Channel-Aware Behavioral Power Modeling of CMOS OOK Transceivers for Wireless Network-on-Chip Systems

TL;DR

This paper introduces a channel-aware behavioral power modeling framework for CMOS OOK transceivers in Wireless Network-on-Chip systems, enabling energy-efficient operation across frequencies.

Contribution

It develops a frequency-dependent power model incorporating channel loss, revealing power trade-offs and optimal operating points for CMOS OOK transceivers.

Findings

Power dominance shifts from oscillator and ED at low frequencies to PA and LNA at high frequencies.

The energy-per-bit landscape has identifiable sweet spots and a global minimum.

The model guides energy-efficient design of high-frequency wireless links in WNoC systems.

Abstract

Wireless Network-on-Chip (WNoC) systems enable low-latency communication in many-core platforms through short-range wireless links. However, the power consumption of integrated transceivers (TRXs), dominated by that of the RF front-end circuitry, remains a major challenge. Moreover, the optimal operating frequency is still unclear, as bandwidth, energy efficiency, and technology maturity must be balanced. This work presents a channel-aware behavioral modeling framework to estimate power consumption and identify energy-efficient operating points in non-coherent On-Off Keying (OOK) TRXs over a wide frequency range. The approach leverages survey data from CMOS implementations to derive frequency-dependent power models for key TRX sub-blocks, including the power amplifier (PA), oscillator, mixer, low noise amplifier (LNA), and envelope detector (ED). By incorporating the frequency-dependent channel loss into the TRX power budget, the model captures system-level power trade-offs across operating regimes. The analysis reveals a frequency-dependent shift in power dominance between the transmitter and receiver: oscillator- and ED-dominated regimes at lower frequencies transition to PA- and LNA-dominated behavior at higher frequencies. Furthermore, the energy-per-bit landscape exhibits sweet spots and a model-based global minimum, indicating that optimal operation cannot be achieved by optimizing transmitter or receiver independently. Overall, the proposed framework enables rapid and physically grounded exploration of power scaling with frequency and channel conditions, providing practical guidelines for energy-efficient design of high-frequency wireless links for WNoC systems and beyond.