Digital Compensation of Frequency Dependent Imperfections in Direct Conversion I-Q Modulators Antonio Cantoni University of Western Australia Perth, Australia cantoni@watri.org.au John Tuthill QRSciences Pty Ltd Perth, Australia jtuthill@qrsciences.com Abstract— In a direct-conversion transmitter system, the non- ideal transfer characteristics of analog subsystems in the I and Q channels adversely affect the performance of the system. The study of static errors, including gain imbalance, phase errors, and DC offsets has received considerable attention in the literature. Much less attention has been paid to the effect of frequency-dependent imperfections that arise in the analog reconstruction filters commonly employed in direct-conversion structures. This paper presents results of research into robust methods for automatic digital compensation for these non ideal frequency-dependent characteristics. The design and implementation of FIR and Eigenfilter structures to achieve compensation are presented. As a specific application the paper considers a direct-conversion system used to generate continuous-phase frequency-shift keying (CPFSK) signals. The performance achievable using the proposed digital compensation structures is presented, including measurement results taken on a DSP-based modulator platform. I. INTRODUCTION A flexible and cost effective approach that is widely used to implement the exciter section of a radio transmitter is to synthesize base-band inphase (I) and quardrature (Q) signals using a DSP (Digital Signal Processor) followed by a vector modulator to directly up-convert the base-band signals to RF as shown in Fig . 1. Figure 1. DSP based IQ Modulator In this paper we considered, specifically, IQ modulators for CPFSK (Continuous Phase Frequency Shift Keying). The CPFSK modulation format is of current interest, [1]–[4], and is attractive for a number of reasons, including its narrow radio frequency (RF) bandwidth requirements and its constant RF envelope The constant-envelope property allows a highly nonlinear but power-efficient RF power amplifier (PA) to be used in the output section of the transmitter, obviating the need for expensive linear PAs [1]. The reconstruction filters in the IQ modulator attenuate the unwanted image components in the base-band signal spectrum before transmission. In a practical IQ modulator implementation, the transfer characteristics of the reconstruction filters and errors in their implementation result in the pass-band characteristics departing from constant magnitude, linear phase. Furthermore, there will also be some mismatch between the I and Q channel filter frequency responses. These shortcomings may result in significant degradation of the performance of the transmitter system. In the case of CPFSK signals, these practical limitations result in the introduction of amplitude ripple in the envelope of the vector modulator output RF signal [5]. This produces undesirable side-lobes in the RF signal spectrum when the vector modulator output signal passes through the RF power amplifier (PA) [6]. Increases in side-lobe levels and out-of- band emissions can result in unacceptable interference with other adjacent frequency channels and the consequent failure to comply with transmission standards. Digital compensation for deficiencies in the analog subsystems of IQ modulators and demodulators has received considerable attention in the literature [6]–[22]. Much of this work [7]–[16] centers on static non-ideal characteristics in IQ modulators and demodulators, and a number of techniques have emerged that have been shown to be effective in overcoming the influence of static non-ideal characteristics on modulator performance. One digital-compensation scheme for compensating for the frequency dependent errors in the IQ modulator is shown in Fig. 2. The approach provides frequency dependent compensation for: 1) the frequency-dependent non-ideal characteristics in the analog reconstruction filters; 2) the magnitude characteristic of the zero-order-hold digital-to- analog (D/A) converters. The design of optimal compensation FIR filters for the frequency-dependent characteristics of the D/A D/A Vector modulator RF power amplifier f c I Q DSP Analogue reconstruction filters Bandpass filter Data Input LPF 1 LPF 2 LO 269 1-4244-0921-7/07 $25.00 © 2007 IEEE.