RF circuits
Double biquad and other DIY Wi-Fi antennas, sized from the wavelength
By Bulan Sarkar, founder of ElectronicsInfoline (since 2000) · About a 10-minute read
A home-made Wi-Fi antenna is a sheet of metal and some wire, cut to the wavelength. Get the lengths right and it works. Get them wrong by a few millimetres and it is decoration.
Below are the biquad, the double biquad, the tin-can feed and the helix, each sized for the 2.4 GHz band from its formula. We also give the most gain each one can honestly deliver and what the link and the rules in India allow.
For 2.4 GHz Wi-Fi, design at 2442 MHz (channel 7), where a wavelength is 122.8 mm. A biquad is two squares of wire with 30.7 mm sides, 15.3 mm in front of a 123 mm square reflector. A double biquad is four such squares on a 246 mm × 123 mm reflector.
Their reflector areas cap the gain at about 11 dBi and 14 dBi. A pair of 10 dBi antennas on a 100 mW router reaches about 2.5 km in clear line of sight, inside India's 36 dBm EIRP limit (Use of low power Equipment in the frequency band 2.4 GHz to 2.4835 GHz (Exemption from Licensing Requirement) Rules).
Start from the wavelength
Every dimension on this page is a fraction of one number. Wi-Fi channel n in the 2.4 GHz band is centred on 2407 + 5n MHz, so channel 1 is 2412 MHz and channel 13 is 2472 MHz. The middle, channel 7 at 2442 MHz, is the frequency to design for. An antenna cut for it works across the whole band.
Wavelength is the speed of light divided by frequency: 299,792,458 m/s ÷ 2442 MHz = 122.8 mm. At the band edges it is 124.3 mm and 121.3 mm, a spread of under 2.5 %. That spread is your tolerance budget. Cut a part 1 mm long on a 30 mm side and you have moved the antenna by about 3 %, more than the whole band.
Our take: measure with vernier callipers, not a ruler, and measure to the centre of the wire. On a 30 mm side, the wire's own diameter is already a 5 % error if you measure to its edge.
The biquad
A biquad is two square loops sharing a feed point, standing on their corners. Each side is a quarter wavelength, 30.7 mm, so the whole element is 246 mm of wire bent in eight places. Copper wire of about 1.5 mm diameter holds its shape and solders well.
The coax comes through the middle of the reflector. The two ends of the wire meet at the centre with a small gap between them: solder the braid to the reflector and to one side of that gap, and the centre conductor to the other. Mount the element 15.3 mm in front of the reflector. That is λ/8, our choice from the 0.1 to 0.15 λ range builders use; small changes here trade gain against how well it matches 50 Ω.
How much gain? There is a quick ceiling. An aperture of area A can give at most 4πA/λ² of gain. For a λ × λ reflector that is 4π, or 11.0 dBi. A real biquad loses some of that to spill-over at the edges and to its own losses, so treat 11 dBi as the ceiling, not the result.
Our take: the biquad is the best first antenna to build. It is forgiving and needs no tuning tools. A claim of 14 dBi from a single biquad on a CD-sized plate breaks the aperture limit.
The double biquad
Put two biquads side by side, feed them in phase from one point, and double the reflector: 246 mm × 123 mm. The element is four squares, sixteen sides of 30.7 mm, about 491 mm of wire. Feed it in the centre exactly as before.
Doubling the area adds 3 dB to the ceiling: 14.0 dBi. It also halves the beam in the long direction, so mount the long side horizontal for a narrow horizontal beam and a wide vertical one, which is what a point-to-point link wants. Mount both ends of the link the same way up, or their polarisations will not match.
Our take: build a single biquad first and get it working. The double version is the same skill twice, plus one more joint at the feed that has to be clean.
The can feed (cantenna)
A tin can with a probe in it is a round waveguide. It only passes a signal above its cutoff frequency, which depends on the inside diameter: f = 1.841 c / (πD) for the first mode (D. M. Pozar). For channel 1 to get through, the can must be wider than 72.8 mm. To keep the next mode out at channel 13, it must be narrower than 92.8 mm. Anything between works, and the middle is easiest.
| Inside diameter | Cutoff | Probe from back | Can length, at least | Gain estimate |
|---|---|---|---|---|
| 70 mm | 2.51 GHz | does not work | – | – |
| 76 mm | 2.31 GHz | 95.3 mm | 286 mm | 5.8 dBi |
| 83 mm | 2.12 GHz | 61.6 mm | 185 mm | 6.5 dBi |
| 90 mm | 1.95 GHz | 51.1 mm | 153 mm | 7.2 dBi |
Inside the can the wave is stretched. Its guide wavelength is λ / √(1 − (fc/f)²), so the probe, a quarter of that from the closed end, moves a long way as the can narrows: from 51.1 mm in a 90 mm can to 95.3 mm in a 76 mm one. Close to cutoff the can turns fussy, because a millimetre of diameter moves the probe by several. The probe itself is always a free-space quarter wave, 30.7 mm, soldered to an N or SMA panel socket.
The gain column is the same aperture estimate as for the biquad, using the area of the open end. It is an estimate, not a measurement, but it shows the scale: a single can sits in single digits. Treat any double-digit gain figure for one with suspicion.
Our take: a can feed is a fine feed for a dish and a poor antenna on its own. Measure the inside diameter first; if your can is under 73 mm, no probe position will save it.
The helix
An axial-mode helix is a coil whose circumference is about one wavelength, wound over a ground plane. Kraus's rules: circumference between ¾ λ and 4/3 λ, pitch angle 12° to 14° (J. D. Kraus and R. J. Marhefka). For 2442 MHz with C = λ that gives a 39.1 mm diameter and, at 12.5°, a turn spacing of 27.2 mm. Twelve turns make it 327 mm long, on a ground plane at least 92 mm across.
Kraus's estimate for that helix is 16.0 dBi with a 32° beam. Measurements of long helices came in lower than this simple formula (H. E. King and J. L. Wong), so plan on less. The helix also radiates circular polarisation. Pointed at an ordinary linear antenna, it loses 3 dB at once.
Its input is about 140 Ω, not 50. A quarter-wave section of 84 Ω line (√(50 × 140)) between the coax and the first turn fixes that; a thin copper strip along the first quarter turn is a common way to make it.
Our take: build a helix when you need circular polarisation or want to learn matching. For a plain Wi-Fi link, a double biquad gives similar gain with less work.
Which one to build
| Antenna | Gain, at most | Size | Build |
|---|---|---|---|
| Biquad | 11.0 dBi (aperture) | 123 mm square | Easiest; wire, plate, one socket |
| Double biquad | 14.0 dBi (aperture) | 246 mm × 123 mm | Easy; one more joint at the feed |
| Can feed (83 mm) | about 6.5 dBi (aperture) | 185 mm+ long | Easy if the can is the right width |
| Helix, 12 turns | 16.0 dBi (Kraus, optimistic) | 327 mm long | Hardest; needs a match |
Our take: the double biquad, for most links. It has the best gain for the effort, and it is flat enough to mount on a wall bracket.
The link budget and the law
A worked link: a router putting out 20 dBm (100 mW), a 10 dBi antenna at each end, 2 m of LMR-400 at each end and 0.5 dB for the connectors. LMR-400 loses 22.2 dB per 100 m at 2.5 GHz (Times Microwave Systems), so each end loses 0.94 dB. The transmit end radiates 29.1 dBm EIRP.
Free space costs 100.2 dB over 1 km at 2442 MHz (Recommendation ITU-R P.525-4). The receiver gets -62.0 dBm. Ask for -70 dBm, which leaves 10 dB of fade margin above a −80 dBm receiver, and the link reaches about 2.5 km. Every doubling of distance costs 6 dB.
That is free-space distance. Real links fail on the Fresnel zone, the fat ellipse the signal needs clear around the straight line. At the middle of a 2.5 km path its radius is 8.8 m, and you want at least 60 % of it, 5.3 m, clear of roofs and trees (Recommendation ITU-R P.530). Line of sight is not enough.
Cable matters more than it looks. Every 13.5 m of LMR-400 halves the power, and thinner cable is worse. Put the radio at the antenna if you can and run Ethernet down instead.
In India, 2.4 to 2.4835 GHz needs no licence if the transmitter puts out at most 1 W (30 dBm), the EIRP is at most 4 W (36 dBm), and the antenna is within 5 m of the roof of an existing authorised building (Use of low power Equipment in the frequency band 2.4 GHz to 2.4835 GHz (Exemption from Licensing Requirement) Rules). Our worked link radiates 29.1 dBm, well inside. With these antennas the cap arrives at about 27 dBm of transmitter power. The height rule bites first: the 5.3 m of clearance a 2.5 km link wants is more than a 5 m mast can give over a crowded skyline. Other countries set their own limits, so check your regulator before you point anything at a neighbour.
Our take: gain on both ends beats power on one. Doubling the router's power adds 3 dB to one direction of the link; a better antenna adds its gain to both directions.
Lessons behind the numbers
- Frequency and periodWhere λ = c / f comes from.
- The decibelWhy every link figure here is added, not multiplied.
- Impedance matchingThe 140 Ω helix and the quarter-wave fix.
- ImpedanceWhat the 50 Ω on the coax means.
- ResonanceWhy a quarter-wave element is the right length.
- Cutoff frequencyThe same idea that makes a can a high-pass filter.
- Skin effectWhy cable loss climbs with frequency.
- FM transmitter circuitThe same range model at 100 MHz.
Sources and assumptions
- Use of low power Equipment in the frequency band 2.4 GHz to 2.4835 GHz (Exemption from Licensing Requirement) Rules, 2005, G.S.R. 45(E), 28 January 2005 (India: no licence needed in 2.4–2.4835 GHz with at most 1 W (30 dBm) transmitter output, 4 W (36 dBm) EIRP, and the antenna within 5 m above the roof of an existing authorised building; higher needs SACFA clearance)
- Recommendation ITU-R P.525-4, Calculation of free-space attenuation (2019) (free-space basic transmission loss L = 32.4 + 20 log f(MHz) + 20 log d(km) dB)
- Recommendation ITU-R P.530, Propagation data and prediction methods for terrestrial line-of-sight systems (first Fresnel zone radius F1 = 17.3 √(d1·d2 / (f·d)) m, f in GHz, distances in km)
- Times Microwave Systems, LMR-400 flexible low loss communications coax, datasheet (typical attenuation 22.2 dB/100 m (6.8 dB/100 ft) at 2500 MHz)
- J. D. Kraus and R. J. Marhefka, Antennas for All Applications, 3rd ed., McGraw-Hill 2002, chapter 8 (helical antennas) (axial mode for ¾ < C/λ < 4/3 and 12°–14° pitch; D ≈ 15 (C/λ)² N S/λ, HPBW ≈ 52° / ((C/λ)√(N S/λ)), R ≈ 140 (C/λ) Ω)
- H. E. King and J. L. Wong, Characteristics of 1 to 8 wavelength uniform helical antennas, IEEE Trans. Antennas and Propagation 28(2), March 1980, pp. 291–296 (measured gains of long helices below the simple Kraus estimate)
- D. M. Pozar, Microwave Engineering, 4th ed., Wiley 2012, §3.4 (circular waveguide) (TE11 cutoff f = 1.841 c / (π D), TM01 cutoff f = 2.405 c / (π D); guide wavelength λg = λ / √(1 − (fc/f)²))
- BIPM, The International System of Units (SI Brochure), 9th ed., 2019 (speed of light in vacuum c = 299 792 458 m/s, exact)
Read on 2 October 2026. Our own choices, not taken from any source: channel 7 as the design frequency, λ/8 element spacing, the λ × λ reflector, the 12-turn helix at 12.5°, the ¾ λg minimum can length, the 20 dBm router, 10 dBi per antenna, 2 m of cable and 0.5 dB of connectors per end, and the −80 dBm receiver with 10 dB of margin. The gain figures are upper limits from aperture area (4πA/λ²), not measurements.