Antenna component testing involves radiation pattern measurement (360° azimuth scan, 0.5° step, field strength logged), return loss checks (2.4-5GHz via VNA, target >10dB), gain verification (comparing to 10dBi standard horn in far-field, ratio within ±0.5dB), polarization testing (90° linear probe rotation, signal difference <3dB), and impedance matching (S11 <-10dB at center freq ±10% bandwidth using VNA).
Table of Contents
Checking Impedance and VSWR
Most modern RF systems use 50-ohm impedance as the standard—why? Because it balances power handling (higher than 30 ohms) and loss (lower than 75 ohms) in coaxial cables. If your antenna’s input impedance is 50 + j10 ohms (a common deviation), that “+j10” is reactance, which creates standing waves. VSWR quantifies this: it’s the ratio of the maximum to minimum voltage on the transmission line. A perfect match is 1:1, but in real life, 1.5:1 is often the upper limit for acceptable performance.
Take a 5G small cell antenna we tested last year. At its operating frequency of 3.5GHz, the spec required VSWR ≤1.3:1. Using a vector network analyzer (VNA) with a frequency range of 9kHz to 6GHz and a dynamic range of 120dB, we measured VSWR spikes up to 1.6:1 at 3.45GHz. Digging deeper, we found oxidation on the SMA connector—just 0.01mm of corrosion increased the impedance by 2.3 ohms (from 49.8 + j2.1 to 52.1 + j3.4). After cleaning, VSWR dropped to 1.2:1, cutting reflected power from 4.2% to 1.1%—a big win for signal efficiency.
To avoid these issues, start with proper VNA calibration. We use a 3-step process: short (0 ohms), open (infinite impedance), and load (50 ohms) at the measurement ports. Calibration errors here can throw off results by ±0.3dB—enough to miss a critical mismatch. During testing, sweep the frequency range slowly(100 points per octave) to catch narrowband resonances; fast sweeps might skip over a 100kHz dip in VSWR that causes intermittent failures.
| VSWR | Reflection Coefficient (Γ) | Reflected Power (%) | Power Delivered to Load (%) |
|---|---|---|---|
| 1.0 | 0 | 0 | 100 |
| 1.1 | 0.05 | 0.23 | 99.77 |
| 1.2 | 0.10 | 0.83 | 99.17 |
| 1.3 | 0.15 | 1.94 | 98.06 |
| 1.4 | 0.20 | 3.55 | 96.45 |
| 1.5 | 0.25 | 5.62 | 94.38 |
Notice how a jump from 1.3:1 to 1.5:1 doubles the reflected power—from ~2% to ~5%. For a 100W transmitter, that’s 3W bouncing back instead of radiating. In a cellular system, that wasted power could mean fewer users per cell or higher cooling costs for the base station.
Measuring Radiation Pattern Performance
For a 5G macro cell site, a 1dB drop in peak gain (the strongest direction of radiation) can shrink usable coverage by 15–20%, translating to 500+ fewer users per cell during peak hours. For satellite communication antennas, a poorly controlled side lobe (unintended radiation direction) might let interference drown out your signal, increasing bit error rates by 30%.
A standard chamber is at least 10m long—why? Because at 2.4GHz, a 10m distance ensures the antenna under test (AUT) is in the “far field,” where electromagnetic waves behave predictably (no near-field weirdness like standing waves messing up readings). The chamber walls are lined with RF-absorbing material—usually pyramidal foam or carbon-loaded vinyl. At 10GHz, good absorbers have a reflectivity of ≤-50dB (meaning they bounce back less than 0.001% of the signal). If the absorber is old or damaged (say, a 10cm tear in a foam panel), that reflection jumps to -35dB, adding a 2–5dB “ghost lobe” to your pattern—enough to make your antenna look like it’s radiating toward the ceiling when it should be focused on the ground.
A vector network analyzer (VNA) or a signal generator + spectrum analyzer setup. For 5G mmWave antennas (24–40GHz), we use a VNA with a frequency range of 100MHz to 50GHz and a dynamic range of ≥100dB—critical for distinguishing the main lobe from weak side lobes. The AUT mounts on a precision turntable that rotates with an accuracy of ±0.1°—miss by 0.5°, and a 28GHz antenna’s 3dB beamwidth (the angle where signal drops by half) could be mismeasured by 10%, throwing off coverage predictions.
Calibration is non-negotiable. Before testing, we use a standard gain horn antenna (with a known gain, say 20dBi at 28GHz) to calibrate the chamber. The process involves measuring the horn’s pattern at 10+ positions across the turntable, then applying corrections to account for chamber imperfections (like slight absorber misalignment). Without this, your measured gain could be off by ±1.5dB—enough to pass a spec that should fail, or fail one that should pass.
The spec called for a -10dB beamwidth of 60° (meaning 90% of the signal energy stays within a 60° cone) and side lobes ≤-20dB (signal 1/100th as strong as the main lobe). Initial scans showed the main lobe at 55° (-10dB) and a side lobe at -17dB—close, but not quite. Digging into the data, we noticed the turntable’s rotation speed was too fast (5°/second instead of the recommended 1°/second). At higher speeds, the VNA’s sampling rate (set to 100 samples/second) missed fine details, averaging out small directional variations. Slowing to 1°/second and doubling the samples to 200/second fixed it: the -10dB beamwidth stabilized at 61°, and side lobes dropped to -21dB—meeting spec.
Antenna materials (like Rogers RO4003C PCB) expand when heated, shifting the phase center (where the signal “originates”) by 0.02mm/°C. Over a 25°C to 40°C temperature swing (common in outdoor testing), that’s a 0.3mm shift—enough to tilt the main lobe by 1–2°, which can push coverage away from a target area (like a busy street corner). We always pre-condition antennas in the chamber for 1 hour before testing, letting them stabilize at 25°C ±1°C.
Testing Gain and Efficiency
You can have a powerful engine (high gain), but if half the fuel is wasted (low efficiency), you’re not going far. In RF terms, a 5G base station antenna with a gain of 18 dBi sounds impressive, but if its radiation efficiency is only 60%, you’re losing 40% of your input power as heat—wasting thousands of dollars annually in electricity and potentially overheating the system. For a satellite uplink antenna operating at 30GHz, a 2dB gain drop (from 45dBi to 43dBi) could force a 60% increase in transmit power just to maintain the link, slashing battery life and boosting operational costs by $15,000/year per terminal. These numbers aren’t abstract; they’re why we test.
- Gain: How directionally focused the antenna is (comparing its peak radiation to a theoretical isotropic radiator). Measured in dBi.
- Efficiency: The percentage of input power that actually radiates (not lost as heat or reflections). A 90% efficient antenna radiating 100W only loses 10W.
Here’s how it works in practice: we place a standard gain horn (with a precisely known gain value, e.g., 15.2 dBi at 28GHz) in our anechoic chamber and measure its received power from a fixed source. Then, we replace it with our antenna under test (AUT) and measure its received power under identical conditions. The gain difference is calculated from the power difference. The accuracy of this method hinges entirely on the standard gain antenna. If its calibration is off by 0.5 dB (a common tolerance for older horns), your AUT’s gain will be wrong by the same 0.5 dB. For a mmWave antenna, that error could mean misjudging cell coverage radius by 10-15 meters.
We once tested a batch of Wi-Fi 6E dipole antennas where simulations predicted 5.2 dBi gain at 6GHz. Our initial measurements showed 5.8 dBi—a suspiciously high value. The issue? The chamber’s absorbers were aging. At 6GHz, their reflectivity had degraded from -50dB to -42dB. These weaker reflections constructively interfered with the main signal, artificially boosting the measured gain by ~0.6 dB. Replacing the absorber panels brought the measurement down to 5.1 dBi, aligning with simulation.
We use the Wheeler Cap method for electrically small antennas or reverberation chambers (mode-stirred) for larger arrays. The Wheeler Cap is a hollow, lossy metal enclosure that fits over the antenna. You measure the impedance (and thus the return loss) with and without the cap. The difference in these measurements reveals the power lost inside the antenna itself. A typical 2.4GHz PCB trace antenna might have 85% radiation efficiency measured in free space, but drop to 75% when mounted on a plastic housing because of new dielectric losses. Reverberation chambers are more accurate for complex antennas, using a mechanical paddle to stir EM fields and collect hundreds of samples, calculating efficiency from the statistical distribution of received power. This method can achieve an accuracy of ±0.4 dB (~10% absolute efficiency).
A 4×4 MIMO antenna array for a cellular router can see its element efficiency drop from 75% to 60% when a user’s hand is placed 3cm away from the radiated element—a 20% performance loss that directly translates to slower download speeds. We always test gain and efficiency across the entire operating temperature range (-30°C to +60°C for automotive apps). A antenna’s parasitic elements can expand with heat, detuning the structure. We observed a 0.3 dB decrease in peak gain for an outdoor antenna when its temperature increased from 25°C to 55°C.
Evaluating Under Different Conditions
An antenna’s datasheet specs are measured in a perfect lab at 25°C. Reality is never perfect. A 5G mmWave antenna might hit its target 28 dB gain in a temperature-controlled chamber, but when mounted on a sun-exposed rooftop in Phoenix, where surface temperatures hit 70°C, its plastic substrate expands, detuning the elements and slicing gain by 2.5 dB. That’s a 45% drop in effective radiated power. Or consider a vehicle-mounted antenna: vibration from driving can loosen a connector by just 0.3 mm, increasing VSWR from 1.4 to 2.1 and reflecting 17% more power back into the transmitter.
- Environmental Factors: Temperature, humidity, and physical stress.
- Operational Factors: Nearby objects (like a human hand) and frequency agility.
- Lifetime Factors: Material degradation over time.
A copper trace on an FR-4 PCB expands at about 17 ppm/°C. Over a 50°C temperature swing (from -10°C to 40°C), a 100mm long trace will expand by 0.085mm. This tiny change shifts the resonant frequency of a 3.5GHz antenna by approximately 15 MHz. If your 5G band is only 100MHz wide, you’ve just drifted 15% off center. We test this by placing the antenna in a thermal chamber, cycling it from -40°C to +85°C over 6 hours, and measuring S11 at 5°C intervals. It’s common to see the resonant frequency drift linearly by 0.3 MHz/°C.
For a consumer device, we perform repeated 5kg pressure tests on the plastic housing near the antenna. A flex of just 2mm can deform the antenna pattern, reducing efficiency by up to 12%. We also test under vibration. A automotive antenna must withstand 5-500 Hz random vibration at 0.02 g²/Hz for 4 hours per axis. We’ve seen SMA connectors fracture after 3 hours of this, increasing return loss from -15 dB to -7 dB.
A smartphone antenna’s efficiency can drop by 40% when held in a user’s left hand versus their right hand, due to the way the body absorbs RF energy. We test phones in multiple handgrip positions using realistic phantoms filled with simulated body fluid (≈ 2/3 water, 1/3 sugar). The same applies to installation. A GPS antenna specified for a car roof must be tested on a 1m x 1m ground plane, as its performance on a smaller ground plane (like a dashboard) will be severely degraded, with gain patterns distorting by up to 5 dB in lower elevations.
| Test Condition | Parameter Measured | Typical Performance Change | Failure Impact |
|---|---|---|---|
| Temperature (25°C to 75°C) | Resonant Frequency | +0.35 MHz/°C drift | Falls out of band |
| Hand Grip (Right Hand) | Total Radiated Power | -25% to -40% | Dropped calls |
| Vibration (4 hrs, 5-500Hz) | Connector Resistance | Increase from 0.2Ω to 2.1Ω | VSWR > 2.0:1 |
| Humidity (95% RH, 96 hrs) | PCB Loss Tangent | Increase by 15% | Efficiency -8% |
An antenna might have a VSWR < 2.0:1 at 2.45GHz, but that can creep up to 2.8:1 at the band edges (2.4GHz and 2.5GHz for Wi-Fi), especially under load. For a frequency-hopping system, this edge-of-band performance is critical. We test at a minimum of 20 frequency points per 100MHz of bandwidth to map these performance cliffs accurately.
Verifying Connectors and Cables
A 5 connector fails. A single poorly torqued SMA connector can introduce an insertion loss of 0.3 dB and increase VSWR from 1.2:1 to 1.8:1 at 6 GHz. For a 100W transmitter, that’s 3 watts of power lost as heat at the connection point. Over a year of continuous operation, that wasted energy adds over 260 kWh in electricity costs, enough to power a small lab setup for a month. In a phased array system, a 30-degree phase imbalance between cables due to length tolerances can distort the beam pattern, reducing effective gain by 4 dB and shrinking 5G cell coverage by 35%.
We perform a 2-port calibration (open, short, load, through) using high-precision calibration kits traceable to NIST standards. After calibration, the residual directivity of the VNA should be better than 40 dB, meaning it can measure return losses down to -40 dB with confidence. We then connect the cable assembly under test. The first critical metric is insertion loss per meter. For a premium low-loss coaxial cable like LMR-400, the spec at 3 GHz is approximately 0.22 dB/m. We routinely test batches where a slight kink or imperfect dielectric centering increases this loss to 0.28 dB/m. Over a 15-meter cable run, that extra 0.06 dB/m loss adds up to 0.9 dB of lost signal—enough to require a high-power amplifier to compensate, adding $500+ in unnecessary component costs per link.
We perform a 90-degree bend test on cables, measuring VSWR before and after. A good cable will show a VSWR change of less than 0.1 after 5,000 bend cycles at a 25mm radius. A poor-quality cable might see VSWR degrade from 1.3:1 to 1.9:1 after just 1,000 cycles. For the phase stability critical in array systems, we measure the phase shift vs. temperature. A standard RG-58 cable has a phase stability of approximately 80 ppm/°C. A 10°C temperature change in a 2-meter cable at 10 GHz can induce a 9.6-degree phase shift, enough to missteer a beam by 3 degrees. We test this by placing cables in a thermal chamber and measuring S21 phase from -40°C to +85°C at 10°C intervals.
An SMA connector requires 8 in-lbs of torque for proper coupling. Under-torquing to 5 in-lbs can increase contact resistance, raising insertion loss by 0.15 dB. Over-torquing to 12 in-lbs can crack the connector dielectric, creating an impedance discontinuity that spikes VSWR to 3.0:1 above 4 GHz. We use a calibrated torque wrench and measure VSWR at each torque setting. We also perform durability cycling, mating and unmating connectors to their specified limit (e.g., 500 cycles for SMA). After 500 cycles, a low-quality connector’s contact resistance can increase from 2 milliohms to 15 milliohms, and its VSWR can degrade permanently.
| Parameter | Test Method | Acceptable Range | Typical Failure Mode & Impact |
|---|---|---|---|
| Insertion Loss | VNA S21 measurement, 1m length | < 0.25 dB/m @ 6 GHz | Kinked dielectric: Loss increases to 0.35 dB/m, wasting 10% of power. |
| VSWR | VNA S11 measurement, full band | < 1.35:1 up to 18 GHz | Under-torqued connector: VSWR rises to 1.8:1, reflecting 8% of power. |
| Phase Stability | S21 phase measurement, -40°C to +85°C | < 100 ppm/°C | Poor dielectric: Phase shifts 150 ppm/°C, distorting beams in arrays. |
| Durability | Mate/demate for 500 cycles | VSWR change < 0.2 | Worn center pin: VSWR degrades from 1.3:1 to 2.1:1 after 300 cycles. |
We mount antennas on a vibration table and shake them at 5-500 Hz with a 2 Grms profile for 2 hours per axis while monitoring VSWR and insertion loss in real-time. A common failure we catch is a slightly loose connector that, under vibration, develops an intermittent disconnect, causing VSWR to spike to 4.0:1 for 200-millisecond durations, enough to cause dropped packets in a data link.
