On-line condition monitoring and real-time fault detection are essential to protect power transformers from unscheduled outage of services. The core earth signals (CES) are valuable for fault detection and diagnosis of power transformers, as they are related to multiple transformer defects and faults, such as core multi-point earthing fault and core deep saturation. Moreover, the CES also contain valuable information of transient impulses intruded into power transformers such as very fast transient overvoltage (VFTO) impulses. This paper develops a novel power transformer condition monitoring system (NCMS), which makes full use of CES to detect and diagnose the external impulses and internal defects and faults. Firstly, the NCMS carries out the wide-band measurement of CES. Secondly, the NCMS obtains data from supervisory control and data acquisition system (SCADAS), dissolved gases analysis (DGA) system, and other monitoring systems, and correlates these data using grey relational analysis (GRA). Finally, the NCMS detects the impulses and diagnoses defects and faults based on the correlation relationships, then presents the monitoring results to the users. This novel system has been applied to the field and successfully diagnosed many defects and faults. It has high accuracy (< 3%) at the power and harmonic frequencies and possesses the capability to capture VFTO-pulses. INDEX TERMS Correlation analysis, core earth signal (CES), condition monitoring, industry applications, power transformer, wide-band measurement.