Frequency Domain Self Interference Cancellation for Full Duplex Radios

13 concepts12 papers1 proposal

In-Band Full-Duplex (IBFD) wireless communication is often described as the holy grail of spectral efficiency. By allowing a radio to transmit and receive data simultaneously on the very same frequency band, it theoretically doubles network throughput and drastically cuts latency.

The fundamental obstacle to achieving this is self-interference (SI). Because the transmit antenna sits directly adjacent to the receive antenna on the same device, the radio’s own transmitted signal leaks into its receiver. This self-interference can be up to 100 dB (ten billion times) stronger than the weak, incoming signal from a distant node. If this massive interfering signal reaches the receiver’s Analog-to-Digital Converter (ADC), it saturates the hardware, causing digital clipping and permanently destroying the desired incoming signal.

To make full-duplex communication viable, a radio must suppress this self-interference down to the receiver noise floor before it ruins the incoming data.

The Multi-Stage Architecture

Because no single mechanism can provide 100 dB of attenuation on its own, practical full-duplex systems distribute the suppression burden across multiple sequential layers. As established by Duarte et al. and further detailed by Smida et al., this partitioning spans three distinct domains:

  1. Passive Isolation: Physical antenna placement, cross-polarization, or specialized components like non-magnetic CMOS circulators (Nagulu et al.) provide the first line of defense, attenuating the signal as it propagates.
  2. Active Analog/RF Cancellation: An analog circuit taps the transmit path, manipulates the signal to mirror the leakage, and subtracts it from the receive path before the ADC. This stage is critical because its primary job is to prevent receiver hardware saturation.
  3. Digital Cancellation: Once the remaining signal is safely digitized without clipping, digital signal processing (DSP) algorithms execute fine-grained mathematical filtering to eliminate residual interference.
[Tx Signal] ---> (Passive Isolation) ---> (Analog/RF Cancellation) ---> [ADC] ---> (Digital Cancellation) ---> [Clean Rx]
                        |                           ^
                        +---[Cancellation Path]-----+

The Wideband Challenge and the Failure of Delay Lines

In a perfectly controlled environment, the self-interference path would be flat and uniform. You could cancel it simply by taking a copy of the transmit signal, inverting its phase, matching its amplitude, and injecting it back into the receiver.

However, real-world environments are frequency-selective. The antenna interface, device chassis, and nearby physical objects cause multi-path reflections. This means that different frequencies within the signal experience different amounts of attenuation and time delay. For wideband communication (such as 20 MHz WiFi or 5G channels), the self-interference channel becomes a highly complex, undulating waveform in the frequency domain.

To cancel a wideband signal in the analog domain, the RF cancellation circuit must precisely replicate this frequency-selective behavior. Historically, this was achieved using time-domain delay lines—physical, coiled transmission lines designed to mimic the exact time delays of the reflection paths. Unfortunately, physical delay lines are bulky, rigid, and impossible to integrate onto a silicon microchip. This constraint long restricted full-duplex radios to large, experimental testbeds, making them impractical for small-form-factor mobile handsets or tablets.

Frequency-Domain Equalization in the RF Path

To bypass the physical limitations of time-domain delay lines, researchers shifted the problem from the time domain to the frequency domain. In a single-author work, Zhou demonstrated that an analog cancellation path can be channelized using parallel, reconfigurable bandpass filters instead of physical delays. This technique is known as Frequency-Domain Equalization (FDE).

Instead of trying to delay the entire wideband signal at once, an FDE RF canceller splits the signal path into parallel frequency segments using integrated filter taps.

$$\text{RF Input} \longrightarrow \begin{bmatrix} \text{Filter Tap 1 } (f_1, Q_1) \\ \text{Filter Tap 2 } (f_2, Q_2) \\ \vdots \\ \text{Filter Tap } N (f_N, Q_N) \end{bmatrix} \longrightarrow \text{Synthesized SI Reflection}$$

By independently tuning two core parameters of each filter tap—the center frequency and the quality factor ($Q$)—the circuit can precisely manipulate the amplitude and the group delay of that specific frequency bin.

As shown by Chen et al. and Kohli et al., adjusting the $Q$ factor changes how long a signal dwells within the filter, effectively emulating a time delay without requiring physical distance. Because these reconfigurable bandpass filters can be fabricated using standard integrated circuit (IC) components, they deliver wideband self-interference cancellation within a highly compact footprint.

Deployments in large-scale wireless testbeds have confirmed that this hybrid approach—using compact RF-FDE to protect the ADC from saturation, followed by digital refinement—can reliably suppress over 95 dB of self-interference across 20 MHz bandwidths (Chen et al.).

Coordinating the Hybrid Interface

While the RF FDE stage handles the heavy lifting, the final precision relies entirely on how effectively the analog and digital layers work in tandem. This interface introduces unique hardware and algorithmic challenges.

First, active analog cancellation is highly vulnerable to hardware imperfections. For example, Sahai et al. identified local oscillator phase noise as a primary bottleneck. Because phase noise introduces random, rapid phase fluctuations between the transmitted signal and the cancellation circuit, it creates a fundamental floor past which active analog cancellation cannot penetrate.

Second, the analog and digital stages must be dynamically coordinated. If the digital stage attempts to cancel residual interference based on outdated assumptions about what the analog RF circuit is doing, the two layers can make conflicting adjustments, degrading the overall signal-to-interference-plus-noise ratio (SINR).

To address this, Balaji et al. explored sample-efficient digital cancellation using in-context learning, highlighting the ongoing need to cleanly coordinate learned digital filters with active analog hardware. Other advanced paradigms, such as the Semantics-Division Duplexing proposed by Niu et al., attempt to bypass traditional digital filtering limits altogether by processing and isolating signals within high-dimensional semantic spaces.

Ultimately, by pairing compact frequency-domain equalization in the RF front-end with intelligent digital back-ends, modern full-duplex architectures successfully bridge the gap between extreme mathematical isolation requirements and real-world hardware constraints.

Key Foundational Work

Design and Characterization of a Full-Duplex Multiantenna System for WiFi Networks (2014) — contributed 1 concept elements to this direction.
Understanding the impact of phase noise on active cancellation in wireless full-duplex (2012) — contributed 1 concept elements to this direction.
History & citationVersion 1 · 2026-10-04

Current · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_frequency_domain_self_interference_cance_937fb216,
  title = {frequency-domain self-interference cancellation},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/frequency-domain-self-interference-cancellation/rdr_937fb216c0614b7a86cc294a9a71bc99},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-19},
}

This exact version (version 1, frozen text):

@misc{lacuna_frequency_domain_self_interference_cance_937fb216_v1,
  title = {frequency-domain self-interference cancellation},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/frequency-domain-self-interference-cancellation/rdr_937fb216c0614b7a86cc294a9a71bc99/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-19},
}
About

These directions overlap with this one through shared papers.

See something wrong? Report a mistake