Abstract

Indoor millimeter wave propagation at the frequency of 24 GHz is studied by experimental methods. Measurements are performed to obtain temporal and spatial channel model using a channel sounder and rotating antennas in a corridor. The measured impulse responses are processed to obtain compact channel model following Saleh-Valenzuela’s model. The responses are compared with those of 5.3 GHz for the same test sites. Angular spread of 24 GHz is found to be smaller than that of 5.3 GHz, while echoes of 24 GHz are found to be longer than those of 5.3 GHz.

1. Introduction

With increasingly large demand for mobile internet data, unused frequency bands are being explored and frequency bands with reduced demand are reallocated to accommodate new wireless communication services. Among those frequency bands, millimeter wave frequencies have been given much interest due to the availability of large chunk of bands. For the introduction of new services such as fifth-generation (5G) services; however, the properties of electromagnetic waves at those bands should be known in various environment to simulate and estimate the performance of the new services, which necessitates accurate channel modeling [1].

In contrast to the wave propagation at SHF frequencies, the scattering directions at millimeter wave frequencies have narrow angles due to short wavelengths. The received signal strengths change rapidly with viewing angles of antennas. Obstacles with the dimensions comparable to or larger than the wavelength induce scattering along the specular or incoming directions [2]. Indoor or outdoor environments which encircle the base stations and user terminals are composed of many facets generating complicated scattering. Although simple radio propagation models such as a free space model, a ground reflection model, and statistical models such as Hata’s or Okumura’s are useful, those models cannot give precise information on the millimeter wave band which varies rapidly with positions and directions of antennas [3].

In this paper, a procedure to obtain an indoor wave propagation model at millimeter wave band is presented. Measurement results are given at the frequency of 24 GHz, which are the impulse responses as a function of delay and angles. A channel sounder with chip rate of 300 MHz and transmitting/receiving RF front ends is built and used to obtain impulse responses [4]. The measured data of 24 GHz band are compared with that of 5.3 GHz. Modified Saleh-Valenzuela’s model [5] parameters are extracted which catch important features of impulse response by exponential functions. The obtained channel model enables measurement data to be processed such that receiving voltages can be retrieved as a function of time and angle. The measurement setup is given in Section 2, measured data is given in Section 3, and modified S-V model parameters are extracted in Section 4.

2. Measurement Setup

To obtain the impulse response of 24 GHz, a sliding correlator channel sounder is built which generates 9-bit maximal pseudorandom binary sequence as in Figure 1. The chip rate of the transmitter is 300 MHz and that of the receiver is 300.1 MHz. The pseudorandom noise (PN) generator on the transmitter spreads the 24 GHz continuous wave signals into wideband ones, and those on the receiver despread them. The difference of the chip rates makes the cross-correlated output on the receiver show channel impulse responses with elongated time scale, whose period is the inverse of the difference frequency. The number of bits is chosen to maximize signal-to-noise ratio on the receiver side. The transmitter and the receiver are synchronized with the external reference signal of 10 MHz which is not shown in the figure. A 300 MHz PN generator of the receiver synchronized with that of the transmitter and acts as trigger signals which indicate the start of impulse responses. The PLLs attached to the oscillators 300 MHz, 300.1 MHz, and 24 GHz are all fed by the same 10 MHz reference clock. A programmable logic device is used to facilitate the change of specifications.

The RF front ends of the channel sounders are configured to measure 24 GHz and 5.3 GHz impulse responses at many positions in the indoor environment as in Figure 2. Impulse responses are recorded as the receiver is moved from the nearby position of the transmitter to far places. Table 1 shows coordinates of positions of the transmitter and the receiver. The height of ceiling of the corridor from the floor is 2.57 m.

Figure 3 shows the photos of the test site, which is a corridor 60 m long and 2.5 m wide. Impulse responses are obtained with the positions of the receiver changed. The amplitude variations of the received signals along the receiver positions are measured separately by a spectrum analyzer due to the narrow dynamic range of the oscilloscope. To extract spatial and temporal responses, impulse responses are recorded for 36 arrival angles by rotating the receiving antenna by 10 degrees after each measurement at a given position. The angular measurements are performed at 13 positions as shown in Figure 3. The transmit power is 10 dBm, and total cable losses for the transmitter and the receiver are 10 dB.

The antennas used for 24 GHz are standard gain horns with 21 dBi gain and waveguide size WR42 as shown in Figure 4. The radiation pattern for the antennas is also shown. For the frequency of 5.3 GHz, antennas are replaced with patch arrays with the same peak gain whose radiation pattern is also shown.

3. Measured Results

The measured impulse responses are given in Figure 5 as a function of arrival angles and time delays. The time axis is adjusted to take into account the time dilation of the sliding correlator which is due to the chip rate difference between the PN generator of the transmitter and that of the receiver. The output time scale of the sliding correlator is elongated by the ratio of the chip rate to the difference of the chip rates. The colors show relative strengths of the received signals normalized by the maximum value at each receiver’s position for the range of 50 dB. Impulse responses at the frequencies of 24 GHz and 5.3 GHz are shown side by side to compare them. The reference direction for angles of arrival is defined in Figure 5(a). Although the center frequencies are different, the time delay of peak received signals is similar.

The measured impulse responses are compared with one another. They are sorts of bistatic radar images in that the time axis is proportional to the distances to obstacles in the corridor as shown in Figure 6 which is the impulse response at the position of RX1. During the first-round trip time for the transmitted signals (0~0.42 μs), the impulse response at the angle of 180 degree shows time history of scattered waves from the obstacles along the corridor. As the corridor is relatively long with narrow widths, it can be considered as a one-dimensional scattering problem. Large reflections or scatterings are observed from the walls at either end of the corridor, side walls of lockers, and steel doors of the rooms. The attenuation of the signal level through the wall is measured to be about 50 dB, which means that the scattering in the corridor can be assumed to be nearly one-dimensional scattering. The peak positions in the impulse responses can be identified with the scattering point in the corridor as shown in Figure 6(b), which helps to form impulse responses at other positions.

Figure 7 shows the variation of the signal strength as the receiver is moved from the position near the transmitter to the right end of the corridor. One received signal can be retrieved by the product of the signal strength and impulse responses in Figure 5.

4. Extracted Channel Model Parameters

Although the measurement data are important enough, their sizes are cumbersome to utilize in simulations for wireless communications. To alleviate the computational burden, curve fitting is used like Saleh-Valenzuela’s channel model [5, 6] to simplify the impulse responses which are composed of clusters of delta functions. Clusters of delta functions are approximated collectively by decaying exponential functions. Besides, the angular dependence of the impulse responses is also taken into account by modifying S-V model. The modified S-V model at one receiver position is as follows:where

The model parameter is th peak delay times, is the slope parameter prior to , and is the slope parameter after that time. Likewise, the model parameter is th azimuthal angle in which direction a peak power is observed. The coefficients represent relative strengths of the peak powers. The procedure to extract model parameters is summarized by the flow graph in Figure 8.

Following the procedure, channel parameters for the position of RX1 are extracted as shown in Figure 9 with the dynamic range of 25 dB. The responses of 24 GHz signals have longer echoes than those of 5.3 GHz in time in that higher order reflections with considerable strength can be observed. Although the beam widths of the 24 GHz and 5.3 GHz antennas are similar, the angular spread of the 5.3 GHz impulse response is larger than that of 24 GHz.

Table 2 summarizes channel parameters extracted at the position of RX1. The amplitudes A’s are normalized by peak value at each receiver position. As shown in Figure 9, the delay parameters ’s of 24 GHz are more widely spread than those of 5.3 GHz, whereas the spread of angular parameters of 5.3 GHz is larger than that of 24 GHz.

5. Conclusion

In this paper, angular and temporal characterization of 24 GHz millimeter wave propagation is made experimentally. Comparison with 5.3 GHz case is also made to emphasize the properties of millimeter wave signals. To obtain impulse response as a function of time and angle, a channel sounder with rotating antenna is used. The measured data show that angular spread of 24 GHz signal is smaller than that of 5.3 GHz, while echoes of 24 GHz are longer than those of 5.3 GHz. A procedure to extract modified S-V model parameters is also presented to facilitate simulation of millimeter wave communication systems.

Competing Interests

The authors declare that there is no conflict of interests regarding the publication of this article.

Acknowledgments

This work was supported by the ICT R&D program of MSIP/IITP [B0101-16-0222, Development of Core Technologies to Improve Spectral Efficiency for Mobile Big-Bang].