Active Cauer-Elliptic filters?

[BobEllis]

Looking at Jon's filters got this active filter die hard thinking. Amplifiers, opamps and vector board I have aplenty. So rather than spend the bucks on passive XO's...

Is there any advantage to using an active elliptical filter that approximates an 8th order L-R over just doing the 8th order L-R?

I've modeled Jason's active circuit with 5 spice, and the bounce comes up to -20 db before resuming its fall. With an RS180 the breakup modes will be barely 30 db down, which doesn't seem like enough.

I also tried adding another 2nd order section to Jason's circuit, which gets around 7th order initial rolloff with the bounce staying >30 db down. Of course the HF needs to be inverted and the filter frequencies adjusted to avoid a big ripple at XO frequency.

I haven't tried modeling the elliptic sections in Lancaster's "Active Filter Cookbook" which combine a state variable notch filter and a second order section.

So my question boils down to what type of filter would you recommend for an active Modula type speaker (pushing tweeter low end limits and midwoofer with nasty breakup issues)? ~7th order elliptical (3 op amps), Lancaster style ~8th order inital rollof Elliptical (4 op amps) or straight 8th order L-R (4 op amps)?

[Mark K]

Looking at Jon's filters got this active filter die hard thinking. Amplifiers, opamps and vector board I have aplenty. So rather than spend the bucks on passive XO's...

Is there any advantage to using an active elliptical filter that approximates an 8th order L-R over just doing the 8th order L-R?

I've modeled Jason's active circuit with 5 spice, and the bounce comes up to -20 db before resuming its fall. With an RS180 the breakup modes will be barely 30 db down, which doesn't seem like enough.

I also tried adding another 2nd order section to Jason's circuit, which gets around 7th order initial rolloff with the bounce staying >30 db down. Of course the HF needs to be inverted and the filter frequencies adjusted to avoid a big ripple at XO frequency.

I haven't tried modeling the elliptic sections in Lancaster's "Active Filter Cookbook" which combine a state variable notch filter and a second order section.

So my question boils down to what type of filter would you recommend for an active Modula type speaker (pushing tweeter low end limits and midwoofer with nasty breakup issues)? ~7th order elliptical (3 op amps), Lancaster style ~8th order inital rollof Elliptical (4 op amps) or straight 8th order L-R (4 op amps)?

Hi Bob,

You should be able to keep the bounce under 40-50 dB. You can try Jon's trick and use 8th order LR targets, or, try John K's spreadsheets and target slightly different targets. (See his posts on the MAD board.) I've been curious about some of these as well. )

Don't ask me anything about op amps though... :Z

[noah katz]

What do you think of a Behringer DCX2496 with 48 dB/oct and a notch filter at the bass/mid's breakup freq?

I read where a guy used additional parametrics to steepen the slopes near XO to effect an elliptic filter.

I'm parroting, what *is* an elliptic filter, anyway? I gather it's steep.

Thanks

[oneoldude]

Bob

Have you tried modelling the LR8th and the 4thCE and checked to see where those troublesom breakup modes would be placed?

Then a peak at phase and group delay might be worth while. Although I believe that SL says it really dosen't matter because those variations are inaudible. At least that is what I understand he says. I surely could be wrong there though.

If you do the above I hope you show us what you get.

Perhaps Jon will venture an opinion?

[JonMarsh]

The alignment I use in the Modulas, Arvos, and other projects has less group delay than a classic LR-8, and I prefer it for that reason.

Cauer elliptic filters are most often used in applications like microwave or RF filters. I used one in a Class D amplifier (output filter) paper presented at an AES conference (1991) and published in the AES journal. They are sometimes called equal ripple filters, because the voltage magnitude of the ripple in the pass band and stop band is the same. As Mark notes, I usually optimize the stopband ripple for the peak to be -50dB or better.

Duplicating this type of filter with linear active circuits or with the Behringer or other digital means has been discussed in a few threads here, so the search engine will turn those up for those interested.

Being the crotchety luddite that I am, I'm still going passive...

~Jon

[BobEllis]

Thanks all. Lower group delay sounds like the ticket. I hadn't figured out how to display that (if possible in 5Spice)

I am guilty of posting before searching. Sorry.

[ThomasW]

Cauer elliptic filters are most often used in applications like microwave or RF filters.

I opened up one of my microwaves and didn't see anything that looked like a CE filter .... :B

[noah katz]

"Duplicating this type of filter with linear active circuits or with the Behringer or other digital means has been discussed in a few threads here, so the search engine will turn those up for those interested."

Will do, thanks.

[JonMarsh]

Thanks all. Lower group delay sounds like the ticket. I hadn't figured out how to display that (if possible in 5Spice)

I am guilty of posting before searching. Sorry.

OK, as Moderator, I guess I have to administer the punishment of two lashes with a wet noodle, self inflicted. Take that and some egg nog with a little Rum, and call us back tomorrow morning.

[JonMarsh]

I opend up one of my microwaves and did see anything that looked like a CE filter .... :B

Be careful about opening up your microwave... if you let out all the magic smoke, it won't work anymore!

~Jon

And no, I haven't been into the egg nog yet, though I did pick some up. Forgot to get the rum.

[Dennis H]

Short version, start with a relatively high-Q 4th order crossover -- Butterworth is the only choice with the digital boxes, it's variable with an analog circuit. You may need a parametric filter at Fc to adjust the Q with a digital box. Add a notch filter (several with a digital box to get the depth) a bit more than an octave above or below the nominal Fc. Optimize all the F, Q and gain numbers with a program like LspCAD until you've got an amplitude response like LR8 an octave above and below Fc and the stopband ripple (bounceback after the notch) is at least -50dB.

[john k...]

Thiele presented the topology for an active 6th order which would yield an LR8 like initial response. I intended to add that to my spread sheets but I haven't gotten to it yet. Basically two 3nd order S&K filter with different Q's are cascaded with a third active stage (for the notch, which has a bridged T topology. A simple notch filter will give a "look alike" type response, but not the complimentary HP/LP pairs as per Thiele.

FYI, I have combined the spread sheets for the 4th, 5th and 6th order passive topologies and hopefully it will be posted at the FRDC site in the near future.

[EdL]

Has anyone considered mechanical damping of the higher frequency? Theil Audio uses an aluminum diaphragm similar in size to the RS180. It includes a foam (styrofoam?) attachment to the backside. It appears to vary in thickness: thicker at the voice coil, tapering to the surround.

edit: spelling