Peak-to-average power ratio reduction versus digital pre-distortion in OFDM based systems Charles Nader ∗†‡ , Per Niklas Landin ∗†‡ , Wendy Van Moer ‡ , Niclas Björsell ∗ , Peter Händel † , Magnus Isaksson ∗ ∗ Center for RF Measurement Technology, University of Gävle, Gävle, SE-801 76, Sweden † Signal Processing Lab, Royal Institute of Technology, Stockholm, SE-100 44, Sweden ‡ Department ELEC, Vrije Universiteit Brussel, Brussels, B-1050, Belgium Email: charles.nader@hig.se Abstract—In this paper we evaluate the effect of applying peak-to-average power ratio (PAPR) reduction and digital pre- distortion (DPD) on radio frequency power ampliﬁers when an orthogonal frequency division multiplexing (OFDM) signal is used. The PAPR reduction method, presented as a convex problem, is based on reshaping the time domain signal by redistributing its energy in the frequency domain, with respect to constraints on in-band and out-of-band errors. The DPD method consists of modeling the behavior of the power ampliﬁer using a parallel Hammerstein model, and then extracting its inverse parameters based on the indirect learning approach. The cases where PAPR and DPD are applied separately and combined, are studied and investigated. Power ampliﬁer ﬁgures of merit are evaluated. A good performance is shown when combining both pre-processing techniques up to a certain operating point where DPD performance deteriorates. Solutions to improve the DPD performance at strong compression are suggested. Index Terms—OFDM, power ampliﬁers, peak-to-average power reduction, digital pre-distortion, green radio. I. I NTRODUCTION Orthogonal frequency division multiplexing (OFDM) is widely used in today’s wireless systems due to its high bandwidth efﬁciency and robustness against frequency fad- ing. However, applying such modulated signals to a radio frequency (RF) link presents a major challenge because these signals are characterized by high peak-to-average power ratios (PAPR). Hence, the RF base station power ampliﬁer (PA) needs to be operated at lower power level compared to its saturation point in order to avoid compression and hence clipping of the signals peaks, which generates in-band and out- of-band distortions. However, backing-off the PA’s operating power by a number of dB’s proportional to the signals PAPR will reduce the efﬁciency of the ampliﬁer as a large amount of supplied power will be dissipated as heat. Hence, a trade-off exists between nonlinearity and efﬁciency. RF system designers strive to overcome such trade-off by correcting for its causes digitally. Some of them target to reduce the nonlinearity effect by digitally pre-distorting (DPD) the signals [1]; while others aim to reduce the PAPR, hence the back-off margin [2]-[5]. In general, PAPR reduction and DPD are applied separately. Only a few researchers have experimentally studied their overall impact on the PA power performance when combined together, e.g. [6]. Hence, a high interest arises on evaluating such combined application, with focus on the latest state-of- art PAPR reduction technique, as a substantial improvement in the linear behavior and power efﬁciency of the PA can be achieved. This paper evaluates the effects of combining signal shaping for PAPR reduction and DPD on the overall performance of the PA. Do PAPR reduction and DPD complement each other’s performance? What is to improve? Is there an optimum strategy to combine those digital techniques for a maximum performance improvement? Answers will be given through the evaluation process. The PAPR reduction technique adopted is based on [4], [5], and described in Section II. The structure of the DPD with memory, along with its parameters identiﬁcation is presented in Section III. The excitation signal, device under test, and measurement setup are presented in Section IV. Power am- pliﬁer performance is evaluated in Section V for the cases where PAPR reduction and DPD are applied separately, and combined together. Suggestions on how to optimize the PA performance when combining PAPR reduction and DPD is given in Section VI. Conclusions are drawn in Section VII. II. PAPR REDUCTION In this section, OFDM PAPR reduction using convex opti- mization, as formulated in [4] and [5], is brieﬂy reviewed. The optimization method is a novel art in shaping signals that reduces the peaks by redistributing the power spectrum, while satisfying constraints on in-band errors and out-of-band emissions. Such PAPR reduction technique is advantageous compared to other methods, e.g. clipping and ﬁltering or coding, due to its low out-of-band emission and data rate reduction. In the following, the PAPR minimization problem is brieﬂy described. A WLAN OFDM signal with spectrum formed by 48 data, 4 pilot and 12 power-free sub-carriers is used [7]. The data/pilot sub-carriers are modulated with 16-quadrature amplitude mod- ulation (QAM). In addition, 128 symbols are considered. Consider c 0 =( 0,1 ,..., 0, )  a complex-valued vector of length  to be the ideal frequency constellation. During the minimization process, c 0 will be modiﬁed by a factor  ∈ ℂ  such that c = c 0 + , with c =( 1 ,...,  )  . Hence, its 978-1-61284-757-3/11/$26.00 C2011 IEEE