Report BRAVE D2.0: Propagation channel model and RF impairments definition has been released by the BRAVE project.

This deliverable gives a state-of-the-art and first BRAVE outcomes related to the characterization and modelling of the sub-THz physical layer (above 90 GHz). This will serve the identification of appropriate radio modulation schemes and efficient waveforms for sub-THz beyond-5G communications; and will contribute to the performance assessment in terms of coverage range, throughput, capacity, signal processing complexity and power consumption.

First, the Deliverable D2.0 explores the wireless channel properties inside a typical office building, for ranges up to 20 meters, in line-of-sight (LoS) and non-line-of-sight (NLoS) situations. The path-loss, delay spread, angular spread and coherence bandwidth are evaluated based on ray simulations. The channel properties in frequency range 90 to 200 GHz are compared to those simulated in same conditions at 2 GHz or within the 5G millimetre-wave spectrum. The path-loss and the channel spreading parameters are severely affected by the presence of obstructing partitions, but the reflections remain a very powerful propagation mechanism. Actually, a major factor that impacts the wireless channel is the antenna beam-width. The channel is observed to be “flat” over a 100 MHz bandwidth or even wider, when the beam-width at one terminal is 6°.

A similar analysis is conducted for lamppost-to-lamppost links (up to 200 meters range) in an urban outdoor scenario. The results are representative of the physical channel that would be experienced by a sub-THz mesh xhaul infrastructure feeding a dense small-cell access network or a fixed wireless access (FWA) infrastructure. The antenna beam-width at each terminal is supposed to be thin enough such that only the dominant path is captured. The blockage by buildings plays obviously the major role. But the attenuation by trees is another critical propagation component; it is simulated based on accurate 3D representation given by LiDAR point clouds technology. A path-loss exponent between 2.3 and 3.0 is observed in situations where the direct link is obstructed by some vegetation. The exact value of the path-loss exponent varies with the vegetation loss that is considered, which actually depends on branch and leave density. Today, the appropriate values for sub-THz vegetation loss remain unknown.

Both the in-office and outdoor wireless simulations are synthetized into approximate analytical models, where the path-loss and spreading properties are only depending on very few parameters. However simplified, such models are often convenient for integration of the physical layer into link or system-level studies.

Second, the Deliverable D2.0 investigates the generation of phase noise in oscillators and its effect on the communication systems performance. As the frequency increases, the phase noise becomes stronger, and its mitigation more complex. A sub-THz phase noise model is proposed beyond the current state-of-the-art (i.e. Gausian, Wiener and Leesson models). Phase noise is modelled as the sum of a frequency-flat (or white) component partly due to the reference crystal and several coloured components from the free-running oscillator that respectively depends on 1/f³, 1/f², 1/f and f0. Below the nominal bandwidth g of the PLL (Phased Locked Loop), the phase noise by the crystal prevails and is referred to as the near-carrier phase noise, while above the cut-off frequency g, the VCO (Voltage Controlled Oscillator) is the main contribution. This model is derived into a versatile simulation architecture, which may be directly calibrated upon oscillator spectral measurements. The choice of the phase noise model depending on the oscillator spectral properties and application will be investigated later in BRAVE project.

Today, the mathematical convenience of the phase noise Gaussian model is exploited to represent the poor performance of sub-THz systems, to optimize constellations and to improve the demodulation performance. An optimized sub-THz demodulator is derived for M-QAM modulations when high-SNR (Signal-to-Noise Ratio) signal is corrupted by gaussian phase noise. The symbols modulation and the optimal demapper are defined using polar metrics instead of the traditional I/Q metrics. The polar metrics are shown to greatly enhance the performance of coded systems. The optimal signal processing as well as some performance assessments are given.

The research and conclusions reported in this document will be completed in future deliverable D2.1 that is due in mid-2020. This future BRAVE delivery will provide deeper analysis of the physical layer, along with further description and evaluation of the proposed modulation schemes and waveforms.