The excitation windings of the synchronous generators are generally reliable. However, electrical, mechanical, and thermal stresses that the machine is exposed to during its operation leads to the inter-turn and ground insulation faults. On-line condition monitoring could provide a valuable real-time assessment of the synchronous generator. It could discriminate gradual aggravating defects at an incipient stage before it leads to irreversible and costly damages. On-line monitoring of the air gap magnetic field has been used in different types of electrical machines. However, some concerns should be considered when applying this method to the salient pole synchronous generators (SPSG), especially in the SPSG, with a large number of turns per poles that show less or lack of sensitivity to a fault. This difficulty could be solved by wise locating of the Hall effect sensor, choosing precise sampling rate, re-sampling the data and signal processing approach. In this paper, detailed online monitoring of the air gap magnetic field under the excitation winding defect is proposed. A procedure that could detect the severity and location of the fault-based on different analyzing methods of flux density in no-load and full load is proposed. The nominated approach is examined by using numerical modeling and a experimental test rig with a 100 kVA SPSG. It is proven that the air gap magnetic field spectrum could provide a reliable assessment of the machine under the short circuit fault of the rotor field winding.