Sinteza i realizacija uskopojasnih rekurzivnih digitalnih filtara nepropusnika opsega
Докторанд
Krstić, Ivan B.Ментор
Nikolić, Saša V.Чланови комисије
Stančić, GoranMilovanović, Gradimir
Nikolić, Tatjana
Lekić, Predrag
Метаподаци
Приказ свих података о дисертацијиСажетак
New IIR multiple-notch filters design methods, whose utilization
lead to complete satisfaction of specifications regarding positions
of notch frequencies and notch-bandwidths defined for an arbitrary
attenuation value at cutoff frequencies, are investigated in
this thesis. Most of the proposed design methods offer the possibility
of further pass-bands magnitude response improvement
by increasing the order of the filter. On the other hand, since
proposed design methods are based on the assumption that IIR
multiple-notch filter with desired magnitude response can be obtained
by parallel connection of the delay line and the stable allpass
filter, pass-bands magnitude response improvement is followed
by the improvement of linearity of the pass-bands phase
response. Comparison with the existing IIR multiple-notch filter
design methods showed that proposed design methods could result
in shorter duration of transient response to sum of sinusoidal
interferences, as well as in im...provement of pass-bands phase response.
Sensitivity of the IIR multiple-notch filter’s magnitude response
to coefficient quantization is considered from the point of view
of the right choice of existing (standard) and development of the
new realization structures. Algorithm for determination of the
multipliers coefficients’ quantized values such that needed number
of bits for theirs representation is minimized, as well as the
algorithm for determination of the optimal representations of the
all-pass filter transfer function (characterizing the IIR multiplenotch
filter), realized by cascade connection of the first- and
second-order nonstandard all-pass sections, are proposed. Efficiency
of the proposed algorithms is verified through several examples.
General conclusion is that significant reduction in total
number of bits needed for representation of multipliers coefficients
can be achieved if nonstandard realization structures are used.