Printed Resistor AC Impedance: Resistance, Reactance and Fixture Boundaries

1. Define what approximately resistive means for the circuit

Record whether the application limits impedance magnitude, its real component, phase angle or the resulting transfer function. These quantities are not interchangeable. A nearly unchanged magnitude can coexist with a phase error significant to a bridge or pulse-measurement circuit. Retain the frequency band and the reference DC value used for normalization.

Specify the electrical amplitude and any permitted DC bias as well as temperature. The frequency sweep should remain within the element's linear and thermally controlled region. A change observed only after dissipating appreciable power is not necessarily an intrinsic high-frequency parasitic effect; the resistance may have changed with its operating state.

2. Define the physical connection plane

Draw the measurement boundary at named terminals and show the surrounding ground or return conductor. Decide whether connecting traces, bond wires and landing pads belong inside that boundary. A model extracted at a coaxial connector includes a different physical network from one referred to the printed element's terminal edges.

Keep mounting height, substrate support and nearby conductive surfaces consistent during comparisons. Moving a metal support or changing the return path can change capacitance and inductance while leaving the printed paste untouched. If the intended installed response includes those structures, retain them in the model instead of subtracting them merely to obtain a smaller parasitic number.

3. Start with a declared low-order equivalent circuit

One possible two-terminal approximation places a series inductance ahead of a parallel resistance and capacitance. It can represent a limited frequency interval when its terms adequately describe the observed response. It is not a universal internal structure for every thick-film resistor. Additional ground capacitance, distributed paths or frequency-dependent material behavior may require a different network.

Use the simplest model that answers the specified circuit question and explains the complex data within their uncertainty. A more elaborate fit can reduce residuals while making its individual coefficients poorly identifiable. Keep the intended topology with the coefficient list; the same labels R, L and C can describe different connections and therefore different responses.

4. Examine phase even when magnitude changes little

For an assumed 1 kΩ resistor with 1 pF parallel capacitance and negligible series inductance, the frequency at which ωRC equals one is approximately 159.15 MHz. At one tenth of that frequency, about 15.915 MHz, the impedance is approximately 990.10 − j99.01 Ω. Its magnitude is about 995.04 Ω, only 0.496% below the DC value, while its phase is approximately −5.71 degrees.

A magnitude-only one-percent comparison would not expose a five-degree phase requirement failure in this example. At the same 15.915 MHz, an assumed 10 nH series inductance adds j1 Ω. The resulting imaginary component is about −98.01 Ω, not zero. These assumed values illustrate the stated circuit model within its declared lumped-network boundary.

5. Separate fixture compensation from changing the specimen boundary

Follow the chosen instrument's calibration and compensation method at the intended reference plane. Open, short and suitable load conditions serve different error-model functions. Their physical implementation must represent the fixture sufficiently well for the required frequency band. An open created far from the specimen contacts is not automatically equivalent to an open at those contacts.

Record the standards, plane, fixture arrangement and compensation state with the sweep. Check an independent known device where practical and revisit the correction after reconnecting or rearranging the fixture. A smooth corrected trace does not establish that the residual error is negligible. Retain the uncorrected and corrected records internally when they help expose how much of the apparent response came from the fixture.

6. Test whether the fitted parameters are actually distinguishable

At frequencies where ωRC is much less than one, expanding the simple model gives an approximate imaginary part of ω(Ls − R²Cp). The measured low-frequency slope therefore constrains a combination of inductance and capacitance. It may not separately determine both with useful precision.

For an assumed 1 kΩ resistance, R² times 1 pF equals 1 microhenry in this first-order expression. A very small apparent reactance slope could result from partial cancellation rather than from both reactive terms being absent. Use enough justified frequency span and additional boundary information to discriminate the terms. Do not extrapolate a narrow-band fit into an unmeasured resonance merely because the fitting software returns precise coefficients.

7. Choose the next comparison from the complex response

Compare real and imaginary components as well as normalized magnitude. Their patterns help separate a changed resistive baseline from fixture movement or an inadequate topology, but they do not identify a physical mechanism without further evidence.

8. Recheck the response after a relevant construction change

A revised terminal, trim geometry, protection layer or mounting arrangement can preserve DC resistance while changing the measured AC network. Compare those revisions using the same defined plane and operating state. Keep any process-specific trend tied to the actual structures measured rather than assigning one parasitic model to every resistor made from a common paste family.

For the engineering handoff, provide the usable measured band, the complex data, uncertainty or verification limits and the fitted topology if one is required. State separately whether the requirement belongs to a standalone element or its installed assembly. This supplies the customer with a reproducible electrical boundary and prevents an attractive model fit from becoming an unsupported broadband performance claim.