No abstract

The signal and threshold quantization and noise level estimation eflects on the pe@omance of two power-law detectors are studied. The other detector is the conventional total-power radiometer whereas the other peflorms better if the signal to be detected is narrowband. The goal is to define the suficient number of quantization bits and noise samples to maintain an acceptable pelformance when the noise level is estimated from the discretized samples. The threshold is calcu1;ltecl using a Cornish-Fischer approximation. We study also the impacts of the dynamic range. The simulations illustrate that the selection of the dynamic range and the noise estimate aflect the perfwmance enormously. If the dynamic range is correctly set, the adequate number of quantization bits is 4. A too narrow dynamic range causes noise saturation, which decreases the false alarm probubilizy drastically. This is due to changed statistic, i.e., the noise becomes non-Gaussian. The performance degradation cannot be compensated for by increasing the number of quantization bits and noise estimation samples. The suflcient number of noise level estimation samples is about 1000.

No abstract

No abstract