Cross ambiguity processing is comm only used for target detection in passive bistatic radar system. However, its th eoretical detection performance on passive bistatic radar is rarely encountered in previous published papers. This paper addresses the theoretical problem of de tecting the presence of a target echo in passive bistatic radar when the bandwid ths of opportunity illuminators are unknown. Analytical expressions are obtaine d for the characteristic function and probability density function (PDF) of norm alized cross ambiguity function (NCAF) when the data samples in the target surve illance and reference channels are contaminated by narrowband uncorrelated white Gaussian noises. The closed form of detection probability and false alarm proba bility for NCAF are derived. They are used to determine which cell in the amb iguity surface exceeds a preset threshold. Results are presented in the form of curves showing the values of input SNRs required to achieve a specified detectio n probability for different system parameters. The detection performance shows that target signal can be detected when SNR in target surveillance channel is large than -20 dB and SNR in direct path channel is large than 5 dB. Furthermore, compared to the detection performance in matched sampling case, the detection probability will be decreased about 0.1 if the mismatch coefficient of sampling is less than 2.