Bilinear Transform of besself coefficients leading to instability

Hey -

Creating a low pass filter with besself, and then using the bilinear transform to design a discrete time equivalent has been leading to instability.

[b a] = besself(4,80);
[num den] = bilinear(b,a,48000);

freqs(b,a) displays a stable, usable filter. However,

freqz(den,num,8192,48000) does not.

A quick check of the denominator poles,

abs(poles(den))

ans
0.998905200088155
0.998905200088155
0.998493252853120
0.998493252853120

indicates multiple poles extremely close to the unit circle. Can anyone offer suggestion as to what might be going wrong, or any alternative method known to produce usable results? Thanks. Start a New Thread