Second Order Filters

[soundchaser]

Hi David;

I have had very good results with the canned crossover in Audio Weaver but was wanting to experiment with the SOF component. I'd like to give the Harsch crossover a try. BW4 lowpass with a BES2 high pass plus a half cycle of delay at the crossover frequency for the highpass.

I cannot for the life of me get a SOF to work in any kind of state; frequency response in Audio Weaver looks fine but I do not see any change in the response of the woofer. I typically am trying to use the variable Q LPF; cascade or even two singles; with Q's of .5412 and 1.3065 to make a 4th order BW. The filter has no effect on the measured FR (ARTA). Is there  box I have to check or some kind of gotcha?  I have attached the file. Maybe it will be obvious.

I know that I can use the canned BW filter (it works no problem) but at some point I have to make Bessel high pass filters and the SOF would allow that. Otherwise it's biquads and I have had some success with them but still ironing out some issues.


That’s the left woofer channel that is connected.

Thanks
Really enjoying the preamp.
Matt

[David Samborsky]

I believe the problem is that you need to place the interpolator after the filter and not after interleaver and you need to interpolate back up to 192kHz.

If you want to run everything at 96kHz you can force the hardware to run at 96kHz or 48kHz.

regards
david

[soundchaser]

Thanks David;

It's not that; although I do admit that putting it there might be incorrect. I cannot get second order filters to work no matter what. I was also having the same issue with biquads too but I got that sorted.

It turns out I was making an error in specifying the file type in the server when loading on to the target; I wasn't selecting "compiled script" which is what now makes the biquads work and all frequency responses are as predicted even with 192kHz sample rates. I am not doing any really low frequency operations; a shelf at 400Hz for baffle step is about the lowest so no decimation.

I also have to (thanks to your post on the AW forum) invert the signs of a1 and a2 and make sure that I use CSV or text files to set the coefficients in order to get sufficient precision. I view that as a complete miss on the part of DSP Concepts that they cannot simply be typed in to 8-10 decimal places but it's a small issue.

The file type change did not resolve the SOF issue. I had even resorted to using AW 8.XX to create a design but 6.18.02 Server to upload in case there was a bug in the AW models (probably not).

As for the above issue with file type; I don't recall consciously doing that when loading a canned 3-way crossover design based on the canned AW Butterworth/Linkwitz-Riley crossover macro model. All of those variations in crossovers worked right off the bat. So did the millisecond delays.

Anyways; as long as I have my Excel spreadsheet for computing the biquad coefficients all seems well. I can even load up multiple files and via the AW Server Flash Memory Manager switch between several different designs with a mouse click for measurement and listening comparison; there's a beep when you switch live but nothing too awful.

It's a lot of fun to experiment with different crossovers so quickly. This is a absolutely fantastic product.

Thanks and stay safe.

Matt

[David Samborsky]

Hi Matt

It's good to see you have made some progress. I didn't know about that Execute function so you learn something new everyday I knew you could load up different scripts but didn't know how you could utilize them. Some people want to have many different scripts and be able to select them from the front panel but the last time I inquired I don't think this can be done with this version of Audioweaver which isn't the latest btw but is the latest I have for the UP.

Regarding filter coefficients, I had an issue trying to implement LW transforms for low frequency EQ of sealed box enclosure. It seems that at 192k sampling rates there are too much round off errors at the lower frequencies so the filters aren't as accurate as they are at 48kHz. The solution is to calculate the coefficients at 48k and then down-convert from 192k to 48k and then later up-convert back to 192k again for the bass channel only !! The other channels should be able to operate at 192k without issues.

regards
david