
Antenna Efficiency vs Gain: The One-Line Answer
Gain equals directivity multiplied by radiation efficiency: G = ηrad × D. A datasheet that quotes only gain hides how much of that number comes from focusing the beam (directivity) and how much power the antenna actually radiates versus what it wastes as heat. Two antennas can share the same 6 dBi gain figure yet differ by 2 dB in efficiency—meaning one converts 40 % less input power into radiated fields. Understanding the split is essential for link-budget accuracy, thermal management, and battery life.
Definitions Engineers Actually Need

Directivity (D)
Directivity describes how tightly an antenna focuses energy relative to an isotropic radiator, assuming zero loss. It is a purely geometric property of the radiation pattern. A half-wave dipole has a directivity of about 2.15 dBi; a 20-element patch array might reach 18–20 dBi. Directivity tells you nothing about whether the antenna is lossy or lossless.
Radiation Efficiency (ηrad)
Radiation efficiency is the ratio of power radiated into space to power accepted by the antenna’s input port. Losses that reduce ηrad include ohmic loss in copper traces, dielectric loss in the substrate (quantified by the material’s dissipation factor, Df), and surface-wave excitation that never leaves the board. Typical PCB patch antennas on FR-4 at 2.4 GHz achieve ηrad around 60–75 %; the same geometry on Rogers RO4350B (Df ≈ 0.0037 at 10 GHz) can reach 85–92 %.
Total Efficiency (ηtotal)
Total efficiency folds in mismatch loss: ηtotal = ηrad × (1 − |Γ|²), where Γ is the reflection coefficient at the feed. A return loss of 10 dB (|Γ| = 0.316) costs an additional 0.46 dB on top of whatever radiation efficiency already takes. Many datasheets quote “peak gain” measured at the frequency of best match, which flatters the number by minimizing mismatch loss at that single point.
Realized Gain
Realized gain is the metric that matters for a link budget: Grealized = ηtotal × D. It accounts for both radiation loss and mismatch loss. IEEE Std 145-2013 defines this term precisely, and it is the quantity most simulation tools (HFSS, CST) report by default. If a datasheet says “gain” without qualification, you cannot be sure whether mismatch loss is included.
The Relationship in Numbers
| Parameter | Symbol | Unit | What It Captures |
|---|---|---|---|
| Directivity | D | dBi | Pattern shape only (lossless) |
| Radiation efficiency | ηrad | % or ratio | Ohmic + dielectric + surface-wave loss |
| Mismatch efficiency | 1 − |Γ|² | % or ratio | Reflected power at feed |
| Gain | G | dBi | D × ηrad |
| Realized gain | Grealized | dBi | D × ηrad × (1 − |Γ|²) |
In decibels the multiplication becomes addition: G(dBi) = D(dBi) + 10 log₁₀(ηrad). An antenna with 8 dBi directivity and 70 % radiation efficiency has a gain of 8 + 10 log₁₀(0.70) = 8 − 1.55 ≈ 6.45 dBi. That missing 1.55 dB is heat. For a deeper treatment of how gain, directivity, and efficiency interrelate, see antenna gain versus directivity.
What Datasheets Commonly Hide (or Obscure)
1. Quoting Gain Without Stating Efficiency
The most frequent omission: a chip-antenna datasheet lists “peak gain 2.5 dBi” at 2.45 GHz but never mentions radiation efficiency. A monopole-type chip antenna on a small ground plane may have directivity around 4–5 dBi. If the quoted gain is 2.5 dBi, the implied efficiency is roughly 55–65 %. That is not bad for a chip antenna—but you need to know it to close a link budget accurately.
2. Measurement on a Reference Board, Not Yours
Antenna vendors measure on their own evaluation board with a specific ground-plane size, keep-out zone, and substrate. Your product’s ground plane, nearby plastics, and radome dielectric effects will shift both gain and efficiency. A 1–3 dB drop from datasheet to product is routine if the integration guidelines are not followed.
3. Gain Measured at a Single Frequency
Peak gain is typically quoted at the center of the band. At band edges, return loss degrades and realized gain can fall 1–2 dB. Always request or measure the gain-versus-frequency curve across your full operating bandwidth.
4. Confusing Gain and EIRP
Some module datasheets list EIRP (effective isotropic radiated power) and call it “gain.” EIRP is the product of transmit power and antenna gain; quoting it as a gain figure inflates the number by however many dBm the PA delivers. Read the fine print and check the units: dBi is gain, dBm is power.
5. Not Specifying Polarization
A circularly polarized antenna has roughly 3 dB less gain in any single linear polarization than its total (co-pol + cross-pol) gain. If the datasheet quotes total gain but your receiver is linearly polarized, you lose that 3 dB in practice.
Where the Loss Actually Goes on a PCB Antenna
Understanding the loss mechanisms lets you predict efficiency before building hardware and guides substrate material selection.
Conductor Loss
At microwave frequencies, current crowds into a skin depth of roughly 0.66 µm at 10 GHz for copper. Surface roughness on standard electrodeposited copper (Rz ≈ 5–8 µm) can increase effective resistance by 30–60 % compared to the smooth-surface model. Rolled-annealed or very-low-profile (VLP) copper foils reduce this penalty. For millimeter-wave designs above 60 GHz, conductor loss often dominates.
Dielectric Loss
Dielectric loss scales with frequency and the substrate’s Df. FR-4 (Df ≈ 0.02 at 1 GHz, rising with frequency) is acceptable for 900 MHz ISM-band antennas but increasingly costly above 2.4 GHz. PTFE-based laminates such as Rogers RT/duroid 5880 (Df ≈ 0.0009) keep dielectric loss below 0.1 dB for typical patch geometries at 10 GHz. The trade-off is cost and processability; hybrid stackups that place high-frequency laminate only on the antenna layer are a common compromise.
Surface-Wave Loss
Thick, high-Dk substrates excite surface waves that propagate laterally and never radiate usefully. This is a hidden efficiency killer in arrays: it lowers element efficiency and raises mutual coupling. The threshold scales with substrate electrical thickness (h/λ₀). Keeping h/λ₀ below about 0.09 for Dk ≈ 3.5 suppresses the dominant TM₀ surface wave. For a discussion of how substrate thickness affects bandwidth and loss, see the linked guide.
Feed-Network Loss
In an array, the corporate or series feed network can consume 1–3 dB before power even reaches the radiating elements. This loss reduces array gain without changing element directivity. Substrate-integrated-waveguide feed networks can cut feed loss at millimeter-wave frequencies by confining fields inside a low-loss dielectric channel rather than open microstrip.
How to Extract Efficiency from a Gain-Only Datasheet
- Simulate the ideal (lossless) structure. Model the antenna geometry in a full-wave solver with PEC conductors and lossless dielectric. The resulting gain equals directivity.
- Compare with the datasheet gain. The difference in dB is 10 log₁₀(ηrad). If lossless simulation gives 7.5 dBi and the datasheet says 6.0 dBi, ηrad ≈ 10−1.5/10 ≈ 71 %.
- Verify with a Wheeler-cap or reverberation-chamber measurement once you have hardware. These methods separate radiation efficiency from mismatch loss directly. Details on both approaches are covered in antenna efficiency measurement techniques.
For production antennas, tracking realized gain across frequency in an anechoic chamber or near-field scanner is the most reliable way to confirm that efficiency has not degraded due to manufacturing variation.
Practical Scenarios: When Efficiency Matters More Than Gain

Battery-Powered IoT Nodes
A sensor node transmitting at +10 dBm through an antenna with 50 % efficiency wastes 5 mW as heat and radiates 5 mW. Improving efficiency to 80 % radiates 8 mW—equivalent to raising transmit power by 2 dB—without touching the PA or the battery. Over millions of duty-cycled transmissions, this extends battery life measurably.
Phased Arrays with Tight Thermal Budgets
A 256-element active phased array at 28 GHz, each element fed with 10 dBm, dissipates 256 × 10 mW × (1 − ηrad) in the aperture. At 70 % efficiency that is 768 mW of waste heat spread across the array face; at 50 % efficiency it is 1.28 W. In a sealed radome with no forced airflow, that difference can push junction temperatures past reliability limits.
Automotive Radar at 77 GHz
At 77 GHz every fraction of a dB matters. Conductor loss on standard copper is severe; dielectric loss on anything other than low-Df laminates (e.g., Rogers RO3003, Df ≈ 0.0010) becomes the dominant efficiency limiter. A 1 dB efficiency improvement translates directly into 1 dB more SNR at the receiver, extending detection range by roughly 12 % (range scales as the fourth root of SNR in radar).
Improving Efficiency Without Changing Directivity
- Lower-loss substrate. Moving from FR-4 to a ceramic-filled hydrocarbon (e.g., RO4350B) at 5.8 GHz typically recovers 0.5–1.5 dB of radiation efficiency.
- Smooth copper. Specify VLP or hyper-VLP foil for the antenna layer. The improvement is most pronounced above 10 GHz.
- Thinner substrate (within bandwidth constraints). Reduces surface-wave excitation and dielectric volume loss, at the cost of narrower impedance bandwidth.
- Cavity-backed or SIW-backed elements. Placing a metallic cavity behind a patch suppresses back-radiation and surface waves, recovering 0.5–1 dB of efficiency in dense arrays.
- Minimize feed-line length. Every millimeter of microstrip between the connector and the radiating edge dissipates energy. Direct-probe or aperture-coupled feeds shorten the lossy path.
Improving Gain by Increasing Directivity
If you need more gain and efficiency is already high, the path is to increase directivity—focus the beam tighter. Approaches include adding array elements, using a reflectarray aperture to collimate the beam from a smaller feed, or applying superstrate layers. Each approach trades beamwidth, bandwidth, or physical size for higher directivity, and none of them fix a lossy substrate.
Quick Checklist: Reading an Antenna Datasheet
- Is the quoted figure “gain,” “realized gain,” or “directivity”? Check the IEEE definitions in Std 145-2013 if the datasheet is ambiguous.
- Is radiation efficiency stated? If not, estimate it from the gain-directivity gap or request it from the vendor.
- At what frequency and over what bandwidth is gain specified?
- What is the reference board size, ground-plane geometry, and substrate?
- Is mismatch loss included (realized gain) or excluded?
- Is the gain total gain or co-polarized gain only?
- Are any radome or housing effects included in the measurement?
Frequently Asked Questions
Can an antenna have high gain but low efficiency?
Yes. A large array with many elements can achieve high directivity (and therefore high gain in absolute terms) while each element is individually lossy. The gain may still be impressive, but the system wastes significant power as heat. This is common in low-cost FR-4 arrays above 5 GHz.
Does improving efficiency always improve gain?
Yes, if directivity stays constant. Gain = ηrad × D, so any increase in ηrad adds directly to gain in dB. However, some design changes that improve efficiency (such as thinning the substrate) may also alter the pattern and thus change directivity slightly.
What is a “good” radiation efficiency for a PCB antenna?
It depends on the application and frequency. Above 80 % (−1 dB) is generally considered good for discrete patch antennas on low-loss laminates. Chip antennas and electrically small antennas on FR-4 commonly fall in the 40–70 % range. At millimeter-wave frequencies, 60–70 % on a PCB is respectable.
How do I measure radiation efficiency separately from gain?
The Wheeler-cap method, the reverberation-chamber method, and the gain-directivity comparison method are the three most common approaches. Each has trade-offs in accuracy, cost, and frequency range. The Wheeler cap is simplest for single elements below about 6 GHz. See introductory antenna measurement resources for background on test setups.
Why do simulation tools sometimes show higher efficiency than measurements?
Simulators use idealized material models. Real copper roughness, solder-mask dielectric loading, connector transitions, and manufacturing tolerances (etching undercut, Dk variation) all introduce losses not captured in a default simulation. Calibrating the simulation’s conductor roughness model (e.g., Huray or Hammerstad in HFSS) and including solder-mask layers closes most of the gap.
Ready for a quotation?
Send your requirements for a material, stackup and manufacturing review. Price and schedule are confirmed for your project.
