![]() | Only 14 pages are availabe for public view |
Abstract The measure~ent of spectral density of signals is subjected to random and bias errors • The truncation of data is a source of bias errors which must be considered carefully when making ~easurements • The thesis contains a review of types of errors encountered in spectrum measurement, and the different data windows used • The effect of data windo·11 choice is investigated and it was found that the proper choice of data ’llinc’.o·;r depends on the shape of the SJe::tru:n of the si(P.’lal to be e1easured • Different ’llindo·;:s were tested using dif£’ere!1t types of signals • The rectan(r-<lar window ·>~as found to be sui t3- ble for measuring spectrn of periodic signals while the Ha~~ing window was found to be suitable for measuring smooth spect..,-a.· • A dual - channel spectrum analyzer •lfhich consists of two parallel cha.nnels is proposed One channel having a HaMming window suitable for measuring smooth spectr~~s and the other having rectang’c~lar window sui table for measuring spectra’ with impulses • It is shoW!’! that the measured spectrwn has less bias error than that obtained by usi~’S conventional single channel spectr’xn an .1lyzer having r ”ct:a!’.gular or Harr~ing wincow separately • Moreover no deep ~owlecge is required to make th2 proper ch0ice of the data ·rindo’ll suitable to the signal, as it is the case in singl~ ch~~nel spectrum anal::,”zer but it is done automatically using the proposed dual-channel spectrum analyzer • The dual - ch~~el spectrum analyzer has been simulated using digital computer and satisfactory results were obtained |