Four-Square vs Half-Square: Compare the Complete Array
Four-Square vs Half-Square: Compare the Complete Array
A four-square can steer a strong null; a half-square can deliver a useful fixed broadside pattern with much less hardware. The defensible choice comes from element currents, loss, ground, geometry and evidence—not an unsupported gain ranking.
The honest answer is conditional. A four-square offers four switchable headings and can produce substantial rejection behind the selected beam, but only when the installed element currents and feed network are correct. A half-square is simpler and normally bidirectional broadside; it avoids a ground-plane return system, but its pattern, loss and feed impedance still depend on geometry, surroundings and soil.
Comparison rule: never rank antennas from isolated dBi numbers. Use the same frequency, ground model, installation, polarization, accepted power and gain reference. State whether feed-line, matching, switching, conductor, ground and mismatch losses are included, and label every result as theoretical, modeled or measured.
Start with the Current Distribution
Antenna names describe geometry, not guaranteed performance. The far field is set by the magnitude and phase of current along every conductor, combined with the conductor positions and their environment.
| Array | Nominal geometry | Pattern control | Current return and practical dependency |
|---|---|---|---|
| Half-square | Two roughly quarter-wave vertical legs joined by a roughly half-wave top wire | Usually a fixed, bidirectional broadside pattern | The wire structure supplies the RF return path; soil and nearby objects still change loss and pattern |
| Bobtail curtain | Three vertical legs coupled by two top-wire sections; total span is roughly one wavelength | Usually fixed and bidirectional broadside | Self-contained wire current path, but feedline, height, soil and surroundings still matter |
| Four-square | Four vertical elements at the corners of a square, often near quarter-wave side spacing | Relative element currents select a beam heading and rear rejection | Ground-mounted versions depend strongly on each element's radial/ground system and on the phasing network |
In an idealized half-square, the two vertical sections carry currents that reinforce broadside and tend to cancel in the plane of the array. The top wire is part of the antenna's distributed current system; calling it a non-radiator is only an approximation. Bending the wire, changing leg length, tilting a support or coupling to the feedline changes the current distribution and therefore the pattern.
A bobtail curtain can often be interpreted as three coupled vertical radiators, and some idealized treatments produce a center-to-outer current relationship near 2:1. That is not a construction guarantee. The current magnitudes and phases follow from the complete wire, feed arrangement, height and environment. “1:2:1” should be a measured or modeled result for a stated design, not a universal property.
A conventional corner-firing four-square is often described with equal current magnitudes and nominal phase steps such as 0°, −90°, −90° and −180°. That shorthand does not tell a power divider what voltages to deliver. Mutual coupling changes each driven impedance, so equal source voltage, equal feeder power or equal cable length does not guarantee the required element currents.
The engineering quantity to verify is the complex current at each element base. Amplitude and phase must be checked with all four elements connected and in each selected direction. A low common-port SWR proves only that the network presents an acceptable input impedance; it does not prove the intended pattern.
Gain, Directivity and Realized Gain Are Not Interchangeable
Directivity describes how concentrated a radiation pattern is. Gain also includes radiation efficiency. Realized gain additionally includes mismatch at the stated input reference. A phasing box can be outside the antenna model, inside it, or partly represented by ideal sources. Those choices move loss across the reference plane and can change the published number by decibels without changing the metal in the air.
The same caution applies to dBi and dBd. dBi uses an isotropic reference; dBd uses a half-wave dipole reference. This article uses dBi only where the source does, and it does not convert an unstated figure by assumption.
A useful 21 MHz case study—properly labeled
Garth Swanson, G3NPC, documented a practical 21 MHz four-square in the ARRL paper A 21 MHz Four Square Beam Antenna. It used four 10.95 ft monopoles, an 11.64 ft square and eight ground-level quarter-wave radials per element. Its result is valuable precisely because the paper separates calculation from measurement:
- Modeled, not measured: forward gain on the practical hybrid ground was estimated at 7–8 dBi; calculated headings using the measured currents ranged from 7.00 to 7.36 dBi.
- Measured: front-to-back ratio exceeded 20 dB in field-strength measurements around the array.
- Modeled: the main-lobe elevation was close to 20°, with about 35° vertical and 90° horizontal half-power beamwidth.
- Measured network loss: about 2 dB from the common feed point to the element inputs, including about 1.2 dB in its transformers. A 2 dB loss passes only 10−2/10 = 0.631, or 63.1%, of the input power.
- Design-specific bandwidth: modeling predicted only small pattern changes across a 0.45 MHz operating span at 21 MHz—about 2.1% fractional bandwidth.
Those numbers describe that antenna, ground treatment, current set, network and reference plane. They do not establish that every four-square delivers 7–8 dBi, more than 20 dB front-to-back, a 20° elevation peak or a 0.45 MHz bandwidth.
Phasing changes the answer even before construction loss
Al Christman, K3LC, modeled quarter-wave four-square arrays in Half Wavelength versus Quarter Wavelength Vertical Antennas. One 3.65 MHz case used 60 buried quarter-wave radials per element over specified “average” soil. Traditional equal-amplitude phasing modeled 5.89 dBi maximum gain, a 23.7° takeoff angle and 17.99 dB elevation-plane front-to-back. A maximum-gain current set modeled 6.36 dBi, 22.0° and 33.96 dB respectively.
The 0.47 dB gain change was modest; the rear rejection changed by almost 16 dB. The higher-phase-shift cases also narrowed the azimuth beam. This is why “the four-square pattern” is not one immutable pattern and why a deep null is more sensitive than broad forward gain to current error.
Do not combine the two case studies into a performance range. One is a practical 21 MHz build with eight radials per element, measured feed loss and measured current error; the other is a 3.65 MHz numerical study with 60 radials per element and prescribed ideal source currents. Different assumptions produced different answers.
Why the Old Half-Square and Bobtail Gain Numbers Were Removed
A statement such as “a half-square has 3–5 dBi” or “a bobtail has about 5 dBi” is incomplete without the model. Height, ground conductivity and permittivity, wire loss, feedline current, geometry, frequency and whether the comparison uses directivity, gain or realized gain all matter. Even the elevation angle at which “maximum gain” is read can shift with ground.
The four-square values above are not valid baselines for subtracting a few decibels and assigning the remainder to a half-square. A defensible comparison requires all candidates in the same model and then an installed measurement at a defined reference plane.
For the same reason, a bobtail curtain is not automatically “closer to a four-square” because it has three vertical legs. A diagonal projection of a four-square can group two elements at the same spatial coordinate, which is a useful array-factor teaching device. It does not turn the two physical antennas into the same current distribution, ground system, loss model or steerable structure.
Front-to-Back Is the Wrong Score for a Bidirectional Wire Array
A symmetrical half-square normally has two broadside main lobes. If one is called “front,” the opposite lobe is not unwanted back radiation—it is the other intended lobe. Front-to-back is therefore near 0 dB by that naming convention and says little about quality. Report broadside gain, front-to-side ratio, azimuth half-power beamwidth and the depth and bearing of the end-fire null instead.
The same applies to a conventional bidirectional bobtail curtain. A four-square, by contrast, is designed to favor one selected direction and reject the opposite sector, so front-to-back is useful. Null depth should still be reported with its angular width: a very deep but narrow modeled null can disappear when current error, terrain or the arrival direction changes slightly.
Feedpoint Voltage: Useful Arithmetic, Not an Antenna Rating
The open ends of a half-square or bobtail can be high-voltage regions. An ideal resistive calculation shows the scale. If a feedpoint were exactly 4000 Ω resistive, then:
At 100 W: VRMS = √(100 × 4000) = 632 V; Vpeak = 894 V
At 1 kW: VRMS = √(1000 × 4000) = 2000 V; Vpeak = 2828 V
At an exactly 50 Ω resistive point, the corresponding values are 70.7 V RMS at 100 W and 223.6 V RMS at 1 kW. The arithmetic is correct, but an antenna feedpoint may be reactive, voltage varies along the wire and matching networks can create higher local voltage or current. Use a field solver or validated measurement to rate insulation, capacitors, connectors and matching components; do not use the simple resistance example as a worst-case guarantee.
A current-rich feedpoint can reduce the impedance transformation and sensitivity to stray capacitance, but it is not inherently more efficient. Efficiency follows from conductor, dielectric, ground, common-mode and network losses. High-current locations also demand low-resistance connections and components with adequate current and heating ratings.
“No Radials” Does Not Mean “No Ground Effect”
A half-square or bobtail can provide its intended RF return through the wire structure rather than a ground-plane radial field. That is a major construction difference from a ground-mounted quarter-wave monopole. It does not isolate the antenna from Earth: reflected fields, soil loss, vertical-leg height, slope and nearby conductors still change the elevation pattern and efficiency.
A four-square can use buried or elevated radial systems, but the design must specify which. Radial count, length, bonding, overlap, soil and element-base loss are part of the array. Swanson's 21 MHz model estimated about 65% radiation efficiency for each element with eight radials; Christman's numerical example used 60 radials per element. “Four verticals” is not enough information to compare ground loss.
Control the Feedline, Then Verify It
A coax feedline can carry common-mode current on its outside surface and become an unintended radiator. That can alter impedance, fill a null, distort the azimuth pattern and make the result depend on cable routing. A current choke at or near the feedpoint is often appropriate, but “use mix 43” or “drop the coax at 45°” is not a complete specification.
Choke impedance varies with frequency, winding geometry, cable, core material, temperature and power. Parasitic capacitance creates resonances, so more turns are not automatically better. The ARRL article A Common-Mode Choke for Coax demonstrates measured frequency-dependent choke behavior. Select and test a design for the band and power, route the line to minimize coupling where practical, and measure outside-shield current rather than inferring success from SWR.
Match Bandwidth and Pattern Bandwidth Are Different
An SWR curve reports input match at one reference plane. It does not show whether a four-square still has the intended element-current ratios, beam bearing or rear null. Frequency changes the electrical lengths of the radiators and feeders, transformer phase, relay and connector parasitics, and the coupled element impedances.
Half-squares and bobtails also have separate match and pattern behavior. Wire diameter, exact proportions, feed position, height, common-mode current and surroundings determine the result. Avoid declaring one topology “broadband” or “narrowband” without defining a criterion such as SWR below a limit, gain within a tolerance, front-to-back above a limit or current phase within a tolerance.
Dimensions Are Starting Points, Not Cut-and-Install Answers
The following free-space dimensions use the exact SI speed of light, c = 299,792,458 m/s, at representative frequencies. They show scale only. Element resonance, end effect, conductor diameter, insulation, height, mutual coupling and the phasing-network design require trimming or modeling. Choose a design frequency that is legal in your jurisdiction and appropriate for the intended operating segment.
| Band | Example frequency | λ/4 scale | λ/2 scale | λ scale |
|---|---|---|---|---|
| 30 m | 10.125 MHz | 7.40 m | 14.80 m | 29.61 m |
| 20 m | 14.150 MHz | 5.30 m | 10.59 m | 21.19 m |
| 17 m | 18.100 MHz | 4.14 m | 8.28 m | 16.56 m |
| 15 m | 21.200 MHz | 3.54 m | 7.07 m | 14.14 m |
| 12 m | 24.950 MHz | 3.00 m | 6.01 m | 12.02 m |
| 10 m | 28.400 MHz | 2.64 m | 5.28 m | 10.56 m |
NIST documents the exact speed-of-light constant used in this calculation. For a half-square, λ/4 legs and a λ/2 top section are coupled starting dimensions, not three independent resonant wires. For a bobtail, a λ total span is likewise only a topology scale. For a four-square, λ/4 side spacing is common, but other spacings and current sets are possible; the feed network must be designed for the chosen geometry.
A Fair Model-and-Measurement Protocol
- Define the mission. State the desired azimuths, arrival-angle range, transmit duty cycle, receive-noise problem, available footprint and maintenance tolerance.
- Use one geometry reference. Record coordinates, heights, conductor size, supports, feedline route, nearby conductors and terrain for every candidate.
- Use one ground definition. Record conductivity and relative permittivity or measurement method; identify buried/elevated radials and all loss conductors.
- Model the entire current system. Include top wires, radials, realistic wire loss and, when practical, the feedline/common-mode path. LLNL's official NEC v5 description confirms that NEC can represent wires, ground, transmission lines and networks; the model still needs convergence and geometry checks.
- Separate ideal and realized cases. First prescribe element currents to understand array-factor potential. Then include coupled impedances, matching, phasing, switching and measured insertion loss at a common input reference plane.
- Sweep frequency and error. Vary amplitude, phase, component loss, soil and dimensions. Report forward gain, beam bearing, front-to-back or front-to-side, beamwidth, elevation pattern and input match.
- Commission the installed array. Measure each element's complex current with all elements connected, check every switched direction, measure network loss and common-mode current, then map the pattern with a stable far-field source or suitable reciprocal method.
Keep the evidence labels attached: an NEC plot is modeled evidence; a network-analyzer trace is a network measurement; an element-current probe is an installed-current measurement; an on-air report is operational evidence. None should silently inherit the certainty of another.
Choose by Constraint, Not by Topology Loyalty
| If the station priority is… | Half-square or bobtail may fit when… | Four-square may fit when… |
|---|---|---|
| Directional coverage | Two fixed broadside headings are useful | Four switchable headings and rear-sector rejection justify the network |
| Mechanical simplicity | Suitable high supports and long horizontal span are available | A square footprint, four radiators and their radial systems are practical |
| Predictable maintenance | A passive wire structure and accessible feed arrangement are preferred | Feeders, transformers, relays and current calibration can be tested and maintained |
| Receive interference rejection | The wanted and unwanted directions already suit a broadside/null pattern | Interference arrives from a direction that a verified switched null can reject |
| Efficiency | Wire, support, matching and common-mode losses are low | Ground and feed-network losses are measured and small enough to preserve the array advantage |
Passive linear antennas are reciprocal, so the verified transmit pattern also describes receive directionality. A four-square improves receive signal-to-noise ratio only when unwanted noise or interference has a spatial distribution the array rejects. It cannot remove receiver-generated noise, and a lossy network attenuates wanted signal as well as external noise.
The Defensible Verdict
A four-square is the more capable steerable array, but capability is conditional on the installed currents, radial system and feed-network loss. A half-square is not a lower-grade four-square; it is a simpler fixed bidirectional wire array with different mechanical and ground-return requirements. A bobtail adds aperture and another coupled vertical section, but it still needs its own model and current verification.
For a real station, compare the complete systems at the same reference plane. If the four-square's measured directional benefit survives its ground and network loss, and its four headings solve an actual operating problem, its added complexity can be justified. If two broadside headings are enough and the site favors high wire supports over radials and switching hardware, the half-square may be the better engineered choice.
Primary engineering references
- ARRL: A 21 MHz Four Square Beam Antenna—measured currents, feed loss and front-to-back, plus clearly labeled modeled gain and bandwidth.
- ARRL/QEX: Half Wavelength versus Quarter Wavelength Vertical Antennas—controlled NEC studies of vertical-array phasing, ground and radial systems.
- Lawrence Livermore National Laboratory: NEC v5—official solver scope and modeling capabilities.
- NIST: Meet the Constants—the exact speed of light used for the free-space dimension table.
- ARRL: A Common-Mode Choke for Coax—measured, frequency-dependent choke behavior.
Mini-FAQ
- Does a four-square always have more gain than a half-square? No. The answer depends on geometry, ground, conductor and feed-network loss, match, element-current accuracy and the reference plane used for gain.
- What does front-to-back mean for a half-square? A symmetrical half-square has two intended broadside lobes, so front-to-back is usually not the useful metric; report broadside gain, front-to-side ratio, beamwidth and end-fire null depth.
- Is 1:2:1 current guaranteed in a bobtail curtain? No. It can be a useful idealized description, but the installed current ratios and phases depend on the complete wire, feed, height, surroundings and common-mode path.
- Can good SWR prove a four-square is phased correctly? No. Good SWR only describes the common-port match; verify the complex current at every element and the installed pattern in every switched direction.
- Does a half-square need radials? It does not normally need a monopole-style radial field for RF return, but soil, height, nearby conductors, feedline current and loss still affect its pattern and efficiency.
- Which array is better for receive? The one whose verified pattern rejects the site's dominant noise or interference while preserving wanted signal; topology alone does not guarantee a signal-to-noise improvement.