A MAC Protocol with Time Reversal for Wireless Networks within Computing Packages

Ama Bandara, Abhijit Das, Fátima Rodríguez-Galán, Eduard Alarcón, Sergi Abadal

TL;DR

Researchers have developed TR-MAC, a new communication protocol for wireless networks inside computer chips. By leveraging a technique called Time Reversal (TR), the protocol turns the chaotic "echo chamber" of a chip package into a precision tool, allowing multiple wireless signals to travel simultaneously on the same frequency without crashing into each other.

What Problem Does This Paper Address?

As we pack more cores into processors (think "many-core" architectures and multi-chiplet systems), the internal wiring becomes a massive bottleneck. Traditional wired Networks-on-Chip (NoCs) struggle with long-distance communication and energy efficiency as they scale.

Wireless Network-on-Chip (WNoC) is a promising alternative, but it faces a unique physical challenge: reverberation. Because a chip package is a tiny, enclosed metallic environment, radio waves bounce around like crazy. This creates massive "multipath effects," leading to Inter-Symbol Interference (ISI)—where bits of data overlap and become unreadable—and Co-Channel Interference (CCI), which prevents multiple pairs of nodes from talking at once.

Methods

The authors propose TR-MAC, a Medium Access Control protocol designed specifically to exploit the physics of Time Reversal.

How Time Reversal Works:

In a static environment like a chip package, the path a signal takes from Point A to Point B is unique and unchanging. TR involves "pre-characterizing" this path (the Channel Impulse Response). When a transmitter wants to send data, it pre-codes the signal using a time-reversed version of that path. This causes the waves to converge and "focus" their energy precisely at the intended receiver in both space and time, naturally canceling out echoes and ignoring other nearby devices.

The TR-MAC Protocol:

The protocol operates in three distinct phases:

  1. Phase I (Preamble): The transmitter sends a TR-precoded preamble containing its address. Because of the spatial focusing, only the intended receiver "sees" the clear energy peak.
  2. Phase II (ACK): The receiver sends back a 1-bit acknowledgment (ACK) using TR. If the transmitter detects a specific energy threshold, it knows the path is clear.
  3. Phase III (Data): The actual data packet is sent.

To prevent "hidden terminal" issues, the protocol uses a low-bandwidth busy tone on a separate channel to signal when a transmission is in progress, ensuring no new transmissions start mid-cycle.

Datasets

The study relies on architectural simulations rather than physical hardware datasets. The researchers used:

  • Multi2sim: A standard simulation framework for CPU-GPU computing, modified to include a wireless interconnect model.
  • Synthetic Traffic Models: They used a model with parameters ($H=1, \sigma=0.5$) to simulate "bursty" and "hotspot" traffic, which mimics real-world heavy computational workloads.
  • Configuration: A 64-node system with a baseline data rate of 30 Gbps per link.

Key Insights and Ideas

  • Multipath as a Feature, Not a Bug: Usually, engineers try to eliminate reflections. TR-MAC embraces them. The "richer" the scattering environment (more bounces), the better TR can focus the energy, allowing for more parallel transmissions.
  • Spatial Multiplexing on a Single Frequency: Most MAC protocols avoid collisions by making nodes wait their turn (Time Division) or use different "colors" of radio waves (Frequency Division). TR-MAC allows multiple nodes to use the same frequency at the same time by separating them in space through energy focusing.
  • Static Channel Advantage: Unlike mobile phones moving through a city, the components inside a computer chip never move. This makes the initial "calibration" of the wireless channels a one-time cost that pays off forever.

Key Results

The researchers compared TR-MAC against two industry standards: BRS (a random-access protocol) and Token Passing.

  • Throughput: TR-MAC achieved significantly higher aggregate throughput than BRS. In scenarios with 64 cores, TR-MAC on a single frequency channel performed comparably to Token Passing systems that required multiple frequency channels.
  • Latency: At low traffic loads, TR-MAC maintains the low latency of random-access protocols. As traffic increases, it avoids the "latency explosion" seen in BRS because it can handle multiple parallel streams (NPT - Number of Parallel Transmissions).
  • Scalability: When tested on 128 and 256 cores, TR-MAC showed much better latency stability than Token Passing, which becomes sluggish as the "token" has to travel through more nodes.

Limitations and Future Work

  • Calibration Overhead: The system requires an initial phase to map the channels between every possible pair of nodes. While the paper notes this is done "a priori," the complexity of this calibration grows as the number of cores increases.
  • Hardware Complexity: Implementing TR precoding requires specialized filters at the transceiver level. The authors mention this briefly as a "resource-constrained scenario," but the actual area and power overhead of these filters on-chip remains a point for further hardware-level validation.
  • Future Direction: The authors suggest exploring "over-the-air computing," where the interference patterns themselves are used to perform mathematical operations (like summing values from multiple sensors) directly in the wireless medium.

Summary updated

Abstract

Wireless Network-on-Chip (WNoC) is a promising concept which provides a solution to overcome the scalability issues in prevailing networks-in-package for many-core processors. However, the electromagnetic propagation inside the chip package leads to energy reverberation, resulting in Inter-Symbol Interference (ISI) with high delay spreads. Time Reversal (TR) is a technique that benefits the unique time-invariant channel with rich multipath effects to focus the energy to the desired transceiver. TR mitigates both ISI and co-channel interference, hence providing parallel communications in both space and time. Thus, TR is a versatile candidate to improve the aggregate bandwidth of wireless on-chip networks provided that a Medium Access Control (MAC) is used to efficiently share the wireless medium. In this paper, we explore a simple yet resilient TR-based MAC protocol (TR-MAC) design for WNoC. We propose to manage multiple parallel transmissions with simultaneous spatial channels in the same time slot with TR precoding and focused energy detection at the transceiver. Our results show that TR-MAC can be employed in massive computing architectures with improved latency and throughput while matching with the stringent requirements of the physical layer.
BibTeX
@inproceedings{bandara2024mac,
  author = {Ama Bandara and Abhijit Das and Fátima Rodríguez-Galán and Eduard Alarcón and Sergi Abadal},
  title = {{A MAC Protocol with Time Reversal for Wireless Networks within Computing Packages}},
  booktitle = {NANOCOM},
  year = {2024},
  url = {https://openreview.net/forum?id=czUyTuZJUP},
}

Loading related content...

Figure