Active DSP DIY speakers, for testing a few ideas

[Ra1zel]

Some pieces of actual knowledge that can be extracted out of this all:

1. Use good drivers with low distortion in the first place.
2. Recognize that the distortion reduction seen from using passive notch filters has to do with current source so the argument about some kind of superiority of passive components is complete nonsense.
3. Good speaker and crossover design, like no substantial directivity errors or no whole opera of diffractions, will certainly give you more than lowering HD by some hundreds of a percent on already good drivers.

 

[MAB]

Some pieces of actual knowledge that can be extracted out of this all:

1. Use good drivers with low distortion in the first place.

Or in this case, drivers with such low distortion that the harmonic components of the cone’s breakup mode are the dominant distortion. These Excel drivers have very low distortion, so you can really see the resonance artifacts. This seems to be the case with many drivers these days that have gone to more rigid cones, you may get low distortion but you may have resonances. I think this is what motivated Purifi to write the whitepaper.

2. Recognize that the distortion reduction seen from using passive notch filters has to do with current source so the argument about some kind of superiority of passive components is complete nonsense.

Yes, it was all spelled out in the PuriFi whitepaper. I don't think passive or active are in general superior. I'm certainly not arguing for passive crossovers here! But it does make sense that using a passive notch can provide a type of electromagnetic damping of the driver that an active notch can't. And if you want to control that third order distortion peak the passive notch is the way to go, the active notch only attenuates the frequency of excitation and does basically nothing to tame the distortion byproducts of the breakup mode.

3. Good speaker and crossover design, like no substantial directivity errors or no whole opera of diffractions, will certainly give you more than lowering HD by some hundreds of a percent on already good drivers.

Yeah, these are intrinsically very low distortion drivers so this is like wringing the last little bit of distortion out of them. I do think the biggest problem with these drivers isn't distortion, it's that the tweeter and woofer conspire to have a large directivity mismatch at any reasonable crossover frequency when used as a 2-way which is a pretty common problem.

 

[tktran303]

Hi MAB,

I echo Mark’s sentiment, thanks for publishing your study on this. Another thing on my to-do list, checked off by you. @capslock @5th element Might want to check out this thread.

After Purifi published their white paper, it caused a little stir for me; it reopened opportunities to improving all those ancient stiff coned drivers with the fierce high Q breakup.

Interestingly, Jonmarsh @HTGuide has being using notch filters in series with drivers for a long time. Although he never documented it clearly the way Lars has done, he's used it as far back as his NatalieP design.
That thing is nearly 2 decades old. Recently he's even does it with tweeters such as the Peerless DA25TX in his Calliope speaker design:

Purifi white paper shows that a series notch can decrease breakup distortion - HTGuide Forum

Moving forward with your speaker experiment, you may be interested to trial an acoustic LR2 HP/LP. It will reduce the directivity mismatch, particularly if you're crossing around 2-3KHz; the T25CF002 should be able to handle this. If you must take an LR4; you may need to compromise a bit on the on axis/LW- introducing a mild droop around 1-2dB will allow a smooth PIR response.

I'm convinced that this is why people prefer acoustic LR2- it's about the horizontal dispersion/directivity. I think @augerpro believes it may be about the vertical (floor response) too, though I haven't gotten around to verifying either of these ideas.

@markk
Welcome back Mark! Any chance you can rouse Zaph? Turns out there’s no need to XO at 1/3 the breakup frequency!

 

[MAB]

Hi MAB,

I echo Mark’s sentiment, thanks for publishing your study on this. Another thing on my to-do list, checked off by you. @capslock @5th element Might want to check out this thread.

After Purifi published their white paper, it caused a little stir for me; it reopened opportunities to improving all those ancient stiff coned drivers with the fierce high Q breakup.

Interestingly, Jonmarsh @HTGuide has being using notch filters in series with drivers for a long time. Although he never documented it clearly the way Lars has done, he's used it as far back as his NatalieP design.
That thing is nearly 2 decades old. Recently he's even does it with tweeters such as the Peerless DA25TX in his Calliope speaker design:

Purifi white paper shows that a series notch can decrease breakup distortion - HTGuide Forum

Moving forward with your speaker experiment, you may be interested to trial an acoustic LR2 HP/LP. It will reduce the directivity mismatch, particularly if you're crossing around 2-3KHz; the T25CF002 should be able to handle this. If you must take an LR4; you may need to compromise a bit on the on axis/LW.

I'm convinced that this is why people prefer acoustic LR2- it's about the horizontal dispersion/directivity. I think @augerpro believes it may be about the vertical (floor response) too, though I haven't gotten around to verifying either of these ideas.

@markk
Welcome back Mark! Any chance you can rouse Zaph? Turns out there’s no need to XO at 1/3 the breakup frequency!

Thanks for the thread reference. Yes, very interesting and on topic!
I also started experimenting with these Seas drivers about 20 years ago. Took me longer to figure out what was going on!!!
Yeah, Zaph has great data on these Seas Excel driver's hard cone breakup modes, and remarked on the on the 3rd harmonic peak at 1.7kHz on the W18EX001 (sister to this driver).

Zaph|Audio

www.zaphaudio.com

It's like 1% 3rd order distortion right in the critical band of 1.5-2kHz on an otherwise extremely low distortion and wide-band driver. It's completely eliminated with the series notch, DSP has zero impact... So the passive notch stays! Plus, I have bins full of passive parts that need to be used somewhere!!!

I will try the LR2 crossover you suggest. Yeah, the TC25CF002 goes really low, I used them in the even lower but the heat dries out the ferrofluid if you drive them hard. I'll get my act together and try this!

 

[kipman725]

Interesting results but fully active current output amplifier + DSP would surpass the notch filter from my understanding of loudspeaker distortion mechanisms. Very stiff light cones with poor internal damping seem common for high end drivers. I'm not sure they are better although frequency response and harmonic distortion look fine once filters have been applied. Are you able to generate waterfall plots with and without the notches? if they also look fine then it would seem the poorly damped cones can be tamed.

 

[cavedriver]

Will definitely be reading back through this thread carefully as I'm working on using a miniDSP and a couple amps to direct drive a pair of JBL Stage A190's. Dennis Murphy already made an effective passive crossover for the A180 to solve the cone breakup and distortion problems with the cheap drivers they used and the original 2.5 way crossover design, but I don't want to redo a passive crossover for the A190 as a one-off so it's DSP time.

 

[jamescarter1982]

Really great work, analysis and presentation, highly appreciated! I would guess the 1.5 kHz HD peak is a third harmonic which maps to the reduced 4.5 kHz distortion, could you maybe check it?

how do you like the dewalt thicknesser next on my list for my little workshop not sure if I should get this one or the smaller one .

 

[AnalogSteph]

2. Recognize that the distortion reduction seen from using passive notch filters has to do with current source so the argument about some kind of superiority of passive components is complete nonsense.

In theory, the same effect should be achievable with an amplifier that provides an output impedance peak then, i.e. frequency-selective current driving. I think this could be done by replacing R3 in Fig. 6 by a parallel RLC? Hmm. I may have to get SPICEy later...

 

[MAB]

In theory, the same effect should be achievable with an amplifier that provides an output impedance peak then, i.e. frequency-selective current driving. I think this could be done by replacing R3 in Fig. 6 by a parallel RLC? Hmm. I may have to get SPICEy later...

Yes, I believe you are 100% correct.