PaperlayerView exact revision on arXiv

Source-linked AI summary

Simultaneous Transmission and Reception: Algorithm, Design and System Level Performance

Yang-Seok Choi, Hooman Shirani-Mehr

arXiv:1309.5546v1cs.NI

TL;DR

The paper addresses strong transmit echoes and interference challenges in simultaneous transmission and reception by developing analogue-domain cancellation and evaluating STR protocols and cellular applications. It reports pre-LNA echo suppression, throughput improvements in single- and multicell CSMA networks, and methods for reducing BS-BS and UE-UE interference.

  • Problem

    Strong transmit echoes can overwhelm the receive chain, while STR cellular operation introduces BS-BS and UE-UE interference that can degrade performance.

  • Method

    The paper develops passband analogue echo cancellation using delayed replicas and Hilbert transforms, then studies STR protocols in CSMA networks and interference-reduction methods for cellular systems.

  • Results

    The technique provides sufficient pre-LNA echo suppression and is robust to RF impairments without additional antennas; STR protocols improve throughput, while proposed methods reduce BS-BS and UE-UE interference.

  • Takeaways & Limitations

    Analogue cancellation can make STR practical at the receive chain, but CSMA and cellular deployments require protocol and interference-management techniques.

Abstract

from arXiv · show

Full Duplex or Simultaneous transmission and reception (STR) in the same frequency at the same time can potentially double the physical layer capacity. However, high power transmit signal will appear at receive chain as echoes with powers much higher than the desired received signal. Therefore, in order to achieve the potential gain, it is imperative to cancel these echoes. As these high power echoes can saturate low noise amplifier (LNA) and also digital domain echo cancellation requires unrealistically high resolution analog-to-digital converter (ADC), the echoes should be cancelled or suppressed sufficiently before LNA. In this paper we present a closed-loop echo cancellation technique which can be implemented purely in analogue domain. The advantages of our method are multiple-fold: it is robust to phase noise, does not require additional set of antennas, can be applied to wideband signals and the performance is irrelevant to radio frequency (RF) impairments in transmit chain. Next, we study a few protocols for STR systems in carrier sense multiple access (CSMA) network and investigate MAC level throughput with realistic assumptions in both single cell and multiple cells. We show that STR can reduce hidden node problem in CSMA network and produce gains of up to 279% in maximum throughput in such networks. Finally, we investigate the application of STR in cellular systems and study two new unique interferences introduced to the system due to STR, namely BS-BS interference and UE-UE interference. We show that these two new interferences will hugely degrade system performance if not treated appropriately. We propose novel methods to reduce both interferences and investigate the performances in system level.

I. INTRODUCTION

STR can improve capacity and network operation, but practical deployment requires analogue-domain echo cancellation and careful treatment of cellular interference.

  • STR benefits: STR can potentially double capacity by transmitting and receiving simultaneously in the same frequency and time.It also supports hidden-node reduction, lower multihop delay, cognitive-radio sensing, device discovery, and interference cancellation among co-located radios.
  • Implementation challenge: High-power leakage and reflected signals can overwhelm the receive chain, making echo suppression before the LNA imperative.A 46 dBm transmit output with 25 dB circulator isolation produces a 21 dBm echo before additional antenna-mismatch and environmental echoes.
  • Prior approaches: Existing analogue cancellation approaches are limited by narrowband operation, additional antennas, or open-loop sensitivity to impairments.Reported prior suppression includes about 20 dB for 5 MHz signals, 37 dB with separate antennas, 50 dB in a hardware experiment, and 31–32 dB or 24 dB in other pre-LNA approaches.
  • Proposed method: The proposed adaptive closed-loop analogue technique supports wideband signals without additional antennas and is scalable to MIMO systems.The authors motivate closed-loop operation as more robust to impairments than open-loop techniques.
  • Network applications: In cellular STR, BS-BS and UE-UE interference introduce distinct challenges, whereas CSMA networks can use STR to reduce hidden nodes without these interferences.The paper proposes BS-BS cancellation and a non-cooperative method for reducing UE-UE interference, while studying STR protocols in single- and multicell CSMA networks.

II. ECHO CANCELLATION IN ANALOGUE DOMAIN

The paper models passband echo cancellation with delayed signal replicas and derives analogue-domain design insights using Wiener and MMSE formulations.

  • System model: The receive signal is modeled as an echoed transmitted signal plus the desired signal and additive white Gaussian noise.The echo has unknown gain and delay, and cancellation is performed by estimating it from delayed replicas of the transmitted signal.
  • System model: The canceller estimates the echo with delayed replicas and Hilbert-transform components, then subtracts the estimate from the corrupted receive signal.Two taps use x(t−τ1), x(t−τ2), and their Hilbert transforms to provide phase rotation and scaling.
  • Wiener solution: Minimizing canceller-output power provides an MMSE design objective because the desired signal and noise are uncorrelated with transmitted replicas.The Wiener solution is used for performance and design analysis even though direct analogue implementation is difficult.
  • Wiener solution: Smaller normalized tap-delay differences and more taps improve suppression, with nearly 90 dB achievable using two taps at a normalized difference of 0.01.The cited condition corresponds to a 1 ns delay difference in a 10 MHz system; suppression tends toward infinity as tap-delay difference approaches zero.
  • Multiple echoes: Echoes closer to a tap are easier to estimate, making a single echo midway between neighboring taps the worst-case simulation condition.The discussion extends this observation to arbitrary numbers of echoes with arbitrary powers between neighboring taps.
  • Multiple echoes: With equal total echo power, residual echo power decreases as a second echo moves toward the second tap.Fig. 2 illustrates the two-echo comparison described by the analysis.

B. Adaptive Echo Cancellation in Analogue

The adaptive analogue canceller compensates phase-shifter imbalance with multiple phase shifters and updates its weights using a low-pass-filtered steepest-descent procedure.

  • Phase-shifter implementation: Multiple phase shifters per tap allow delayed transmitted replicas to be rotated and scaled despite phase imbalance.With M = 3, the paper states that phase imbalance below 30° keeps the three phases nonidentical.
  • Phase-shifter implementation: The analogue canceller forms a weighted combination of phase-shifter outputs and represents each tap weight with a real-valued coefficient.The resulting complex-baseband model uses the phase-shifter outputs as the cancellation basis.
  • Hardware structure: The proposed block diagram uses variable attenuators and fixed attenuation to implement the weights and small step-size in analogue hardware.The authors conclude that the overall structure is implementable.
  • Adaptive update: Steepest descent updates the weights from the cost function, with the step-size controlling the adaptation rate.The approach is selected because direct analogue implementation of the Wiener filter is difficult or impractical.
  • Adaptive update: Low-pass integration filters noisy channel estimates before they update the analogue weights.The noise arises from the random transmitted signal, receiver noise, and hardware impairments.

C. Impairments

The analogue canceller uses impairment-aware adaptive processing, with convergence preserved despite phase, amplitude, attenuator, downconverter, and transmit-chain impairments.

  • Phase and attenuator impairments: The adaptive technique remains convergent when phase differences between downconverters stay below 45° and attenuator phase distortion is absorbed into those differences.The phase difference is defined between downconverters for Z(t) and Xk,m(t).
  • Receiver impairments: The method trains weights on impaired downconverter outputs without attempting explicit compensation for unknown phase and amplitude imbalances or DC offset.The study assumes zero DC offset achievable with a DC-blocking capacitor.
  • Noise and nonlinearity: The impairment analysis includes component noise and assumes a 0 dBm input to the variable attenuator to limit nonlinear distortion.The variable attenuator’s reported input third-order intercept point is 50 dBm.
  • Transmit-chain impairments: The adaptive analogue canceller is immune to transmit-chain impairments when signal bandwidth is unchanged, unlike digital cancellation that omits some RF impairments.If bandwidth changes, tap delay spacing must be adjusted to preserve the normalized delay difference.

D. Simulations

Simulations evaluate adaptive analogue cancellation with OFDM signals, noise, finite attenuation range, transmit noise, and combined analogue-digital processing. Results show strong ideal suppression, slower convergence and higher floors under noise, and residual echo below thermal noise with joint cancellation.

  • Simulation setup: The simulation uses 10 MHz random OFDM signals with 102 active subcarriers, QPSK symbols, 20 dB-above-thermal transmit noise, two echoes, and a three-tap canceller.Three taps can cancel echoes with at least 5 ns delay spread.
  • Analogue cancellation: More than 110 dB of suppression is possible with unlimited attenuator range and no noise, approaching the Wiener-filter solution.The simulated residual in-band echo is far below the -104 dBm thermal-noise power.
  • Analogue cancellation: Smaller normalized tap-delay differences improve suppression but slow convergence because the adaptive filter’s input correlation matrix has greater eigenvalue spread.This trade-off is demonstrated in the Wiener-filter and analogue-cancellation analysis.
  • Noise and dynamic range: Noise causes slower convergence and a higher residual floor, while a 35 dB attenuator range produces noticeable fluctuation during weight sign changes.Reducing the step size allows residual echo power below -90 dBm at 1.5 × 10^5 symbols.
  • Spectrum: Transmit noise above thermal noise can also be cancelled, although out-of-band reduction is weaker because the cost function is dominated by high-power in-band components.The reported residual echo power is -90 dBm, while nearby out-of-band power reaches thermal noise.
  • Combined cancellation: Joint analogue and digital cancellation produces residual echo power below thermal noise by disabling analogue weight updates after the residual reaches a threshold.Digital cancellation estimates and subtracts echo separately on each OFDM subcarrier.

E. Phase noise and Time varying channel

The analogue echo canceller remains robust to phase noise and tracks severe time-varying echo channels, although unpredictable channel variation makes higher residual echo power unavoidable.

  • Phase noise: White phase noise is not a limiting factor for convergence under the simulated conditions.The assumed phase noise has 2-degree rms, and convergence is nearly identical to the noise-free case.
  • Phase noise: Low-pass filtering suppresses phase-noise variation in echo-channel estimation, preventing phase noise from directly appearing in the adaptive weights.The resulting lower echo-estimation power makes baseband noise more prominent, rather than directly destabilizing the weights.
  • Time varying channel: Higher residual echo power is unavoidable because time-varying channels cannot be predicted and the second echo power increases.When the second residual echo becomes dominant, fixed supersession levels increase total residual power; a larger step-size enables faster tracking.
  • Time varying channel: With a 45 dB attenuation range, the adaptive technique converges well even in the severe time-varying-channel simulation.The wider attenuation range avoids glitches caused by limited-range sign changes and improves tracking smoothness.

F. Discussions

The discussion positions the proposed analogue canceller as a wideband, impairment-tolerant design whose practical performance depends on tap configuration, convergence, and implementation complexity.

  • Tap configuration: Smaller normalized tap-delay differences and more taps provide better echo suppression, with nearly 90 dB achievable using two taps.Tap count should cover the delay spread and keep neighboring normalized delay differences sufficiently small.
  • Wideband operation: The canceller minimizes echo power over the whole band rather than creating a null at one frequency.Consequently, out-of-band residual echo power is not higher than in-band residual echo power.
  • Convergence: Residual echo powers reach -70 ∼-77 dBm within 20 ∼30 symbols, or 0.23 ∼0.35 msec.An additional 10 ∼20 dB suppression requires 1 ∼2 ×10^5 symbols; variable step-size can accelerate convergence.
  • Implementation scope: The analogue canceller remains feasible with RF impairments, but MIMO extension can increase cancellation complexity exponentially because of cross-talk from adjacent transmit antennas.Digital baseband cancellation may supplement the analogue method when baseband noise is high or attenuator dynamic range is limited.

III. MAC LEVEL SIMULATIONS IN CSMA NETWORKS

STR protocols are evaluated in single-cell and multiple-cell CSMA networks under realistic packet arrivals and sensing delays, showing hidden-node benefits alongside delay-related collision limits.

  • Motivation and assumptions: STR can completely resolve the single-cell hidden-node problem by sending a dummy packet when a node receives a packet while having one to transmit.The study assumes equal coverage radius and examines both single-cell and multiple-cell settings.
  • Limitations: Collisions caused by non-zero sensing or header-decoding delay h can overshadow STR’s throughput gain.This limitation does not apply to a single-cell scenario with only one active terminal.
  • STR protocols: i-STR uses a dummy packet as an implicit ACK, reducing feedback delay to h instead of at least one packet duration for a traditional ACK.The protocol is introduced to improve collision avoidance and release the channel earlier when collisions are detected.
  • Simulation setup: The simulations use non-slotted nonpersistent CSMA, Poisson packet arrivals, constant packet length, zero propagation delay, and randomly distributed terminals.The zero-delay assumption isolates hidden-node effects from contention increases caused by propagation delay.
  • Multiple cells: A seven-cell layout shows that neighboring cells increase collisions and create exposed-node effects, making throughput decline unavoidable.The central cell is evaluated with six surrounding cells whose centers are separated by 2r.
  • Results: With h = 0%, s-STR and i-STR match ideal CSMA throughput because their dummy packets eliminate hidden-node collisions.i-STR outperforms s-STR in both single-cell and multiple-cell cases by releasing the channel earlier after detected collisions.
  • STR protocols: d-STR supports asynchronous packet arrivals by allowing a newly arriving information packet to overwrite the dummy packet.Its ideal throughput approaches 2 as offered load G tends to infinity when downlink probability p is neither 0 nor 1.
  • Results: In a single cell, d-STR achieves ideal throughput at h = 0% and incurs negligible loss when h > 0%.In multiple cells, neighboring-cell packet holding and collisions alter throughput, while reducing b to 25% lowers simultaneous-transmission opportunities.

IV. SYSTEM LEVEL SIMULATIONS IN CELLULAR SYSTEMS

The cellular evaluation models BS-BS and UE-UE interference under 3GPP-inspired conditions and applies elevation-angle null forming to mitigate BS-BS interference. Null forming can bring STR capacity close to twice non-STR capacity, although signal-power loss remains at cell edges.

  • Simulation assumptions: The simulations use a 19-site wrap-around hexagonal grid with three sectors per site and blockwise flat fading.BS-UE links follow a 3GPP path-loss model with 20 dB penetration loss; BS-BS links use a dual-slope LoS model.
  • BS-BS interference reduction: BS-BS interference can overwhelm weak uplink signals, so the design splits each 4λ antenna into eight vertically stacked 1/2λ elements.The resulting configuration is a two-dimensional 4 × 8 array with unchanged effective antenna size and gain.
  • BS-BS interference reduction: MMSE null forming maintains the main beam toward θ0 while creating nulls toward selected angles, with ε controlling null depth.The vertical weight vector is normalized so that no element magnitude exceeds 1.
  • BS-BS interference reduction: Wide nulls from 89° to 91° accommodate neighboring base-station height variation, while applying null forming to both transmit and receive beams relaxes null-depth requirements.The normalization is practical for transmit beamforming, though some transmit power loss is expected.
  • System-level results: Null forming produces some cell-edge signal-power loss, but the loss can be offset by reducing cell size or increasing antenna elements.The beam-pattern evaluation shifts the null to 90° and compares null width with small- and large-cell depth targets.
  • System-level results: With null forming, the STR capacity CDF approaches twice the non-STR capacity in the large cell.The large cell is evaluated because BS-BS interference is particularly harmful there, especially at the cell edge.

C. UE-UE Interference

UE-UE interference depends on relative path loss and can dominate in small cells, where non-cooperative resource scheduling reduces its impact by trading some uplink capacity for downlink gains. In large cells, STR with null forming can approach or exceed twice non-STR capacity.

  • Interference conditions: UE-UE interference is more severe than BS-UE interference in the small cell but weaker in the large cell for 25% < p < 99%.In the large cell, BS-UE interference is more dominant and UE-UE interference can be ignored in the reported evaluation.
  • Interference conditions: At p = 25% and 115 m, the UE-UE path loss is 75 dB versus 113 dB for BS-UE, illustrating why proximity can make UE-UE interference dominant.The small-cell radius is 70.6 m, and UE-UE links can remain LoS over relevant distances.
  • Interference conditions: The study varies p because correlated BS-UE and UE-UE channel models were unavailable for representing different propagation environments.The simulations therefore use multiple p values to emulate environments with different UE-UE LoS likelihoods.
  • Non-cooperative mitigation: Coordination could schedule strongly interfering UE pairs on orthogonal resources, but the proposed approach avoids coordination and instead turns off selected uplink transmissions.This design improves downlink capacity at the cost of uplink capacity.
  • Non-cooperative mitigation: The resource-block structure divides time into DL-centric and UL-centric portions, selecting STR users whose additional uplink gain exceeds the resulting downlink loss.Scheduled UEs send uplink pilots so interference-inclusive downlink SINR can be estimated before data transmission.
  • Non-cooperative mitigation: The optimization maximizes the sum of downlink and uplink gains for both cell-edge and cell-average capacity.The objective explicitly combines four gains: DL edge, UL edge, DL average, and UL average.
  • Results: More than 100% DL cell-average capacity gain is achieved in the large cell with STR and null forming, while uplink gain is smaller.The gains beyond 100% arise from reduced DL co-channel interference; cell-edge capacity remains below twice non-STR capacity.
  • Results: In small cells, addressing UE-UE interference yields positive STR gains with 12 antennas by sacrificing uplink capacity.Without mitigation, cell-edge users experience serious downlink-capacity loss.

V. CONCLUSIONS

The paper concludes that analogue echo cancellation enables STR, STR protocols improve CSMA throughput and hidden-node behavior, and cellular STR requires separate treatment of BS-BS and UE-UE interference.

  • Conclusions: The proposed analogue echo-cancellation technique suppresses echoes before the LNA and remains robust to RF impairments without additional antennas.The technique is presented as a closed-loop analogue-domain method.
  • Conclusions: STR can reduce hidden-node problems and improve throughput in single- and multiple-cell CSMA networks.For cellular systems, the paper identifies BS-BS and UE-UE interference and provides a complete BS-BS cancellation solution plus a non-cooperative UE-UE study.

APPENDIX A

The appendix defines correlation quantities and derives the Wiener-based echo-cancellation expressions, including residual echo power and suppression level.

  • Correlation definitions: Normalized correlation is introduced using the average power of X(t), followed by a normalized autocorrelation definition.The appendix then forms cross-correlation and autocorrelation quantities for the cancellation derivation.
  • Wiener solution: The derivation expresses the cross-correlation vector through a diagonal matrix and likewise constructs the autocorrelation matrix.These quantities feed the Wiener solution for the echo-cancellation weights.
  • Suppression calculation: The final expressions determine residual echo power and the resulting echo-power suppression level.The appendix connects the residual-power denominator to the reported suppression calculation.

APPENDIX B

Appendix B develops an analogue adaptive echo-channel estimator and examines how signal, phase-noise, and receiver impairments affect convergence. Low-pass integration stabilizes noisy estimation, while shared-oscillator phase noise can remove the associated scale loss.

  • The ideal weight converges toward the in-phase component |H|cos(θh) of the echo-channel response.Here H = |H|e^jθh denotes the echo channel response.
  • The raw echo-channel estimate is inaccurate unless normalized by |X(t)|^2 and is further corrupted by noise, impairments, and OFDM’s high PAPR.A small-step-size integrator low-pass filters this noisy estimate before it directly affects the adaptive weight.
  • The estimator uses downconverted transmitted and received signals to update the in-phase cancellation weight according to the sign of Re{X*(t)Z(t)}.The update moves in the correct direction when this sign agrees with the sign of |H|cos(θh).
  • Phase noise and the transmitted signal’s random phase cause the echo-channel estimate to fluctuate, but low-pass integration averages these time-varying effects.The analysis assumes stationary phase noise and transmitted-signal phase when interpreting integration as an ensemble average.
  • With a common oscillator driving all downconverters, the phase-noise difference becomes zero and the scale factor e^-σ2 disappears.With separate phase noise, the scale reduction makes performance more sensitive to noise.
Loading 1309.5546v1…