Anyone here have any insights on asymmetrical crossovers? I'm looking to build a new set of 2-ways specifically for my living room with the idea being similar to the GedLee speaker lines Earl Geddes used to make and sell. I found the waveguide thread and software by mabat, and will be utilizing that in the design. But the topic is about the XO.
In my imagination, mating a high efficiency 15" woofer with a compression driver requires a low XO point. I don't think I want too steep of crossovers based on past experience w/r/t off-axis transition, though I suppose if I'm matching off-axis response of the woofer to the waveguide at the XO freq, I may not care as much. Since any compression driver will be so much more sensitive than any high efficiency woofer, I'm not concerned about running the CD lower than often happens. With that in mind, if I wanted to combine a higher order low-pass for the woofer with a lower order high-pass for the CD, the CD does that have any design merit? The idea is to roll off the woofer as it starts getting to "beamy," and letting the CD contribute a bit down lower in conjuction with the woofer. For example, 18 dB/octave LP with a 6 dB/octave high pass. I haven't researched this much, and it's been many years since I was designing speakers, but if I'm remembering correctly, those two filters should allow me to swap polarity on the CD and keep the signals in phase while they're summing together. I know that I won't be able to sum to a completely flat response in doing so, but can choose two frequencies to minimize any bumps/dips to be acceptable.
Anyway, does anyone have any experince with something like this? Things to look out for? Reasons it's a terrible idea?
In my imagination, mating a high efficiency 15" woofer with a compression driver requires a low XO point. I don't think I want too steep of crossovers based on past experience w/r/t off-axis transition, though I suppose if I'm matching off-axis response of the woofer to the waveguide at the XO freq, I may not care as much. Since any compression driver will be so much more sensitive than any high efficiency woofer, I'm not concerned about running the CD lower than often happens. With that in mind, if I wanted to combine a higher order low-pass for the woofer with a lower order high-pass for the CD, the CD does that have any design merit? The idea is to roll off the woofer as it starts getting to "beamy," and letting the CD contribute a bit down lower in conjuction with the woofer. For example, 18 dB/octave LP with a 6 dB/octave high pass. I haven't researched this much, and it's been many years since I was designing speakers, but if I'm remembering correctly, those two filters should allow me to swap polarity on the CD and keep the signals in phase while they're summing together. I know that I won't be able to sum to a completely flat response in doing so, but can choose two frequencies to minimize any bumps/dips to be acceptable.
Anyway, does anyone have any experince with something like this? Things to look out for? Reasons it's a terrible idea?
While asymmetrical crossovers are often used, it isn't usually to make sure the result is also asymmetrical. Why is it you want the two ways to play together this way, and are you prepared for handling the overlap?
The horn on the high end will usually add some phase shift. This usually changes what you would expect from a symmetrical crossover or the expected rules go out the window.
Best approach is to get a measuring setup to see if your crossover approach works.
REW is really useful for tuning crossovers in this case.
Best approach is to get a measuring setup to see if your crossover approach works.
REW is really useful for tuning crossovers in this case.
An abrupt cross-over will be audible as a "resonance", so they are to be avoided. But High frequency drivers are compromised when exposed to low frequencies, so you want to keep the high-pass fairly steep. High frequencies are no threat to a woofer, so there is no urgency in the low-pass. Steep analog filters also have a lot of phase shift. But 6dB down the bandpass, the effect on the total sound is minimal, so the presents of some signal 2x the lowpass or 1/2 the high pass is minimal. It is worth noting that a 12dB low-pass may sound brighter than a 6dB (inductor) because the capacitor is a low impedance voltage source, vs a limited current from the series inductor. A cross-over should be designed to "equalize" your drivers, not a pretty symmetric pair of filters.
The shortest version comes from my original post:
To elaborate on this a bit, the GedLee speakers were designed with very intentional off-axis roll-off. The idea was to match the waveguide's rolloff to the woofer's natural roll off as frequencies crossed over, to give as constant a directivity as possible. I'm not sure I can make a waveguide to match how narrow the woofer's dispersion will get at higher XO frequencies.
Agreed, but that's not always practical, especially when dealing with a wife and family, and what crosses the "this annoys wifey way too much" threshold. Also, the need to have multiple values of components on hand can be a bit wasteful/costly. If I were still single and didn't have to be mindful of the "too damn annoying" factor nor the costs w/r/t raising a kiddo, I'd be 100% on board. The measurement thing can be worked around with enough planning so it only takes a few hours to figure things out, but part of that is the purpose of this post. Going in blind means things are setup in the living area for too long (at least when I'm blindly running the show), and also means too many times trying to figure things out. Finding out what I can ahead of time to best prepare is definitely the win on this stuff.
Good to know. Thanks! And thanks for responding.
Understood, and for the most part, can agree. That's what I've done in the past, and easily agree on most dome tweeters. But given the sensitivity differences between the woofer and the CD, driving the CD lower doesn't feel like a concern to me since whatever a CD's physical excursion limit should be unlikely to be hit, even at lower frequencies, until we're sending crazy amounts of power through the system. I'm pretty sure Earl Geddes ran his XOs pretty low on the models he sold, though I could be wrong about that. When we're comparing something like a 97dB/W/m woofer to a 108dB/W/m, We're talking about around 16x the power needed to get a woofer to produce the same volume as the CD. Because of such a discrepancy, I'm less concerned about running the compression driver a little lower than the recommended XO freq from the manufacturer since it'll be extremely unlikely I'll end up reaching the physical limits of the CD, even at such a low frequency.
Yeah, my thinking was a 3rd order low-pass would have a 270* phase shift, the 1st order high-pass would have a 90* phase shift, so reversing the polarity on the HP (or LP) would keep them in phase when overlapping. But I'm going on memory from years and years ago (just getting back into this hobby), so I could be wrong on that.
Oh, good call. I'll have to think about this. Thanks for the response!
There should only be one frequency where the woofer matches the waveguide. Yes, the woofer continues to narrow but that's why you cross it. If you use different slopes, phase will vary around the cross.
Asymmetrical XO's are really quite common, depending on the drivers used,
but combining 18dB and 6dB is relatively rare > especially the way you have thought.
Using those slopes, most would have 6dB on woofer, and 18dB on the highs, not vice versa.
If you first determine R/C impedance correction of the woofer, a good 12dB filter will work well.
Not only should it do a good job re. 'beaming', but also avoids the insertion loss of a second coil.
It would only make sense to also use 12dB on the highs > giving much higher 'power safety'.
As I have mentioned in another thread - I have found 12dB/second order filters to be very 'musical'.
but combining 18dB and 6dB is relatively rare > especially the way you have thought.
Using those slopes, most would have 6dB on woofer, and 18dB on the highs, not vice versa.
If you first determine R/C impedance correction of the woofer, a good 12dB filter will work well.
Not only should it do a good job re. 'beaming', but also avoids the insertion loss of a second coil.
It would only make sense to also use 12dB on the highs > giving much higher 'power safety'.
As I have mentioned in another thread - I have found 12dB/second order filters to be very 'musical'.
Agreed, there should only be one freq that matches exactly, but there should be a range that's very close and acceptable.
Yeah, and that was actually my thinking around the 3rd order and 1st order. As long as I'm remembering this correctly, a 3rd order low-pass would have a 270* phase shift, the 1st order high-pass would have a 90* phase shift, so reversing the polarity on the HP (or LP) would keep them in phase when overlapping.
Thanks for contributing to the post, and understood. I totally understand it's always the other way and for the safety of the higher frequency driver. I address this in another comment, but I'm guessing you were likely typing your reply when that comment posted, so you wouldn't have seen it. Essentially, I'm not worried about the CD because the sensitivity mismatch between a 97 dB/W/m woofer and a 108 dB/W/m CD (typical values on drivers of these types, no specific drivers were chosen yet) mean it's really unlikely to reach any physical limitations on the CD, even at lower frequencies.
And it may just be worth it to do a symmetrical 2nd order. But I'd like to continue to explore this more before I decide to throw in the towel because it's easy.
I did understand your high efficiency CD horn and the headroom after attenuation,
but why even remotely stress its lower end? You even risk 'tickling' resonance at high SPL's.
but why even remotely stress its lower end? You even risk 'tickling' resonance at high SPL's.
They cross and then separate, potentially still within the bounds of the cross.
Last edited:
Don't forget about the roll off from the waveguide. Are you talking acoustic or electrical?? A 6db single cap with 12db wave guide roll-off will get you an acoustically 18 db high pass.
Rob 🙂
Rob 🙂
If you manage to measure raw FR from your loudspeaker elements, you can simulate and build the XO with VituixCAD Loudspeaker simulator:
https://kimmosaunisto.net/ This can save you a lot of trial and error.
https://kimmosaunisto.net/ This can save you a lot of trial and error.
This might be of interest:
https://www.diyaudio.com/community/...et-for-horns-particularly.345388/post-5979025
https://www.diyaudio.com/community/...et-for-horns-particularly.345388/post-5979025
I keep reading these simplistic "back-of-the-envelope" calculations. Sorry, it DOES NOT work that way.
It only would if the unfiltered "raw" drivers both had perfectly linear frequency response (and hence also linear phase) from 0Hz to infinity.
Obviously, that's a physical impossibility.
With real-world, bandwith-limited drivers, the natural roll-off / roll-on of the unfiltered "raw" drivers will cause the driver's phase response to "bend" from linearity starting well above/below the crossover frequency (Fx).
This means that, at Fx, the actual phase shift of each filtered driver will be MUCH LARGER than that of the electrical filter only.
Given the limited bandwidth of drivers and their differences in off axis responses, the issue is to blend them together on a given xo frequency, such that on axis response and off axis responses have desired result, usually power and directivity index are important indicator of a good blend. This implies the acoustical behaviour at xo and at least one octave below and above the xo frequency are of importance, not the electrical xo order ( i am talking passive xo here)
In vituixcad f.i. you can evaluate all important parameters, and create and compare variations and alternatives. When you have a design you think is ok, build it and listen to it with a variety of music and/ot testsignals. Then change it to a variation and listen again. Etc.
In vituixcad f.i. you can evaluate all important parameters, and create and compare variations and alternatives. When you have a design you think is ok, build it and listen to it with a variety of music and/ot testsignals. Then change it to a variation and listen again. Etc.