18 February 2013
Researchers at the University of Calgary in Canada have experimentally demonstrated distortions arising from dual-band concurrent operation and modulator imperfections in a transmitter. To address this, they have proposed a neural network based digital pre-distortion (DPD) scheme, which will mitigate the effect of the distortions and that of the hardware imperfections in a single step. As well as improving performance, this one step solution will lead to significant savings in the necessary digital signal processing.
The introduction of modern communication standards has motivated the design and implementation of modern dual-band power amplifiers (PA), especially in connection with base station transmitters, to provide savings in terms of both hardware and reconfigurability. However, maintaining optimum efficiency over the band is complicated in the case of PA design, especially when the bandwidth is intended to cover the two communication standards, which are far apart in terms of their carrier frequencies. For such cases, a dual-band transmitter provides an optimum solution over a limited range of bandwidth around these two frequencies.
Unfortunately, when both the frequency bands are used concurrently, cross-modulation products are generated. The distortion resulting from the cross-modulation products adds to the distortion resulting from the intermodulation products due to power amplifier nonlinearity, which increases in-band distortion leading to increased bit-error-rate and out-of-band distortions interfering with the signals of neighbour channels. Moreover, if the transmitter modulator is not perfect, the input RF signal develops gain and phase imbalances, and D.C. leakages, which adds to in-band distortions for direct conversion transmitters.
In order to overcome these issues (to achieve high linearity) DPD has been extensively used, due to its accuracy, flexibility and reconfigurable properties. The DPD technique involves the introduction of an element that is the exact inverse of a PA (predistorter) before the PA, which compensates for nonlinearity. It is, therefore, important to obtain an accurate inverse model of the wireless transmitter, capable of capturing all imperfections of the hardware. The Calgary group have proposed the use of neural networks for this inverse modelling.
For the transmitter under consideration, complex baseband input to complex baseband output modelling is required. To account for any phase imbalance and time skew between the real (I) and imaginary (Q) components of the baseband complex signal, it is preferable to use a model capable of modelling two real inputs and two real outputs, instead of a fully complex model. Dr. Meenakshi Rawat, one of the authors of the research, explained that “in the group’s example of dual band concurrent applications, the model complexity will increase to be able to model the relationship between at least four inputs and four outputs accounting for both the bands”. According to the mathematical theory of neural networks, the universal approximation theorem states that the standard multilayer feed-forward network containing a finite number of neurons in its hidden layer is a universal approximator among continuous functions - “This universal approximation is the motivation for using a feedforward MIMO neural network to model complex dual band concurrent transmitter” said Rawat.
Dr Rawat and her colleagues are associated with the iRadio Lab at the University of Calgary. The iRadio group is engaged in multi-disciplinary research projects, and currently the group is focussing on multi-band/broad-band transmitter and receiver designs; signal processing and RF design for MIMO radio systems, devices and circuits; systems behaviour modelling, all-digital and digitally-enhanced reconfigurable RF transceivers; and system integration using heterogeneous nanotechnologies (CMOS, GaN, SiGe, LTCC etc.).
Such focussed research facilities are increasing in importance, as Rawat explained: “Recently, the report from the U. S. President’s Council of advisors on Science and Technology indicated that spectrum should be shared between Federal agencies and commercial agencies by freeing up spectrum for commercial purposes at other times and places, respecting the needs of the Federal system”. She expanded by saying “moreover, from a technical and social perspective, the deployment of low-power low-cost wireless data communication systems operating in the license-free industrial, scientific and medical bands (ISM) is likely to expand dramatically in the near future”.
Because of this, there has been a strong trend towards research related to multiband and wideband device design for transmitters and receivers to accommodate many frequencies in one transceiver system. New designs of such devices and circuits, in turn, will require innovation in software algorithms and architectures. “Indeed, with interdisciplinary and co-dependent interests, the field of multiband, multi-standard and multi-mode transceiver system design and related signal processing have immense scope for future research” said Rawat. These multi-band concurrent radios are key components for a shared wireless infrastructure (the base station and repeaters). This disruptive technology will reduce the number of base stations and hence lower the overall energy consumption of wireless communication networks and ultimately their negative impact on the environment: electromagnetic pollution and the CO2 footprint.
This article is based on the Letter: Joint mitigation of nonlinearity and modulator imperfections in a dual-band concurrent transmitter using neural networks (new window).
The report from the U. S. President’s Council of advisors on Science and Technology: http://arstechnica.com/information-technology/2012/07/bold-plan-opening-1000-mhz-of-federal-spectrum-to-wifi-style-sharing/ (new window)
iRadio Lab: http://www.iradio.ucalgary.ca/ (new window)
A PDF version (new window) of this feature article is also available.
Browse or search all papers in the latest or past issues of Electronics Letters on the IET Digital Library.