In radar systems of automatic detection, an estimate of background clutter power is used to set the detection threshold. An interference saturated environment is frequently encountered in these systems (multiple target situations). Therefore, the detection of signals in such an environment becomes one of the most important problems to be solved. The double-threshold algorithm is one of the more interesting detectors used in these situations. While the first threshold operation ensures that the calculation of the detection (second) threshold is based on a set of samples which is free of strong interferers and is therefore much more representative of the noise level, the second threshold is used to declare the presence or the absence of the radar target. The object of the present paper is to analyze the performance of such type of CFAR schemes when the radar receiver contains a noncoherent integrator amongst its basic elements. It is found that the processor detectabil ity loss is very low and the performance degradation, caused by interferers is quite small even if the number of outlying targets is large, given that the first threshold is properly chosen. In interference systems of automatic detection, an estimate of background clutter power is used to set the detection threshold. An interference saturated environment is frequently encountered in these systems (multiple target situations). the most important problems to be solved. The double-threshold algorithm is one of the more interesting detectors used in these situations. While the first threshold operation ensures that the calculation of the detection (second) threshold is based on a set of samples which is free of strong interferers and is therefore much more representative of the noise level, the second threshold is used to declare the presence or the absence of the radar target. The object of the present paper is to analyze the performance of such type of CFAR schemes when the radar receiver contains a noncoherent integrator amongst its basic elements. It is found that the processor detectabil ity loss is very l ow and the performance degradation, caused by interferers is quite small even if the number of outlying targets is large, given that the first threshold is properly chosen.