Experimental Evaluation of Fuzzy-Integral and Classical controls for Power Management in a 24 GHz mmWave 5G Transceiver

TL;DR

The paper tackles maintaining PA linearity and low EVM in a 24 GHz mmWave transceiver under temperature- and cable-induced gain variations. It compares PID, pure integral, and a novel fuzzy-integral (FI) controller implemented on hardware with an RMS power detector feeding a DSA-based attenuation loop; FI combines fuzzy inference with integral action to handle nonlinearities and eliminate steady-state error. Experiments show FI achieves faster settling, greater stability, and lower EVM deviations than the classical controllers, enabling reliable operation in the linear region. The work demonstrates practical feasibility on resource-constrained hardware and points to future integration with DPD to extend linear operation at higher power levels.

Abstract

The deployment of 5G millimeter-wave (mmWave) systems poses significant challenges in maintaining power amplifier linearity and efficiency under varying conditions, such as temperature-induced gain variations that degrade error vector magnitude (EVM). This paper presents a comparative study of three control strategies-PID, pure integral, and fuzzy-integral (FI)-for adaptive power management in a 24 GHz mmWave transceiver. The FI controller integrates fuzzy logic for handling nonlinearities with integral action for zero steady-state error. Experimental results show the FI controller outperforms others in settling time, stability, and EVM minimization.

Figures (6)

• Figure 1: Block diagram of the 24 GHz mmWave transceiver with fuzzy-integral power control.
• Figure 2: Experimental setup showing the SDR, a power amplifier, a 6-bit variable attenuator, a local oscillator (LO) at 9.6 GHz, an up-converter mixer with an output at 24.2 GHz, a down-converter mixer to 5 GHz, a power detector, and reception in the SDR.
• Figure 3: Characterization of the RMS power detector showing output voltage versus input power with linear regression fit ($R^2=0.998$).
• Figure 4: Experimental power-linearity trade-off of the 24.2 GHz transmit chain. Higher gain settings achieve greater output power at the cost of accelerated non-linear EVM degradation.
• Figure 5: Temporal response of classical and fuzzy integrator controllers managing the output power of the 24.2 GHz RF system. The chain is subject to abrupt variations in system gain. The left subplot illustrates the transient response following a step gain reduction from 10 dB to 5 dB. Conversely, the right subplot displays the response to a step gain increase from 5 dB to 10 dB. Both control architectures utilize an identical integral gain ($K_i = 2$). In both scenarios, the fuzzy integrator demonstrates a significantly faster transient response compared to its conventional counterpart.