TUTCRIS - Tampereen teknillinen yliopisto

TUTCRIS

Spectrum Sensing under RF Non-Linearities: Performance Analysis and DSP-Enhanced Receivers

Tutkimustuotosvertaisarvioitu

Yksityiskohdat

AlkuperäiskieliEnglanti
Sivut1950-1964
Sivumäärä15
JulkaisuIEEE Transactions on Signal Processing
Vuosikerta63
Numero8
DOI - pysyväislinkit
TilaJulkaistu - 15 huhtikuuta 2015
OKM-julkaisutyyppiA1 Alkuperäisartikkeli

Tiivistelmä

Intermodulation products arise as a result of low noise amplifier (LNA) and mixer non-linearities in wideband receivers. In the presence of strong blockers, the intermodulation distortion can deteriorate the spectrum sensing performance by causing false alarms and degrading the detection probability. We theoretically analyze the impact of third-order non-linearities on the detection and false alarm probabilities for both energy detectors and cyclostationary detectors under front-end LNA non-linearities. We show that degradation of the detection performance due to nonlinearities of both energy and cyclostationary detection is strongly dependent on the modulation type of the blockers. We then propose two DSP-enhanced receiver architectures to compensate the impact of nonlinearities. The first approach is a post-processing technique which compensates for nonlinearities effect on the test statistic by adapting the sensing time and detection threshold. The second approach is a pre-processing method that compensates by correcting received samples prior to computing the test statistic. This approach is based on adaptively estimating the intermodulation distortion, weighting it by a scalar constant and subtracting it from the subband of interest. We propose a method to adaptively compute the optimal weighting coefficient and show that it depends on the power and modulation of the blockers. Our results show that the pre-processing sample-based compensation method is more effective and that clear dynamic range extension can be obtained by using intermodulation compensation without resorting to increasing the sensing time. We also study the impact of uncertainties about the knowledge or estimates for nonlinearity parameters.