Electronics Guide

Trueness, Precision, and Accuracy

Trueness describes how closely the average of many measurements approaches the true value; a systematic offset degrades trueness. Precision describes how closely repeated measurements agree with one another, irrespective of whether they cluster around the correct value. Accuracy is the combination of the two: an accurate result is both true and precise. An instrument with a stale calibration can be highly precise and badly untrue, returning the same wrong answer every time.

This distinction shapes measurement strategy. Random scatter can be reduced by averaging, so poor precision is often affordable if enough readings are available. Systematic error is not reduced by averaging at all; it must be removed by calibration, by correction, or by a technique such as reversal that cancels the offending term. Recognizing which kind of error dominates determines whether more readings or a better reference will improve the result.

Resolution, Repeatability, and Reproducibility

Resolution is the smallest change an instrument can display or discern, and it is not a claim of accuracy. An eight-and-a-half-digit multimeter resolves parts per billion of full scale while specifying accuracy in parts per million; the extra digits are useful for detecting change and for null measurements, not for asserting absolute value. Confusing resolution with accuracy is among the most common errors in reported measurements.

Repeatability describes agreement among measurements made under unchanged conditions—same operator, same instrument, same short interval. Reproducibility describes agreement when conditions deliberately vary, such as measurements taken in different laboratories or months apart. Reproducibility is almost always the harder specification to meet, and it is the one that matters when a result must be defended outside the laboratory that produced it.

Josephson Voltage Standards

The Josephson effect provides an intrinsic voltage standard based on fundamental physical constants. When a superconducting tunnel junction is irradiated with microwave energy of frequency f, its current–voltage curve develops quantized steps—Shapiro steps—at voltages V = nf/K_J, where n is an integer step index and K_J is the Josephson constant. Since the 2019 revision of the SI fixed the Planck constant h and the elementary charge e, the Josephson constant K_J = 2e/h takes the exact value 483 597.848 4… gigahertz per volt. The realized voltage therefore depends only on a frequency measurement and on defined constants, both of which can be known extremely well.

A single junction produces a very small voltage: at a drive frequency near 70 gigahertz, the first step corresponds to roughly 145 microvolts. Practical standards therefore connect junctions in series, using arrays that range from thousands of tunnel junctions in conventional standards to hundreds of thousands of junctions in programmable systems, to synthesize outputs up to 10 volts. Programmable Josephson voltage standards divide the array into independently biased segments so that the output can be switched rapidly among selected quantized levels, which permits stepwise-approximated AC waveforms and direct substitution against a device under test. These standards provide the primary realization of the volt for national metrology institutes, and comparisons between independent Josephson systems agree to within parts in 10¹⁰ or better.

Solid-State Voltage References

A Josephson standard requires liquid helium or a cryocooler, microwave instrumentation, and skilled operation, so it is impractical as a day-to-day working reference. That role belongs to solid-state DC reference standards built around temperature-stabilized buried-Zener diodes, which offer low noise and long-term drift on the order of a part per million per year. These instruments typically provide a 10 volt output and a divided 1.018 volt output, the latter chosen for historical continuity with the saturated Weston standard cells that solid-state references displaced during the late twentieth century.

Because Zener references drift predictably rather than randomly, laboratories track them over years and fit the drift, then correct readings against the fitted trend between calibrations. Maintaining a small group of references rather than a single unit allows the group mean to be monitored; a unit that departs from its neighbors reveals itself before it corrupts downstream work. Periodic comparison against a Josephson standard anchors the group to the SI and resets the drift model.

Precision DC Voltage Measurement

Precision digital voltmeters reach accuracies of a few parts per million for DC voltage. Eight-and-a-half-digit instruments specify one-year DC voltage accuracy on the order of ten parts per million, improving to a few parts per million over 24 hours relative to a recent calibration in a temperature-controlled laboratory. Such instruments combine a stable internal voltage reference, a high-linearity integrating or multi-slope analog-to-digital converter, and internal self-calibration routines that periodically re-measure gain and offset against the internal reference.

Integration time is a central setting. Integrating converters average the input over an aperture usually specified in numbers of power line cycles; choosing an integer number of cycles places nulls in the converter response at the line frequency and its harmonics, strongly rejecting mains-borne interference. Longer apertures also reduce random noise, so precision work trades reading rate against resolution and rejection.

Measurement technique matters as much as the instrument. Kelvin (four-wire) connections keep lead resistance out of the measured quantity, guarding intercepts leakage currents before they reach the input, and low-thermal copper connections and consistent connector materials suppress thermoelectric voltages. Reversal techniques, in which the source polarity is inverted and the two readings are averaged, cancel offsets that remain constant through the reversal. Allowing the instrument and the item under test to reach thermal equilibrium before measurement prevents slowly settling thermal gradients from masquerading as drift.

AC Voltage Measurement

Precision AC voltage measurement presents challenges beyond those of DC, including frequency response, waveform dependence, and the distinction among peak, rectified-average, and true root-mean-square values. The reference method is AC–DC transfer: a thermal converter compares an AC voltage with a DC voltage that produces the same heating in a resistive element, reducing an AC measurement to a DC measurement that traceability already supports well. The figure of merit is the AC–DC difference, and thermal transfer standards achieve differences of a few parts per million across audio and low radio frequencies. Digital sampling voltmeters offer greater flexibility and wider bandwidth at somewhat lower accuracy, and they extend naturally to distorted and non-repetitive waveforms.

True RMS measurement responds correctly to any waveform, computing the root-mean-square value that represents the actual heating power of the signal. Instruments that rectify and scale for a sine wave produce large errors on distorted waveforms, a frequent source of trouble when measuring the current drawn by switching power supplies or phase-controlled loads. True RMS instruments carry their own limits: a maximum crest factor, typically specified at full scale and improving on lower ranges, and a bandwidth beyond which high-order harmonics are no longer counted. Understanding these characteristics ensures appropriate application and correct interpretation of results.

Current Shunts

Precision current shunts are resistors designed specifically for current measurement, with stable, accurately known resistance values and low temperature coefficients. Four-terminal (Kelvin) construction, in which separate current and potential terminals are provided, keeps lead and contact resistance out of the measurement. Thermal design ensures that self-heating does not change the resistance appreciably; a shunt dissipating watts must move that heat to a case or heat sink whose temperature rise is bounded and repeatable.

Shunt selection trades sensitivity against burden voltage. A higher resistance produces a larger output voltage and better signal-to-noise ratio, but it inserts a larger voltage drop that can alter the circuit under test and increases self-heating for a given current. Low-resistance shunts minimize burden but demand more sensitive voltage measurement and expose thermoelectric voltages that may be comparable to the signal. At higher frequencies the shunt's series inductance introduces a phase error and a rising magnitude error, so wideband shunts use coaxial or foil geometries that minimize enclosed loop area.

Current Transformers and Transducers

Current transformers provide isolated measurement of alternating current, stepping high currents down to levels a shunt or instrument can accept. Precision current transformers reach small ratio and phase errors through careful core material selection, tight winding symmetry, and controlled burden, and two-stage or compensated designs cancel much of the magnetizing current that causes ratio error. Zero-flux current transducers extend accurate operation to direct current and low frequencies by using a feedback winding to drive the net core flux to zero, so the ratio is set by turns rather than by core properties.

Hall-effect and magnetoresistive sensors provide non-contact current measurement, typically with lower precision than direct methods but with excellent isolation and no insertion loss. These sensors are valuable for high-current or high-voltage applications, in switching converters where the conductor cannot be broken, and wherever safety requires galvanic separation. Their principal error terms are offset drift, gain drift with temperature, and sensitivity to external magnetic fields.

Current Comparators

Current comparators detect the balance between two currents flowing in opposing windings on a common magnetic core, sensing the residual flux with a detection winding. Because the balance condition depends only on the ampere-turns and therefore on turns ratios, which are exact integers, the comparator sidesteps the stability limits of resistive dividers. Cryogenic current comparators, which use a superconducting shield to define the flux linkage, achieve current ratio accuracies at the level of parts in 10⁹ or better. These instruments form the basis for precision resistance and capacitance measurement systems in which ratio measurement is far more accurate than absolute measurement, and they are the standard means of transferring the quantum Hall resistance to practical resistance values.

Quantum Hall Resistance Standards

The quantum Hall effect produces resistance values that depend only on fundamental constants. When a two-dimensional electron system is subjected to a strong magnetic field at cryogenic temperature, its transverse (Hall) resistance becomes quantized on plateaus at values R_H = R_K/i, where i is an integer plateau index and R_K is the von Klitzing constant. Following the 2019 SI revision, R_K = h/e² is exact, with the value 25 812.807 45… ohms; the widely used i = 2 plateau therefore lies at approximately 12 906.403 7 ohms.

Conventional implementations use gallium arsenide heterostructures and require temperatures near 1 kelvin and magnetic fields of roughly 10 tesla. Epitaxial graphene on silicon carbide has since made quantized Hall resistance accessible at markedly higher temperature and lower magnetic field, which allows table-top systems cooled by closed-cycle refrigerators rather than by liquid helium baths. Independent quantum Hall realizations reproduce one another at the level of parts in 10⁹ or better, and a cryogenic current comparator bridge transfers that value to the 100 ohm and 1 ohm working standards that laboratories actually use.

Four-Wire and Guarded Measurement

Four-wire (Kelvin) measurement is essential for precision resistance work. Current is forced through one pair of leads while voltage is sensed through a separate pair carrying negligible current, so the resistance of the leads and of the contacts falls outside the measured quantity. The benefit is largest for small resistances: ordinary test leads and connections contribute on the order of tens to hundreds of milliohms, which is a negligible fraction of a megohm but a large error on a 1 ohm standard. As a working rule, four-wire connection is required whenever the resistance is low enough that lead and contact resistance is comparable to the target uncertainty.

Guarding extends precision measurement in the opposite direction, to high-value resistors where leakage currents rather than lead resistance set the limit. A guard conductor held at nearly the same potential as the high terminal presents almost no voltage across the leakage path, so leakage current is intercepted before it can join the measured current. Precision instruments drive the guard automatically, and high-resistance work also demands attention to humidity, surface contamination, and insulator choice, since the leakage paths in question run across physical surfaces.

Resistance Bridges and Ratio Transfer

Bridges compare an unknown resistance to a standard through a null-balance measurement, which is inherently more accurate than absolute measurement because it depends on a ratio and on the sensitivity of a detector near zero rather than on the absolute accuracy of a voltmeter. Modern direct-current comparator bridges and cryogenic current comparator bridges use magnetically defined ratios, reaching uncertainties far below what a direct reading instrument can offer, while digital control and automatic balancing remove the tedium of manual null seeking.

Because the most accurate comparisons are ratio measurements, scaling between decades is itself a metrological operation. Series-parallel (Hamon) transfer resistors, which contain nominally equal elements that can be connected either in series or in parallel, provide an accurate 100:1 step that depends on the equality of the elements rather than on their absolute values. Chains of such transfers propagate the quantum Hall value across the resistance decades from milliohms to gigohms.

Temperature and Time Effects

Resistance varies with temperature, so precision measurement requires either controlling temperature or correcting for its effect. Temperature coefficients of precision resistors are specified in parts per million per kelvin, and the best standard resistors and bulk metal foil elements achieve coefficients below 1 ppm/K near their reference temperature. Even so, a 1 ppm/K coefficient turns a 2 kelvin laboratory excursion into a 2 ppm error, which is significant at this level of work, so primary standard resistors are commonly maintained in stirred oil baths or air enclosures held to better than 0.01 kelvin.

Resistance also changes over time through aging, typically drifting at a decelerating rate that can be modeled once enough history exists. Precision resistors are artificially aged before final adjustment to move past the fastest part of that curve. The practical consequence is that a standard resistor is characterized by its drift rate as much as by its nominal value, and calibration intervals should follow observed stability rather than a default schedule. Mechanical shock, moisture ingress, and excessive power dissipation all cause step changes that invalidate a fitted drift model, which is why standards are handled deliberately and their history is recorded.

Capacitance Measurement

Precision capacitance measurement uses bridges, resonance methods, or LCR meters, according to the accuracy required and the value being measured. Transformer ratio-arm bridges achieve the highest accuracy by comparing an unknown capacitor against a standard through a ratio defined by turns on a common core, reaching part-per-million uncertainties and better. Fused-silica and gas-dielectric capacitors serve as the stable transfer standards in the 10 picofarad to 1000 picofarad range that such bridges compare. At the primary level, the calculable cross-capacitor exploits a geometric theorem that makes a capacitance calculable from a single length measurement, historically providing a route to the farad and, through impedance bridges, to the ohm; laboratories now more commonly derive impedance standards from the quantum Hall effect using quadrature bridges.

Parasitic capacitance from cables, fixtures, and nearby objects can dominate low-capacitance measurements, since a hand near an open fixture is worth a fraction of a picofarad. Guarding, three-terminal connection schemes, and open and short compensation of the fixture remove most of this error. Measurement frequency also affects the result, because dielectric permittivity and loss vary with frequency and because series inductance eventually resonates with the capacitance under test. Standards work frequently uses 1 kilohertz or 1592 hertz, the latter corresponding to an angular frequency of 10⁴ radians per second, but component measurements should be made at a frequency relevant to the intended application.

Inductance Measurement

Inductance is the least accurately realized of the common electrical quantities, and inductance standards are usually derived from capacitance and resistance standards through bridges such as the Maxwell–Wien configuration rather than from a calculable inductor. Practical measurement is complicated by the strong frequency dependence of real inductors and by their resistive losses. LCR meters measure complex impedance and resolve it into reactive and resistive parts, reporting inductance together with a quality factor or an equivalent series resistance; benchtop instruments offer basic accuracy in the range of roughly 0.05 to 0.1 percent under favorable conditions.

Magnetic cores make the measurement conditions part of the answer. Permeability depends on flux density, so a ferrite or powdered-iron inductor measured at millivolt excitation may show a different inductance than the same part carrying its rated current, and approaching saturation reduces inductance sharply. Precision characterization therefore specifies both the test frequency and the test current or applied volt-seconds. At high frequency, distributed winding capacitance combines with the inductance to produce a self-resonant frequency above which the component behaves capacitively; characterizing that frequency is essential for radio-frequency and switching-converter design.

Impedance Measurement

General impedance measurement characterizes a component without assuming a simple equivalent circuit. Impedance analyzers sweep frequency while measuring magnitude and phase, revealing the full behavior of a real part: the equivalent series resistance minimum of a capacitor, the self-resonance of an inductor, and the multiple resonances of a physically large component. Vector network analyzers extend the same idea to radio and microwave frequencies through reflection coefficient measurement, with calibration standards that move the measurement reference plane to the device terminals.

These measurements feed directly into simulation. A model fitted to measured impedance across frequency predicts circuit behavior far better than a nameplate value, and the discrepancies between the model and the measurement often expose fixture problems or unexpected component behavior that would otherwise appear only as a puzzling result late in development.

Frequency Standards

Atomic frequency standards use the invariant resonance frequencies of atoms as references. The SI second is defined by fixing the cesium-133 ground-state hyperfine transition frequency at exactly 9 192 631 770 hertz, and cesium fountain clocks realize that definition with fractional uncertainties of a few parts in 10¹⁶. Rubidium standards and hydrogen masers offer excellent short-term stability, smaller size, and lower cost while requiring periodic steering to remain accurate over long intervals. Optical lattice and trapped-ion clocks, which interrogate transitions tens of thousands of times higher in frequency, now reach systematic uncertainties near the 10⁻¹⁸ level and are the leading candidates for a future redefinition of the second.

Crystal oscillators provide stable references for most electronic applications. Temperature-compensated oscillators (TCXOs) achieve stability of roughly one part per million over their operating temperature range, while oven-controlled oscillators (OCXOs) hold the crystal at a fixed elevated temperature and reach parts per billion, at the cost of warm-up time and continuous heater power. Because quartz ages, an OCXO drifts steadily even in a constant environment, so long-term accuracy comes from disciplining. A GPS-disciplined oscillator steers a quartz or rubidium oscillator to satellite-broadcast time, combining the excellent short-term stability of the local oscillator with frequency accuracy approaching parts in 10¹² when averaged over a day, and it holds over gracefully when satellite reception is lost.

Frequency Measurement Techniques

Conventional frequency counters count input cycles over a gate interval derived from a reference oscillator. Resolution improves in proportion to gate time but is limited by the ±1 count ambiguity, which is severe at low frequencies: counting a 10 hertz signal for one second yields one digit of resolution. Reciprocal counters instead measure the period of an integer number of input cycles against a high-frequency internal clock, delivering the same number of significant digits regardless of input frequency and measuring low frequencies quickly.

Whatever the architecture, a counter can be no better than its timebase, so a counter reading a signal to twelve digits against an undisciplined internal crystal is reporting resolution rather than accuracy. For the finest work, heterodyne and phase-comparison techniques beat the unknown against a reference and measure the small difference frequency directly, so that the full resolution of the measurement is applied only to the discrepancy. This approach resolves fractional frequency differences far below what direct counting allows, and phase-tracking measurements of the accumulated phase difference are the standard method for comparing high-performance oscillators.

Time Interval Measurement

Time interval measurement determines the elapsed time between two events, supporting propagation delay characterization, pulse width measurement, radar and lidar ranging, and verification of network synchronization. A counter clocked at a few hundred megahertz would resolve only nanoseconds by direct counting, so precision instruments add interpolation: analog ramp or time-to-digital converter circuits measure the fractional clock periods at the start and stop events, extending single-shot resolution into the low picoseconds.

At that resolution the trigger circuit and the cabling become the limiting factors. Trigger-level timing error caused by noise on a slowly rising edge, channel-to-channel skew, and the propagation delay of the cables themselves all enter the result, and each must be characterized or removed by a reversal or self-calibration measurement. Averaging many measurements of a repetitive interval reduces random jitter, but it cannot remove a fixed delay offset.

Characterizing Stability

Standard deviation is a poor description of oscillator performance, because the noise processes in real oscillators include random-walk and flicker components for which the ordinary standard deviation does not converge as more data are added. Frequency metrology therefore uses the Allan deviation, computed from the differences between successive average frequencies, which remains well behaved in the presence of these processes.

Plotted against averaging time, the Allan deviation separates the noise mechanisms of a source: it typically falls with increasing averaging time while white noise dominates, flattens across a flicker floor, and rises again once drift and random walk take over. The minimum identifies the averaging time that yields the best possible measurement, and the shape tells the designer whether a better oscillator, a longer measurement, or a better environment would help. The same plot shows why a disciplined oscillator is configured with a long control time constant: the local oscillator is superior at short intervals, and the reference is superior at long ones.

Thermal, Shot, and Flicker Noise

Every resistance generates Johnson–Nyquist noise by virtue of its temperature alone. The open-circuit noise voltage in a bandwidth B is the square root of 4k_BTRB, where k_B is the Boltzmann constant and T the absolute temperature. At room temperature a 1 kilohm resistor produces about 4 nanovolts per square root hertz, and the noise rises as the square root of resistance, so a 1 megohm source produces about 130 nanovolts per square root hertz. This sets an immediate limit on high-impedance measurement and explains why source impedance is a design variable in precision analog front ends.

Shot noise arises from the discrete nature of charge crossing a barrier and scales as the square root of the current, so it dominates in photodiode and low-current measurements where the signal current is small. Flicker noise, with a power spectral density rising as 1/f toward low frequency, is the practical enemy of DC precision: it merges seamlessly into what an experimenter calls drift, and it is why the best low-frequency amplifiers use chopper or auto-zero techniques that translate the measurement to a higher frequency where flicker noise is negligible and translate it back afterward.

Averaging, Bandwidth, and Drift

For white noise, averaging N independent readings improves the signal-to-noise ratio by the square root of N, which is the justification for long integration times in precision instruments. The returns diminish quickly: reducing noise tenfold requires a hundred times as many readings, and the required measurement duration eventually collides with the stability of the apparatus and the patience of the operator.

Averaging also fails outright against 1/f noise and drift, because longer measurements admit more low-frequency content rather than less. This is the fundamental reason that precision measurement favors ratio, substitution, and reversal methods: comparing an unknown against a standard within a few seconds rejects the slow variation that would corrupt two absolute measurements made minutes apart. Narrowing the measurement bandwidth around a modulated signal achieves the same end by different means, which is the principle exploited by synchronous detection in lock-in amplifiers.

Temperature Control

Temperature affects virtually every aspect of precision measurement: component values, instrument gain and offset, thermoelectric voltages, and physical dimensions. Metrology laboratories work at the standard reference temperature of 20 degrees Celsius, commonly held within ±1 degree and, for the most demanding work, within a few hundredths of a degree in a controlled enclosure. Airflow and radiant heating deserve as much attention as the setpoint, because a temperature gradient across an instrument matters more than the mean room temperature. Items brought in from elsewhere must be allowed to equilibrate, which for a massive standard resistor in an oil bath can take many hours.

Thermoelectric voltages arising from the Seebeck effect at junctions of dissimilar metals introduce microvolt-level errors in precision DC measurements, and a copper-to-solder junction with a one-degree gradient across it is enough to matter at the microvolt level. Countermeasures include using consistent connector materials, preferring low-thermal copper-to-copper connections, shielding connections from drafts, minimizing gradients by keeping heat sources away from the signal path, and reversing the source polarity so that the averaged reading cancels the thermal offset.

Electromagnetic Environment

Electromagnetic interference corrupts precision measurements through both conducted and radiated coupling. Shielded enclosures, filtered mains supplies, and isolation transformers protect the most sensitive measurements, while twisted pairs, coaxial cabling, and short signal paths reduce the loop areas that pick up magnetic fields. Mains-frequency interference is best attacked at the instrument as well, by choosing an integration aperture that is an integer number of power line cycles.

Ground loops form when multiple ground connections create a closed conductive loop that couples magnetic interference and carries stray currents; the resulting voltage appears in series with the signal. Single-point grounding, guarded inputs, differential measurement, and galvanic isolation of instrument interfaces all prevent or break such loops. Optical isolation of digital control links is a common practical measure, since the automation cabling that makes a precision measurement convenient is often what injects interference into it.

Vibration and Mechanical Stability

Mechanical vibration affects measurements through microphonic effects in cables and components, through triboelectric charge generation as cable dielectrics flex, and through displacement of sensitive elements. Vibration isolation tables, low-noise cable constructions, and cable routing that avoids unsupported spans minimize these effects. Some measurements require scheduling around building activity, elevator use, or traffic, and a night measurement is occasionally the most economical solution to a vibration problem.

Measurement Procedures

Documented measurement procedures ensure consistent technique and make results comparable over time and across operators. A procedure should specify instrument settings, connection details, warm-up and equilibration times, measurement sequence, environmental requirements, and acceptance criteria. Specifying the sequence matters more than it appears to: a procedure that brackets each unknown between two standard readings automatically cancels linear drift, whereas one that measures all standards first and all unknowns afterward does not.

Uncertainty Budgets

Formal uncertainty analysis identifies every significant error source and combines them into a statement of how well the result is known. International practice follows the Guide to the Expression of Uncertainty in Measurement, which classifies contributions as Type A, evaluated statistically from repeated observations, or Type B, evaluated by other means such as calibration certificates, manufacturer specifications, or physical reasoning about a bounded effect.

Each contribution is expressed as a standard uncertainty, converting a stated interval into an equivalent standard deviation through an assumed distribution—a rectangular distribution for a specification limit divides the half-width by the square root of three. The contributions are then combined in quadrature, weighted by sensitivity coefficients that describe how strongly each affects the result, and the combined standard uncertainty is multiplied by a coverage factor, conventionally k = 2 for an interval of approximately 95 percent confidence. The budget is as valuable for what it reveals as for the number it produces: quadrature summation means the largest term dominates, so the table immediately shows where effort would improve the measurement and where it would be wasted.

Measurement Assurance and Traceability

Measurement assurance programs verify that a measurement system continues to perform as its uncertainty budget claims. Check standards measured on a regular schedule and plotted on control charts detect drift or degradation before it affects delivered results, and they distinguish ordinary variation from an assignable cause that warrants investigation. A calibration certificate describes performance on one day; a control chart describes performance continuously.

Traceability supplies the external anchor. An unbroken chain of documented comparisons, each with a stated uncertainty, links a working measurement to a national or international realization of the SI unit, and calibration laboratories commonly demonstrate their competence through accreditation to ISO/IEC 17025. Interlaboratory comparisons provide an independent check that laboratories customarily trusting their own procedures actually agree, and they occasionally uncover a systematic error that no amount of internal consistency would have exposed.