A nonparametric allpass filter design method is presented for matching a desired group delay as a function of frequency. The technique is useful in physical modeling synthesis of musical instruments and emulation of audio effects devices exhibiting dispersive wave propagation. While current group delay filter design methods suffer from numerical difficulties except at low filter orders, the technique presented here is numerically robust, producing an allpass filter in cascaded biquad form, and with the filter poles following a smooth loop within the unit circle. The technique was inspired by the observation that a pole-zero pair arranged in allpass form contributes exactly 2 radians to the integral of group delay around the unit circle, regardless of the (stable) pole location. To match a given group delay characteristic, the method divides the frequency axis into sections containing 2 total area under the desired group-delay curve, and assigns a polezero allpass pair to each. In this way, the method incorporates an order selection technique, and by adding a pure delay to the desired group delay, allows the trading of increased filter order for improved fit to the frequency-dependent group delay. Design examples are given for modeling the group delay of a dispersive string (such as a piano string), and a dispersive spring, such as in a spring reverberator.
[1]
Steve Winder.
Introduction to digital filters
,
2002
.
[2]
Davide Rocchesso,et al.
A Dispersive Resonator in Real Time on MARS Workstation
,
1992,
ICMC.
[3]
Trevor J. Terrell.
Introduction to Digital Filters
,
1980
.
[4]
Julius O. Smith,et al.
Bark and ERB bilinear transforms
,
1999,
IEEE Trans. Speech Audio Process..
[5]
Zhongqi Jing,et al.
A new method for digital all-pass filter design
,
1987,
IEEE Trans. Acoust. Speech Signal Process..
[6]
G. R. Reddy,et al.
Digital all-pass filter design through discrete Hilbert transform
,
1990,
IEEE International Symposium on Circuits and Systems.
[7]
Markus Lang.
Allpass filter design and applications
,
1998,
IEEE Trans. Signal Process..
[8]
Autoregressive Modeling.
Design of Recursive Group-Delay Filters by Autoregressive Modeling
,
.
[9]
Stephen A. Dyer,et al.
Digital signal processing
,
2018,
8th International Multitopic Conference, 2004. Proceedings of INMIC 2004..
[10]
M. Lang,et al.
Simple and robust method for the design of allpass filters using least-squares phase error criterion
,
1994
.
[11]
Vesa Välimäki,et al.
Tunable dispersion filter design for piano synthesis
,
2006,
IEEE Signal Processing Letters.