6G • RAN
Neural Receiver
Deep Learning End-to-End Demodulation & Decoding
A deep neural network running on GPU/NPU hardware that replaces traditional channel estimation, equalization, and demapping modules in base stations.
Technical Explanation
Traditional cellular receivers contain a rigid sequence of hand-engineered mathematical blocks: channel estimation, equalization, noise covariance estimation, LLR calculation, and LDPC decoding. A 6G Neural Receiver replaces these separate blocks with a single deep convolutional or transformer neural network. Trained on millions of real-world wireless channels, the neural receiver accurately decodes signals distorted by non-linear power amplifiers, phase noise, and hardware imperfections.
Key Functions
- Replacing multiple hand-engineered receiver blocks with an end-to-end deep neural network
- Robust decoding across non-linear amplifier distortion and severe oscillator phase noise
- Achieving 2-4 dB gain in signal-to-noise ratio over optimal classical baseline receivers
- Continuous online self-learning adapting to unique local multipath environments
- Hardware acceleration on edge tensor processing units (TPU/NPU) at line rate
Specifications
IEEE Transactions on Wireless Communications, NVIDIA 6G Research, 3GPP Rel-19/20
Interfaces
Neural-RX-CoreTensor-PHY
Related 6G Concepts
Want to memorize 6G concepts like this one?
Study it with SuperMemo SM-2 spaced repetition flashcards.