Using a Parallel Notch Filter

[robertm]

Provided Link: Parallel Filter Calculator


I have a set of speakers with a known frequency response curve. Problem is, it has a hump at a very nasty location which is quite audible and annoying. I found this calculator which might assist in addressing this issue. The nasty peak is around 13KHz. Using this calculator with F1 at 9KHz and F2 at 14KHz (it drops off steeply after the peak), I get C=2.31uF, L=.06mH, R=13.78ohm. If I change F1 and F2 (which affects the steepness of the hump), I get varying values of R, the others are constant, so C is somehow related to the peak frequency I pick. I also notice that .06mH is quite small, is this a negligible value, so what I end up with is really just a cap and resistor in parallel? Nothing that small seems to be available, other than surface mounted type for PC boards. Does it matter if I put this circuit before the existing crossover for the speaker or after? Is this approach even valid? I am looking for something simplistic, not much technical knowledge here, just looking to fool around and learn by doing.

Calculator at:
<A HREF="http://www.carstereo.com/help/Articles.cfm?id=19">http://www.carstereo.com/help/Articles.cfm?id=19</A>

[Pete Schumacher ®]

> I have a set of speakers with a known
> frequency response curve. Problem is, it has
> a hump at a very nasty location which is
> quite audible and annoying. I found this
> calculator which might assist in addressing
> this issue. The nasty peak is around 13KHz.
> Using this calculator with F1 at 9KHz and F2
> at 14KHz (it drops off steeply after the
> peak), I get C=2.31uF, L=.06mH, R=13.78ohm.
> If I change F1 and F2 (which affects the
> steepness of the hump), I get varying values
> of R, the others are constant, so C is
> somehow related to the peak frequency I
> pick. I also notice that .06mH is quite
> small, is this a negligible value, so what I
> end up with is really just a cap and
> resistor in parallel? Nothing that small
> seems to be available, other than surface
> mounted type for PC boards. Does it matter
> if I put this circuit before the existing
> crossover for the speaker or after? Is this
> approach even valid? I am looking for
> something simplistic, not much technical
> knowledge here, just looking to fool around
> and learn by doing.

> Calculator at:
>
> <A HREF="http://www.carstereo.com/help/Articles.cfm?id=19">http://www.carstereo.com/help/Articles.cfm?id=19</A>

While the inductance is small, it's not negligible, and if you don't include it, you won't get the filter characteristic you're after.

[damkor]

I am looking for
> something simplistic, not much technical
> knowledge here, just looking to fool around
> and learn by doing.

I would put it right before the driver. You will probably need to do a little tweaking. It's much easier to equalize a wide peak than a narrow one.

Start with a 0.1mH inductor, and a 3uF cap. Have a 2.2uF cap handy. And you can simply unwind the inductor a turn at a time. You should have 8, 10, 15, and 20 ohm resistors on hand to try.

The larger the resistor, the deeper the notch. The larger the cap, the lower in frequency the effect. Ditto the inductor. If you don't think a small inductor has much effect, just try putting it in series with a tweeter on its own. The cap is small too, you still need it.

Compare those values with this site:

<A HREF="http://www.mhsoft.nl/ParallelNotchFilter.asp">http://www.mhsoft.nl/ParallelNotchFilter.asp</A>

[dlneubec]

> I have a set of speakers with a known
> frequency response curve. Problem is, it has
> a hump at a very nasty location which is
> quite audible and annoying. I found this
> calculator which might assist in addressing
> this issue. The nasty peak is around 13KHz.
> Using this calculator with F1 at 9KHz and F2
> at 14KHz (it drops off steeply after the
> peak), I get C=2.31uF, L=.06mH, R=13.78ohm.
> If I change F1 and F2 (which affects the
> steepness of the hump), I get varying values
> of R, the others are constant, so C is
> somehow related to the peak frequency I
> pick. I also notice that .06mH is quite
> small, is this a negligible value, so what I
> end up with is really just a cap and
> resistor in parallel? Nothing that small
> seems to be available, other than surface
> mounted type for PC boards. Does it matter
> if I put this circuit before the existing
> crossover for the speaker or after? Is this
> approach even valid? I am looking for
> something simplistic, not much technical
> knowledge here, just looking to fool around
> and learn by doing.

> Calculator at:
>
> <A HREF="http://www.carstereo.com/help/Articles.cfm?id=19">http://www.carstereo.com/help/Articles.cfm?id=19</A>

Well, PE sells 18 gauge Jantzen air core inductors at values of .025, .05. .10 and up. I have used all of these in a notch filter I'm working on and they all have a significant effect. Personally, I would pick up both a .05 and .1, 18 gauge to try.

[robertm]

> Well, PE sells 18 gauge Jantzen air core
> inductors at values of .025, .05. .10 and
> up. I have used all of these in a notch
> filter I'm working on and they all have a
> significant effect. Personally, I would pick
> up both a .05 and .1, 18 gauge to try.

You're right, I was looking at a dated catalog...

[curt_c]

would be to input the driver .frd and .zma files into the PCD crossover program and design your conjugate filter there. You will in all likelyhood be able to optimize the filter function far better than the textbook calculator approach. You can also specify a standard value inductor, and adjust the cap value to suit. Paralleling caps is easier than unwinding inductors...

C

[Deward Hastings]

Both the L and C are necessary . . . they are the critical components of a notch filter, and what determines its center frequency. The R determines the depth of the notch, and can often be omitted entirely. "Tuning" the notch requires some measurement/testing to make sure it sits on top of the resonance you wish to remove. As Curt suggests PCD makes it much easier to visualize what is happening that the simple calculators do . . . and it is not necessary to load driver data to "see" the effect of the filter. Of course it is better if you have the tweeter FRD . . . it makes it easier to "tune" the notch to the frequency you want.

But . . .

You can get close, and probably close enough to get the result you seek, with the following (note the 0.1 inductors on the "Special Sale" page at Madisound) "brick wall" filter which knocks out everything above 11-12 kHz while having essentially no effect at all below 10 kHz:

0.1 mh in series with (followed by) 0.1 mh in series with 1.5 mfd, those two in parallel with the tweeter.

,, 0.1 mh
---wmw----------
,,,,,,,,,,, |
,,,,,,,,,,, s , 0.1 mh
,,,,,,,,,,, s
,,,,,,,,,,, |
,,,,,,,,,,, = , 1.5 mfd
,,,,,,,,,,, |
--------------------

I hope that comes out right . . . it looks right in "preview" . . . the commas are there to force formatting . . .
You can further "tune" the notch (if you wish to) by "adjusting" either the capacitor or the second inductor, whichever is more convenient for you. Mount the inductors at right angles to each other (and as far apart as possible) to avoid interaction/coupling between them.

[robertm]

> Both the L and C are necessary . . . they
> are the critical components of a notch
> filter, and what determines its center
> frequency. The R determines the depth of the
> notch, and can often be omitted entirely.
> "Tuning" the notch requires some
> measurement/testing to make sure it sits on
> top of the resonance you wish to remove. As
> Curt suggests PCD makes it much easier to
> visualize what is happening that the simple
> calculators do . . . and it is not necessary
> to load driver data to "see" the
> effect of the filter. Of course it is better
> if you have the tweeter FRD . . . it makes
> it easier to "tune" the notch to
> the frequency you want.

> But . . .

> You can get close, and probably close enough
> to get the result you seek, with the
> following (note the 0.1 inductors on the
> "Special Sale" page at Madisound)
> "brick wall" filter which knocks
> out everything above 11-12 kHz while having
> essentially no effect at all below 10 kHz:

> 0.1 mh in series with (followed by) 0.1 mh
> in series with 1.5 mfd, those two in
> parallel with the tweeter.

> ,, 0.1 mh
> ---wmw----------
> ,,,,,,,,,,, |
> ,,,,,,,,,,, s , 0.1 mh
> ,,,,,,,,,,, s
> ,,,,,,,,,,, |
> ,,,,,,,,,,, = , 1.5 mfd
> ,,,,,,,,,,, |
> --------------------

> I hope that comes out right . . . it looks
> right in "preview" . . . the
> commas are there to force formatting . . .
> You can further "tune" the notch
> (if you wish to) by "adjusting"
> either the capacitor or the second inductor,
> whichever is more convenient for you. Mount
> the inductors at right angles to each other
> (and as far apart as possible) to avoid
> interaction/coupling between them.

I have no idea as to the "why", but I'll add it to the different approaches to try.

[robertm]