Effect of driver resonance on crossover

[Alex M]

I am designing a passive crossover for my three-way speakers, to replace the active crossover I have been using between bass and midrange, and am nearly there. I have modelled the driver response and impedance, designed impedance equalisation networks, and now have a crossover network that promises to do almost everything that I want.

The one thing that is holding me back is the predictions of the low-frequency response of the system. I'm using an Audax HM210Z0 woofer, which has an f0 of 30Hz, and am using a Linkwitz-Riley fourth-order filter at 422 Hz. Here is the low-pass filter circuit



This gives a near perfect response into a resistive 6.3 ohm load.

Now I have modelled the complex impedance of the HM210Z0 to give a good approximation to the measured impedance magnitude above about 50 Hz, and calculated a compensation network which produces an impedance that only varies by about 5% across the frequency range. This plot shows the impedance compensated first for the inductance only (green dashes), and then for the inductance and the resonance (green solid line):



You might think that, since the resonance frequency is less than 1/10 of the crossover frequency, its effect on the driver impedance would be negligible, and I expected that I would be able to get away with only a Zobel network to equalise the inductance, leaving the low-frequency peak untouched. However it seems to interact with the crossover to give a peak of about 4 dB at 100Hz:



This surprised me when I first saw it, but actually the part-equalised impedance (the dashed green line in the second plot) has risen to 7.9 ohms at 100 Hz, or 25% higher than the nominal impedance of 6.3 ohms. The combined impedance of the series inductors in the filter at 100 Hz is about 4.2 ohms, so with hindsight this is not so surprising.

The compensation circuit which cancels the resonant peak (giving the solid green curve above) has huge component values (15 mH and 1,200 uF), and it would seem difficult to achieve these values with high-quality parts.

The thing is, I've never heard anyone mention this effect before, so I start to question whether my models are correct. I'm sure that the HM210Z0 is not particularly unusual in its parameters. Has anyone seen this before? Is there an obvious solution that I am missing?

Thanks,

Alex

[slogan2112]

There is some discussion of this in the following threads.

PRELUDE

and

HTGuide Forum - Ardent Speaker Camp

it's mentioned in post #821.

[P.Lacombe]

Yes, a simple solution exists, but I cann't explain or calculate.

Forget the Zobel network. Decrease the inductors, and/or capacitors.
Cut and try. I admit this is laborious.... but it works, and very well !

Regards

[Pano]

I suggest Jeff Bagby's most excellent "Passive Crossover Designer". It runs in MS Excel.

With imported impedance curves, you can tweak and tweak virtual crossover parts to your heart's content. You may find a combination of values that gives you just what you need.

[bjorno]

Originally Posted by Alex M

..The compensation circuit which cancels the resonant peak (giving the solid green curve above) has huge component values (15 mH and 1,200 uF), and it would seem difficult to achieve these values with high-quality parts.

The thing is, I've never heard anyone mention this effect before, so I start to question whether my models are correct. I'm sure that the HM210Z0 is not particularly unusual in its parameters. Has anyone seen this before? Is there an obvious solution that I am missing?..

Hi Alex,

Have a look at my picture where you will find other component values for the resonance compensation:
Note that the filter you proposed doesn't include any DCR values in series with the coil inductors but is a must when simulating real filters.

b

[bentoronto]

Why would anybody willingly go from active to passive crossover?

I certainly may not not know what I am talking about, but the Zobel circuit helps the amp see a constant impedance. Not sure that makes any difference.

You are looking at speaker output as if it depended on voltage input. In fact, the cone has various kinds of corrective feedback. Adding elements to the crossover impairs the damping from the amp even if it provides some kind of really sanitary (linear) drive voltage.

The aim is to get feedback to the cone to fix that weak-link element, not to "purify" the drive signal.

[Ian Millar]

Having gone from 3-way passive single amp to active tri-amp and experiencing a total transformation of the sound from mush to thrilling, I agree that is seems a retrograde approach. Anyway perhaps he wants one of his amps for something else and is prepared for the inductor-mush in this system.

[Pano]

There are lots of active crossover fans on this forum. Understandable.

But some of the very best, world class systems I've heard used passive crossovers. They just happen to have been done right. Doing it right is easier with active, that's all.

[AllenB]

Alex M,
I think you'll be OK with a cheaper high DCR inductor as there'll be a series resistor in your resonance peak filter anyway. Just subtract DCR from the needed resistance, and make sure the structure can dissipate the power that will present across this resistance.

This is where a circuit simulator of some description could come in handy, or just use your hands and feel.

You can use an electrolytic capacitor as this filter doesn't work at high frequencies, and components here won't affect the sound too much.

BTW, you may be able to get the same result without the impedance compensation using different values in your filter, but you should manage good results this way too.


bentorino, put simply would you rather have a capacitor or a transistor in your path? Then, if I didn't sound controversial enough already I doubt I'd enjoy the sound of a valve amp if damping factor was such a problem

[bentoronto]

OK, let's get serious.

As best as I can imagine, you look at the speaker and believe it is having too much motion at resonance. So you want to create a crossover which gives some kind of inverted input to the driver.

As people who try to similarly iron-out room resonances can tell you, the resonance is the issue and cutting the input is a fairly pale approach to taming it.

What your Zobel concept (as Wikipedia seems to indicate) brought to my mind is the right way to handle untoward cone motions: motional feedback using a model of the driver as one leg in a Wheatstone Bridge which provides the feedback voltage (which is esp. friendly to do with tube amps, if that's your inclination as it seems to be).