Digital distortion compensation for wideband direct digitization RF receiver

Abstract

This paper deals with the study of a new correction structure for the distortion compensation in direct digitization receivers. In that purpose, a modified post-distortion algorithm including an efficient adaptive filter is used. Performance have been demonstrated on measurements from a state of the art wideband TI-ADC and from an entire direct-digitization RF receiver. On the TI-ADC, with a two-tone input signal, the algorithm has improved the SFDR by ∼ 25dB. Besides, Besides, on the entire receiver, the algorithm has been able to achieve ∼ 16dB of improvement on the SFDR.

Frequency Identification of a Memory Polynomial Model for PA Modeling

“…We will be interested here in the modeling of the gain of the power amplifier under test (frequency response H (0) ), as well as the inter-modulation products of orders 3, 5 and 7 (IMD3, IMD5 and IMD7). These correspond respectively to H (1) , H (2) and H (3) frequency responses.…”

“…The most comprehensive one is the Volterra model [6], [7], and a significant part of the literature deals with the simplification of this model [8]. Mention may in particular be made of the models of Wiener, Hammerstein [9]- [11], Wiener-Hammerstein [3], memory polynomial [12], [13], or generalized memory polynomial [14]. Our approach is based on the memory polynomial model, for reasons of complexity and accuracy [8], but also because the structure of this model allows the design of the proposed identification method, as will be seen further.…”

“…In order to construct the frequency responses H (0) , H (1) , H (2) , and H (3) , of the filters modeling the amplifier, the coefficients a, ϕ − , and ϕ + are deduced from the Fourier transform of the reference signal, and the coefficients H k± , from the Fourier transform of the distorted signal. This means that two FFT have to be computed, one for the signal at the input of the PA, and one for the signal at its output.…”

“…In time domain, modeling filters impulse responses are finally obtained by IFFT of the frequency responses H (0) , H (1) , H (2) and H (3) . These filters h (k) , k ∈ 0; 3 , allow in the end to reproduce the behaviour of the amplifier.…”

“…The objective of power amplifier behavioral modeling is to model the distortion induced by amplification, in order to correct it, to linearize the component. This correction can be performed for example by pre-distortion or post-distortion techniques [1]- [3].…”