A Hacker Built a 20 MHz Software-Defined Radio for Under $50

Hackaday Europe 2026: High Performance SDR On The Cheap

A firmware engineer walked into a packed conference room in Europe and proved that building a capable software-defined radio no longer demands a fat wallet or mysterious black-box components. Anders Nielsen’s latest creation pulls in the entire FM broadcast band simultaneously, yet the bill for parts stayed firmly under $50.

The talk, delivered at Hackaday Europe 2026, arrived at a moment when off-the-shelf SDR platforms dominate the hobbyist landscape. Plenty of affordable options exist: Nielsen himself pointed to the Zync/AD9363 combo as one ready-made example. But he wanted something different. Something transparent. Most commercial units bury the signal chain beneath layers of abstraction and still carry a price tag that stings budget-conscious experimenters.

So he assembled his own, selecting each component to expose rather than conceal the radio’s inner workings.

A TLV3253 handles quadrature sampling, prized for its rapid switching and minimal on-resistance. The local oscillator duties fall to a Silicon Labs 5351, which can phase-shift two clocks from a single PLL: same frequency, 90 degrees apart. An ADA4891 op-amp buffers the signal between the sampling detector’s capacitors and the ADC stage. Testing began humbly, with a computer soundcard delivering a mere 44.1 kHz of bandwidth. From there, Nielsen graduated to an STM32 microcontroller offering dual ADCs for 200 kHz, then ultimately landed on the HT9201, a dual 20 MHz, 10-bit converter. An FX2LP clone shuttles data over USB 2.0 at 480 megabits per second, allowing the builder to choose between 20 MHz of bandwidth at 8-bit resolution or 10 MHz at a cleaner 10 bits.

Obstacles surfaced throughout development. Flat analog frequency response refused to cooperate, demanding careful PCB layout and deliberate amplification and filtering choices. Clock jitter and synchronization headaches joined the list, along with front-end bandwidth constraints.

None of that prevented a live demonstration. Nielsen fired up the rig, captured the entire FM broadcast spectrum, and played a commercial station for the audience: crunchy, unmistakably real, and proof the whole contraption worked. He also walked attendees through methods for spotting and suppressing mirror signals when they inevitably appear.

The roadmap extends further. A wideband front-end mixer, a low-noise amplifier, and a smarter filtering architecture all sit on the horizon, promising gains in both flexibility and raw performance. This radio will never outclass professional lab equipment. But it already succeeded at its true purpose: forcing a builder, and potential adopters, to understand every link in the chain. That knowledge, more than any spec sheet number, constitutes the real win.