FIR filter behaviour

[EA Support]

EA Support

27 June 2025 at 6:54 pm #1076

I have a question surrounding the practical implementation of an FIR filter in an audio processor.

If a filter is created utilizing FIR Designer with stopbands defined at 500 and 20000hz; is there any expectation that the filter, once implemented into a processor, will affect the ranges outside the defined stopbands?

I haven’t yet set this up in a lab to perform a transfer function measurement; however, listening (and instrumenting) to the results through a system appear to reveal something is going on below the 500hz filter stopband.

Thoughts?

Possibly. It really depends on the design. For example, on the Mag Adjust tab, add a minimum-phase 500 Hz Butterworth high-pass filter prototype, then on the Export tab, set the filter delay to 0 and the filter length to 100 samples. Then on the Export tab, set the “FIR Filter Frequency Response” plot “Level : min” to -100 dB.

You will see that the FIR filter doesn’t match the ideal below 100 Hz and levels out at roughly -40 dB below 50 Hz. Then if you turn on “DC Compensation”* the FIR filter LF improves but doesn’t perfectly match the ideal.

IIR high-pass filters work very low in frequency due to the recursion in the filter topology. FIR filters are somewhat limited in LF reach by their finite length, and so can’t always achieve the ideal (before truncation and windowing) at low frequencies. (The longer the filter, the better for LF!)

* Optional “DC Compensation” is a process, in the FIR filter calculation, where:
– if any high-pass filter prototypes are used on the Mag Adjust tab, the FIR filter is guaranteed to be -infinity dB at DC, or
– if no high-pass filter prototypes are used on the Mag Adjust tab, the FIR filter is guaranteed to be 0 dB at DC.

Also, take a look at the FIR filter guide…

FIR Filter Guide

Regards,
EA Support