FIR Filter, impulse response - am I doing it wrong

[bluecipher]

Hi all,

am I am pretty new to things like REW, DSPs, FIR filters etc., so please be patient...

I am using a Mini-DSP 2x4 HD to create a digital crossover for my two-way DIY speaker. After doing a "normal" XO I thought I give FIR filters a try, just for the fun of it. After doing a few rounds using REW for measurement and rePhase for filter design, I actually got something working that does sound better than before. I get a pretty decent flat frequency and phase reponse, however the impulse graph is looking a bit weird..

.

So basically I get a larger negative peak before the actual "smaller" positive reponse. Am I doing something terribly wrong here? Is that a sign of some huge mistake I'm making? Thanks for any hints!

 

[skid00]

Something in your signal chain may be inverting the impulse.

My Carver speakers are wired inverted...

You can invert the image in the settings.

 

[Foly83]

Hello blue
It would be interesting if you would upload your working steps possibly with pictures.
Maybe then more people can help you better.

In the step response, for example, there is always a pre-swing. This is normal for an FIR filter.

An analog or IIR filter usually produces a serious delay at the cutoff frequency, both in the transient response and in the decay response, which means that signals at the cutoff frequency enter with a delay, but then decay.
This effect should be called post-ringing.

Post ringing, on the other hand, is relatively "harmless" because there are no instruments/noise sources anyway.

I myself have played a lot with my Mini DSP and rePhase.
I would be happy if I can learn something here as well.
Because so really many instructions in relation to Mini DSP and rePhase I have not found.

Also look forward to correction if I'm wrong

many greetings

 

[bluecipher]

It might have been a matter of time aligning the drivers. But I have re-done my workflow and created some screenshots - both as a cry for help and to document where a newbie like myself might start off. So this is not meant as a "how to" (probably more a "how-not-to"...), but others might make the same mistakes as myself and might profit from helpful comments and corrections just like me.

Let me start off with what I am trying to do. As stated above, I have built a pair of speakers (to be precise the "Small Thor" enclosure suggested in an almost classic thread over at diyaudio. This is a two-way d'Appolito speaker using Seas Excel drivers



I've ripped out the passive crossover, so apart from a protective capacitor in line with the tweeter it's just the drivers, and I am trying to design a digital XO using FIR filters and a MiniDSP 2x4 HD.

As for the measurement setup, I am using a Behringer ECM8000 mic. I don't have a calibration file for it, so I just use it as-is and hope for the best. The following measurements were done with the mic at the same height as the tweeter and at a distance of 0.5 meters. I am using REW for the measurements and rePhase for the filter design.

I started with the woofer. I disabled any PEQ, XO etc. settings in th MiniDSP and ran a sweep with the following settings:



The raw data gives the following result



I then used Controls->IR Delay



set the IR windows (the following scrrenshot is actually not fully correct, I was using a right window of 50ms)



and applied a 1/6 smoothing. This resulted in the follwong SPL & Phase graph



(the magnesium membrane of the driver has a pretty breakup frequency between 4 and 5 kHz, easily visible in the graph) and a corresponding impulse response



to be continued...

 

[bluecipher]

I exported the measurement (see below) and imported it into rePhase



Following the "The rePhase FIR tool" tutorial by the MiniDSP folks, I first applied some "Paragraphic Gain EQ" adjustments to flatten the frequency response (even though I'll later choose a cutover frequency at 2 kHz I brought the breakup frequency around 4.5 kHz down a little). Under "Filter Linearization" I then applied a "vented high Q" rolloff at 35Hz and tried to somewhat flatten the remaining pahse curve using the "Paragraphic Phase EQ" tab. Generated the filter, exported as a bin file, imported into the MiniDSP and measured in REW again .





The woofer impulse response show (part of) that behavior I was mentioning in my first post. I then added a "2nd order" low-pass at 2000 Hz, re-generated the filter etc. and measured again





to be continued...

 

[bluecipher]

On to the tweeter... same workflow, only with a different gating (right 4ms) and a sweep starting at 200 Hz



giving an initial measurement of (see attachment)




Exported, imported in to rephase



flattened frequncy response, then phase. Exported, imported to MiniDSP, REW measurement





(see again the hefty negative jump pre-signal). And then added the high-pass 2nd order at 2000 Hz





just one to go...

 

[bluecipher]

Finally, I activated both channels and measured them together





At this point, the drivers were not time aligned, you see the negative pre-jump and (more importantly) some extinction around 3 kHz. I then delayed the tweeter response (somewhat hand-wringing it) by 0.15ms in the MiniDSP, which gave the following result





Interestingly, not only is the frequency response pretty flat now, but also the impulse reponse looks pretty nice to me. So probably in my initial try I wasn't aligning the drivers properly. Still, I would b very interested in any comments, tips, hints for approval, grave mistaked I made etc. For example - should I try to flatten the phase in the > 4kHz region, or is that not audible? What about the phase below 200 Hz, and how would you approach that?

Thanks so much (and sorry for the lengthy series of posts...)

 

[Foly83]

Hi blue
Thank you for the detailed explanation of how you proceeded.
Do you still have your two measurements from the tweeter and woofer before you processed them in rePhase?
Or maybe the complete measurement series as .mdat before and after correction?
Unfortunately I did not work with the Xo in rePhase. I had only adjusted the phase. Would try with the original measurement my method as soon as I have time

many greetings

 

[bluecipher]

The original measurements (txt files exported from REW) can be found in the posts above (woofer_unfiltered.txt and tweeter_unfiltered.txt), and I have added the REW measurements and rePhase filed below.

Another question: Using the "Box rolloff" to correct the phase in the base results in significant loss of SPL - is there a better way to do this?

 

[Foly83]

Hi blue
Wish you a happy new year
Have a look at the method that you have followed.
And created a correction based on your original measurement.
Would you load my two files into your DSP for the tweeter and woofer and make a measurement. I would be interested in what comes out.
The correction is without Xo
Here is a link with the files

Blue FIR - Google Drive

drive.google.com

And what I noticed.
You don't have a callibration file stored in REW. what kind of microphone are you using?
The "Box rolloff" I have not yet needed. I have a DBA configured as bass and make the Xo at 110 Hz, so I do not need this correction.
But can like to check that times with my D3.

many greetings