• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

FIR filters via REW?

Tim Link

Addicted to Fun and Learning
Forum Donor
Joined
Apr 10, 2020
Messages
806
Likes
685
Location
Eugene, OR
Is that a typo? There are minimum phase FIR crossovers, but with FIR lin phase is more typical.
Not a typo. In REW you can apply parametric EQ and minimum phase crossover filters to your measurement and get a new predicted measurement. This is the way to do what I did. I wanted to be able to model my driver's phase response when it's hitting its target curve, so using this function in REW allows REW to make a modeled response that will actually have the phase that the driver will have when the driver's frequency response is matching the modeled frequency response. REW also lets you make a measurement from the EQ filters that were used to hit the target. So far, all that's been done is what could have been done with regular IIR filters. The phase correction comes in with the inversion of the minimum phase version of the target response to create what I think is called an all pass filter. It's basically another minimum phase filter that creates the opposite phase response, but the magnitude is unchanged by not allowing it to invert the frequency response during the inversion process. Instead the frequency response just ends up flat, so it has no effect.

So take that phase only impulse response and multiply it by the IIR EQ and crossover impulse response and now you have an impulse response that will create the desired crossovers and equalization to hit the target while also causing the driver to have flat phase.

For the woofer I made a target that assumed no bottom end roll off. That way the minimum phase group delay in the lower bass is maintained since the target won't try to correct for it.

If you apply a linear phase crossovers or filters to the driver in advance, the Hilbert transform won't work anymore since it's a mixed up thing at that point, making it more complicated to model. You could work around it but why bother when you're going to flatten the phase collectively anyways. The resulting FIR filter has all the desired phase corrections, crossovers, and EQ built in to it.
 
Last edited:
Top Bottom