We are implementing a chain of FIR filters, all with the config in the attachement but different coefficients!
Sadly after the first FIR filter (if we test with a "chain" of just one filter) we are experiencing an attenuation of approximately factor 32 which would mean that something with the fixed point arithmetics is off by 5 bits. But no matter how I try, I cannot seem to find the mistake ...
Can anybody with some more experience help me spot the hopefully obvious mistake?
It is probably from bit growth in the FIR. The core calculates bit growth based on the data/coef types and also the coefficient values (or worst-case in the case of reloadable coefficients). The core guarantees no overflow via bit growth.