For dual probe heat pulse method (DPHP), probe deflection causes error in measured heat capacity (c) of soil.Recently, by introducing two thermistors in the longitudinal direction of a temperature probe, Liu et al (2013) developed a self-calibrated DPHP sensor for in situ calibrating probe spacing. However, their analysis was based on the assumption that the probes were coplanar and linear deflected. In this study, we carried out experiments in fine sand and silica sand (dry and saturated) with more realistic deflections (load at the free end of the probes, load at the base of the probes, and noncoplanar deflection). We define θ as the inclination angle of probe deviation from the vertical direction. For the coplanar inward deflection with load on the base and 2.58º < θ < 2.77º, the error in c is between − 17.1% and −17.5%; after in situ calibration, it is between 3.5% and 4.6%. For noncoplanar outward deflection with 3.42º < θ < 3.57º, the error in c is in the range from 26.5% to 29.2%; after in situ calibration, the error reduces to from – 2.3% to − 1.3%. Our results indicated that the coplanar linear deflection of Liu et al (2013) can significantly eliminate the error caused by the more general noncoplanar and nonlinear probe deflections in reality.