We proposed the design and evaluation of low-overhead user and beam selection protocols for enabling downlink and uplink multi-user multi-stream communication in S-T WLANs.
Transmissions in S-T bands needs to be directional and hence both transmitter and receiver have to discover the right beam configuration to be able to communicate. Brute force beam discovery for a multi-stream multi-user transmission incurs prohibitively large training overhead that limits the achievable gains of such simultaneous transmissions. More specifically, due to the sparse multipath nature of wireless channel in S-T bands, we found beam selection to be the key to providing sufficient diversity to separate streams and manage inter-stream interference. The key activity for PIs Knightly and Koutsonikolas over the past year was to design and experimentally evaluate efficient and low-overhead beam selection mechanisms that can successfully enable multi-user downlink transmission exploiting the limited reflections available in NLOS paths.
Another aspect that needs to be taken into account when moving to the S-T bands is the design of adequate modulations that can make the most out of the much larger (up to tens of GHz) bandwidth. The high cost, energy consumption and power requirements of Digital to Analog and Analog to Digital Converters (DACs and ADCs, respectively) make the use of multiple transmission chains prohibitive. In this direction, ways to multiplex users while still utilizing one transmission chain are needed. The key activity for PI Jornet over the past year has been to design and numerically evaluate the performance of bandwidth hierarchical modulations, as a way to both make the most out of the distance-dependent bandwidth at THz frequencies and to enable the transmission of separate information flows at the same time and frequency to users in the same space.
In addition, we have worked towards the development of the NEST platform, by merging the two testbeds available to the team at UB, namely, the X60 and the TeraNova testbed.
In order to find the multi-stream beam configuration that supports the maximum multiplexing gains, one needs to exhaustively search over all possible combinations. Instead, we proposed a low-complexity and low-overhead system that identifies dominant paths between the AP and each client in order to efficiently steer 60 GHz beams over diverse or ideally orthogonal paths, such that undesired channel correlations are minimized. Our design takes advantage of GHz-scale sampling rates to distinguish multipath components and exploit the knowledge of predefined beam patterns to build an angular profile for each path.
Second, using the estimated multipath profile for each client, we designed a multi-stream beam selection protocol to maximize the aggregate rate for concurrent transmissions to a given group of clients. Exploiting different analog beams impacts the effective directional S-T channel as each beam, with its unique radiation pattern, amplifies certain paths and weakens others. Knowing the multipath profile and the radiation pattern, we can predict the achievable multiplexing gain under different analog configurations without any signaling and with zero overhead. Our strategy is based on the observation that multiplexing independent streams should avoid common paths as it will otherwise incur throughput degradation due to channel correlation and inter-stream interference
Third, we analyzed the fundamental limits of multiplexing in S-T bands for both single-user and multi-user communications. We studied the maximum multiplexing gains under the perfect user and beam selection for single-user and multi-user multi-stream communications. We exploit the aforementioned beam selection protocol as a mechanism to provide sufficient diversity and hence achieve the maximum aggregate rate.
When it comes to the design of bandwidth hierarchical modulations, we proposed for the first the time the concept of bandwidth hierarchical modulations. Partially related to the concept of hierarchical or concatenated modulations, the fundamental idea in this case is to embed multiple binary information streams on the same carrier signal by manipulating the symbol time. More specifically, for users over short distances, in which the available bandwidth is larger and the path-loss much lower, symbol duration can be made shorter than that for users over longer distances.
We devised the implementations for the modulator and demodulator, we analytically investigated the performance of the proposed scheme in terms of achievable data rate and we compared it to that of traditional hierarchical modulations. In addition, we derived the symbol error rate by starting from the new defined constellations.
Finally, in terms of the development of the experimental platform, we acquired the NI PXIe-3620 IF modules required to up-convert to an intermediate frequency (12 GHz) the baseband signal generated by the X60 baseband (2 GHz of bandwidth), as required for the TeraNova 1 THz up-converters as well as for the new 240 GHz up-converters to be acquired. The software was correspondingly adapted to support the new configuration.
We performed extensive measurements using the X60 testbed and presented the first experimental evaluation of MIMO beam steering in 60 GHz networks. First, we explored the impact of analog beam steering on the achievable multi-user rates. Our results revealed that selecting beams that provide high SNR at target users is not always the best approach. In particular, we showed that exploiting such a strategy, the aggregate rate of two-user transmission might even fall below the single-user rate. This is due to the irregularity of beam patterns that might capture the same physical path under different beams and cause high channel correlations. We demonstrated that in such cases, digital precoding methods such as zero-forcing are of little help. That is, digital precoding cannot compensate for a bad choice of analog beams that obtain low stream separability in the analog domain due to an ill conditioned channel matrix.
Second, we implemented the key components of our path discovery and beam selection platform and performed experiments in different indoor environments and configurations. We demonstrated that our method achieves 90% of the maximum aggregate rate for both single-user and multi-user MIMO, with only 0.04% of the training overhead compared to exhaustive search. Hence, our system approximates the PHY throughput of exhaustive search, while searching over only a few beams with diverse paths.
Third, we evaluated the fundamental limits of multi-stream communications in 60 GHz bands using over-the-air measurements. We found that the aggregate rate scales linearly with the number of users for a small number of streams. However, by further increasing the number of spatial streams above four, the network’s aggregate rate is below from ideal case which would linearly scale with number of spatial streams. This implies that the MIMO multiplexing gains do not endlessly increase proportionally with the number of streams because of undesired channel correlations and limited efficacy of zero forcing.
When it comes to the bandwidth hierarchical modulations, we collected extensive numerical results, obtained by utilizing an analytical channel model that captures the peculiarities of the THz band, previously developed in our group and experimentally validated with the TeraNova testbed. We demonstrated that hierarchical bandwidth modulation outperforms the existing modulation schemes by approximately 25% and conventional modulations by approximately 50%, in terms of data-rates for a fixed target bit error rate. The next steps include to experimentally test the system.
In relation to the NEST, we experimentally characterized the performance of the different communication blocks in the transmitter and the receiver in charge of time, frequency and phase synchronization. The results showed that the current TDMA-type physical layer, with only two synchronization slots for every 100 slots, is not able to support our target multi-Gigabit-per-second transmissions at THz frequencies. While the slots immediately after the synchronization slots are properly received, the latter slots lose synchronization and, thus, cannot be properly decoded. At this point, we are working towards changing the frame structure as well as on replacing the entire TDMA physical layer with an OFDM-based system.
Publications
Y. Ghasempour, M.K. Haider, C. Cordeiro, D. Koutsonikolas and E. Knightly, “Multi-Stream Beam-Training for mmWave MIMO Networks,” in Proceedings of ACM MobiCom 2018, New Delhi, India, October 2018.
Z. Hossain and J. M. Jornet, “Hierarchical Bandwidth Modulation for Ultra-broadband Terahertz Communications,” in Proc. of the IEEE International Conference on Communications (ICC), Shanghai, China, May 2019.