Wireless Communications and Mobile Computing

Volume 2018, Article ID 8086073, 9 pages

https://doi.org/10.1155/2018/8086073

## A Novel Simulation Model for Nonstationary Rice Fading Channels

Correspondence should be addressed to Qiuming Zhu; nc.ude.aaun@gnimuiquhz

Received 28 September 2017; Revised 7 December 2017; Accepted 19 December 2017; Published 18 January 2018

Academic Editor: Carlos A Gutierrez

Copyright © 2018 Kaili Jiang et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

#### Abstract

In this paper, we propose a new simulator for nonstationary Rice fading channels under nonisotropic scattering scenarios, as well as the improved computation method of simulation parameters. The new simulator can also be applied on generating Rayleigh fading channels by adjusting parameters. The proposed simulator takes into account the smooth transition of fading phases between the adjacent channel states. The time-variant statistical properties of the proposed simulator, that is, the probability density functions (PDFs) of envelope and phase, autocorrelation function (ACF), and Doppler power spectrum density (DPSD), are also analyzed and derived. Simulation results have demonstrated that our proposed simulator provides good approximation on the statistical properties with the corresponding theoretical ones, which indicates its usefulness for the performance evaluation and validation of the wireless communication systems under nonstationary and nonisotropic scenarios.

#### 1. Introduction

The channel simulator has the ability to reproduce the statistical properties of propagation channels and has become a very important software-assisted tool for the performance evaluation of wireless communication systems. There are a number of research papers on designing accurate and efficient simulators [1–19]. For the good fitting with plan-wave propagation channels, the sum-of-sinusoids (SoS) [1, 2] or sum-of-cisoids (SoC) [3] based simulators and their derivatives [4–8] have gained the most widespread acceptance.

It should be highlighted that the traditional SoC simulators are only suitable to reproduce wide-sense stationary (WSS) fading channels, which means the statistical properties of output channels are time-invariant. However, measurement campaigns have proved that the WSS assumption is only valid for a short time interval [20]. Therefore, several modified SoC-based simulators were proposed in [9–19] to generate nonstationary fading channels. For example, the WINNER+ model [9] was simulated by generating several independent channel segments with different simulation parameters. Simulation parameters were updated according to the fixed trajectories of transceivers and clusters in [10, 11, 13, 14], which improves the continuity of adjacent channel states.

However, we have found that the generated channel phases of these simulators in [9–11, 13, 14] cannot guarantee a smooth transition between adjacent channel states, which makes the output Doppler frequency shifts not very accurate. To overcome this shortcoming, a new method, namely, sum of frequency modulation signals [12, 15–18] or sum of chirp signals [19], was proposed to simulate nonstationary Rayleigh channels very recently, but it lacked implementation details and performance analyses. To fill this gap, this paper develops a new simulator based on this idea to generate the nonstationary Rice fading channels under nonisotropic scattering scenarios, as well as the upgraded computation methods of simulation parameters. Moreover, the time-variant statistical properties of the proposed simulator, that is, the probability density functions (PDFs) of the envelope and phase, autocorrelation function (ACF), and Doppler power spectral density (DPSD), are derived in detail and also verified by numerical simulations.

The rest of this paper is organized as follows. In Section 2, we introduce the reference model and traditional SoC simulators for stationary fading channels. Section 3 proposes a new simulator for the nonstationary Rice channels, as well as computation methods of the time-variant simulation parameters. The theoretical results of PDF, ACF, and DPSD for the proposed simulator are derived in detail in Section 4. In Section 5, simulation results are given and compared with the corresponding derivation results. Finally, conclusions are drawn in Section 6.

#### 2. Reference and Simulation Models for Stationary Fading Channels

Under the nonstationary condition, the reference model for Rice fading channels can be written as [3]where and are the mean power and Rice factor, and the non-line-of-sight (NLOS) component denotes a normalized zero-mean complex Gaussian process. In (1), represents the line-of-sight (LOS) component, where and are the Doppler frequency and phase of the LOS component, respectively. From (1), we can easily obtain the reference model for Rayleigh fading channels by omitting the LOS component ( = 0).

The characteristics of reference model are determined by its first- and second-order statistical properties, that is, PDF, ACF, and DPSD. Given the envelope and phase , the reference PDFs can be expressed as [3]where and denote the modified Bessel function of the first kind of order zero and complementary error function, respectively. The ACF of is defined by , where and denote the statistical expectation and complex conjugation operator, respectively. Finally, the DPSD can be obtained by the Fourier transform of ACF as . For nonisotropic scattering scenarios, the von Mises (VM) distribution is widely accepted to describe the angle of arrival (AoA) [8] and is given bywhere controls the angular spread and denotes the mean value. Under the stationary VM scattering scenario, the references ACF and DPSD have been derived as [8]where denotes the maximum Doppler frequency and , , and are the movement speed, light speed, and carrier frequency, respectively.

The traditional SoC simulator can be expressed as [8]where is the number of simulation paths, , , , and are the time-invariant simulation parameters, mean Doppler frequencies, initial phases of the NLOS, and LOS component, respectively. Note that the frequency parameters and are determined by the ACF or DPSD of the reference model, while the phase and are generated randomly and uniformly over .

#### 3. A New Simulator for Nonstationary Fading Channels

Under the nonstationary condition, the statistical properties of channels would change over time [20], and this makes channel parameters time-variant. In order to apply the classic SoC method, some modified simulators [9–11, 13, 14] directly use and to substitute and in (7), respectively. However, it can be proved that the output phases of these models are not accurate and the output Doppler frequencies do not agree with the theoretical ones [18]. In this paper, we use and to replace and , respectively, and the new simulator can be written as

A key issue for the proposed simulator is to find a proper set of simulation parameters , which can guarantee that the output statistical properties are close to the desired ones. Several parameter computation methods have been proposed in the literature [8, 21–23], such as the Riemann sum method (RSM) [8], the extended method of exact Doppler spread (EMEDS) [21], the generalized method of equal areas (GMEA) [22], and the -norm method [23]. However, these methods are only suitable for the time-invariant parametric simulator. Therefore, an upgraded computation method is required to make the proposed simulator realizable.

Since the WSS assumption is still valid for the nonstationary channels during a short time interval [20], it is reasonable to assume that the statistical properties keep unchanged during a short time duration , namely, the updating interval of channel states. Thus, the parameters and within the th interval can be denoted as and , where () and means the integer part. The corresponding DPSD within the th interval can be defined bywhere and are the DPSDs of the NLOS component and LOS component, respectively, and can be obtained by (6) under the VM scattering scenario or measured under other realistic scenarios. Then, by using aforementioned methods [8, 21–23], the frequency parameters of the th interval, denoted as , can be calculated. Hence, the time-variant frequency parameters can be computed by and . Although this computation method is straightforward and has been applied to simulate WINNER+ model [9] and 3GPP-3D model [24], it does not consider the fact that channel properties in reality usually change smoothly along time. On the other hand, measurement campaigns [20] have shown that the stationary interval is very short, that is, 9 ms in 80% of the case and 20 ms in 60% of the case under the high speed train (HST) scenario. In this paper, we assume that the Doppler frequency changes linearly within the stationary interval, so can be calculated byTaking the fact into account that the simulations of MIMO channels often demand for generating multiple uncorrelated Rayleigh or Rice fading waveforms, the new calculation method of is given aswhere , , and denote the initial value, the slope, and the small random offset of the frequency parameter of the th path, respectively. is generated randomly and uniformly over , while holds the value at the end of the previous interval. The slope is calculated bywhere refers to the number of period within each interval, and the upper boundary and low boundary are used to keep the frequencies changing within valid ranges, which are defined byNote that the parameter computation method in [19] can be viewed as a special case of (11) with . To visualize and demonstrate the proposed computation method, let us consider the reference DPSDs of and under VM scattering environment as shown in Figure 1. The parameters are configured as follows, ms, , and . The frequencies and boundaries can be calculated by using (10)–(13) and are also shown in Figure 1. It is clear that the frequencies are continuous and change linearly within the upper and lower boundaries. The frequencies also include a small and random offset, which is to guarantee the generated multiple fading channels are uncorrelated.