How do you calculate patch antenna dimensions?
About half a guided wavelength long and a little wider. For 2.45 GHz on 1.6 mm FR-4 (εr 4.4), W = (c ÷ 2f) √(2 ÷ (εr + 1)) = 37.2 mm, and after the fringing correction L = 28.8 mm. Feeding it 10.7 mm in from the edge gives an estimated 50 Ω match.
Microstrip patch antenna design equations
The transmission-line model treats the patch as a low-impedance line with a radiating slot at each end. Design runs in this order: width, effective permittivity, fringing extension, length.
W = (c ÷ 2f) × √(2 ÷ (εr + 1))
- W
- patch width, across the feed direction
- c
- speed of light
- f
- resonant frequency
- εr
- substrate relative permittivity
εeff = (εr + 1) ÷ 2 + ((εr − 1) ÷ 2) × (1 + 12 h ÷ W)^(−1/2)
- h
- substrate thickness
- εeff
- effective permittivity, for W/h above 1
ΔL = 0.412 h × (εeff + 0.3)(W/h + 0.264) ÷ ((εeff − 0.258)(W/h + 0.8))
- ΔL
- fringing extension at each radiating edge
L = c ÷ (2f √εeff) − 2ΔL
- L
- physical patch length, along the feed direction
Edge resistance comes from the radiation conductance G1 of one slot and the mutual conductance G12 between the two slots, with X = k0 W. The inset depth y0 follows from a cos² law along the patch:
G1 = (−2 + cos X + X Si(X) + sin X ÷ X) ÷ 120π²
G12 = (1 ÷ 120π²) ∫ [sin((X ÷ 2) cos θ) ÷ cos θ]² J0(k0 L sin θ) sin³θ dθ
Redge = 1 ÷ (2 (G1 + G12)) y0 = (L ÷ π) arccos √(Z0 ÷ Redge)
- Si
- the sine integral
- J0
- the Bessel function of the first kind, order zero
- k0
- 2π ÷ λ0
- θ
- the angle being integrated over, 0 to π
- Redge
- resistance at the radiating edge, at resonance
- Z0
- feed line impedance
The plus sign in Redge is for odd modes, where the voltage has opposite signs at the two slots, which is the dominant mode sized here. Even modes take a minus sign. G12 uses the physical length L, not L + 2ΔL.
Worked examples
Example 1: 2.45 GHz on 1.6 mm FR-4 (εr 4.4 assumed)
c ÷ 2f = 61.18 mm and √(2 ÷ 5.4) = 0.6086, so W = 37.23 mm. Then εeff = 4.081, ΔL = 0.739 mm and L = 28.81 mm. G1 = 969.3 µS and G12 = 586.2 µS, so the estimated edge resistance is 321 Ω and a 50 Ω feed goes in 10.7 mm from the edge, on a 3.08 mm wide line. Leaving out G12 would give 516 Ω and an inset of 11.5 mm.
Example 2: 5.8 GHz on 0.813 mm RO4003C (design Dk 3.55)
W = 17.13 mm, εeff = 3.293, ΔL = 0.387 mm and L = 13.47 mm. The inset estimate is 4.89 mm.
Example 3: what frequency does a 30 by 38 mm FR-4 patch resonate at?
With L = 30 mm, W = 38 mm, h = 1.6 mm and εr 4.4, εeff = 4.086 and ΔL = 0.739 mm, so fr = c ÷ (2 √εeff (L + 2ΔL)) = 2.356 GHz.
Example 4: checking the edge resistance against a published example (Balanis, Example 14.2)
L = 9.06 mm, W = 11.86 mm, 10 GHz, with c rounded to 3 × 10⁸ m/s as the example does. Numerical integration gives G1 = 1.573 mS and G12 = 616.8 µS, so Redge = 228.4 Ω and the 50 Ω inset is 3.126 mm. EMPossible slide 36 prints G1 = 1.57 mS and G12 = 616.8 µS for this patch. FOSSEE’s Python notebook for the example keeps more digits of G1 and prints 228.35 Ω and 3.126 mm. Without G12 the same patch would read 318 Ω.
Patch antenna dimensions at common frequencies
Width by length, from the formulas above. For 2.4 GHz patch antenna dimensions, read the 2.45 GHz column, or enter your exact frequency in the calculator at the top of this page.
| Substrate | 915 MHz | 1.575 GHz | 2.45 GHz | 5.8 GHz |
|---|
| FR-4, 1.6 mm (εr 4.4 assumed) | 99.7 mm × 77.9 mm | 57.9 mm × 45.1 mm | 37.2 mm × 28.8 mm | 15.7 mm × 11.7 mm |
| RO4003C, 1.524 mm (Dk 3.55) | 109 mm × 86.7 mm | 63.1 mm × 50.2 mm | 40.6 mm × 32.1 mm | 17.1 mm × 13.1 mm |
| RT/duroid 5880, 1.575 mm (Dk 2.20) | 130 mm × 110 mm | 75.2 mm × 63.4 mm | 48.4 mm × 40.5 mm | 20.4 mm × 16.5 mm |
Input resistance of an inset-feed microstrip patch antenna
The cos² law for any patch. Multiply the edge resistance by the factor; a depth of L/2 is the center of the patch, where the resistance goes to zero.
| Inset y0 / L | Factor | From 300 Ω edge | From 500 Ω edge |
|---|
| 0 | 1.000 | 300 Ω | 500 Ω |
| 0.05 | 0.976 | 293 Ω | 488 Ω |
| 0.1 | 0.905 | 271 Ω | 452 Ω |
| 0.15 | 0.794 | 238 Ω | 397 Ω |
| 0.2 | 0.655 | 196 Ω | 327 Ω |
| 0.25 | 0.500 | 150 Ω | 250 Ω |
| 0.3 | 0.345 | 104 Ω | 173 Ω |
| 0.35 | 0.206 | 62 Ω | 103 Ω |
| 0.4 | 0.095 | 29 Ω | 48 Ω |
What moves a real patch off these numbers
Resonance is very sensitive to L and to the actual εr. On the FR-4 example above, 1 % more length lowers the resonance by 0.94 % and 1 % more εr lowers it by 0.48 %, and a simple patch has at most a couple of percent of bandwidth to absorb that. Use the laminate’s design Dk where one is published: Rogers lists 3.55 for RO4003C against a process value of 3.38. FR-4 has no single permittivity, so measure it or trim after a first board, and expect its loss to cost efficiency at microwave frequencies. The model also assumes a ground plane larger than the patch and a uniform field through the substrate, which is why it needs h well under a tenth of a wavelength. Above about 8 GHz the quasi-static εeff needs a dispersion correction. Like any microstrip antenna calculator built on this model, this one gives a starting point for an EM simulation or a bench trim.
The mutual conductance is not a small correction here. The McMaster notes cited below say G12 is normally much smaller than G1, but it is 39 % of G1 in the textbook example above and 60 % in the FR-4 example, because a higher εr brings the slots closer in free-space wavelengths. G1 and G12 describe radiation only: copper and dielectric loss, which matter on FR-4, and the inset notch itself are not in the formula.
Glossary
- Patch antenna
- A flat metal rectangle over a ground plane, about half a guided wavelength long, that radiates from its two opposite edges.
- Relative permittivity (εr, Dk)
- How much the substrate slows a wave compared with free space.
- Effective permittivity (εeff)
- The single value that accounts for fields being partly in the substrate and partly in the air.
- Fringing extension (ΔL)
- The extra electrical length added by the field spreading past each radiating edge.
- Inset feed
- A microstrip feed recessed into the patch to reach a point of lower input resistance.
- Design Dk
- The permittivity a laminate maker recommends for circuit design, measured in a working transmission line.
Frequently asked questions
How do you calculate the length and width of a patch antenna?
Set the width with W = (c ÷ 2f) √(2 ÷ (εr + 1)), find the effective permittivity εeff from W and the substrate, compute the fringing extension ΔL, then set the length with L = c ÷ (2f √εeff) − 2ΔL.
How do you design a patch antenna for a given frequency?
Choose the substrate, then apply the patch antenna formulas in order: width, effective permittivity, fringing extension, length. For a 50 Ω feed add the inset depth, and treat the result as the starting point for an EM simulation or a bench trim.
Can I use this as a PCB antenna calculator?
Yes, for a rectangular patch on your PCB substrate over a ground plane larger than the patch, which is what the model assumes. For dipoles and quarter-wave whips use the antenna length calculator.
What size is a 2.45 GHz patch antenna on FR-4?
With εr 4.4 and 1.6 mm thickness, about 37.2 mm wide and 28.8 mm long. FR-4 permittivity varies between laminates and with frequency, so expect to trim the length after a first measurement.
Why is the patch shorter than half a wavelength?
The electric field fringes past the two radiating edges, which makes the patch electrically longer than it is physically. The model adds ΔL at each edge, 0.739 mm in the FR-4 example, and takes it off the physical length.
How do I match a patch antenna to 50 Ω?
Feed it from a point inside the edge. Input resistance falls as cos²(π y0 ÷ L) with inset depth y0, so y0 = (L ÷ π) arccos(√(50 ÷ Redge)). The FR-4 example has an estimated edge resistance of 321 Ω and an inset of 10.7 mm.
How accurate is the edge resistance?
It is a model estimate, not a measurement. The calculator uses Rin = 1 ÷ (2 (G1 + G12)), which includes the mutual conductance of the two radiating slots. On the FR-4 example G12 lowers the result from 516 Ω to 321 Ω, about 38 %. The formula counts radiation only, so copper and dielectric loss and the feed still move the real value: tune the inset in an EM simulator or on the bench.
How thick should the substrate be?
Patch substrates usually run from 0.003 to 0.05 free-space wavelengths. Thicker substrates give more bandwidth and efficiency but invite surface waves. The model needs h below a tenth of a wavelength.
What bandwidth does a patch antenna have?
Simple rectangular patches are narrowband, at most a couple of percent. A thicker substrate or a lower εr widens that, and it is why the patch length has to be close to right and the laminate permittivity well known.
More free PCB tools
See all free electronics tools →
- dBm to watts converterConvert dBm, dBW, watts and milliwatts, plus the RMS and peak-to-peak voltage in 50 Ω or any impedance.
- Fresnel zone calculatorRadius of the first Fresnel zone at any point on a radio link, the 60% clearance height, and earth bulge over distance.
- VSWR calculatorConvert between VSWR, return loss, reflection coefficient and mismatch loss, or compute them from a load impedance.
- LoRa airtime calculatorTime on air of a LoRa packet from spreading factor, bandwidth, coding rate and payload, and how often you may send under a duty-cycle limit.
- Free-space path loss calculatorFree-space path loss between two antennas from distance and frequency, in dB, with the distance for a given loss.
- Cascaded noise figure calculatorTotal noise figure and gain of up to six cascaded stages with the Friis formula, noise temperature and receiver sensitivity.
Sources
- N. Nikolova, McMaster University, Lecture 21: Microstrip Antennas Part II (design procedure, eq. 21.1 to 21.14)
- N. Nikolova, McMaster University, Lecture 20: Microstrip Antennas Part I (εeff and microstrip Z0, eq. 6 and 10)
- EMPossible, EE-4382/5306 Antenna Engineering, Topic 5: Microstrip Patch Antennas (slide 29 sign rule, slide 30 G12 integral, slide 36 published G1 and G12)
- FOSSEE textbook companion to Balanis, Antenna Theory, chapter 14: Example 14.2 worked in Python (a third party’s worked version of the book’s example, not the book itself)
- Rogers RO4000 series laminates data sheet (RO4003C and RO4350B design Dk)
- Rogers RT/duroid 5870 and 5880 data sheet
Design the whole board, not one number
ProtoFlow is free AI PCB design software: describe a circuit and get a schematic with real parts, PCB layout with an autorouter, SPICE simulation and Gerber files. Native KiCad, Altium and EAGLE import and export. Free on macOS, Windows and Linux.