Why do we actually talk about Butterworth or LR filters. Butterworth has a peak of 3 dB at the XO frequency. But unless this peak has to compensate for a dip, then you want to compensate for this, so i think, you probably just move the highpass a bit up, and the lowpass a bit down so that it fits. And everyone writes that you can't cakculate the XO anyway and have to adjust the components, since a drivers is not a pure resistor but an impedance that depends on the frequency. When you have adjustet the filter and it and measured and/or sounds good, you still end up with something that does not fit with Butterworth or LR. So why this distinction at all.
I have a 2-way where there is 2. order on the tweeter. Here is a 8.2 µF capacitor and a 0.22 mH coil. It doesn't suit either Butterworth or LR. And it must be because the designer has taken into account some characteristics of the tweeter and perhaps also the the bass. So what kind of filter is it?
I have a 2-way where there is 2. order on the tweeter. Here is a 8.2 µF capacitor and a 0.22 mH coil. It doesn't suit either Butterworth or LR. And it must be because the designer has taken into account some characteristics of the tweeter and perhaps also the the bass. So what kind of filter is it?
You are perfectly right that the variable load impedance seen by the filter networks results in a deviation of the "pure" theory behind Butterworth or LR filters (and others). The "Butterworth", "LR" , etc. terminology therefore does indeed make very little sense with passive loudspeaker xovers.
The fact is that what matters it the acoustical slope, and not the electrical one. Unless you cross a driver at a frequency where both FR and impedance are flat in an extended region beyond and after that frequency, applying a textbook crossover will result in an acoustic slope different from the electrical textbook slope. I have successfully designed LR4 crossovers with 2nd order and/or 3rd order electrical networks.
As for the original question, you can't know what acoustical filter is without a FR measurement. You can more or less derive it if you know the mfg published impedance and FR.
Ralf
+1
I concur, it is the acoustic magnitude and phase response that defines the filter type and order.
regards.
Peter
It makes as much sense with a passive crossover. The driver/speaker variations figured in... Any reason it shouldn't?
To answer your original question, what type of filter is it you would need supporting measurements of the electrical response. It is second order electrical, but in terms of what classical 2nd order filter slope it matches roughly matches is an unknown. You could make educated guesses by assuming 4-, 8-, or 16-ohm driver impedance and applying the representative filter equation's,
As it is electrical 2nd order the acoustical response of the filter and the driver combined could actually be 3rd or 4th as the electrical filter and driver acoustic Rolloff of the driver will combine.
I can only assume the designer chose the values to give the best compromise in terms of the measured combined response with the combined response of the following mid or tweeter. Modern software allows you to add or remove filter components easily, so after half an hour of running many different sims the original idea of using 2nd ,3rd or 4th order filters may have changed several times.
I do not worry too much about filter orders, only that the simpler ones are cheaper to implement. Hence many speakers from years gone by have relatively simple filters, helped by less stiff or heavy cones that rolled off nicely without too much need of a steep filter slope.
As it is electrical 2nd order the acoustical response of the filter and the driver combined could actually be 3rd or 4th as the electrical filter and driver acoustic Rolloff of the driver will combine.
I can only assume the designer chose the values to give the best compromise in terms of the measured combined response with the combined response of the following mid or tweeter. Modern software allows you to add or remove filter components easily, so after half an hour of running many different sims the original idea of using 2nd ,3rd or 4th order filters may have changed several times.
I do not worry too much about filter orders, only that the simpler ones are cheaper to implement. Hence many speakers from years gone by have relatively simple filters, helped by less stiff or heavy cones that rolled off nicely without too much need of a steep filter slope.
Tanks for resopns.
I read "the crossover design cookbook" by Mark Lawrance last night. It was when I read about the different XO types and their Q that I became totally confused. For Butterworth, Q was 0.7 and for LR 0.5. He describes that you must not go higher than 0.7, as you get an uneven frequency curve with a lot of "ringing". But what is Q when the XO components in the right value are neither Butterworth, LR, Bassel nor one with high Q anyway.
I know that the electrical order and the acoustic order do not have to be the same. As I understand it, the electrical order will simply add what the unit has naturally, so if you cut with the 2nd order filter, where the unit itself rolls off with e.g. 12 dB then it is acoustically seen as a 4th order, as I understand it.
I read "the crossover design cookbook" by Mark Lawrance last night. It was when I read about the different XO types and their Q that I became totally confused. For Butterworth, Q was 0.7 and for LR 0.5. He describes that you must not go higher than 0.7, as you get an uneven frequency curve with a lot of "ringing". But what is Q when the XO components in the right value are neither Butterworth, LR, Bassel nor one with high Q anyway.
I know that the electrical order and the acoustic order do not have to be the same. As I understand it, the electrical order will simply add what the unit has naturally, so if you cut with the 2nd order filter, where the unit itself rolls off with e.g. 12 dB then it is acoustically seen as a 4th order, as I understand it.
Butterworth is typically used for odd order crossovers (1st order, 3rd order, etc). This is because for odd order crossovers the drivers are 90degrees out of phase at the crossover frequency. -3dB + -3dB = +3dB, but if the drivers are 90degrees out of phase this results in -3dB so on-axis you get 0dB. Listening off-axis may result in a +3dB peak if the path length difference between each driver and the listener causes the drivers to then become in-phase at the listening position.
Linkwitz-Riley is used for even order crossovers (2nd order, 4th order, etc). The drivers are in phase at the crossover frequency. -6dB + -6dB = 0dB
Speaker drivers are minimum phase devices like passive crossovers, so if you are using a Butterworth electrical filter to "compensate for a dip", the acoustic response and phase alignment that results may no longer be Butterworth at all. It is very valid that you may use a "2nd order Butterworth" electrical filter, in order to achieve a "2nd order Linkwitz-Riley" acoustic response in that scenario.
There are few cases that you should use a Butterworth alignment over an Linkwitz-Riley alignment. One case is using a Butterworth alignment for an MTM arrangement to lessen nulls in the off-axis response.
Linkwitz-Riley is used for even order crossovers (2nd order, 4th order, etc). The drivers are in phase at the crossover frequency. -6dB + -6dB = 0dB
Speaker drivers are minimum phase devices like passive crossovers, so if you are using a Butterworth electrical filter to "compensate for a dip", the acoustic response and phase alignment that results may no longer be Butterworth at all. It is very valid that you may use a "2nd order Butterworth" electrical filter, in order to achieve a "2nd order Linkwitz-Riley" acoustic response in that scenario.
That's absolutely fine. Just worry about targeting the desired acoustic Q and slope. The slope and Q of the electrical filter will be whatever it needs to be.
There are few cases that you should use a Butterworth alignment over an Linkwitz-Riley alignment. One case is using a Butterworth alignment for an MTM arrangement to lessen nulls in the off-axis response.
Last edited:
As I wrote above, the variable impedance of the drivers is part of the xover filter function. I don't see how a typical passive xover system could follow the theoretical filter function of a Butterworth or LR filter, which assume constant (or very high) load impedance.
I could ask the same question with regards to what to do about acoustic variations... which any crossover will have to deal with. The needed filters will not look like the target.
There really are three ways that a person can take the original question asked in this thread, regardless of which way Henrik meant...
1. Using a text book filter designed to work into a resistance, with a real speaker impedance, doesn't make much sense.
2. Using whatever filter gives the right electrical response into a real speaker load is a step in the right direction, but ignores the driver acoustic response variations.
3. Making the driver acoustic response into a perfect LR or Butterworth is closer to what this discussion should be about... except even that ignores the speaker acoustic variations (such as power).
So add a fourth if you will, but whatever the case, we need to be on the same page.
There really are three ways that a person can take the original question asked in this thread, regardless of which way Henrik meant...
1. Using a text book filter designed to work into a resistance, with a real speaker impedance, doesn't make much sense.
2. Using whatever filter gives the right electrical response into a real speaker load is a step in the right direction, but ignores the driver acoustic response variations.
3. Making the driver acoustic response into a perfect LR or Butterworth is closer to what this discussion should be about... except even that ignores the speaker acoustic variations (such as power).
So add a fourth if you will, but whatever the case, we need to be on the same page.
That's not quite true for frequency response... unless all you care about is the on axis response. All Butterworth crossovers have a 3dB peak to one side (e.g. above), and a null to the other side (e.g. below), in the vertical plane. They also have the property that, when you reverse the phase of one driver, the location (above or below) of the peak and null simply swap - the pattern flips "upside down" from the normal in-phase phase one. When you integrate over all axes the contributions from the peak and dip happen to cancel out, which is why the Butterworth responses have constant power (power takes into account radiation in all directions). This is true for all orders.
Also, keep in mind that these effects are only for the vertical axis. In the horizontal plane there is no change caused by the crossover itself as you move off axis, no matter what the crossover type, assuming the drivers centers are aligned along a vertical line on the baffle.
In contrast, the LR crososvers (AKA Butterworth squared) sum to flat on axis, but both above and below in the vertical plane fall off - the LR response shape in the vertical plane is like a large, broad lobe.
The vertical axis responses come about because there is a different delay of the wavefront from each driver at the observation point as you move up or down from the "on axis" plane. The claims made above assume that the two "drivers" have infinite bandwidth and have no time or phase difference at their acoustic center. This is never the case for a real loudspeaker, and in a real loudspeaker the vertical axis response family can be more complicated. I wrote an Excel spreadsheet to study all these effects at one time, and I could post some examples here if anyone is interested in more gross detail. It is similar to what you can read about on this web page:
https://www.tonmeister.ca/wordpress...o-build-a-good-loudspeaker-part-1-crossovers/
I prefer LR responses overall, consider both the frequency and time domain responses. But I would not really lose too much sleep over these issues. The effects are not all that important in domestic settings and home audio loudspeakers.
My point was that the frequency response is NOT FLAT for Butterworth types, off axis, even for "perfect" models. In fact it is not even flat on axis for most loudspeakers because the drivers are not like a resistor in terms of phase/delay. Since most DIY builder never bother to think about or measure anywhere else except at one "on axis" point, they might never realize it.
Please read the link I shared above for more info.
A response peak is much more obtrusive than a response dip, at any frequency. When a 3dB peak is happening at 2kHz or so, that can be a real and audible problem. I prefer no peaks. Thus it's a "no thanks" to Butterworth for me.
I am glad there are so many wise minds in this forum who have responded to my question. I can see that my question is not quite easy. Much of what you are saying is at a higher level than what I currently understand and there is also disagreement, I can see. In Danish you would say "I'm confused on a higher level" I don't know if that saying goes in English.
I had some thoughts ambut changing my tweeter Scanspeak D2905/970000 with a D3004/ 662000. But I wanted to understand my XO first, soI don't have to redesign the whole thing. I did not design my XO myself. It is from 1998, but I think it was made by a master in the field. When I read the thing about Q value, which I have not yet found out how XO affects it, I became curious since most XOs still deviate from the pure BUT and pure LR and you might end up somewhere in the middle in between, what is Q. I don't know the XO frequency and I am aware that you can't just calculate it from the 2 components 8.2 µF and 0.22 mH as it depends on how the unit measures before the XO comes on.
Now I have actually just tried to enter the values into a calculator program and there the 2 values will fit quite nicely on a Chebychev with a Q of 1, and according to what I read, it should give a terrible sound with many peaks and a lot of ringing. It is not characteristic of the speaker at all.
I had some thoughts ambut changing my tweeter Scanspeak D2905/970000 with a D3004/ 662000. But I wanted to understand my XO first, soI don't have to redesign the whole thing. I did not design my XO myself. It is from 1998, but I think it was made by a master in the field. When I read the thing about Q value, which I have not yet found out how XO affects it, I became curious since most XOs still deviate from the pure BUT and pure LR and you might end up somewhere in the middle in between, what is Q. I don't know the XO frequency and I am aware that you can't just calculate it from the 2 components 8.2 µF and 0.22 mH as it depends on how the unit measures before the XO comes on.
Now I have actually just tried to enter the values into a calculator program and there the 2 values will fit quite nicely on a Chebychev with a Q of 1, and according to what I read, it should give a terrible sound with many peaks and a lot of ringing. It is not characteristic of the speaker at all.