Allpass filter for lowpass to M-band transformation
[AllpassNum,AllpassDen] = allpasslp2mb(Wo,Wt)
[AllpassNum,AllpassDen] = allpasslp2mb(Wo,Wt,Pass)
[AllpassNum,AllpassDen] = allpasslp2mb(Wo,Wt) returns
AllpassNum, and the denominator,
Mth-order allpass mapping filter for performing
a real lowpass to real multipassband frequency transformation. Parameter
the number of times an original feature is replicated in the target
filter. This transformation effectively places one feature of an original
filter, located at frequency Wo, at the required
target frequency locations, Wt1,...,WtM.
By default the DC feature is kept at its original location.
[AllpassNum,AllpassDen] = allpasslp2mb(Wo,Wt,Pass) allows
you to specify an additional parameter,
chooses between using the "DC Mobility" and the "Nyquist Mobility."
In the first case the Nyquist feature stays at its original location
and the DC feature is free to move. In the second case the DC feature
is kept at an original frequency and the Nyquist feature is movable.
Relative positions of other features of an original filter do not change in the target filter. This means that it is possible to select two features of an original filter, F1 and F2, with F1 preceding F2. Feature F1 will still precede F2 after the transformation. However, the distance between F1 and F2 will not be the same before and after the transformation.
Choice of the feature subject to this transformation is not restricted to the cutoff frequency of an original lowpass filter. In general it is possible to select any feature; e.g., the stopband edge, the DC, the deep minimum in the stopband, or other ones.
This transformation can also be used for transforming other types of filters; e.g., notch filters or resonators can be easily replicated at a number of required frequency locations without redesigning them. A good application would be an adaptive tone cancellation circuit reacting to the changing number and location of tones.
Design the allpass filter changing the real lowpass filter with
the cutoff frequency of
a real multiband filter with band edges of
defined. Plot the phase response normalized to π, which is in
effect the mapping function
Please note that the transformation works in the same way for both
positive and negative frequencies:
Wo = 0.5; Wt = [1:2:9]/10; [AllpassNum, AllpassDen] = allpasslp2mb(Wo, Wt); [h, f] = freqz(AllpassNum, AllpassDen, 'whole'); plot(f/pi, abs(angle(h))/pi, Wt, Wo, 'ro'); title('Mapping Function Wo(Wt)'); xlabel('New Frequency, Wt'); ylabel('Old Frequency, Wo');
Frequency value to be transformed from the prototype filter
Desired frequency locations in the transformed target filter
Numerator of the mapping filter
Denominator of the mapping filter
Frequencies must be normalized to be between 0 and 1, with 1 corresponding to half the sample rate.
Franchitti, J.C., "All-pass filter interpolation and frequency transformation problems,"MSc Thesis, Dept. of Electrical and Computer Engineering, University of Colorado, 1985.
Feyh, G., J.C. Franchitti and C.T. Mullis, "All-pass filter interpolation and frequency transformation problem," Proceedings 20th Asilomar Conference on Signals, Systems and Computers, Pacific Grove, California, pp. 164-168, November 1986.
Mullis, C.T. and R.A. Roberts, Digital Signal Processing, section 6.7, Reading, Massachusetts, Addison-Wesley, 1987.
Feyh, G., W.B. Jones and C.T. Mullis, An extension of the Schur Algorithm for frequency transformations, Linear Circuits, Systems and Signal Processing: Theory and Application, C. J. Byrnes et al Eds, Amsterdam: Elsevier, 1988.