FM transmitter circuit: one transistor, every value worked out

Fourteen parts plus the battery. Nothing is exotic: the coil is wound by hand, and every other part is a stock value. The transistor needs to work well at 100 MHz. The 2N3904's transition frequency is at least 300 MHz (onsemi 2N3904 datasheet), which leaves enough gain at the FM band. A BC547 will often work, but it has less in reserve.

The oscillator. The coil and the capacitors across it form a parallel tuned circuit on the collector. It has one frequency where it resonates, and that is where the circuit runs. The 4.7 pF capacitor from collector to emitter feeds a slice of the collector swing back to the emitter, in phase. That is the Colpitts arrangement; the transistor oscillators lesson builds it up from the loop-gain condition.

Why the base has 1 nF to ground. At 98 MHz, 1 nF is 1.6 Ω, so the base sits at RF ground and the transistor works as a common-base amplifier. At 1 kHz the same capacitor is 159 kΩ, so to audio the base is still free to move. That one capacitor lets RF and audio share the base.

The modulation. Sound moves the base voltage by a few millivolts. That shifts the transistor's operating point, and with it the junction capacitances that sit across the tank. A change in capacitance is a change in frequency, and a frequency that follows the audio is FM. It works on the same principle as a varactor, only without the varactor, and it is also why this circuit is so sensitive to everything else.

The antenna. It takes a little signal from the collector through 4.7 pF. A small capacitor keeps the antenna from loading the tank too much. Every time you touch the antenna, you also retune the oscillator a little.

Bias. The 22 kΩ and 10 kΩ divider puts 2.81 V on the base from 9 V, with a source resistance of 6.88 kΩ. Taking a current gain of 200 (our assumption; the part varies), the emitter current is (2.81 − 0.7) V ÷ (470 Ω + 34 Ω) = 4.2 mA. The circuit draws about 38 mW from the battery, so an alkaline PP3 lasts a few days of continuous use. The BJT biasing lesson explains why the divider-plus-emitter-resistor scheme holds the current steady while the gain varies.

The coil. Five turns of 0.7 mm (22 SWG) enamelled wire, wound on a 5 mm drill bit and stretched to 5 mm long, gives a mean diameter of 5.7 mm. Wheeler's formula for a single-layer air coil gives 106 nH. The formula is good to about 1% for a coil this shape. The hand-wound coil itself will not be that accurate, which is why the trimmer is there.

The tuning capacitance. For 98 MHz with 106 nH, C = 1 / ((2πf)² L) = 25.0 pF in total across the coil. We allow 5 pF for the transistor, the feedback network and the wiring (an estimate), and the fixed capacitor is 10 pF, so the trimmer sits near 10.0 pF. Over its 4–20 pF travel the circuit tunes from 82.8 MHz to 112.3 MHz, which covers the whole band with a little either side. The parallel RLC and resonance lessons have the formula's background.

Why the fixed capacitor. A 5–30 pF trimmer on its own would swing this coil from below the band to past 150 MHz, into the aircraft band at 118–137 MHz. The fixed part narrows the trimmer's range to the FM band.

The antenna length. At 98 MHz the wavelength is 3.06 m, so a quarter-wave whip would be 76 cm. Most published versions of this circuit tell you to fit one. Read the legal limit section before you do.

Our take: if a circuit diagram gives no coil dimensions, it has not given you a frequency. The coil sets it, and on this circuit the coil is also the least precise part.

On this tank, adding 1 pF moves the carrier down by 1.9 MHz. A broadcast FM channel is 200 kHz wide, so 1 pF is about 10 channels. A hand near the coil adds roughly that much capacitance.

The battery does the same thing more slowly. As the voltage falls, the bias current falls, the junction capacitances change, and the carrier walks. So does the temperature. Nothing in the circuit pulls it back, because a free-running LC oscillator has no reference to lock to.

Fixes that help: put the whole thing in a metal box with only the antenna out; add a buffer stage between oscillator and antenna; use NP0/C0G ceramic capacitors in the tank (see ceramic capacitors for why); and run the oscillator from a regulated supply rather than straight off the battery.

Our take: for anything beyond a bench experiment, use a PLL transmitter chip. KT0803L and Si4713 modules lock the carrier to a crystal and set the power in software. The single transistor teaches you how FM works, but it will not stay on one frequency.

The old page at this address promised 4 km. To test that, we need three numbers: what the transmitter radiates, what the path takes away, and what the radio needs.

What it radiates. Our estimate for this circuit into a quarter-wave whip is 10 mW, i.e. +10 dBm. That is generous for 4 mA at 9 V.

What the path takes. In free space the loss at 1 km and 98 MHz is 72 dB. But a transmitter on a table and a radio in someone's hand are both about 1.5 m off the ground, and the wave reflected off the ground cancels most of the direct one. Past about 9 m the loss rises 40 dB per decade instead of 20. At 1 km that plane-earth loss is 113 dB; at 4 km it is 137 dB. Walls, trees and bodies only add to it.

What the radio needs. We take 2 µV across 75 Ω, about -103 dBm, as the level for listenable mono on a decent tuner. A phone with the earphone lead as its antenna needs more.

Put together, the 10 mW bug falls below that level at about 990 m. Reaching 4 km at the same height would need about 2.7 W, some 270 times more. That takes a proper power amplifier and a mast, which is a broadcast station and needs a licence. The decibel lesson covers the arithmetic.

Our take: range claims on hobby transmitter circuits are usually measured with the receiver on a hill, or not measured at all. At ground level, a few hundred metres is a good result.

India. The 2018 short-range device rules exempt 87.5–108 MHz from licensing for "high duty cycle or continuous transmission" devices, which covers wireless microphones and personal audio transmitters, at up to 50 nW e.r.p. (Use of Low Power and Very Low Power Short Range Radio Frequency Devices (Exemption from Licensing Requirement) Rules). That is 50 billionths of a watt measured against a half-wave dipole, or 82 nW EIRP. Europe and the UK use the same 50 nW figure.

US. FCC Part 15 allows 250 µV/m at 3 m, inside a 200 kHz bandwidth (47 CFR § 15.239). Converted with E = √(30 × EIRP) / d, that is 18.8 nW EIRP, even less than India allows.

Our 10 mW estimate is about 1,22,000 times the Indian limit. At the legal limit, Figure 2 gives a range of about 53 m (about 37 m under the FCC rule), about what a car-stereo FM adaptor manages.

The practical rule: build and test it with a 10–20 cm wire or no antenna at all, next to the radio. A transmitter with a quarter-wave whip and no filtering also puts out harmonics at 196 and 294 MHz, and an unstable one can wander into the aircraft band.

Layout. Keep every lead in the tank and the feedback path as short as the parts allow; at 100 MHz, 1 cm of wire is about 10 nH, a tenth of the coil. Build on a scrap of copper-clad board with the ground as the copper, rather than on a breadboard, whose strips add several pF each.

Finding the carrier. Put an FM radio 1 m away, tuned to a quiet spot near 98 MHz, and turn the trimmer slowly with a plastic tool; a metal screwdriver retunes it as you turn. When you hit the carrier, the radio's hiss goes silent. Tap the microphone to confirm.

Coarse tuning with the coil. Pull the turns apart to lower the inductance and raise the frequency, or squeeze them together to go down. On this coil, stretching it from 5 mm to 7 mm long takes it to 84 nH and moves the carrier from 98 MHz to about 110.2 MHz. The inductor types lesson covers why air cores are used here.

The microphone. A two-lead electret capsule has a FET inside that needs bias; 10 kΩ from 9 V does it. The 100 nF coupling capacitor and the bias network behind it pass audio down to about 231 Hz, which is fine for speech. The microphones glossary entry has the capsule types.

• Transistor oscillatorsColpitts, Hartley and the loop-gain condition behind Figure 1.
• ResonanceWhy one L and one C pick one frequency.
• Parallel RLCThe tank on the collector, analysed.
• Q factor and bandwidthHow sharp the tank is, and why that helps stability.
• BJT biasingThe divider and emitter resistor that set 4 mA.
• VaractorThe proper way to turn a voltage into a frequency.
• Ceramic capacitorsWhy the tank wants NP0/C0G.
• The decibelAdding up power, path loss and sensitivity.

• Use of Low Power and Very Low Power Short Range Radio Frequency Devices (Exemption from Licensing Requirement) Rules, 2018, G.S.R. 1047(E), Table III (87.5–108 MHz, 50 nW e.r.p., high duty cycle devices including wireless microphones and personal audio streaming)
• 47 CFR § 15.239, Operation in the band 88–108 MHz (emission inside 200 kHz, field strength at most 250 µV/m at 3 m)
• onsemi 2N3904 datasheet (fT at least 300 MHz (IC = 10 mA, VCE = 20 V, f = 100 MHz))

Read on 30 September 2026. Our own choices, not taken from any source: the component values, the 200 current gain, the 5 pF stray estimate, the 10 mW output estimate, the 2 µV tuner sensitivity and the 1.5 m antenna heights. Change any of them and the numbers move; the method stays the same.