Assignment 3 – Ambient Readings

From manual observation to machine sensing: I’m collecting local, situated data around a Chinese balloon flower using a microcontroller + sensors.

中国桔梗花

Chinese balloon flower

ESP32-based ambient sensing rig for a Chinese balloon flower: BME280 (temperature/humidity/pressure), BH1750 (light), and a capacitive soil-moisture probe are wired and fixed on the windowsill, logging every 5 minutes to a CSV. The setup translates this small, situated corner into data for Assignment 3—so we can visualize changes over time and reflect on what the machines notice and what they miss.

ESP32 传感装置配备 BME280、BH1750 和电容式土壤湿度探头，用于读取当地环境温度

Device connection diagram

With the sensing rig deployed, the capacitive soil-moisture probe is inserted into the pot of the Chinese balloon flower while BME280 and BH1750 sample the surrounding microclimate. Driven by the ESP32, the system now logs temperature, humidity, pressure, light, and soil moisture at 5-minute intervals to a CSV (seen on the serial monitor), kicking off the ambient dataset for Assignment 3.

基于 ESP32 的传感装置测量桔梗的微气候 - BME280（温度/湿度/压力）、BH1750（光照）以及插入花盆的电容式土壤湿度探头 - 每 5 分钟记录一次数据。

Ambient readings at the plant, captured by ESP32 + sensors

Lux over time — a step-up after ~20:05 (≈12–17 lux) indicates a lighting change near the plant.

Temperature stayed nearly constant around 24–25 °C with minor fluctuations (<0.5 °C).

Relative humidity hovered around 48–50% with a small mid-series peak.

Barometric pressure remained steady near ~1006 hPa across the hour.

Reflection

I set out to sense the microclimate around a single, situated subject—the Chinese balloon flower on my windowsill—and translate that place into data. Using an ESP32 with BME280 (temperature/humidity/pressure), BH1750 (light), and a capacitive soil-moisture probe, I logged readings every five minutes for about an hour on a cloudy evening.

What worked: the digital sensors were stable and easy to synchronize. NTP time on the ESP32 made the CSV instantly usable, and the charts were legible without heavy cleaning. The BH1750 captured small, timely shifts in illuminance; around ~20:05 the lux stepped up (≈12–17), likely from a lighting change nearby. Temperature and humidity, by contrast, changed slowly and within a narrow band (~24–25 °C, ~48–50% RH), which made the different “temporal textures” of light vs. heat/moisture very clear.

What didn’t: the soil-moisture channel initially stayed flat because I had not completed calibration. I plan a simple two-point reference (air = dry, immersion = wet) and a fixed insertion depth to get comparable percentages. Minor friction points included cable management and remembering that the ESP32 is 3.3 V only; once wiring was secure, I²C behaved.

What surprised me or felt meaningful: how quickly the machine registered small human actions (curtain, screen brightness, a lamp switching on) while thermal variables barely moved. The sensor’s “field of view” felt intensely local—usefully so, but also limiting. Smell after rain, plant vigor, and my embodied comfort remain unreadable. This gap—between what the machine notices and what I feel—will guide my next steps: a longer 24-hour run, event annotations (window/lighting changes), and a calibrated soil trace to connect plant behavior to the ambient swings I recorded.