Full-duplex integrated sensing and communications often begins with a difficult receiver problem: a transmitter’s own signal can overwhelm the echo or data stream that the same platform is trying to receive. One response is to estimate that leakage and cancel it in radio-frequency, analog, or digital processing. A newly published open-access prototype tests a different operating point—separate the two functions in frequency and accept the resulting spectrum and hardware cost.
The FR3-DUALIS platform places bistatic communications at 24.25 GHz and monostatic sensing at 25.175 GHz. Each channel occupies 360 MHz in the upper third frequency range (FR3), leaving 925 MHz between their carrier frequencies. Dedicated transmit and receive chains serve the two functions, while a shared OFDM baseband structure preserves some common processing.
Frequency separation becomes the isolation mechanism
The central design decision is architectural rather than algorithmic. By assigning sensing and communications to different bands, the prototype shifts cross-function isolation into spectrum allocation instead of relying on a digital self-interference cancellation filter.
In the reported QPSK-OFDM tests, frequency-separated operation produced zero observed bit errors over the tested communication frames, matching the communications-only baseline. Running the two functions at the same frequency produced a bit-error rate of about 0.05. These values describe the tested frames and hardware arrangement; they are not a general reliability guarantee for arbitrary channels, powers, or deployments.
A common baseband can still use function-specific waveforms
Frequency separation does not require the two tasks to use identical symbols. The platform keeps a unified OFDM baseband architecture but uses Zadoff–Chu sequences for sensing and QPSK symbols for communications. Relative to QPSK-based sensing in the reported experiment, the sensing sequence improved the peak-to-sidelobe ratio by 2.49 dB and the integrated-sidelobe-level ratio by 4.43 dB.
The hardware study also exposes a processing boundary. The authors report that stable ranging required at least 8,192 subcarriers because common phase error becomes harder to control at smaller configurations, and that 360 MHz was the maximum bandwidth sustained by their real-time processing setup. With 1,000 OFDM symbols per frame, the prototype continuously tracked velocity at 10 Hz with a reported resolution of 0.21 m/s.
Lower cancellation complexity is purchased, not free
Moving isolation into the frequency plan removes the tested need for computational digital cancellation, but it consumes two bands and duplicates radio-frequency chains. The result is therefore best read as a measured point in a three-way design space: signal-processing load, spectrum occupancy, and hardware count.
That trade-off may fit spectrum-abundant experimental or specialized deployments better than congested systems. Wider field trials would still need to examine adjacent-channel leakage, oscillator and calibration behavior, mobility, multipath, regulatory allocations, and coexistence with other users. The contribution is not that one architecture dominates every setting, but that a working wideband prototype makes the costs measurable.
Research notes
FR3-DUALIS: from signal processing to spectrum allocation for real-time wideband full-duplex ISAC—a prototype validation at FR3
Authors: Bixing Yan and Yang Miao.
Status: Open-access article published in npj Wireless Technology on 4 September 2026.
What the public evidence establishes: The prototype separates 360 MHz sensing and communications channels by 925 MHz in upper FR3, reports zero observed communication bit errors in the tested frequency-separated frames versus about 0.05 for same-frequency dual-function operation, and demonstrates real-time velocity tracking at 10 Hz with 0.21 m/s resolution.
Limits: The reported results are specific to the tested prototype and do not establish performance across arbitrary propagation, interference, regulatory, or deployment conditions. Frequency separation also requires additional spectrum and duplicated RF chains.