Single sideband (SSB) is the most common voice mode for analog communications over HF radio. Its widespread adoption is due in part to its spectral efficiency, with a typical SSB signal occupying ~3 kHz of bandwidth. This efficiency, however, comes with a tradeoff: because only part of the frequency range of human speech is transmitted, received SSB audio sounds less natural than speech heard in person. Improving speech quality while maintaining the characteristics that make SSB effective over HF remains a challenge in amateur radio.

This is exactly the problem that ARDC grantee FreeDV set out to solve. Since late 2023, they have been working to create a digital voice mode that could compete with SSB across varying signal-to-noise conditions on HF while providing more natural-sounding speech. In June of 2025, they released Radio Autoencoder (FreeDV RADE or RADE), their flagship open-source digital voice mode. Building on that milestone, in May 2026, the team worked with radio manufacturers to integrate the mode directly into their radios.

Capabilities

FreeDV RADE combines machine learning with classical digital signal processing (DSP) and can operate at signal-to-noise ratios (SNRs) as low as -2 dB. Its design also addresses spectrum use, speech quality, and power efficiency, all important considerations for voice communications over HF.

Whereas a typical SSB signal occupies about 3 kHz, FreeDV RADE requires about 1.5 kHz, cutting the amount of spectrum used by half. At the same time, it can transmit about 8 kHz of voice audio, capturing a wider range of human speech frequencies, resulting in more natural-sounding speech.

This difference is particularly noticeable when people hear RADE for the first time. When FreeDV demos RADE at hamfests, amateur radio operators who are familiar with DMR, D-STAR, and other forms of digital voice on UHF/VHF are amazed by the audio quality and how closely it resembles natural human speech.

“In south east Australia I conduct a weekly net using FreeDV RADEV1,” says Peter Marks VK3TPM. We get stations from VK3, VK2, VK5 and VK7. Many of these stations live in inner city locations where HF noise has become so bad that listening to sideband signals is difficult and tiring. Stations report how easy FreeDV RADE is to listen to compared to analog modes and it’s helped to bring them back to HF.”

Power efficiency was another consideration. FreeDV RADE has a peak-to-average power ratio (PAPR) that allows for the efficient use of power amplifiers, resulting in less heat and lower power consumption during transmission. The narrower bandwidth and the PAPR of RADE give users a way to experiment with digital voice while making more efficient use of both spectrum and power.

These capabilities do require additional computing resources. FreeDV RADE uses more memory and CPU than the team’s previous efforts, a tradeoff for improved speech quality and the ability to operate across different conditions on HF. Despite these requirements, RADE runs well on a Raspberry Pi 4 or 5.

Open source was also an important requirement for FreeDV RADE. The team wanted to make the technology available for others to learn from, experiment with, and build upon by avoiding patent and licensing barriers.

Implementation

Support from ARDC has helped the team continue building on its earlier work. Funding has allowed the team to address shortcomings in previous modes, improve the FreeDV application, fix bugs, and add new functionality. As part of that effort, the team has also worked with radio manufacturers, such as FlexRadio, to integrate FreeDV RADE directly into their radios.

For users with FlexRadio 6000, 8000, or Aurora series radios, that integration makes trying RADE more straightforward. A module can be downloaded from the FreeDV website (https://freedv.org/download/) and installed on the radio, creating FreeDV as a mode in the radio itself, similar to selecting SSB. Operators can then use RADE without needing a separate computer to run the software.

Looking Ahead

Users from around the world, including in Europe, Japan, South America, and the United States, are experimenting with FreeDV RADE. Their interest reflects one of the project’s original goals: giving people another way to experiment with high-quality digital voice over HF.

A C library for RADE is now available (https://github.com/freedv/rade_c), which is helping to get the mode integrated into more radios and SDR programs. A current list of third-party clients is available at https://freedv.org/alternative-freedv-rade-v1-software/.

Development is also continuing with FreeDV RADE v2, where the team is working on additional improvements to speech quality and aiming to narrow the bandwidth further, with a goal of getting it under 1 kHz.

For the FreeDV team, RADE is about exploring a different approach to voice communications over HF, one that combines efficient spectrum use, improved speech quality, and an open-source implementation that others can study, experiment with, and build upon. With FreeDV RADE v2 already in development, that work is continuing.

Learn More

To learn more about FreeDV, visit https://freedv.org to download the software, explore the getting started guide, and find installation instructions. You can also visit the FreeDV Reporter at https://qso.freedv.org/, where you can see a near real-time look at FreeDV activity.