Abstract
Transmission of complex-valued symbols using filter bank multicarrier systems has been an issue due to the self-interference between the transmitted symbols both in the time and frequency domain (so-called intrinsic interference). In this paper, we propose a novel low-complexity interference-free filter bank multicarrier system with QAM modulation (FBMC/QAM) using filter deconvolution. The proposed method is based on inversion of the prototype filters which completely removes the intrinsic interference at the receiver and allows the use of complex-valued signaling. The interference terms in FBMC/QAM with and without the proposed system are analyzed and compared in terms of mean square error (MSE). It is shown with theoretical and simulation results that the proposed method cancels the intrinsic interference and improves the output signal to interference plus noise ratio (SINR) at the expense of slight enhancement of residual interferences caused by multipath channel. The complexity of the proposed system is also analyzed along with performance evaluation in an asynchronous multiservice scenario. It is shown that the proposed FBMC/QAM system with filter deconvolution outperforms the conventional OFDM system.
Introduction
Orthogonal Frequency Division Multiplexing (OFDM) is the current standard for wireless communication, but it suffers from high out-of-band radiation (OoBR). This spectral leakage makes it difficult to efficiently pack diverse services—such as low-latency IoT and high-speed mobile broadband—side-by-side in the spectrum.
Filter Bank Multicarrier (FBMC) systems address this by applying high-quality filters to each subcarrier. However, the Balian-Low theorem dictates that well-localized filters cannot simultaneously achieve perfect orthogonality and Nyquist-rate transmission for complex symbols. Consequently, most FBMC systems rely on Offset-QAM (OQAM) and real-valued signaling to avoid "intrinsic interference." This paper proposes a novel "Type II" FBMC/QAM system that utilizes filter deconvolution to enable standard complex-valued signaling while eliminating self-interference.
Proposed Filter Deconvolution Method
The core innovation is the introduction of an inverse filter matrix at the receiver. By modeling the entire transceiver chain—including prototype filters and the multipath channel—within a matrix framework, the authors derive an inverse matrix $R$ based on the autocorrelation of the transmitter's filters.
Applying this inverse filter at the receiver effectively "undoes" the overlapping effects of the filter bank. This deconvolution step pushes intrinsic interference (ISI and ICI) down to negligible levels (from -13 dB to nearly -310 dB in simulations), allowing the FBMC system to behave as if it were perfectly orthogonal. Standard one-tap Zero Forcing (ZF) or Minimum Mean Square Error (MMSE) equalizers are then used to handle remaining wireless channel distortions.
Complexity and Matrix Structure
A critical advantage of this approach is its computational efficiency. The autocorrelation matrix $G$ is band-diagonal, meaning most of its entries are zero. This structural property is inherited by its inverse, $R$.
By exploiting this sparsity and ignoring negligible elements, the authors demonstrate that the receiver's complexity remains within the same $O(N \log_2 N)$ range as conventional OFDM. This makes the proposed FBMC/QAM system a viable candidate for real-time hardware implementation.
Key Results and Performance
The proposed system was evaluated against conventional OFDM and standard FBMC/QAM in simulated Rayleigh fading channels and asynchronous multi-user scenarios:
- SINR Improvement: While standard FBMC/QAM hit an interference floor at an MSE of -11.2 dB, the deconvolution method reached -31 dB, effectively eliminating the intrinsic interference barrier.
- Asynchronous Robustness: In scenarios mimicking 5G "grant-free" access with timing offsets up to 50%, the proposed system significantly outperformed OFDM. FBMC's tighter filters maintain signal integrity where OFDM's performance degrades rapidly.
- Noise Enhancement Trade-off: The inversion process introduces a slight noise enhancement factor of approximately 1.32x. However, the massive gains in SINR and the ability to use complex QAM symbols far outweigh this minor penalty.
Conclusion
By combining filter bank multicarrier techniques with a sparse inverse filter at the receiver, this paper presents a robust solution for complex-valued symbol transmission. The proposed FBMC/QAM system with filter deconvolution offers the spectral benefits of FBMC without the complexity of OQAM-based interference management, outperforming OFDM in the challenging asynchronous environments expected in future wireless networks.
Summary updated