Purify / Low Distortion Filter for (Rigid Cone) Mid-Woofers

[fineMen]

Here you might find some info on how to prevent to emboss the distortion profile of rigid cone woofers by the common resonant peaks.
See article "Low Distortion Filter ... " at https://purifi-audio.com/tech/

I never thought that would be possible, often told people otherwise, but here you are. I tested it with a small metal-cone midrange driver. The cone's resonance lies around 6666Hz. In short, Purify's filter it works perfectly. Thank's to Purify, great job!

Distortion without serial notch filter:

With filter, alas the peak wasn't hit perfectly. But the reduction in third HD is already astounding. I would say even with this slight miss it gave anyway fully satisfactory results!

Here the filter's response, measured acoustically.

at

 

[fineMen]

Second take. I only wonder why the previous post was read 100s of times, but wasn't considered comment worthy. One 'like', appreciated!

I performed 'for good measure' a more clean investigation. Additional results.

The driver is a coax, an open tweeter with about 2kHz of basic resonance in the center. I didn't short-cut the coil. Maybe there's something else to be optimized.

Anyway, the mid was (a) driven without a passive filter, but was taylored with a digital equalizer to have a very similar frequency response to the (b) passive filtered same driver. The comparison looks like the following. The match isn't perfect, but the filtered one is louder, hence should show more distortion. The absolute level isn't right, as the measurements were performed with 50cm distance to the cone; substract 6dB.

Comparison of HD with digitally equalized versus passive filter:

The distortion component HD3 is reduced to roughly a third (watch the logarithmic scale!) with the passive serial notch filter. This may or not keep it well under the perception limit. HD4 is presumably contaminated with noise. So I won't argue with it further.

Comparison of intermodulation IM with digitally equalized versus passive filter:

There's virtually no effect worth discussing. Maybe the IM is mostly Doppler. All those nasty peaks, though, won't really fit the naive imagination. This little driver has some private life ... which for the time being remains a secret.

ps: the distortion products around 1,3kHz are amplified by a mild, broader resonance around 4kHz. I could lower it with the same technique, but it came out to be a bit ineffective.

 

[jamescarter1982]

really interesting is there a link to more info about the purifi filter ?

 

[fpitas]

Second take. I only wonder why the previous post was read 100s of times, but wasn't considered comment worthy. One 'like', appreciated!

It's interesting mainly if one designs with metal cone drivers.

 

[fineMen]

really interesting is there a link to more info about the purifi filter ?

Sure, I gave it in the first post? And it's actually not Purify's invention. Or, kind of. To point to the merits of the serial notch saved a few drivers I would have dismissed otherwise.

It's interesting mainly if one designs with metal cone drivers.

Not just, e/g the SB Acoustics 6" NRX driver would also benefit from that serial notch.

 

[Tangband]

Here you might find some info on how to prevent to emboss the distortion profile of rigid cone woofers by the common resonant peaks.
See article "Low Distortion Filter ... " at https://purifi-audio.com/tech/

I never thought that would be possible, often told people otherwise, but here you are. I tested it with a small metal-cone midrange driver. The cone's resonance lies around 6666Hz. In short, Purify's filter it works perfectly. Thank's to Purify, great job!

Distortion without serial notch filter:
View attachment 274598

With filter, alas the peak wasn't hit perfectly. But the reduction in third HD is already astounding. I would say even with this slight miss it gave anyway fully satisfactory results!
View attachment 274610

Here the filter's response, measured acoustically.

View attachment 274606
at

Interesting tests ! It has also been discussed in a thread by MAB.

Active DSP DIY speakers, for testing a few ideas

FFLOTSAM (Fabricated From LeftOver Things Studio Active Monitor) Active DSP speakers, for testing a few ideas I had some stuff lying around; a few random Baltic Birch-ply panels, some countertop material, a few cans of partially used brown spray paint, and some not-too clumpy spar varnish (that...

www.audiosciencereview.com

Conclusion is : The distortion with passive notch filtering often gets lower in such a case, but in the Swedish expert-forum faktiskt.io the conlusion by i-or is like this : - better use a good DSP crossover and use a steeper slope without any notch filtering at all and get the frequency response absolutely flat with the help of peq and eq . This will in most cases anyway be a better sounding option overall compared to an all passive solution. If one parameter gets better with a passive notch filter then the active dsp crossover will give you two or three parameters thats superior to the passive crossover technique, - one of the advantages is the ability to use steeper slopes if its needed and get a ruler flat frequency response with the help of peq with the correct Q.

…One interesting option is to use a DSP crossover and combine this with a passive notch filter. Maybe a good solution for hard coned active speakers ?

Better still is (maybe) to use drivers without any serious breakups ( à la Genelec )

My own experience in this is that metal cone drivers or other hard cone drivers that needs a steeper crossover slope than 4:th order ( acoustically ) really benefit from active dsp crossing , i.e its to much fuss to make a passive crossover to sound good in such a case. This is ofcourse a very different story with well behaved drivers and lower order crossovers where a passive crossover will behave much better.

 

[jamescarter1982]

series notch meaning all components worked in the positive cable in a row before the driver rather than across its terminals would be parallel is that correct . interesting if some parts of a passive crossover reduce distortion because this is a much simpler approach than active dsp crossover requiring more amps etc which has always been touted as the better sounding option

 

[fpitas]

series notch meaning all components worked in the positive cable in a row before the driver rather than across its terminals would be parallel is that correct . interesting if some parts of a passive crossover reduce distortion because this is a much simpler approach than active dsp crossover requiring more amps etc which has always been touted as the better sounding option

It works, as Bruno describes, by reducing the coil current at resonance.

 

[jamescarter1982]

am I correct in my description of a series notch filter though I'm new to crossover design

 

[fpitas]

am I correct in my description of a series notch filter though I'm new to crossover design

Well, it's a parallel LCR trap used in series.

 

[kemmler3D]

To @Tangband 's point, I think you actually need the electrical filter even if you use DSP, for the reason @fpitas mentions. If and when I get around to building any decent speakers, I intend to (try to) implement this.

From what I have ready on DIYAudio, this "trick" really helps with high quality metal drivers, which tend to have a sharp but controlled break-up region. Cross over well below the breakup and notch out the resonant peak, and you magically end up with less distortion below the filter frequency. The fact that this works is somewhat beyond my understanding of electronics, but it seems intuitively plausible anyway...

 

[Tangband]

To @Tangband 's point, I think you actually need the electrical filter even if you use DSP, for the reason @fpitas mentions. If and when I get around to building any decent speakers, I intend to (try to) implement this.

From what I have ready on DIYAudio, this "trick" really helps with high quality metal drivers, which tend to have a sharp but controlled break-up region. Cross over well below the breakup and notch out the resonant peak, and you magically end up with less distortion below the filter frequency. The fact that this works is somewhat beyond my understanding of electronics, but it seems intuitively plausible anyway...

One can always combine a dsp crossover with a passive notch , this time maybe without cons.

With that said, the lesser distortion with a passive notch is probably less noticable than a frequency response thats variable +-2 dB , vs ruler flat .

 

[Tangband]

To @Tangband 's point, I think you actually need the electrical filter even if you use DSP, for the reason @fpitas mentions. If and when I get around to building any decent speakers, I intend to (try to) implement this.

From what I have ready on DIYAudio, this "trick" really helps with high quality metal drivers, which tend to have a sharp but controlled break-up region. Cross over well below the breakup and notch out the resonant peak, and you magically end up with less distortion below the filter frequency. The fact that this works is somewhat beyond my understanding of electronics, but it seems intuitively plausible anyway...

A passive notch is cheap and easy to implement

 

[fpitas]

One can always combine a dsp crossover with a passive notch , this time maybe without cons.

With that said, the lesser distortion with a passive notch is probably less noticable than a frequency response thats variable +-2 dB , vs ruler flat .

Also, if the peak is far into the stopband of the crossover, it's probably less important to do this.

 

[fineMen]

Interesting tests ! It has also been discussed in a thread by MAB.

Active DSP DIY speakers, for testing a few ideas

FFLOTSAM (Fabricated From LeftOver Things Studio Active Monitor) Active DSP speakers, for testing a few ideas I had some stuff lying around; a few random Baltic Birch-ply panels, some countertop material, a few cans of partially used brown spray paint, and some not-too clumpy spar varnish (that...

www.audiosciencereview.com

Uups, I didn't look that up. But now you have another verification. I was so excited to be now able to use that (dirt cheap) otherwise nearly KEF-excellent coax in a real quality speaker, sorry.

Thank yo

 

[Tangband]

Uups, I didn't look that up. But now you have another verification. I was so excited to be now able to use that (dirt cheap) otherwise nearly KEF-excellent coax in a real quality speaker, sorry.

Thank yo

Is it the Omnes cx3.3 youre experimenting with ?

 

[tmuikku]

To @Tangband 's point, I think you actually need the electrical filter even if you use DSP, for the reason @fpitas mentions. If and when I get around to building any decent speakers, I intend to (try to) implement this.

From what I have ready on DIYAudio, this "trick" really helps with high quality metal drivers, which tend to have a sharp but controlled break-up region. Cross over well below the breakup and notch out the resonant peak, and you magically end up with less distortion below the filter frequency. The fact that this works is somewhat beyond my understanding of electronics, but it seems intuitively plausible anyway...

Think this: you have DSP low pass filter at 1kHz and play a sine tone at 1kHz. There is cone resonance showing peak in drivers datasheet at 3kHz, well above low pass filter. The resonance peak is 10db. Important thing to understand here is that the peak is not the amplifier boosting it, its the driver (cone) that amplifies any signal, so any 3kHz current through voice coil gets additional 10db boost in acoustic domain. You can reduce the input signal by 10db with DSP, and it gets boosted back by the driver in acoustic domain, and we measure smooth response with a microphone for the input signal, cool.

Now, the driver motor has varying parameter like voice coil inductance Le(x) which changes with excursion, and as you crank up the 1kHz tone the voice coil moves, Le changes and third order distortion component at 3kHz is induced into the voice coil as current. Current in the voice coil turns into voice coil movement at the very moment, is amplified by the cone resonance and you'd see high 3rd order distortion for 1kHz in distortion plot. This all happens after the DSP and it can do nothing about it. The distortion is probably similar with any frequency and similar excursion, thing here is that at 3kHz the cone resonance works as amplifier and elevated distortion reading ensues.

Driver motor non-linearity makes this distortion in impedance, which means with constant voltage input that current in the circuit varies with the nonlinear impedance, the distortion appears as current and emits acoustically right away. Reduction in distortion in acoustic domain can happen if current at 3kHz is reduced in the voice coil. This is another key thing to understand, the parallel notch filter needs to be in series with voice coil, same current flows through both simultaneously. The parallel notch filter in series with the driver increases circuit impedance at 3kHz reducing current through voice coil at 3kHz, including distortion current originating in the driver motor. Now also any distortion products that appear in the electric domain are EQ:d.

In general, one could increase circuit impedance also with a series coil or just with a resistor, or use high output impedance amplifier, what ever, to reduce distortion originating from driver motor. The parallel notch is handy for the breakup peak.

Not sure how audible this is, yet to listen this myself. I've got active system, having series coil as low pass in addition to DSP but not sure I hear any difference, its pretty low distortion system anyway. You can find out more about the subject by looking into current-drive. Read the Purifi paper again as well, also their other material available on the website. Hope it helps

edit.
As disclaimer I'm just a hobbyist and above text might have errors but this is how I understand the basic mechanism behind distortion reduction with circuit impedance, based on Purifi papers and other resources.

 

[Tim Link]

or use high output impedance amplifier, what ever, to reduce distortion originating from driver motor.

Maybe this is why some people like tube amps. Also, perhaps I should be using a resistor on my midrange horn instead of turning down the volume Digitally to match the woofer and tweeter.

 

[Tim Link]

perhaps I should be using a resistor on my midrange horn instead of turning down the volume Digitally to match the woofer and tweeter.

It occurs to me that a simple resister on the driver or a high output impedance amp might not work, reason being that in terms of distortion, it would be bringing the fundamental down as much as the distortion, so there'd be no net gain. You have to bring a higher frequency peak down relative to a fundamental. Am I getting this right? Or would the resistor still have a damping effect on all resonances? I just put a 10 ohm resistor on my midrange and turned the volume up to match before and after. The distortion peaks went from around 0.1 percent to 0.05 percent. Not sure if I trust that measurement because the noise floor went up a little with the resistor measurement.

 

[tmuikku]

Yes, you would have to crank put your amplifier output voltage, for the higher impedance load, to get same current through voice coil and same volume displacement and acoustic output as without the resistor. Some power would now be wasted as heat, but the load is also for the motor distortion current so some distortion ought to drop. Motor distortion depends on basically the excursion, not what voltage your amplifier has to do to make it happen. Only specific distortion though, not all, mostly 3rd order originating from motor non-linearities.