Author: Dr. Younes Jabrane, École Nationale des Sciences Appliquées (ENSA) de Marrakech

Orthogonal frequency-division multiplexing (OFDM) is one of the most widely used modulation techniques in telecommunication technologies. Its strong orthogonality properties make it resilient to inter-symbol interference. Nevertheless, OFDM signals suffer from a major problem known as peak to average power ratio (PAPR), which causes distortion in high power amplifiers (HPA). Several studies have been conducted in reducing the PAPR

[1,2,3]. Active constellation extension (ACE-AGP) is one method that has achieved good results [4]. On the other hand, artificial intelligence, based on neural networks (RN), has become very useful in the approximation of phenomena with a known evolution.

This allowed us to use neural networks to create a system of PAPR reduction in OFDM signals (Temporal architecture and temporal frequency architecture), while relying on algorithms ACE-AGP [5]. Indeed, the inputs and outputs of this algorithm will be taken as a learning data RN.

Our first proposal to reduce the OFDM signals envelope fluctuations is performed in the time domain. We train our ANN by using the signals with low envelope fluctuations obtained by the ACE-AGP algorithm. This way, this ANN learns what are the characteristics for a signal with low envelope fluctuations.

Unfortunately, the main problem with this time domain training scheme is that the neural network is not able to learn which regions of the constellation are allowed and which ones are forbidden. Thus, a second neural network, working in the frequency domain, is proposed to be concatenated to the time-domain scheme.

The performance of the developed system, in term of PAPR reduction, constellations and BER, is important and its complexity is negligible in comparison with other methods. The results are presented in the following figures, which show a comparison of the actual time complexity between these two approaches [5].


1st case: QPSK

1st case: QPSK

Cubic Metric Comparison

BER Comparison. N = 1024

2nd case: 16-QAM

2nd case: 16-QAM

Cubic Metric Comparison

BER Comparison. N = 1024

Our development and prototyping platform

Software-defined radio (SDR) is becoming the standard rapid prototyping platform for scientists, researchers, and R&D engineers to use when performing proof-of-concepts for new algorithms. As such, we implemented our solution (ACE-AGP and neural networks) on a Nutaq SDR platform.

SDR platforms typically consist of an FPGA board on which the developed algorithm can be programmed, tested, and validated. Good software tools can drastically reduce the development time. The Nutaq SDR platform provided us with the ability to choose between two approaches when programming: VHDL coding or model-based design tools for automatic code generation. IP cores that manage the on-board peripherals like storage and communication links are provided for both approaches.

The Nutaq BSDK consists of a VHDL development environment and the Nutaq MBDK. It is fully integrated with Simulink and Xilinx System Generator and forms a model-based development environment. It allowed us to rapidly prototype the algorithm and target the hardware. Moreover, the Nutaq SDR platform is equipped with configurable state-of-the-art RF transceivers and receivers to enable the up/down conversion of the baseband signals to/from the RF domain for transmission and reception. The developed algorithm could then be tested and its performance benchmarked in a real-world environment (rather than in a simulated environment with test data). The result is a tested and validated ACE-AGP algorithm in the form of an FPGA design and a synthesized netlist. It can then be reused on other FPGA hardware or even serve in the design of an integrated circuit (IC).


[1] M. Breiling, S. H. Muller-Weinfurtner, and J. B. Huber, "SLM peakpower reduction without explicit side information," IEEE Commun. Lett., vol. 5, no. 6, pp. 239–241, June 2001.

[2] J.-C. Chen, "Partial transmit sequences for peak-to-average power ratio reduction of OFDM signals with the cross-entropy method," IEEE Signal Process. Lett., vol. 16, no. 6, pp. 545–548, June 2009.

[3] S. H. Han and J. H. Lee, "An overview of peak-to-average power ratio reduction techniques for multicarrier transmission," IEEE Wireless Commun., vol. 12, no. 2, pp. 56-65, Apr. 2005.

[4] B. S. Krongold and D. L. Jones, "PAR reduction in OFDM via active constellation extention," IEEE Trans. Broadcast., vol. 49, no. 3, pp.258–268, Sep. 2003.

[5] Y. Jabrane, V. P. G. Jiménez, A. G. Armada, B. A. E. Said, and A. A. Ouahman. Reduction of power envelope fluctuations in OFDM signals by using neural networks. IEEE Commun. Lett., 14 (7): 599-601, Jul. 2010


Biotel is the official distributor of Nutaq’s Software Defined Radio platforms in Morocco.