Source-linked AI summary
Analysis of Oscillator Phase-Noise Effects on Self-Interference Cancellation in Full-Duplex OFDM Radio Transceivers
Ville Syrjala, Mikko Valkama, Lauri Anttila, Taneli Riihonen, Dani Korpi
TL;DR
The paper examines how oscillator phase noise limits self-interference cancellation in full-duplex direct-conversion OFDM transceivers. It derives closed-form subcarrier-wise residual self-interference power for independent and common oscillators, finding that phase noise can seriously compromise cancellation, particularly with independent oscillators, while multipath phase-noise spread becomes limiting when high cancellation is pursued.
Problem
Phase-noise effects on self-interference cancellation in practical full-duplex direct-conversion transceivers remain insufficiently analyzed, especially beyond narrowband and small-phase-noise assumptions.
Method
The paper derives closed-form subcarrier-wise self-interference power expressions including isolation, multipath propagation, analog and digital cancellation, and independent or common oscillators.
Results
Phase noise can seriously compromise self-interference cancellation, especially with independent oscillators, while multipath-component phase-noise spread limits cancellation after RF/analog suppression.
Takeaways & Limitations
High-quality oscillators or phase-noise estimation and mitigation are needed when high self-interference cancellation is pursued, and a common oscillator is beneficial.
Abstract
from arXiv · showhide
This paper addresses the analysis of oscillator phase-noise effects on the self-interference cancellation capability of full-duplex direct-conversion radio transceivers. Closed-form solutions are derived for the power of the residual self-interference stemming from phase noise in two alternative cases of having either independent oscillators or the same oscillator at the transmitter and receiver chains of the full-duplex transceiver. The results show that phase noise has a severe effect on self-interference cancellation in both of the considered cases, and that by using the common oscillator in upconversion and downconversion results in clearly lower residual self-interference levels. The results also show that it is in general vital to use high quality oscillators in full-duplex transceivers, or have some means for phase noise estimation and mitigation in order to suppress its effects. One of the main findings is that in practical scenarios the subcarrier-wise phase-noise spread of the multipath components of the self-interference channel causes most of the residual phase-noise effect when high amounts of self-interference cancellation is desired.
I. INTRODUCTION
Full-duplex radios can improve spectral efficiency but face severe self-interference and unresolved oscillator phase-noise challenges, especially in practical low-cost transceivers. This paper models direct-conversion architectures with analog and digital cancellation, multipath coupling, and either independent or shared oscillators.
- Full-duplex operation can theoretically double spectral efficiency compared with TDD and FDD when self-interference is solved.
- The main implementation challenge is self-interference caused by imperfect transmitter–receiver isolation, addressed through antenna isolation and active analog and digital cancellation.Reported antenna isolation is typically 20–40 dB, while analog cancellation subtracts an aligned transmit waveform from the receiver input.
- Phase noise is identified as a performance-limiting issue for low-cost full-duplex direct-conversion transceivers, while prior analyses mainly considered narrowband signals and small phase-noise assumptions.
- A practical LTE UE example leaves approximately −87 dBm of self-interference after 30 dB antenna isolation, 30 dB analog cancellation, and 50 dB digital cancellation.This remains above the cited −110 dBm LTE receiver reference sensitivity level.
- The analysis covers both independent transmitter and receiver oscillators and a shared oscillator, without imposing a small phase-noise assumption.
- The transceiver model includes antenna isolation, multipath propagation, oscillator phase noise, analog linear cancellation, and digital cancellation while omitting the useful received signal.
III. SUBCARRIER-WISE SELF-INTERFERENCE POWER DUE TO PHASE NOISE IN OFDM FULL-DUPLEX RADIO
The paper derives closed-form subcarrier-wise self-interference power for OFDM full-duplex radios under independent and common oscillator configurations. These expressions support analysis of analog cancellation, multipath propagation, digital cancellation, and comparisons between oscillator scenarios.
- The study analyzes subcarrier-wise self-interference power in closed form for independent oscillators and for a common shared oscillator.
- The derived formulas are used to evaluate the effects of analog cancellation, multipath propagation, and digital cancellation on residual self-interference power.
A. Subcarrier-wise Self-Interference Power before Digital Linear Cancellation
The paper constructs a sampled OFDM self-interference model with phase noise and analog cancellation, transforms it to subcarrier-domain power, and derives closed-form expressions under stated statistical assumptions.
- The sampled self-interference model includes phase noise and analog cancellation before receiver-side DFT processing.
- The OFDM formulation uses N subcarriers and represents subcarrier self-interference through DFT-domain signal components.
- The derivation assumes independent subcarrier-related terms and multipath components following Bello’s WSSUS model, with zero means and defined second moments.
- Closed-form expressions are derived for both independent and common transmitter–receiver oscillator cases using a free-running oscillator model.
1) Independent Oscillators Case:
For independent oscillators, the closed-form subcarrier-wise self-interference power depends on oscillator phase-noise bandwidth and key OFDM, cancellation, and multipath parameters.
- The independent-oscillator self-interference expression depends on phase-noise 3-dB bandwidth, analog cancellation, subcarrier count, multipath profile, and subcarrier spacing.Numerical illustrations are used to evaluate these dependencies.
2) Common Oscillator Case:
The common-oscillator formulation derives residual self-interference power while accounting for phase-noise-induced subcarrier spreading and channel-estimation errors. This spreading cannot be removed by linear digital cancellation alone and limits achievable suppression.
- Common oscillator case: The common-oscillator expression gives residual self-interference power as a function of phase noise, coupling propagation, and the OFDM waveform.The expression can be evaluated for arbitrary system configurations.
- Subcarrier-wise effects: Phase noise spreads self-interference across subcarriers through intercarrier interference, which digital linear cancellation cannot remove without phase-noise estimation.Common phase error may be partly included in the effective self-interference channel and mitigated through channel estimation.
- Digital cancellation: Channel-estimation errors prevent even the linear self-interference terms from being perfectly suppressed at the digital cancellation output.The analysis incorporates their joint effect with phase-noise-induced subcarrier spreading.
Q C QCE
This section relates antenna separation and analog cancellation to the main multipath component while retaining unsuppressed reflected components. It also defines attainable suppression limits and notes the scope of the single-path analog-cancellation model.
- Multipath propagation: Antenna separation attenuates only the main multipath component, because reflections from farther paths are not essentially reduced in small full-duplex transceivers.Therefore, the main component must be suppressed more than the whole-signal antenna-isolation target when multipath remains.
- Analog cancellation: Analog cancellation is modeled as suppressing only the main multipath component, although its reported suppression is commonly specified for the whole self-interference signal.The required main-component suppression is therefore adjusted to account for the other multipath components.
- Attainable suppression: Negative derived suppression factors indicate that the requested antenna separation or analog cancellation is unattainable for the considered coupling channel.The corresponding expressions define the maximum attainable antenna separation and analog-cancellation factor.
- Model scope: The analysis uses a single-path analog/RF cancellation scheme, while multipath cancellation schemes are acknowledged as possible but outside this low-cost, simple-RF focus.The single-path concept is described as common in existing literature and demonstrations.
IV. SIMULATION SCENARIOS, RESULTS AND ANALYSIS
The study verifies its analytical phase-noise results with full-waveform OFDM simulations using practical LTE-like parameters and two cancellation scenarios. The simulations compare practical and ideal analog/digital cancellation settings while varying phase-noise conditions.
- Simulation setup: The simulator generates a 1024-subcarrier OFDM waveform with 300 active subcarriers on each side of DC, 16QAM data, 15 kHz spacing, and a 63-sample cyclic prefix.The waveform closely resembles a 3GPP LTE downlink signal, while the analysis is stated to apply to arbitrary OFDM signals.
- Simulation setup: The simulated multipath coupling channel uses powers of -30 dB, -65 dB, -70 dB, and -75 dB at delays of 0, 1, 2, and 4 samples, respectively.The profile is selected as a practical model for handheld or portable full-duplex devices.
- Reference scenarios: The Practical scenario assumes 30 dB analog cancellation and 50 dB digital cancellation without phase noise, for 80 dB total suppression.These values are close to reported achievable results but may be slightly optimistic because of laboratory-scale equipment.
- Reference scenarios: The Ideal scenario assumes perfect suppression of the main self-interference component and perfect digital cancellation without phase noise.Additional studies vary either analog or digital cancellation from the reference cases.
C. Results and Analysis
Phase noise substantially limits full-duplex self-interference cancellation, with common oscillators generally outperforming independent oscillators. The residual effect becomes especially important as oscillator bandwidth, multipath strength, or digital cancellation increases.
- Principal spectral illustrations: At 50 Hz oscillator bandwidth, common-oscillator phase noise raises the Practical-case noise floor by about 4–5 dB, whereas independent oscillators leave only 48 dB of SI cancellation.The common-oscillator result remains near the 80 dB phase-noise-free cancellation level; independent oscillators produce a much more severe residual SI effect.
- Principal spectral illustrations: In the Ideal case, independent oscillators still leave high residual SI, while the common oscillator reaches an approximately 77 dB inband SI-suppression limit.This floor occurs despite perfect linear SI-channel knowledge.
- Validation: Analytical and simulated SI results match practically perfectly, including subcarrier-by-subcarrier spectral behavior.This agreement supports the analytical treatment across the illustrated Practical and Ideal cases.
- Effect of oscillator bandwidth: Independent-oscillator phase noise rises rapidly with oscillator bandwidth, while common-oscillator interference begins increasing steadily after approximately 5 Hz.In the Practical case, independent oscillators already exceed −65 dB residual SI at an oscillator bandwidth of around 1 Hz.
- Effect of multipath powers: When multipath components strengthen, Practical-case independent oscillators remain limited by the main component, whereas common-oscillator and Ideal-case behavior increasingly reflects non-main multipath components.The Practical curves terminate near a 3 dB relative channel change because the desired 30 dB analog cancellation can no longer be achieved.
- Effect of digital cancellation: With fixed analog cancellation, digital-cancellation gains eventually floor at phase-noise limits: around 15 dB for independent oscillators and 40 dB for the common oscillator.The floor is set by phase noise from multipath components once digital cancellation becomes sufficiently strong.
4) Effect of Varying DLC Gain:
The results examine how analog and digital cancellation, together with TX–RX oscillator-interface delay, shape residual SI under Practical and Ideal conditions. Phase-noise floors constrain further cancellation, while common-oscillator delay introduces an additional performance dependence.
- Effect of analog cancellation: Non-ideal analog cancellation heavily limits independent-oscillator performance, while increasing analog cancellation eventually reaches a phase-noise floor.The floor is caused by phase noise in multipath components, which analog cancellation affects less directly than the main component.
- Effect of analog cancellation: Beyond a certain analog-cancellation level, improving digital cancellation no longer provides additional benefit because multipath phase noise sets the performance limit.This behavior appears in both Practical and Ideal cases, with the Ideal curve starting slightly lower because of perfect digital cancellation.
- Effect of TX–RX delay: The delay study varies oscillator-interface delay from approximately 0.67 ns to 65 ns, corresponding to roughly 20 cm to 19.5 m.The analysis isolates delay from the accompanying antenna-separation and propagation-loss effects.
- Effect of TX–RX delay: For independent oscillators, changing TX–RX oscillator-interface delay does not change the statistical phase-noise effect; average inband cancellation is −48.6 dB in Practical and −52.1 dB in Ideal cases.The corresponding curves are therefore straight lines and are omitted from the figure.
7) Effect of Varying TX-RX Distance:
Antenna distance strongly affects phase-noise-limited self-interference cancellation. Common oscillators outperform independent oscillators, but increasing distance and delay emphasizes phase-noise effects, especially through multipath components.
- Antenna distance has a huge effect on overall ALC+DLC performance for both independent- and common-oscillator cases.
- 82 dB and 108 dB total SI suppression are obtained at 20 cm with independent and common oscillators, respectively.
- 88 dB and 110 dB total SI suppression are obtained at 20 m with independent and common oscillators, respectively.
- The common-oscillator case outperforms the independent-oscillator case by around 26 dB at 20 cm and around 22 dB at 20 m.
- Higher distances and especially longer delays emphasize phase-noise effects, motivating better oscillators or explicit phase-noise estimation and suppression.
- Strong multipaths and long coupling delays make improved multipath-capable ALC essential with practical oscillators having considerable phase noise.
APPENDIX
The appendix derives phase-noise expectation expressions for residual self-interference power under independent- and common-oscillator configurations using a Brownian-motion phase-noise model.
- The appendix derives the residual self-interference power expression for independent and common oscillators in the transmitter and receiver paths.
- The analysis assumes Brownian-motion, or Wiener-process, phase noise generated by a free-running oscillator.
- Differences between two Brownian-motion samples are modeled as zero-mean Normal random variables whose variance depends on bandwidth and time separation.
- For independent oscillators, transmitter and receiver phase-noise processes are statistically independent.
- For the common-oscillator case, the transmitter and receiver share the same phase-noise process, changing the expectation expression.
Q C QCE
This appendix passage describes the algebraic manipulation of the common-oscillator phase-noise expression by exploiting index separation and independent groupings.
- The statistical properties inside the expectation vary with the separation between the summation indices.
- Phase-noise terms are grouped into nonoverlapping pairs so the pairs are statistically independent for variance calculation.
- The combined variances of the separated pairs are summed to obtain the final expression.
BIOGRAPHIES
The biographies describe the authors’ backgrounds in communications engineering, signal processing, and full-duplex radio research.
- Ville Syrjälä is a communications-engineering researcher whose interests include full-duplex radio, transceiver impairments, and signal-processing algorithms.
- Mikko Valkama is a professor whose research interests include communications signal processing and estimation and detection techniques.
- Lauri Anttila researches statistical and adaptive signal processing, digital front-end processing, and full-duplex radio systems.
- Taneli Riihonen and Dani Korpi are researchers associated with communications engineering and full-duplex radio research.