Bug in ss2tf and ss2zp

[Michiel Koot]

The problem appears to be in tzero.m as the following script demonstrates

% begin m-file

g = 1e16;

num = randn(1,4);
den = [ 1 g*randn(1,4) ];
k   = num(1)/den(1);

[z1,p1,k1] = tf2zp(num,den);

[a,b,c,d] = tf2ss(num,den);

[z2,p2,k2] = ss2zp(a,b,c,d);

[z3,k3] = tzero(a,b,c,d);

% end m-file

For g=1e14 everything works fine, as k=k1=k2=k3, but for g=1e16 k2 and k3
become huge numbers.

I haven't found the exact bug yet, as to me the routine tzero.m is pretty
complex.

Michiel