorises people who are neither engineers nor technicians provided they can operate a meter, testing candidates on diode test, continuity, short-circuit detection and voltage measurement; a step-by-step procedure built on those four operations resolves roughly eighty percent of field faults, because failures concentrate in a small number of over-represented components.[481]
Calibration in practice
Calibration traceability in an automated production setup is maintained by keeping a calibrated meter in series with the instruments under test, so that devices used to calibrate the next generation of devices do not accumulate error from one generation to the next.[640] A domestic laboratory stresses calibration through ordinary temperature cycling, since a building without environmental control runs hot in summer and cold in winter, a real cost of keeping instruments at home rather than in conditioned space.[523] Test equipment at a contract manufacturer cannot be relied on: a volume factory is a lean operation that holds little prototyping equipment because it is not needed once a design is stable, and what exists is heavily used, often with a damaged input and calibration that cannot be trusted, so the working assumption is to bring one’s own instruments or arrange for them in advance.[279]
Market and manufacture
The industry-standard handheld meter has been sold as essentially the same model for decades, introduced in the late 1980s and revised only twice, with the current revision dating from the early 2000s; the commercial logic against replacing it is that existing owners would not upgrade at that price, while the volume comes from organisations whose approved-equipment lists name that specific model, so a new model would forfeit the position rather than extend it.[534] The requirement space is genuinely wide, which is why on the order of a thousand distinct models exist, and why a fully open design would not automatically win the market, since buyers who want a particular feature set will not accept a small-run product that lacks it.[554] Open-source meter projects recur across decades and rarely reach completion, because a meter is a more complex instrument than it appears; the honest justification offered for such a project is curiosity rather than a gap in the market.[449]
Phone- and tablet-connected meters are an additional tool rather than a substitute for a conventional instrument, a position the designers of such products agree with; the error is treating the accessory as a replacement.[218] Products built around a general-purpose converter and marketed as meters are often data loggers: a sixteen-bit converter recording a signal is specified as a percentage accuracy without a digit count, which is the tell that the device was not designed as a meter and is not treated as one by its makers.[199] Combining instruments into one enclosure tends to produce a poor version of each, a pattern that recurs whenever the feature list rather than the use governs the design.[130] The same argument applies to test accessories: a mains socket tester needs only three indicator lamps whose pattern reveals a miswired outlet, and building a measurement chipset, current measurement and non-contact voltage detection into it makes the tool worse rather than better.[660] An open-source instrument driver layer exists that speaks to a wide range of meters, oscilloscopes and logic analysers and exposes them to graphical front ends; it is built for desktop operating systems and carries too much overhead to run on a small microcontroller, so reusing it for an embedded display means extracting only the required parts rather than adopting it whole.[665]
Supply and procurement
Cash flow rather than margin is the constraint in selling instruments: a production run of five hundred units at fifty dollars of build cost ties up twenty-five thousand dollars before anything ships, with distributor credit covering only part of it, and larger runs of a more expensive model reach six figures per shipment.[287] The downside risk on a commodity-popular instrument is unusually low, because unsold stock can be cleared at wholesale cost to buyers who recognise the value; what the inventory actually costs is tied-up capital rather than exposure to loss.[287] Instruments are ordered on lead times of about two months with no stock held by the manufacturer, so a reseller must forecast that far ahead of running out, and those lead times extend further when component supply tightens upstream.[377] The same instrument can cost roughly double in one country compared with another, which changes who buys it, since a company will absorb the difference while an individual will buy a cheaper instrument instead.[9]
Procurement overhead inside companies can exceed the price of the instrument being bought: requiring a written justification and several signatures for a hundred-dollar meter consumes engineering hours worth many times that, which argues for buying enough instruments outright and distributing them rather than rationing them.[200]
Legal and regulatory history
Trade dress is enforced in this market: a manufacturer asserted rights over the yellow holster of its handheld meters against a competitor, who changed the colour to blue rather than contest it, and separately threatened an importer of inexpensive branded meters on the argument that buyers might confuse the two; the counter-argument is that the purchasers are skilled practitioners choosing deliberately between specifications, so confusion would be difficult to demonstrate.[190] Marking a product with patent numbers that have expired can attract a penalty assessed per unit sold, which for a high-volume product is a substantial exposure, so manufacturers remove the marks when a model is revised.[52]