diyAudio

diyAudio (http://www.diyaudio.com/forums/)
-   Multi-Way (http://www.diyaudio.com/forums/multi-way/)
-   -   Deliberately incorrectly doing tweeter impedance compensation (http://www.diyaudio.com/forums/multi-way/212866-deliberately-incorrectly-doing-tweeter-impedance-compensation.html)

sreten 18th May 2012 12:14 PM

Deliberately incorrectly doing tweeter impedance compensation
 
Hi,

The low pass filter in this Zaph design is elegant :Zaph|Audio

Its a 3rd order (1st order low pass + 2nd order notch) elliptical filter
that combines with the driver and box to deliver 4th order L/R.
Seen this used to good effect in a number of designs.

It struck me that if you are forced to compensate the tweeters impedance
peak in a design it allows a further option I've not seen described in a design.

Instead of simply compensating the tweeters impedance peak use this as
a further way of manipulating the actual high pass function of the tweeter.

E.g. say a tweeter needs compensating and 3rd order electrical to meet
an acoustic (typically L/R) 4th order target. It should be possible to
go 2nd order electrical and overdo the compensation (so it starts to
become a real notch filter) for a similar result.

As in the L18 low pass the eventual slopes will of course be different,
but what being discussed is hitting the acoustic targets around the
crossover point, the further away from that point the more you
can wander from your ideal acoustic function.

rgds, sreten.

speaker dave 18th May 2012 02:17 PM

A couple of comments.

I looked at the Zaph audio and rather than a 3rd order elliptical I would call it a first order with notch. A fine distinction but he isn't really trying to hit a real elliptical shape and the notch is well enough removed that, looking at the composite curve, you don't really know it is there.

A notch typically puts a quick transition into the phase curve. Phase swings rapidly approaching the notch and comes back in from the other extreme just beyond. This might give useful phase shift and approximate another order, especially if the dip is mild (gentler phase swing) but wouldn't you have to try it on a case by case basis?

I do think I have done a system in the past where I used a mild stop band notch to give a little better initial slope with less stop band phase shift, since the phase comes back. So it can be useful, but it is still hard to generalize about universal use.

I know there is a craze on this site towards the use of Zobels. They can also be set to overcompensate or undercompensate (at which point I wouldn't call them Zobels). For example, more often than not, I use what I call a damped 2nd order woofer network. L in and R and C in series to ground. People would look at it and call it a Zobel but I would not, since no attempt is made to flatten woofer impedance. I just find that 1, 2, or 3ohms at that juncture is often useful for controlling corner Q. We typically use lots of woofer inductor to set the midrange level (box step correction as some insist on calling it) and then adding the appropriate C value to put the corner at the desired frequency results in too high a Q. Hence the shunt R.

Working with optimizers is always enlightening. You iterate around to find the right topology and once you do the optimizer chugs away until the values are right and the necessary shape is achieved. It teaches you that crossover components, in the end, are merely degrees of freedom for adjustment. Put them in the right topology and the appropriate shape can be had.

"Zobel like" circuit sections can be generally useful, especially when you give up the somewhat artifical need to have flat impedance at some point in the circuit.

David S.

sreten 18th May 2012 02:26 PM

http://www.zaphaudio.com/audio-speak...erfunction.gif

Hi, You can't really see it in the above, but electrically it is an elliptical filter, rgds, sreten.

http://sound.westhost.com/articles/af-f12.gif
Figure 12 - Low Pass Cauer (Elliptic) Filter Response

Wayne Parham 18th May 2012 02:38 PM

I've been using that approach for decades for CD compensation. The first couple octaves above crossover typically need a flat transfer function, since that's where the compression driver is most efficient. Then, at mass rolloff, the slope needs to change to 6dB/octave. So what I do is to set the damping of the core splitter, essentially underdamping it to raise the bottom end of the curve, making it flat up to mass rolloff.http://www.pispeakers.com/Top_Octave_Compensation.gif

speaker dave 18th May 2012 02:44 PM

Do you have a phase curve to go with your figure 12?

DBMandrake 18th May 2012 02:46 PM

Quote:

Originally Posted by speaker dave (Post 3028085)
I know there is a craze on this site towards the use of Zobels. They can also be set to overcompensate or undercompensate (at which point I wouldn't call them Zobels). For example, more often than not, I use what I call a damped 2nd order woofer network. L in and R and C in series to ground. People would look at it and call it a Zobel but I would not, since no attempt is made to flatten woofer impedance. I just find that 1, 2, or 3ohms at that juncture is often useful for controlling corner Q. We typically use lots of woofer inductor to set the midrange level (box step correction as some insist on calling it) and then adding the appropriate C value to put the corner at the desired frequency results in too high a Q. Hence the shunt R.

That's exactly what I've done before in the woofer crossover of a 3 way...just by adjusting that resistor value you can shape the corner Q as you say.

Particularly useful if you've got the woofer deliberately close to the floor and you need to add some additional downwards slope or "droop" near the upper rolloff point to compensate for overlap between floor gain and baffle gain that you can get with a low woofer and a wide baffle, but without resorting to a 1st order rolloff.

Quote:

Working with optimizers is always enlightening. You iterate around to find the right topology and once you do the optimizer chugs away until the values are right and the necessary shape is achieved. It teaches you that crossover components, in the end, are merely degrees of freedom for adjustment. Put them in the right topology and the appropriate shape can be had.

With the right R & C values you have a 1st order with Zobel, with R reduced to zero you have a 2nd order without a zobel, in between those two extremes you have a whole continuum of response shapes, and as you say it is quite interesting to play with the values on a simulator.

Quote:

"Zobel like" circuit sections can be generally useful, especially when you give up the somewhat artifical need to have flat impedance at some point in the circuit.
Yep, I agree. Both Zobels and RLC compensators can be set to deliberately over or under compensate the impedance curve to get the desired acoustic response, I tried to make that point in the phase-alignement based method thread but it seemed to get lost in the noise... the example I gave of an "over compensating" RLC notch for a midrange drivers fundamental resonance in a small enclosure is one which I've used myself with great success when trying to use a large driver for midrange without ridiculous amounts of space devoted to the midrange enclosure...

DavidL 18th May 2012 04:46 PM

Sreten: Thanks for bringing this up. Looks like a very useful circuit

Speaker Dave: Yes optimizers are the way to go for designing. It's beyond my understanding why people still want to do a "trial and error" method for design when there are so many free software and measurement packages available.

DBMandrake: Using a RLC circuit to adjust for midrange drivers in a smaller than normal box is something I've never considered, thanks for the info.

:)

speaker dave 18th May 2012 05:02 PM

Quote:

Originally Posted by DBMandrake (Post 3028120)
Both Zobels and RLC compensators can be set to deliberately over or under compensate the impedance curve to get the desired acoustic response, I tried to make that point in the phase-alignement based method thread but it seemed to get lost in the noise... the example I gave of an "over compensating" RLC notch for a midrange drivers fundamental resonance in a small enclosure is one which I've used myself with great success when trying to use a large driver for midrange without ridiculous amounts of space devoted to the midrange enclosure...

Its a design school: "First I conjugate the drivers, then I add several dip response corectors, then the baffle step correction for the woofer and finaly a first order network. I like the simplicity of first order networks" !!!

David S.

ra7 18th May 2012 05:40 PM

Could you guys point to some free crossover designers/optimizers that you like to use?

speaker dave 18th May 2012 06:29 PM

I'm an old XOPT man myself. Very old.

I've also done a lot with LEAP crossover shop (not free). Passive Crossover Designer is the main free one that I know of. I have dabbled with it but am not proficient.

Is XOPT still available and (somewhat) supported?

David


All times are GMT. The time now is 08:52 PM.


vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright 1999-2014 diyAudio


Content Relevant URLs by vBSEO 3.3.2