Increase EQ filter count?

[Jonathanese]

Hello all. This is my first post. So welcome me or call me a noob or what have you.

So I have been producing EQs in REW to export to EqualizerAPO. However, I found the method of dumping a bunch of seperate filters to be somewhat inefficient. I figured instead of calculating all the filters all the time, you could just run an impulse through the filters and convolve it over your output.

Well sure enough I found the "Export filters impulse response as WAV". Trim it down to 0.05s and it runs well in EqualizerAPO's convolver. (Can you set the export length?)

So the thing is, a convolution of a 0.05s WAV file will always use the same CPU. So the number of filters used to create it is irrelevant. So in that case, is it possible to bump the total filters up above 20?

Yes I'm aware that if you need more filters than this you probably have some very big problems. I just notice the EQ generator will do things like neglect certain resonance peaks in favor of broader changes elsewhere. I'd like to reduce the amount of compromises it has to make.

In fact ideally, you might run an impulse through a transfer function that simply has the inverse of the measured response. Though probably smoothed so your highs don't sound like garbage. That could be the equivalent of thousands of filters, but with convolution it runs exactly the same.

 

[skid00]

You can manually add filters to the .txt file that you might use in eqAPO (if you export the generated filters created by REW, to a text file).

You maybe could generate 20 filters for 10 Hz to 1,000 Hz, export, then generate another 20 filters for 1,000 Hz to 20,000 Hz. Then convolve all 40, maybe?

 

[Jonathanese]

Ohh interesting idea. What it might entail is doing a measurement, creating a correction filter impulse response for 10-1000Hz, then applying this filter and running another test, then having it match 1000-20000Hz.

So in other words, 2 separate tests. REW seems to match your ideal response within the range but it isn't returning to flat response outside that range, this would need to be accounted for.

But I suppose if it's 32-bit FP at 192kHz or even something nuts like 384kHz+, I can convolve the two impulse responses together to create a single impulse response that accomodates both. The extra steps concern me due to potential losses. But it would be interesting to try.

 

[Jonathanese]

OK so quick update regarding this.

I realize now that you can import impulse responses in to REW, multiply(convolve) their responses, and export the resulting impulse response.

So basically, I can do the following:
1. Test with no filter
2. Create a correction EQ
3. Export impulse as WAV
4. Rerun test using this WAV in Equalizer APO
5. Create stage-2 correctional EQ
6. Export impulse as WAV
7. Import stage 1 and stage 2 impulses
8. Do trace arithmetic A*B to convolve their responses
9. Export the A*B result to get a combined FIR based on what would be a total of up to 40 filters.

From there, I tweak the WAV in Audacity as follows:
1. Cut to 50ms before impulse and 100ms after impulse
2. Select all then Effect -> Normalize and normalize to 0dB

The results have been somewhat promising. A 1-stage setup manages bass really well but completely ignores or amplifies frequencies around 1k and 2.5k resulting in it adding "honk". With stage 2, it seems to prioritize these issues a little more and pull them back. Though maybe flat response is just that fatiguing anyway.