A Linearization Strategy for Undersampling Analog-to-Digital Converters

The ability of high performance Radar and Broadband Systems to detect weak targets in presence of strong interferers or clutter is given by their Spurious Free Dynamic Range (SFDR). Although the Signal-to- Noise-Ratio (SNR) necessary for detection may be improved by well-known system processing gains, the dynamic range is ultimately limited by distortion terms caused by nonlinear behaviour of receiver components. The Software Defined Radio (SDR) paradigm assigns the Analog-to-Digital Converter a key role in receiver design. For systems using IF- Subsampling, linearity requirements place a heavy burden on the ADC, as SFDR signifcantly degrades with increasing input frequency. As a consequence, the ADC can only be used at input frequencies fairly below its intrinsic full power bandwidth, restricting the systems IF placement. This contribution discusses the possibility of processing ADC output data in the digital domain to achieve improved linearity. The Volterra series approach of nonlinear systems and its constrained variants are discussed. We will show in detail that for higher input frequencies, dynamic errors cause the harmonic terms to loose their in-phase ability; in higher Nyquist zones a frequency-dependend dynamic phase error has to be considered. Assumptions are backed by an evaluation of coherent data from the LTC2208 (16 Bit, 120 MSPS). A specific correction algorithm incorporating the dynamic phase error will be presented, which yielded 25 dB SFDR improvement in the 7th Nyquist Zone (360-420 MHz). The reproducibility of correction results is considered in some detail.