Due to excellent electrical and mechanical properties, cross-linked polyethylene (XLPE) cables are widely used in power systems. Poor manufacturing techniques in the production and installation of cable joints will cause insulation defects. The interdigital capacitive (IDC) sensor has advantages of simple structure and non-contact with the center conductor and shows great potential for online monitoring on XLPE cables. This paper focuses on the 3D modeling of a fully covered IDC sensor for cable insulation detection. Firstly, a 3D finite element model of the sensor is built, and the electric field distributions are compared with those of the partially covered sensor. For the sensor with more electrode pairs, the sensitivity increases with the sensor length and tends to saturate at the length of 5 cm, while the sensitivity remains constant for the sensor with fewer electrode pairs. Then, the differences between 3D and 2D results are discussed and the sensor parameters are optimized to reduce the influence of the fringe capacitance. The simulation results indicate that air gaps between the sensor and XLPE cable are the main reason of the difference between simulation and experiment. When the electrode width is equal to the gap width, the effects of both the fringing electric field and air gaps are relatively small. Finally, several types of sensors are made and used to detect the cable joint with and without the stress cone dislocation under different excitation voltage frequency. The results show that the measured capacitance decreases with frequency and the capacitance of the cable joint with the defects is smaller than that of the normal cable joint.