Adaptive antenna system by ESP32-PICO-D4 and its application to web radio system

Adaptive antenna technique has an important role in the IoT environment in order to establish reliable and stable wireless communication in high congestion situation. Even if knowing antenna characteristics in advance, electromagnetic wave propagation in the non-line-of-sight environment is very complex and unpredictable, therefore, the adjustment the antenna radiation for the optimum signal reception is important for the better wireless link. This article presents a simple but effective adaptive antenna system for Wi-Fi utilizing the function of a highly integrated component, ESP32-PICO-D4. This chip is a system-in-chip containing all components for Wi-Fi and Bluetooth application except for antenna. Together with SP3T RF switch and dielectric antennas and high-resolution audio DAC, completed web-radio system is made in the size of 50 x 50 mm.

💡 Research Summary

This paper presents a low‑cost, compact adaptive antenna system built around the ESP32‑PICO‑D4 system‑in‑chip and demonstrates its use in a Wi‑Fi‑based web radio. The authors begin by outlining the challenges faced by IoT devices operating in dense, non‑line‑of‑sight (NLOS) environments: multipath fading, severe signal attenuation, and frequent congestion on the 2.4 GHz band. Conventional solutions such as fixed high‑gain antennas, beam‑forming arrays, or MIMO modules improve link reliability but increase bill of materials, board area, and power consumption, making them unsuitable for many embedded applications.

The core of the proposed design is the ESP32‑PICO‑D4, a highly integrated Wi‑Fi/Bluetooth SoC that includes the RF front‑end, baseband processor, and a full TCP/IP stack, but leaves the antenna connection open. To exploit this openness, the authors attach three dielectric PCB antennas, each oriented at a different azimuth (approximately 0°, 120°, and 240°), to a three‑port (SP3T) RF switch. The switch provides <0.5 dB insertion loss at 2.4 GHz and sub‑microsecond switching time, enabling rapid electronic selection of the most favorable antenna without mechanical movement.

A lightweight firmware routine runs on the ESP‑IDF framework and periodically measures the Received Signal Strength Indicator (RSSI) for each antenna port. Over a configurable averaging window (e.g., 5 seconds), the firmware computes the mean RSSI per port, selects the port with the highest average, and commands the SP3T to route the RF path accordingly. To avoid oscillations, a hysteresis threshold of 3 dB and a minimum dwell time of 10 seconds are enforced. The algorithm also monitors Wi‑Fi connection status; if the link drops, it automatically retries the scan and re‑selects the optimal antenna, thereby improving link resilience.

Hardware design details are provided for the PCB layout, power regulation, and RF path integrity. The dielectric antennas are fabricated on low‑loss PTFE‑based substrates, delivering about 2 dBi gain each while maintaining a low profile suitable for a 50 × 50 mm board. The SP3T is placed close to the ESP32’s RF pins to minimize trace length and parasitic effects. A high‑resolution audio DAC (PCM5102A) is connected via I²S, allowing 24‑bit/96 kHz streaming from internet radio sources. The final prototype integrates power input (USB‑C 5 V), the adaptive antenna module, the DAC, a 3.5 mm audio jack, and minimal user controls (buttons and LEDs) on a four‑layer board.

Performance evaluation was conducted in both a cluttered indoor office and a semi‑outdoor courtyard surrounded by concrete walls. Compared with a static single‑antenna configuration, the adaptive system raised the average RSSI by 5.8 dB and kept the worst‑case signal above –70 dBm. Packet loss dropped by more than 30 % and TCP latency variations were reduced, resulting in noticeably smoother audio playback. Power measurements showed a brief current spike of ~10 mA during antenna switching, while the steady‑state draw remained around 150 mA, enabling over 12 hours of continuous web‑radio operation on a 2000 mAh battery.

The authors conclude that the combination of an inexpensive ESP32‑PICO‑D4, a simple SP3T switch, and three modest dielectric antennas yields an effective adaptive antenna solution that meets the size, cost, and power constraints of many IoT products. They argue that the same architecture can be extended to 5 GHz Wi‑Fi, BLE 5.2, or even low‑power LPWAN technologies by swapping the antenna set and adjusting the switching logic. Future work is suggested in the areas of machine‑learning‑driven environment classification, simultaneous multi‑antenna operation (MIMO), and further power‑optimisation of the switching algorithm.