Effective Dielectric Constant Calculator (Microstrip)
Calculate microstrip effective dielectric constant εeff with Hammerstad–Jensen, plus Getsinger dispersion, propagation velocity, delay in ps/mm and ps/inch, and guided wavelength.
Hammerstad–Jensen static εeff · Getsinger (1973) dispersionAPI slug effective-dielectric-constant
Formula
E. Hammerstad & Ø. Jensen, "Accurate Models for Microstrip Computer-Aided Design", IEEE MTT-S 1980; W. J. Getsinger, "Microstrip Dispersion Model", IEEE Trans. MTT-21, 1973
| Symbol | Quantity | Your value |
|---|---|---|
| εeff | Effective dielectric constant seen by the wave | 3.066 |
| εr | Substrate relative permittivity | 4.3 |
| w | Trace width (mm) | 300 µm |
| h | Dielectric height to the reference plane (mm) | 200 µm |
| t | Copper thickness (widens the effective trace) (mm) | 35 µm |
| Δu₁ | Width the copper adds in air, (t/πh)·ln(1 + 4e/((t/h)·coth²√(6.517·w/h))) (Hammerstad–Jensen 1980) | — |
| Δuᵣ | Width the copper adds with the dielectric, ½(1 + sech√(εr − 1))·Δu₁ | — |
| Z₀₁(u) | Hammerstad–Jensen impedance of the strip in air at width ratio u (Ω) | — |
| q | Filling factor, (εeff−1)/(εr−1) | 0.6262 |
| tpd | Propagation delay, √εeff / c (ps/mm) | 5.841 ps/mm |
Assumptions
6 items
- 01Static εeff is Hammerstad–Jensen (1980), the closed form the microstrip impedance calculators use, with copper thickness by their own correction: εeff = εe(u + Δuᵣ)·(Z₀₁(u + Δu₁)/Z₀₁(u + Δuᵣ))², the trace widened by Δu₁ in air and Δuᵣ with the dielectric.
- 02Dispersion follows Getsinger (1973): εeff(f) = εr − (εr − εeff)/(1 + G·(f/fp)²), with fp = Z₀/(2µ₀h), G = 0.6 + 0.009·Z₀, µ₀ = 4π × 10⁻⁷ H/m and Z₀ the static Hammerstad–Jensen value.
- 03The guided wavelength and quarter wave use the dispersive εeff(f); the delay, the velocity and the filling factor q = (εeff − 1)/(εr − 1) use the static value, and q is taken as 1 when εr = 1.
- 04Bare microstrip, quasi-TEM and lossless: air above the trace and no soldermask, one homogeneous substrate over a solid plane, and no conductor or dielectric loss.
- 05Warnings flag w/h outside 0.05–20 or εr above 12, where the fit no longer holds to about 1 %, and a dispersion shift above 2 %, but a result is still returned; a width or height that is not positive, or εr below 1, is refused.
- 06Synthesis solves the trace width or the dielectric height numerically within its range for a target Z₀, or the width for a target static εeff; the material picker sets εr to the laminate’s design Dk.
Worked example
50 Ω line on 0.508 mm RO4003C at 10 GHz
A 10 GHz filter is laid out on 0.508 mm RO4003C (design Dk 3.55) with 35 µm copper. Solve the trace width for a 50 Ω line, then find how far dispersion moves εeff and how long a quarter-wave stub must be.
| Step | Expression | Result |
|---|---|---|
| Trace width for 50 Ω (solved) | Z₀(w) = 50 Ω; h = 0.508 mm, εr = 3.55, t = 35 µm | 1.0974 mm |
| Width the copper adds in air, Δu₁ | 0.035 / (π × 0.508) × ln(1 + 4 × exp(1) × tanh(√(6.517 × 1.0974 / 0.508))² × 0.508 / 0.035) | 0.1111 |
| Width the copper adds with the dielectric, Δuᵣ | ½ × (1 + 1/cosh(√(3.55 − 1))) × 0.1111 | 0.07716 |
| Static εeff (Hammerstad–Jensen) | Hammerstad–Jensen εe(uᵣ, εr) × (Z₀₁(u₁)/Z₀₁(uᵣ))², uᵣ = 1.0974 / 0.508 + 0.07716, u₁ = 1.0974 / 0.508 + 0.1111, εr = 3.55 | 2.738 |
| Getsinger's frequency f_p | 50 Ω / (2µ₀ × 0.508 mm) | 39.16 GHz |
| With dispersion at 10 GHz (Getsinger) | 3.55 − (3.55 − 2.738) / (1 + (0.6 + 0.009 × 50) × (10 / 39.16)²) | 2.79 |
| Propagation delay (static εeff) | √2.738 / 0.299792458 | 5.519 ps/mm |
| Quarter-wave stub at 10 GHz | 0.299792458 / (10 × 10⁻³ × √2.79) / 4 | 4.487 mm |
How it works
A microstrip is a mixed-dielectric structure. The trace sits on the laminate with a reference plane below it, but above the trace is air. The field lines therefore split between two media of different permittivity, and the wave propagates as if it were travelling in a single uniform medium of some intermediate value , strictly bounded by . Every number that depends on velocity — propagation delay, guided wavelength, stub length, electrical length of a matching section — depends on and not on the laminate printed in the datasheet.
Hammerstad and Jensen fitted to the exact quasi-static field solution as a function of a single shape parameter , with an -dependent exponent. Their closed form holds to roughly 1 % over and , which covers essentially all printed-circuit work. Copper thickness enters by Hammerstad and Jensen's own 1980 correction: the trace widens by in air, with , and by with the dielectric, and , because a conductor with real thickness presents extra sidewall to the field, most of it in air; ignoring it overstates noticeably on heavy copper. Against rftools.io's field solver this holds to 0.5 % for up to 0.35, where the 1975 width increase used until November 2026 read up to 5.8 % low.
Two limits confirm the model is doing the right physics. As the trace barely disturbs the air-dielectric interface and , the arithmetic mean of the two half-spaces. As the field is squeezed entirely into the substrate and — the parallel-plate limit, where a microstrip behaves like a stripline. Between them the filling factor states directly what fraction of the field energy sits inside the laminate.
The quasi-static result is frequency-independent, which is not quite true. As frequency rises the field concentrates further into the higher-permittivity substrate and climbs monotonically toward . Getsinger's model captures this as with and . The correction scales as , so it is under 1 % on thin laminate below about 10 GHz and becomes several percent on thick substrate or above roughly 15 GHz.
Practical tips
- Sanity-check any epsilon-eff result against the two bounds: it must lie between (epsilon-r + 1)/2 and epsilon-r. A value outside that range means a units or geometry error, not a marginal model
- Use the filling factor for tolerance analysis — a change in laminate epsilon-r moves epsilon-eff by roughly q times that change, so a line with q = 0.68 passes about two thirds of any laminate variation into your delay budget
- Quote delay in ps/inch when talking to signal-integrity engineers and ps/mm when doing layout arithmetic. Both are the same number scaled by 25.4, and mixing them is a common source of length-matching errors
- Wider traces raise epsilon-eff, which means the delay of a controlled-impedance line is not constant across a board with mixed geometries. Length-match on delay, not on physical length, when the widths differ
- For frequency-critical structures on thick substrate, run the calculation at your operating frequency rather than DC. The dispersion output tells you immediately whether it matters or can be ignored
Common mistakes
- Using the laminate epsilon-r instead of epsilon-eff for stub and wavelength calculations — on FR-4 that gives a quarter-wave stub about 13 percent too short, enough to move a filter edge by hundreds of MHz
- Assuming microstrip and stripline have the same delay on the same laminate — a stripline is fully buried so its epsilon-eff equals epsilon-r exactly, making it roughly 15 percent slower than a microstrip on identical material
- Ignoring copper thickness on heavy copper builds — 2 oz copper adds a substantial effective width, and leaving it out overstates Z0 by several percent
- Applying the quasi-static value at mmWave frequencies — above roughly 15 GHz, or on thick substrate, epsilon-eff climbs measurably toward epsilon-r and the static number under-predicts delay
- Forgetting soldermask — mask is roughly epsilon-r 3.5 and replaces air above the trace, raising epsilon-eff by 2 to 5 percent and lowering Z0 by a similar amount on a typical 50 ohm line
Frequently asked questions
6 questions
Why is epsilon-eff lower than the laminate epsilon-r?
Because a microstrip is only half-buried. Part of the field travels through the air above the trace where permittivity is 1, pulling the average down. A stripline is fully surrounded by dielectric so its epsilon-eff equals epsilon-r exactly, and it is correspondingly slower.
What is a typical epsilon-eff for FR-4?
A 50 ohm microstrip on 4.3 epsilon-r FR-4 usually lands between 3.0 and 3.4 depending on the width-to-height ratio. Wider traces push more field into the substrate and raise epsilon-eff; narrow traces let more field into the air and lower it.
How does epsilon-eff change a quarter-wave stub length?
Physical length scales as 1 over the square root of epsilon-eff. At 2.4 GHz a free-space quarter wave is 31.2 mm; on FR-4 microstrip with epsilon-eff of 3.23 it shrinks to 17.38 mm. Using epsilon-r = 4.3 instead would give 15.06 mm, a 13.3 percent error.
When does dispersion actually matter?
The correction scales with the square of frequency times substrate height. On 0.2 mm substrate it is well under 1 percent up to 10 GHz and can be ignored. On 1.6 mm substrate, or above roughly 15 GHz on any stack-up, epsilon-eff can climb several percent toward epsilon-r and must be used for wavelength-critical structures.
What is the filling factor q used for?
q = (epsilon-eff minus 1) divided by (epsilon-r minus 1) states what fraction of the field energy is inside the substrate. It is convenient for sensitivity analysis: a laminate tolerance of plus or minus 0.2 in epsilon-r moves epsilon-eff by roughly q times 0.2, which propagates straight into your delay and impedance budget.
Does soldermask change epsilon-eff?
Yes. Soldermask is roughly epsilon-r 3.5 and replaces some of the air above the trace, raising epsilon-eff by about 2 to 5 percent and lowering Z0 by a similar amount on a typical 50 ohm line. This calculator models bare microstrip; for a masked line use the embedded-microstrip mode of the controlled impedance calculator.
Shop Components
As an Amazon Associate we earn from qualifying purchases.