PaperlayerView exact revision on arXiv

Source-linked AI summary

All-Digital Self-interference Cancellation Technique for Full-duplex Systems

Elsayed Ahmed, Ahmed M. Eltawil

arXiv:1406.5555v1cs.IT

TL;DR

Full-duplex systems require strong self-interference mitigation to realize their spectral-efficiency potential, while conventional digital cancellation is limited by transmitter and receiver impairments. The paper introduces an auxiliary-receiver-based all-digital cancellation technique with a shared oscillator, and reports residual interference ∼3dB above the receiver noise floor with up to 76% rate improvement over conventional half-duplex at 20dBm transmit power.

  • Problem

    Full-duplex self-interference limits the effectiveness of low-complexity digital cancellation because transmitter and receiver circuit impairments constrain mitigation.

  • Method

    An auxiliary receiver chain obtains a digital-domain copy of the transmitted RF self-interference signal, while a common oscillator alleviates receiver phase-noise effects.

  • Results

    ∼3dB higher than the receiver noise floor, the mitigated self-interference yields up to 76% rate improvement compared to conventional half-duplex systems at 20dBm transmit power.

  • Takeaways & Limitations

    The proposed architecture significantly mitigates self-interference and associated transmitter and receiver impairments using digital-domain processing.

Abstract

from arXiv · show

Full-duplex systems are expected to double the spectral efficiency compared to conventional half-duplex systems if the self-interference signal can be significantly mitigated. Digital cancellation is one of the lowest complexity self-interference cancellation techniques in full-duplex systems. However, its mitigation capability is very limited, mainly due to transmitter and receiver circuit's impairments. In this paper, we propose a novel digital self-interference cancellation technique for full-duplex systems. The proposed technique is shown to significantly mitigate the self-interference signal as well as the associated transmitter and receiver impairments. In the proposed technique, an auxiliary receiver chain is used to obtain a digital-domain copy of the transmitted Radio Frequency (RF) self-interference signal. The self-interference copy is then used in the digital-domain to cancel out both the self-interference signal and the associated impairments. Furthermore, to alleviate the receiver phase noise effect, a common oscillator is shared between the auxiliary and ordinary receiver chains. A thorough analytical and numerical analysis for the effect of the transmitter and receiver impairments on the cancellation capability of the proposed technique is presented. Finally, the overall performance is numerically investigated showing that using the proposed technique, the self-interference signal could be mitigated to ~3dB higher than the receiver noise floor, which results in up to 76% rate improvement compared to conventional half-duplex systems at 20dBm transmit power values.

I. INTRODUCTION

Full-duplex can potentially double spectral efficiency, but conventional digital cancellation is limited by transceiver impairments and residual self-interference. The paper proposes an all-digital architecture using an auxiliary receiver chain to mitigate these effects, achieving near-noise-floor cancellation and substantial rate improvement.

  • Motivation and prior techniques: Conventional digital cancellation is low-complexity but achieves limited self-interference suppression because of hardware imperfections, especially transceiver phase noise and nonlinearities.These impairments are identified as the main performance-limiting factors.
  • Motivation and prior techniques: RF cancellation can mitigate main and reflected self-interference components, but RF FIR filters limit applicability through their size and power consumption.Single attenuator-and-phase-shifter designs mainly mitigate the dominant component and not reflections.
  • Proposed architecture: The proposed all-digital technique uses an auxiliary receiver chain to obtain a digital copy of the transmitted RF self-interference signal and cancel it with associated impairments.The auxiliary chain has components matching the ordinary receiver chain, while a shared oscillator alleviates receiver phase-noise effects.
  • Proposed architecture: All signal processing remains in the digital domain, significantly reducing implementation complexity without requiring highly complex RF cancellation techniques.The architecture is designed to mitigate transmitter and receiver impairments.
  • Analysis and evaluation: Analytical and numerical analyses show that transceiver phase noise and nonlinearities are no longer the main performance-limiting factors with the proposed technique.The evaluation accounts for nonlinearities, phase noise, Gaussian and quantization noise, and channel-estimation errors.
  • Analysis and evaluation: ∼3dB higher than the receiver noise floor, the residual self-interference supports up to 67-76% rate improvement over conventional half-duplex systems at 20dBm transmit power.The overall system combines the proposed digital cancellation with practical passive suppression techniques.

II. SYSTEM MODEL

The proposed architecture uses an auxiliary receiver to capture the transmitted RF self-interference, estimate the channel ratio, and subtract a digitally reconstructed copy from the ordinary receiver output. Shared PLL operation and a PA-output feedback signal allow cancellation of transmitter impairments and mitigation of receiver phase noise.

  • System architecture: An auxiliary receiver chain samples a fraction of the amplified transmit signal as a digital-domain self-interference reference.The feedback is taken from the PA output before auxiliary down-conversion.
  • Receiver chains: The auxiliary and ordinary receiver chains down-convert their respective signals to baseband and share an identical PLL.The wired auxiliary path is represented by Haux, while Hord represents the wireless self-interference channel.
  • Digital cancellation: A channel-estimation block derives Hord/Haux from time-orthogonal training sequences transmitted at each data-frame beginning.The estimate is multiplied with the auxiliary output before subtraction from the received signal.
  • Digital cancellation: The multiplication output is subtracted from the ordinary received signal to obtain an interference-free signal.The system model includes transmitter and receiver phase noise, nonlinearities, ADC quantization noise, and receiver Gaussian noise.
  • Impairment mitigation: Because the feedback signal originates at the PA output, the proposed architecture can significantly mitigate transmitter impairments, while shared PLL operation mitigates receiver phase noise.The signal model explicitly includes transmitter nonlinear distortion, receiver nonlinear distortion, quantization noise, and Gaussian noise.

A. Transceiver Nonlinearities

The transceiver nonlinearity model represents each nonlinear block with a polynomial, retaining odd-order in-band distortion; numerical analysis is limited to third-order nonlinearity.

  • Nonlinear sources: Power amplifiers and LNAs are identified as the main practical sources of system nonlinearity.The PA is at the transmitter and the LNA is at the receiver.
  • Polynomial model: A nonlinear block is modeled as a polynomial whose first term is linear and whose higher-order terms generate distortion.Only odd polynomial orders contribute to in-band distortion.
  • Polynomial model: The baseband model simplifies the polynomial representation for digital input and output signals.The variables g_n and y_n denote the digital baseband input and output of the nonlinear block.
  • Numerical setting: Third-order nonlinearity is the highest order retained in the paper’s numerical analysis.The signal model itself remains valid for any nonlinearity order.

B. Transceiver Phase Noise

The paper models oscillator phase noise with Wiener or Ornstein-Uhlenbeck processes and models receiver noise and quantization in the two chains. Wireless-channel modeling includes Rician LOS and Rayleigh reflective components, with architecture-dependent suppression and coupling boundaries.

  • Oscillator phase noise: The paper distinguishes free-running oscillators, modeled with Wiener phase noise, from PLL-based oscillators, modeled with an Ornstein-Uhlenbeck process.Numerical analysis uses PLL-based oscillators.
  • PLL model: A PLL controls the VCO through a feedback loop containing a phase detector and low-pass filter.The PLL output phase noise is characterized through its autocorrelation and power spectral density.
  • Receiver noise: Receiver Gaussian noise is additive circuit noise characterized through the receiver noise figure, while ADC quantization noise depends on ADC resolution.The two chains have independent Gaussian and quantization noises despite identical components.
  • Wireless channel: The self-interference channel combines a Rician first tap with Rayleigh remaining taps, and its Rician factor represents LOS-to-reflection power ratio.Passive suppression changes both suppression amount and the channel’s Rician factor.
  • Wireless channel: The analyzed system uses two antennas, whereas a circulator-based single-antenna architecture is limited by typically 20dB circulator coupling.The single-antenna channel is expected to have a larger Rician factor because reflected paths make a roundtrip.

III. SELF-INTERFERENCE CANCELLATION ANALYSIS

The cancellation analysis evaluates how receiver impairments and channel-estimation errors limit a digitally reconstructed self-interference copy. It finds that quantization can be reduced with ADC resolution, whereas Gaussian noise becomes the key limitation at high input power.

  • Cancellation principle: The proposed technique copies the transmitted self-interference signal, including transmitter impairments, and uses it for digital-domain cancellation.With impairment-free receiver chains, the architecture should eliminate both the self-interference signal and transmitter impairments.
  • Limiting factors: Perfect cancellation is impossible because receiver impairments and channel-estimation errors constrain cancellation capability.The paper analyzes receiver impairments individually before evaluating overall performance with all impairments.
  • Overall performance: Gaussian and quantization noise form the first performance-limiting factor in the post-cancellation signal model.Because the four noise terms are uncorrelated, their total power is calculated by summing their individual powers.
  • Quantization noise: ADC quantization noise decreases with LNA gain and ADC bits, so increasing ADC resolution can remove its limitation at modest hardware-complexity cost.A 14-bit ADC yields approximately −90dBm quantization-noise power at 0dB LNA gain.
  • Gaussian noise: Gaussian noise typically dominates quantization noise, especially at high input power levels.Reducing Gaussian-noise impact requires lower input power through passive suppression or reduced transmit power.
  • Overall performance: For input power levels ≤−30dBm, the proposed full-duplex system has the same noise floor as half-duplex systems.At high received power, the full-duplex noise floor increases approximately linearly as reduced LNA gain raises the overall receiver noise figure.

B. Impact of Receiver Phase Noise

The proposed cancellation architecture uses shared receiver phase noise and channel-frequency correlation to reduce phase-noise-induced residual self-interference. Larger channel coherence bandwidth produces greater mitigation.

  • The auxiliary and ordinary receiver chains share the same PLL, aligning their receiver phase-noise processes.
  • The residual self-interference after cancellation is caused by receiver phase noise and depends on channel frequency-response correlation across subcarriers.
  • The technique eliminates phase noise from the channel’s line-of-sight component and mitigates part of the non-line-of-sight component according to coherence bandwidth.
  • 15-30dB reduction in residual self-interference power compared to the upper bound is achieved through channel-frequency-response correlation.

C. Impact of Channel Estimation Errors

Channel-estimation errors arise from receiver impairments and channel fading, creating a tradeoff between estimation quality, training overhead, and frame length.

  • Channel-estimation error is directly related to the summed noise components in the auxiliary and ordinary receiver chains.
  • Averaging channel estimates over 4 training symbols reduces degradation to 1dB compared with perfect channel knowledge.
  • Fading is more consequential in full-duplex systems because high self-interference power can make fading-induced channel error dominate other noise components.
  • At high received power and 4% training overhead, a 50-symbol frame with 2 training symbols outperforms a 100-symbol frame with 4 training symbols.

D. Impact of Receiver Nonlinearities

Receiver nonlinearities add residual self-interference, so the proposed architecture estimates and suppresses nonlinear distortion during training and data transmission.

  • Receiver nonlinear distortion limits cancellation, making nonlinearity suppression essential for improved self-interference cancellation.
  • The receiver nonlinearity coefficient α3 is estimated with an additional training symbol and used to reconstruct and subtract distortion during data symbols.
  • The nonlinearity coefficients change slowly enough to be estimated once every several frames, reducing the additional training overhead.
  • 23dB more reduction in residual self-interference power is achieved with nonlinearity suppression than without it.
  • At distortion power levels ≤−45dBm, suppression achieves nearly the linear-receiver performance and eliminates the nonlinearity effect.

E. Overall Cancellation Performance

The overall evaluation examines practical impairment settings and passive-suppression scenarios. The proposed technique removes phase-noise and nonlinearity bottlenecks, reaching residual interference near the half-duplex noise floor in favorable scenarios.

  • The proposed technique reduces phase-noise and nonlinearity effects below the receiver noise floor, unlike conventional digital cancellation.
  • With relatively low passive suppression, receiver Gaussian noise dominates and makes the technique more suitable for low transmit-power applications.
  • With good passive suppression, residual self-interference reaches approximately 3dB above the half-duplex system’s noise floor.
  • Channel fading is the next performance bottleneck after receiver Gaussian noise, but interpolation, pilots, and shorter frames can reduce its effect.

IV. ACHIEVABLE RATE ANALYSIS

The paper evaluates achievable full-duplex rates under transmitter and receiver impairments across three operating scenarios, transmit powers, and SNR values. The proposed cancellation generally outperforms conventional digital cancellation and can improve rates over half-duplex, particularly with sufficient passive suppression and high SNR.

  • IV. ACHIEVABLE RATE ANALYSIS: The evaluation considers full-duplex and half-duplex achievable rates with all transmitter and receiver impairments at 20dBm and 5dBm transmit power.Rates are analyzed across the three operating scenarios and SNR values.
  • IV. ACHIEVABLE RATE ANALYSIS: The proposed technique significantly outperforms conventional digital cancellation in all operating scenarios except at 5dBm in the third scenario.At that operating point, both techniques achieve the same performance because the required 35dB cancellation can be reached conventionally.
  • IV. ACHIEVABLE RATE ANALYSIS: As transmit power increases, proposed cancellation continues to increase while conventional cancellation saturates.This divergence improves the proposed technique’s relative performance at higher transmit powers.
  • IV. ACHIEVABLE RATE ANALYSIS: With relatively low passive suppression in the first scenario, the proposed technique is usable only in low-transmit-power applications.The paper identifies passive suppression as a condition affecting the operating range.
  • IV. ACHIEVABLE RATE ANALYSIS: The proposed technique achieves significant rate improvement over conventional half-duplex systems in the second and third scenarios, especially at high SNR.Average rate gain is calculated over the 0 to 40dB SNR range, while exact gains vary by SNR.

V. CONCLUSION

The paper proposes an all-digital self-interference cancellation technique using an auxiliary receiver chain and impairment-aware processing. Numerical analysis shows mitigation to approximately 3dB above the receiver noise floor and up to 76% rate improvement over conventional half-duplex at 20dBm.

  • V. CONCLUSION: The proposed all-digital technique uses a new full-duplex transceiver architecture to mitigate self-interference and associated transceiver impairments.The conclusion reports significant mitigation of transmitter and receiver phase-noise and nonlinearity effects.
  • V. CONCLUSION: An auxiliary receiver chain obtains a digital-domain copy of the transmitted RF self-interference signal, including transmitter impairments, for digital cancellation.The copy is used to cancel both the self-interference signal and transmitter impairments.
  • V. CONCLUSION: A shared oscillator between auxiliary and ordinary receiver chains alleviates receiver phase-noise effects, while nonlinearity estimation and suppression addresses receiver nonlinearities.The paper also presents analytical and numerical impairment analyses.
  • V. CONCLUSION: The proposed technique is numerically evaluated with practical passive suppression across the three operating scenarios.Both full-duplex and half-duplex performances are considered in the overall evaluation.
  • V. CONCLUSION: ∼3dB above the receiver noise floor is the reported residual self-interference level, yielding up to 76% rate improvement over conventional half-duplex at 20dBm transmit power.The result is reported as the overall performance outcome of the proposed technique.
Loading 1406.5555v1…