[Paper Review] Out-of-Band Radiation Comparison of GFDM, WCP-COQAM and OFDM at Equal Spectral Efficiency

Ali Bulut Üçüncü|arXiv (Cornell University)|Jan 1, 2015

PAPR reduction in OFDM3 references3 citations

TL;DR

This paper fairly compares GFDM, WCP-COQAM, and OFDM in terms of out-of-band (OOB) radiation and carrier frequency offset (CFO) immunity at equal spectral efficiency, applying windowing and guard symbols to all three. It finds that, under equal spectral efficiency, OOB emissions are nearly identical across all three modulation schemes, undermining the claimed OOB advantage of GFDM and WCP-COQAM over OFDM when suppression techniques are uniformly applied.

ABSTRACT

GFDM and WCP-COQAM are amongst the candidate physical layer modulation formats to be used in 5G, whose claimed lower out-of-band (OOB) emissions are important with respect to cognitive radio based dynamic spectrum access solutions. In this study, we compare OFDM, GFDM and WCP-COQAM in terms of OOB emissions in a fair manner such that their spectral efficiencies are the same and OOB emission reduction techniques are applied to all of the modulation types. Analytical PSD expressions are also correlated with the simulation based OOB emission results. Maintaining the same spectral efficiency, carrier frequency offset immunities will also be compared.

Motivation & Objective

• To conduct a fair comparison of OOB emissions among OFDM, GFDM, and WCP-COQAM by equalizing their spectral efficiencies.
• To apply OOB suppression techniques—windowing and guard symbol insertion—uniformly across all three modulation schemes to ensure fairness.
• To evaluate CFO immunity under equal spectral efficiency conditions to validate the fairness of OOB comparisons.
• To derive analytical power spectral density (PSD) expressions and correlate them with simulation results for OOB emission analysis.

Proposed method

• Equating spectral efficiency by increasing the number of subcarriers in OFDM to match GFDM and WCP-COQAM, maintaining the same bandwidth.
• Using analytical PSD expressions derived from signal models to predict OOB emission levels, validated via Monte Carlo simulations.
• Applying windowing and guard symbol insertion to all three modulation types (OFDM, GFDM, WCP-COQAM) to suppress OOB emissions.
• Employing a static COST-207 hilly terrain channel model with perfect channel knowledge and single-tap zero-forcing equalization to assess SER vs. SNR under CFO.
• Using matched filter (MF), zero-forcing (ZF), and interference cancellation (MF-DSIC) receivers for GFDM to evaluate performance across receiver types.
• Comparing analytical AWGN performance of OFDM with simulated results to validate the model and simulation setup.

Experimental results

Research questions

• RQ1 Does GFDM or WCP-COQAM offer a significant OOB emission advantage over OFDM when spectral efficiency is equalized?
• RQ2 How do windowing and guard symbol insertion affect OOB emissions across OFDM, GFDM, and WCP-COQAM under equal spectral efficiency?
• RQ3 Is the CFO immunity of OFDM compromised when its subcarrier count is increased to match GFDM and WCP-COQAM in spectral efficiency?
• RQ4 To what extent do analytical PSD expressions predict the OOB emission behavior observed in simulations?

Key findings

• When spectral efficiency is equalized and suppression techniques are uniformly applied, the OOB emissions of OFDM, GFDM, and WCP-COQAM become nearly identical.
• The OOB emission advantage of GFDM and WCP-COQAM over OFDM—often cited in literature—disappears under fair comparison conditions with equal spectral efficiency and shared suppression methods.
• Analytical PSD expressions closely match simulation results, validating the model’s accuracy in predicting OOB behavior.
• Despite higher subcarrier count, OFDM does not exhibit worse CFO immunity than GFDM or WCP-COQAM under equal spectral efficiency, as ISI from CFO affects all schemes similarly.
• The SER performance of OFDM remains comparable to GFDM and WCP-COQAM under CFO, confirming the fairness of the comparison framework.
• The conclusion holds across different pulse shapes (RRC, Dirichlet, Gaussian), indicating robustness of the findings to pulse shape selection.