An arc fault is an unintentional power discharge which may cause unexpected serious consequences such as fires. Under different voltage, current and loading conditions, along with unplanned locations, the occurrence times of arc faults are always random. Due to the randomness characteristic of arc fault, the measurement of the position and intensity of arc faults in real-time is a challenge to researchers. This brief proposes an analog-digital mixed-signal system applied to the arc fault detection in smart meters. The introduced digital system can automatically identify the load types and the filters of the reconfigurable analog circuit arranged according to the loads. The proposed digital system also implements the arc faults detecting threshold and its respective methods. A field-programmable analog array (FPAA) incorporates the involved analog functions, and an Altera Field Programmable Gate Array (FPGA) realizes the digital system. We test the performance of the proposed mixed-signal system under different load conditions with daily household appliances. The results show that the circuit is able to detect 99.6% of the arc faults during the experiment. The proposed mixed system provides a cost-effective package solution for accurately detecting the arc fault in modern smart meters.