Apparently REW is capable of creating all manner of filters?
I'm not up to speed on DSP and Filters.
Infinite Impulse Response filters are created as Bi-quads, whatever they are. Inifinite means "uses feedback" or recursion - the data loops through the filter (?). Here's one:
b0=0.62337350986264013,
b1=-0.40715651524652960,
b2=0.13734218909046259,
a1=0.24408404926654605,
a2=0.07621183506691347
It's small.
REW can create those for me. They describe one "filter" - think one hump in a response curve. I get like 14 of those in my miniDSP the way it's set up.
Finite Response Filters are created as data to create a corrective impulse response and have a coefficient for every tap (memory location) in the DSP.
Fill the DSP with music samples, apply the filter to that whole chunk of data, to create a single new "filtered" sample. Push the next sample into the DSP memory (the old first one is pushed out) and create the next "filtered sample" to go to the DAC. Repeat the operation at 48kHz.
My miniDSP has 6144 taps so there are 6144 lines of data:
b0 = 0.0,
b1 = 0.0,
b2 = -2.954842926870989e-14,
b3 = -1.8055980430427276e-13,
b4 = -6.127623907413449e-13,
b5 = -1.5591428972547283e-12,
b6 = -3.3313874662349585e-12,
b7 = -6.325315705241197e-12,
...
b417 = -0.00022200765670277178,
b418 = -0.0002233331761090085,
b419 = -0.0002246562798973173,
b420 = -0.00022597686620429158,
b421 = -0.00022729481861460954,
....
b6138 = -4.289302420190655e-12,
b6139 = -1.7101571894717615e-12,
b6140 = -5.128505584615917e-13,
b6141 = -8.538305560626056e-14,
b6142 = 0.0,
b6143 = 0.0,
Plotted as a graph the data looks like this, around the middle of the file, This doesn't include the contribution of the IIR filters (low frequencies), that's another set of data:
An "empty" filter template (do nothing) would have a single value of "1" at tap 3144, the rest "0".
So, for whatever reason, REW doesn't make FIR filters (yet?), probably because it's a little bit new to have them in the hands of the unwashed masses (my guess).
I use AcourateDRC (think Acourate Lite) to do the dirty work of listening to a sweep tone, and coming up with a set of corrective IIR (low frequencies) and FIR (higher frequencies and phase correction) to shove into the convolver (the SHARC DSP in my little miniDSP).
Heres the measured impulse response at the preamp (no filter):
And with the filters for the JBLs applied (left channel):
You muck up the signal going to the amplifiers to "correct" what is received at the listening position.
Measured Preamp Step Response (no filter)
It crosses the zero line at 100ms - because 10Hz (100ms cycle time) was the lowest frequency in the sweep. Nice accuracy there in calculation and reproduction, I think.
It turns into this mess at the preamp output when the FIR and IIR filters are applied:
But it fixes things up for the listener after it goes through the speakers (they are a filter, of sorts).