Honored Contributor
Honored Contributortake your coefficients generated from matlab and perform roots(a) and then roots(b) to derive conjugate pairs. Then poly(a(1:2)) poly(a(3:4)) etc and same for "b". Then match your pairs for optimal shape and lowest noise gain in your system. Then scale so that coeffcients normalised - this may require an extra add with respect to "A"s if any greater than "1" (you will find out what I mean when you try), whereas B can be scaled together. Then chose biquadratic strategy that matches your requirements ... I suggest you research Transpose forms 1 and 2. Try to find one with lowest "summer node" gain.
Now only thing remotely related to FPGA ... good lucj with fxed point ... make sure you properly analyse your filter with respect to stability and regenerative noise gain. FPGAs are not floating point.
