<![CDATA[Low-cost capacitive soil moisture sensors can produce short-term fluctuations that reduce the readability of monitoring displays. This study evaluated the internal signal behavior of Moving Average (MA) and Exponential Moving Average (EMA) filters in an Arduino-based soil moisture monitoring system; it did not assess calibrated soil-moisture accuracy. Three repeated acquisition runs were conducted using fresh 2 kg Regosol samples, a capacitive sensor v1.2, and one-second sampling during staged water addition. The primary comparison aligned Raw ADC, MA5, and EMA with α = 0.2 over samples 5–300, yielding 888 observations. Parameter sensitivity was examined using MA windows of 3, 5, 7, and 10 and EMA coefficients of 0.1, 0.2, 0.3, and 0.5 over 873 aligned observations. MA5 produced lower internal deviation than EMA with α = 0.2, with MAD values of 2.141 versus 3.851 ADC and RMSD values of 6.754 versus 9.367 ADC. MA5 also showed a shorter lag and empirical T50, whereas EMA with α = 0.2 produced slightly smaller point-to-point changes. Sensitivity analysis confirmed systematic trade-offs between smoothing strength, tracking fidelity, and response delay. Therefore, filter selection should be based on the operational objective, while calibration remains necessary before Raw ADC values are interpreted as physical soil-water content.]]>
Copyrights © 2026
