Silicon Physical unclonable functions (PUFs) are regarded as emerging hardware security primitives that can be used for generating secret keys for cryptographic algorithms. Unfortunately, the reliability issues greatly limit the applications of most PUFs since ambient noises and aging issues may alter the output responses of the PUFs. In this paper, a novel error calibration technique (ECT) based on average sampling circuit (ASC) and self-checking circuit (SCC) is proposed for significantly boosting the reliability of PUFs. The primary role of ASC is filtering the ambient noises that may affect the output responses of the PUFs. In contrast, SCC rehabilitates the PUFs once aging issues damage the process mismatches within the PUFs. As shown in the result, a PUF with the proposed ECT is able to achieve a 99% reliability with a 50 Mbps throughput.