Convolutional Modeling and Antenna De-Embedding for Wideband Spatial mmWave Channel Measurement

Utilization of millimeter wave (mmWave) frequencies in the 5G has driven great efforts on measurement and modeling of the propagation channels, which are critical for the system evaluation and deployment. To compensate the large path loss, rotational scanning using high-gain horn antennas is usually employed for spatial channel characterization. However, it is a challenging issue to de-embed the antenna effect from captured channel profiles, especially for the wideband sounding. In this paper, a new convolutional modeling approach is first proposed, by which a synthesized spatial channel response is expressed by the consecutive convolutions of an antenna-free propagation model and directional antenna patterns. Then, based on the convolutional model, a simple two-step antenna de-embedding algorithm is designed. Furthermore, an indoor measurement campaign was performed using a steering receiver antenna and frequency-sweeping from 72.5 to 73.5 GHz. The high similarity between the measured and reproduced spatial channel responses and multipath impulse responses has indicated that the proposed approach can effectively de-embed the antenna effect and mitigate the system noise. It is also illustrated that the sparse impulse propagation model composed of only a few significant paths can sufficiently describe an mmWave channel. The channel angular, frequency, and time responses can be conveniently reproduced by the convolution of antenna patterns and sparse propagation models.