PI Knightly’s and PI Jornet’s research for the past one year focused on single RF chain multi-user beamforming. One of the targeted thrusts in this project is to explore the empirical limits of multi user multi stream communication in S-T bands. It focuses on design and evaluation of low-overhead user and beam selection protocols for enabling downlink multi user multi stream communication in S-T WLANs, leveraging the hierarchical modulation schemes and multi beam codebooks to enable unprecedented simultaneous transmissions. With a dramatic increase to the access point (AP) and user density, multi user multi stream communication is a key factor towards scaling the capacity of future WLANs. Today’s 5GHz WLANs support simultaneous transmissions of multiple data streams to a single client or multiple clients but 60GHz WLANs only supports single stream transmissions. A fundamental limit in this approach is that each RF chain at the AP is restricted to supporting one stream or one user. To overcome this limit, we proposed single RF chain multi user beamforming, a framework for multi-stream multi-user downlink transmission via a single RF chain.
In addition, PI Jornet has been advancing the development of the team’s experimental platform, the Ultrabroadband Networking Systems Testbed. Currently, the testbed supports four different operating frequencies – 60 GHz, 120 GHz, 240 GHz, and 1 THz – and two different signal processing engines, namely, the 32-GHz offline platform from the TeraNova testbed and the 2-GHz real-time platform from the X60 testbed. The platform is currently being used for the experimental validation of solutions developed by the team, including the hierarchical modulations developed during Year 1 of the project as well as the aforementioned works.
In order to support multi-stream multi-user transmission, we introduce a 60 GHz WLAN architecture in which the number of supported streams exceed the number of RF chains. In particular, we introduce single RF chain multiuser beamforming for realizing multi stream multiuser downlink transmission via a single RF chain AP. In this way, we remove the RF chain limitation on multi user scaling enabling more users to be served concurrently. For this, we build on single beamformed transmission via overlayed constellations which employs the same philosophy of hierarchical modulation by not relying on a different space time or frequency resource for simultaneous transmission. We encode different data for different users into a single modulation structure while ensuring that grouped users can share the same transmit beam from the AP.
Second, before a multiuser transmission starts at the AP the key step is to determine for each transmission: which set of users should be grouped, which transmit beam should be used and which modulation level of the overlayed constellation each user should be assigned. For this, we proposed user grouping and beam selection policies that span tradeoffs in data rate, training and computation overhead.
Third, we realize the key components of our design on two testbeds. We employ X60, a fully programmable cross-layer configurable testbed for 60 GHz WLANs. Next, we deploy WARP-60 as a variable beamwidth testbed utilizing WARP for baseband processing, VubIQ for RF functions at 60 GHz, and mechanically steerable horn antennas of different widths. With these two testbeds, we perform over 49,000 measurements spanning multiple scenarios and topologies.
In terms of testbed development, the main focus of this year has been on completing the integration of X60 real-time signal processing engine with the system front-ends. As explained in the report for Year 1, the main challenge was on the physical layer synchronization. The use of higher-quality oscillators (separate and independent at the transmitter and the receiver) has solved the problem. Moreover, we have adopted a 5G NR-like OFDM-based physical layer.
We performed extensive measurements using the X60 testbed and the WARP 60 testbed and presented the first experimental evaluation of multi-user multi-stream transmission on a single RF chain in 60 GHz networks. We implemented the key components of our user and beam selection framework and performed experiments in different indoor environments. First, we begin with a simple yet important baseline case of two simultaneous users and experimentally study the effect of grouping a high SNR user that is close to the AP with a radically aligned user whose distance from the AP is varied from close (high SNR) to farther away (low SNR). The experiments characterize the multi-user gains of our method and enables us to study the impact of the SNR difference between the two AP-user links. The experiments reveal the critical role of receiver SNR spread in realizing gains. Namely both the individual and aggregate throughput increase with increasing SNR difference between the two user’s links to the AP, yielding the aggregate throughput gains of up to 64% compared to single user transmission when the two users are separated by 22.5 meters.
Second, we vary the transmit bandwidth from wide to narrow and find that adapting beamwidth at the AP acts as a knob in controlling the SNR spread and grouping efficiency and thereby the aggregate rate for SIMBA. We explore the tradeoff that wider beams, which ca create Non Line-of-Sight (NLOS) paths even when a LOS path exists, can provide better channel grouping opportunities, as more users will be able to share a beam. Thus, despite its low training overhead and complexity, we demonstrated that our method using 80o beamwidth outperforms the single user transmission by 58% and achieves about 75% of the aggregate rate of exhaustive search.
Third, we explore scaling the number of simultaneous users from 2 to 5 clients. We experimentally showed that beamforming to four simultaneous users using our method results in 2x aggregate rate improvement over single user transmissions. We showed that as group size increases, the multiuser gains do not increase linearly and the number of quantification levels of modulation (and coding) becomes the limiting factor for gains provided sufficient SNR spread exists among the users. Indeed, no throughput gain of multi user beamforming can be observed if only a single modulation and coding scheme is supported. We also studied the latency of our method and found that our best user and beam selection policy has approximately two-fold reduction in total data transmission time compared to single user transmissions, indicating that the latency performance gains exploit grouping of users with high SNR diversity.
Finally, in relation to the testbed, we have demonstrated real-time networking capabilities at S-T frequencies utilizing the X60 signal processing engine at 120 GHz and 240 GHz to stream a video.
Publications
Dasala, J. Jornet, and E. Knightly, “SIMBA: Single RF Chain Multi-User Beamforming in 60 GHz WLANs,” in Proceedings of IEEE INFOCOM 2020, July 2020.
Singh, V. Ariyarathna and J. M. Jornet, “A Plasmonic Array Architecture for Multi-beamSpatial Multiplexing at THz Frequencies,”to appear in Proc. of the 45th International Conference on Infrared, Millimeter, and Terahertz Waves (IRMMW-THz), Buffalo, NY, USA, November 2020.
Ariyarathna, A. Madanayake and J. M. Jornet, “Real-Time Digital Baseband System for Ultra-Broadband THz Communication,” to appear in Proc. of the 45th International Conference on Infrared, Millimeter, and Terahertz Waves (IRMMW-THz), Buffalo, NY, USA, November 2020.