Implementation Aspects in DFT Modulated Filter Bank Transceivers for Cognitive Radio

Abstract

Discrete Fourier transform (DFT) modulated filter banks (FBs) are considered as strong tools used to implement both dynamic spectrum access and spectrum sensing in cognitive radio (CR) systems. High time-frequency (TF) resolution for spectral estimation and effective spectrum access with low complexity transceivers are the basic objectives in CR systems. However, the limitations of self-interference in DFT FBs as well as a primary user interference increase the overall transceiver complexity. In this paper, we design DFT modulated FBs which take into account the aforementioned contradicting requirements of high resolution capabilities, efficient spectrum access and affordable implementation effort for an additive white Gaussian channel. Four simple designs are presented and their performance are investigated and compared for a CR system with basic transmission parameters resembling those of IEEE 802.11g.

Post detection integration in DFT modulated filter bank transceivers for cognitive radio

“…There are different leakage metrics described in [5]. Here, we resort to a metric for measuring the energy of the pulse g outside a given TF window [n, f ] ∈ Ω T × Ω F based on the so-called Levin distribution [14] defined as L g (n, f ) =…”

“…The performance of the SSE increases with the time-frequency (TF) resolution of the employed spectrum estimation [2], [3], [4]. In addition, good resolution properties facilitate the spectrum access (SAC) in view of the limited TF resources to be used by SUs [5]. There are several approaches for optimal designs in the context of FB based SSE and SAC.…”

“…The latter typically prescribe to use a simple detector not requiring explicit channel estimation and may include bit-error rate (BER) thresholds to be achieved in scenarios with predetermined properties of the wireless channel. Thus, a different approach is followed in [5] where a critically sampled FB with optimum TF properties for SSE is designed to achieve the lower BER bound for SAC in an additive white Gaussian noise (AWGN) channel. The critical sampling ensures to have a perfect reconstruction property which is important from a channel capacity point of view and to avoid interference phenomena in SSE.…”

“…In view of the overall system complexity, the SSE and SAC are to be carried out using a common FB for both tasks. While the scheme can be easily extended to π/4-differential quaternary phase-shift keying (π/4-DQPSK) [8], delay dispersion in random fading channels has not been taken into account.…”

“…, g [L − 1]] with L ∈ N to be used in a critically sampled DFT FB which maximizes the pulse's TF resolution expressed by the Levin distribution of the energy of g subject to constraints resulting from system specifications. The latter contain, in particular, the achievable average BER which in turn depends on g. Therefore, the constrained optimization approach of [5], [8] based on analytic expressions of the BER as a function of g has to be extended to fading channels for which analytic BER expressions for arbitrary g with corresponding ISI and ICI contributions are not available.…”