Back in 2000 when I was looking at and extending the B&O filler driver approach to TP crossovers I came across an alignment techniques where by fillers (band pass, BP) responses with any order symmetric roll off could be mated with HP and LP section yielding flat response. I never wrote anything up on it because while interesting I was looking for TP crossovers and just brushed off the higher order 3-ways. Anyway, a couple of years ago the Duelund crossover came to my attention. Steen Duelund had an interesting approach to developing a 3-way crossover with all drivers in phase. Recently I made the connection between Duelund’s approach and what I had done earlier for the case where the filler has symmetric 2nd order slopes. This renewed my interest in these "higher order" 3-ways that I had found. They are not the same as the Duelund in that the higher order filters don't always sum flat with all drivers having the same phase. In fact, only even systems with even order filler (BP) roll offs do that. Those with odd order fillers sum flat in quadrature.
However, the interesting thing is that for cases where the filler has a reasonable wide BP response, all these crossovers can be developed from cascaded LR or Butterworth filter stages making them, or corresponding acoustic targets much simpler to construct than may appear form the Duelund paper. I have posted a web page discussing this approach for the Duelund case where the filler is symmetric 2nd order, and for fillers with symmetric 3rd and 4th order responses.
I though some of you might be interested.
However, the interesting thing is that for cases where the filler has a reasonable wide BP response, all these crossovers can be developed from cascaded LR or Butterworth filter stages making them, or corresponding acoustic targets much simpler to construct than may appear form the Duelund paper. I have posted a web page discussing this approach for the Duelund case where the filler is symmetric 2nd order, and for fillers with symmetric 3rd and 4th order responses.
I though some of you might be interested.
Thanks for your efforts
I second your remark about the usability of the texbook values for the cross-over components. I also only use the theoretical slopes as a target slope.
In my experience for the bandpass additional filter components are only required if the driver is very wideband or needs to do part of the baffle step correction.
In practice the biggest hurdle is to compensate for differences in relative driver position. In my experiments I always had to offset the filler driver to arrive at reasonable results.
I think the original link for the article of Steen was: this one
I second your remark about the usability of the texbook values for the cross-over components. I also only use the theoretical slopes as a target slope.
In my experience for the bandpass additional filter components are only required if the driver is very wideband or needs to do part of the baffle step correction.
In practice the biggest hurdle is to compensate for differences in relative driver position. In my experiments I always had to offset the filler driver to arrive at reasonable results.
I think the original link for the article of Steen was: this one
Hi,
Another interesting related article :
http://www.linkwitzlab.com/crossovers.htm
With another variation on a practical 3-way.
/sreten.
Another interesting related article :
http://www.linkwitzlab.com/crossovers.htm
With another variation on a practical 3-way.
/sreten.Absolutely. I need to get back to more crossover work and had intended to try the Duelund. For some reason, the spreadsheet I have does not let me output the targets to a file.john k... said:
Let me start with targets. I was going to say that if one were to optimize the woofer lowpass and tweeter highpass independently and as long as there were no acoustic offset, then I would think that the filler driver could be created simply by optimizing for the filler with the woofer and tweeter sections locked with a flat summed response as target. However, looking again at your writeup makes me think that this may not be correct, although I'd like to find out. The case where n is odd would be simpler in that the change in rolloff order beyond the original slope may be brought into it by the optimization process, though that may be a stretch given some of the odd results of some optimizers if the starting point is not close to the final target.
I have another comment and a question as well. First, it's interesting to note that for cases where n is odd, the filler driver must have a maximum amplitude equal to that of the other two drivers in the central section of the bandpass due to phase quadrature summation. For the cases where n is even, there can be a degree of "bandpass gain" in that the filler driver can have an optimized sensitivity less than that of the other two drivers. The even case has a bit more flexibility for the absolute sensitivity of the filler driver, with 1 or 2db leeway depending on the targeted response and Fc spread of the crossover.
The question is in regard to your statements:
and
but you state that
In constructing a 3-way, I typically put a midrange driver into a box and use a complementary highpass to effectively shift the pole higher (passively) for a 2nd order highpass above the natural Fcb. This could be done for the case of the filler driver as well, could it not? As long as the pole does not need to be lowered (i.e. a Linkwitz Transform), why would this not work for the filler driver highpass? The tweeter is being adjusted this way, otherwise we'd just have to use a tweeter with its natural highpass. I guess I'm thinking in terms of a higher-Q closed box (Qtc>0.5) with a crossover to yield the desired Qtc=0.5 with a somewhat higher Fcb.
Dave
john k... said:
Hi John
What is the advantage of higher order BP filters over Duelands 1st order approach?
What do you suggest about Duelands loudspeaker equalization networks? ( ie conjugate matching the drivers impedance)
see http://www.steenduelund.dk/download/Loudspeaker impedance correction.pdf
With regard to impedance comp. Completely unnecessary unless needed. When is it needed? Either because the acoustic target can't be met without impedance comp on the driver or because, for what ever reason, the amplifier needs to see flat impedance. These are two very different issues. But ultimately any impedance comp results in power being dissipated in the comp networks robbing power from sound generation.
With regards to higher order its just a matter of protecting the driver better. The point of my write up was really to show how the Duelund and other targets can be easily constructed using the common LR and Butterworth responses when gamma is greater that 1.0. There is little point in a 3-way design with gamma 1.0. Why build a 3-way with gamma = 1 when an LR4 2-way will basically be a better, simpler design? I think gamma = sqrt(2) is probable as low as one would consider a true 3-way. At that point the BP will still cover only 2 octaves.
dlr,
As you know there are lots of ways to make 3-way. Bullock developed another series of 3-way system with symmetric slopes at both crossover points. So unless you specify a specific a target, what the result will be, other than "flat" is kind of up for grabs.
With regard to n = odd, the polarity can also be set so that the HP polarity is inverted relative to the LP. Then the HP and LP will have the same phase and gamma = 1 reduces to, for n = 3, an LR6 2-way. As gamma increase when woofer and tweeter are inverted relative to each other, a dip at the crossover appears and the filler will bring the response to flat, with the filler can be of either polarity.
What I was saying with regard to the Duelund 3-way HP on the mid is that ultimately the mid HP should have to have two poles at Fx of the mid/woofer x-o. A Q = 0.5 box alignment will give you that. On the other hand, alignment with Q < 0.5 will result with two poles, one above the box Fc and one below Fc. In this case, if the higher frequency pole is at the desired Fx the lower pole can be shifted upwards pretty easily using passive components without giving up sensitivity. If you are going to put another high pass filter on the box then you have two set of poles associated with the HP response, one set where they aren't supposed to be. It's not that you can't build a 3-way that way, but you won't achieve any of these alignments that way.
Basically all I was getting at is that with the Duelund 3-way for gamma >1 the HP sections, and the LP for that matter, required that the poles of the acoustic response be located at Fx of the HP and LP crossover points. Since the driver's (mid and tweeter) poles associated with the roll off about the driver Fs aren't like to be there then they will have to be shifted by what ever means necessary.
However, while all this looks very elegant from a mathematical point of view, in reality achieving it acoustically is probably a pipe dream. No matter how you look at it there are still the poles associated with the woofer box alignment that are hanging around and which will influence the woofer phase, not to mention the poles associated with the low pass roll off of the raw midrange and woofer response. I think the best you can hope for is maybe to have all driver in phase from somewhere below the w/m crossover to above the m/t crossover. But I suspect that once the woofer is down 30dB or so the phase will deviate from the target. But if it's down 30 dB, so what?
With regards to higher order its just a matter of protecting the driver better. The point of my write up was really to show how the Duelund and other targets can be easily constructed using the common LR and Butterworth responses when gamma is greater that 1.0. There is little point in a 3-way design with gamma 1.0. Why build a 3-way with gamma = 1 when an LR4 2-way will basically be a better, simpler design? I think gamma = sqrt(2) is probable as low as one would consider a true 3-way. At that point the BP will still cover only 2 octaves.
dlr,
As you know there are lots of ways to make 3-way. Bullock developed another series of 3-way system with symmetric slopes at both crossover points. So unless you specify a specific a target, what the result will be, other than "flat" is kind of up for grabs.
With regard to n = odd, the polarity can also be set so that the HP polarity is inverted relative to the LP. Then the HP and LP will have the same phase and gamma = 1 reduces to, for n = 3, an LR6 2-way. As gamma increase when woofer and tweeter are inverted relative to each other, a dip at the crossover appears and the filler will bring the response to flat, with the filler can be of either polarity.
What I was saying with regard to the Duelund 3-way HP on the mid is that ultimately the mid HP should have to have two poles at Fx of the mid/woofer x-o. A Q = 0.5 box alignment will give you that. On the other hand, alignment with Q < 0.5 will result with two poles, one above the box Fc and one below Fc. In this case, if the higher frequency pole is at the desired Fx the lower pole can be shifted upwards pretty easily using passive components without giving up sensitivity. If you are going to put another high pass filter on the box then you have two set of poles associated with the HP response, one set where they aren't supposed to be. It's not that you can't build a 3-way that way, but you won't achieve any of these alignments that way.
Basically all I was getting at is that with the Duelund 3-way for gamma >1 the HP sections, and the LP for that matter, required that the poles of the acoustic response be located at Fx of the HP and LP crossover points. Since the driver's (mid and tweeter) poles associated with the roll off about the driver Fs aren't like to be there then they will have to be shifted by what ever means necessary.
However, while all this looks very elegant from a mathematical point of view, in reality achieving it acoustically is probably a pipe dream. No matter how you look at it there are still the poles associated with the woofer box alignment that are hanging around and which will influence the woofer phase, not to mention the poles associated with the low pass roll off of the raw midrange and woofer response. I think the best you can hope for is maybe to have all driver in phase from somewhere below the w/m crossover to above the m/t crossover. But I suspect that once the woofer is down 30dB or so the phase will deviate from the target. But if it's down 30 dB, so what?
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.