With research on the technology components for the future 6G wireless communication standard in full swing, the possibilities of an AI-native air interface for 6G are also being explored. In collaboration with NVIDIA, Rohde & Schwarz takes another step forward and presents an enhancement of their recent hardware-in-the-loop demonstration of a neural receiver, showing the achievable performance gains when using trained ML models compared to traditional signal processing, while for the first time also optimizing the transmitter side.
At this year’s Brooklyn 6G Summit, taking place from October 31 to November 2 in Brooklyn, New York, visitors can experience the first demonstration of how a neural receiver paired with a learned custom modulation performs in a 5G NR uplink multiple input multiple output (UL MIMO) scenario. The setup combines premium test solutions for signal generation and analysis from Rohde & Schwarz and the NVIDIA Sionna™ GPU-accelerated open-source library for 5G and 6G link-level simulations. NVIDIA’s Sionna enables rapid prototyping of complex communications system architectures and provides native support to the integration of machine learning in 6G signal processing.
This demonstration builds on the successful showcase at the Mobile World Congress 2023 in Barcelona, where both companies demonstrated the concept of a neural receiver in a hardware-in-the-loop experiment involving two independent companies for the very first time. Academia, leading research institutes, and key industry experts across the globe anticipate that a future 6G standard will use AI/ML for signal processing tasks, such as channel estimation, channel equalization, and demapping. This concept is understood as a neural receiver. Today’s simulations suggest that a neural receiver will improve overall link performance and thus throughput compared to the current high-performance deterministic software algorithms used in 5G NR while keeping computational complexity at a manageable level.
Going a step further, the test bed has been extended to enable the verification of communication systems that apply AI/ML not only in the receiver, but also in the transmitter. The demonstration at Brooklyn 6G Summit showcases the application of learned custom constellations, which researchers are investigating as a stepping stone towards pilotless communication. Instead of relying on well-known, symmetric constellations such as QPSK or QAM modulations, the constellation points are determined through an end-to-end learning process, which jointly optimizes the neural receiver and the constellation mapper of the transmitter while taking the faded mobile radio channel into account.
The test bed setup leverages the R&S SMW200A vector signal generator, which emulates a single user transmitting a 100 MHz (273 PRB) wide signal in the uplink direction at a carrier frequency of 2.14 GHz in a MIMO 2x4 configuration. Fading and noise are applied to the transmission to emulate realistic radio channel conditions. The R&S MSR4 multi-purpose satellite receiver captures the RF signal instantaneously with its independent 4x 200 MHz channels. The digitalized signal is streamed in real-time and transferred to the R&S Server-Based Testing (SBT) framework including R&S VSE vector signal explorer (VSE) micro-services, which performs synchronization and fast Fourier transform (FFT) calculation. The post-FFT data serves as input for a neural receiver implemented and trained using NVIDIA Sionna. As part of the demonstration, the trained neural receiver with custom constellation is compared to the classical concept of a linear minimum mean squared error (LMMSE) receiver architecture, which applies traditional signal processing techniques based on deterministically developed software algorithms. These high-performance algorithms are widely adopted in current 4G and 5G cellular networks.
To verify the performance of such an enhanced neural receiver, including custom modulation, the capabilities of the R&S SMW200A vector signal generator were extended for this demo to allow the user to freely configure each individual constellation point by magnitude and phase in the IQ domain based on the selected QPSK or QAM modulation. The R&S FSW Signal and Spectrum Analyzer and the R&S Vector Signal Explorer (VSE) software today support custom modulation, too. This enables users to assess transmitter performance through well-known key performance indicators such as the error vector magnitude (EVM) even when learned custom constellations are used.
Rohde & Schwarz actively supports 6G research activities across Europe, Asia, and the US, while contributing to research projects, the work of industry alliances, and collaborating with leading research institutes and universities. The company’s test and measurement expertise and solutions help pave the way for the next generation of wireless communications, which is expected to be commercially deployed around 2030.
For further information on Rohde & Schwarz test solutions on the verge of 6G, visit:
https://www.rohde-schwarz.com/6G
At this year’s Brooklyn 6G Summit, taking place from October 31 to November 2 in Brooklyn, New York, visitors can experience the first demonstration of how a neural receiver paired with a learned custom modulation performs in a 5G NR uplink multiple input multiple output (UL MIMO) scenario. The setup combines premium test solutions for signal generation and analysis from Rohde & Schwarz and the NVIDIA Sionna™ GPU-accelerated open-source library for 5G and 6G link-level simulations. NVIDIA’s Sionna enables rapid prototyping of complex communications system architectures and provides native support to the integration of machine learning in 6G signal processing.
This demonstration builds on the successful showcase at the Mobile World Congress 2023 in Barcelona, where both companies demonstrated the concept of a neural receiver in a hardware-in-the-loop experiment involving two independent companies for the very first time. Academia, leading research institutes, and key industry experts across the globe anticipate that a future 6G standard will use AI/ML for signal processing tasks, such as channel estimation, channel equalization, and demapping. This concept is understood as a neural receiver. Today’s simulations suggest that a neural receiver will improve overall link performance and thus throughput compared to the current high-performance deterministic software algorithms used in 5G NR while keeping computational complexity at a manageable level.
Going a step further, the test bed has been extended to enable the verification of communication systems that apply AI/ML not only in the receiver, but also in the transmitter. The demonstration at Brooklyn 6G Summit showcases the application of learned custom constellations, which researchers are investigating as a stepping stone towards pilotless communication. Instead of relying on well-known, symmetric constellations such as QPSK or QAM modulations, the constellation points are determined through an end-to-end learning process, which jointly optimizes the neural receiver and the constellation mapper of the transmitter while taking the faded mobile radio channel into account.
The test bed setup leverages the R&S SMW200A vector signal generator, which emulates a single user transmitting a 100 MHz (273 PRB) wide signal in the uplink direction at a carrier frequency of 2.14 GHz in a MIMO 2x4 configuration. Fading and noise are applied to the transmission to emulate realistic radio channel conditions. The R&S MSR4 multi-purpose satellite receiver captures the RF signal instantaneously with its independent 4x 200 MHz channels. The digitalized signal is streamed in real-time and transferred to the R&S Server-Based Testing (SBT) framework including R&S VSE vector signal explorer (VSE) micro-services, which performs synchronization and fast Fourier transform (FFT) calculation. The post-FFT data serves as input for a neural receiver implemented and trained using NVIDIA Sionna. As part of the demonstration, the trained neural receiver with custom constellation is compared to the classical concept of a linear minimum mean squared error (LMMSE) receiver architecture, which applies traditional signal processing techniques based on deterministically developed software algorithms. These high-performance algorithms are widely adopted in current 4G and 5G cellular networks.
To verify the performance of such an enhanced neural receiver, including custom modulation, the capabilities of the R&S SMW200A vector signal generator were extended for this demo to allow the user to freely configure each individual constellation point by magnitude and phase in the IQ domain based on the selected QPSK or QAM modulation. The R&S FSW Signal and Spectrum Analyzer and the R&S Vector Signal Explorer (VSE) software today support custom modulation, too. This enables users to assess transmitter performance through well-known key performance indicators such as the error vector magnitude (EVM) even when learned custom constellations are used.
Rohde & Schwarz actively supports 6G research activities across Europe, Asia, and the US, while contributing to research projects, the work of industry alliances, and collaborating with leading research institutes and universities. The company’s test and measurement expertise and solutions help pave the way for the next generation of wireless communications, which is expected to be commercially deployed around 2030.
For further information on Rohde & Schwarz test solutions on the verge of 6G, visit:
https://www.rohde-schwarz.com/6G
