there's a follow-up article in this month's AudioXpress on a 2nd order, phase coherent, flat summed response active x-over design by John Kreskovsky. i don't have the original article (5/01) so some details elude me but it looks promising. it's phase coherent like a 1st order but with 2nd order slope and flat response (though you can vary an "overlap" parameter), and requires only one extra stage for equalization before the filters. anyone have any comments? i'm curious how it would sound, perhaps i'll try it sometime.
If you haven't seen it already Johns many projects and papers are on his website:
http://www.geocities.com/kreskovs/John1.html
I haven't built or listened to any of the transient perfect designs, but they are indeed intriguing.
Cheers,
Davey.
http://www.geocities.com/kreskovs/John1.html
I haven't built or listened to any of the transient perfect designs, but they are indeed intriguing.
Cheers,
Davey.
Hi,
I’ve experimented with this and it sounds pretty good. I did the filtering and equalisation at low signal level and used two amps.
There are some drawbacks, you need to be aware of. First the radiation pattern is not symmetrical and it is irregular, as is the case with a normal 1st-order filter. Best results I got with a d’Apolito arrangement. Second the equalisation needed is somewhere between 4 dB and 6 dB, depending on the overlap. This means you need to put around 2 to 4 times the power into the speakers around the crossover frequency. This limits headroom and hence dynamics. Simply stated, If you can normally do with 50W amps, you will need 100W to 200W amps with this approach for the same dynamic range. So to be successful you need at least drivers with high efficiency. Also keep in mind that the speakers need to be capable of handling the extra power, it will put extra demands on especially the tweeters.
And yes navin, the acoustical centres of the drivers need to be in the same plane to get it working. Off-axis I’ve found it more sensitive for displacement errors than LW second order filters. For the experiment I used a stepped baffle. It has the disadvantage that the distance between the unit centres is more than needed. I did not experiment with an electronic delay.
In the past I was rather sceptic about “phase coherence” also because 1st-order filters have too many drawbacks. But after hearing many well designed ones, I must admit that phase coherent speaker systems add to the "naturalness” of the perceived sound.
I’ve experimented with this and it sounds pretty good. I did the filtering and equalisation at low signal level and used two amps.
There are some drawbacks, you need to be aware of. First the radiation pattern is not symmetrical and it is irregular, as is the case with a normal 1st-order filter. Best results I got with a d’Apolito arrangement. Second the equalisation needed is somewhere between 4 dB and 6 dB, depending on the overlap. This means you need to put around 2 to 4 times the power into the speakers around the crossover frequency. This limits headroom and hence dynamics. Simply stated, If you can normally do with 50W amps, you will need 100W to 200W amps with this approach for the same dynamic range. So to be successful you need at least drivers with high efficiency. Also keep in mind that the speakers need to be capable of handling the extra power, it will put extra demands on especially the tweeters.
And yes navin, the acoustical centres of the drivers need to be in the same plane to get it working. Off-axis I’ve found it more sensitive for displacement errors than LW second order filters. For the experiment I used a stepped baffle. It has the disadvantage that the distance between the unit centres is more than needed. I did not experiment with an electronic delay.
In the past I was rather sceptic about “phase coherence” also because 1st-order filters have too many drawbacks. But after hearing many well designed ones, I must admit that phase coherent speaker systems add to the "naturalness” of the perceived sound.
Pjotr said:
yes, we have to pick our poisons don't we?
1st order filters have drawbacks l as you mention, but when properly implemented they still simply sound better than steeper slopes. while i've heard a couple speakers with steeper slopes that sound pretty good the vast majority of my favorite speakers use the simplest 1st order x-overs. they preserve the essential energy and harmonic structure of the sound, i think because of both their phase coherence and lower energy storage/ringing. many high order filters seem to ring like crazy to my ears and screw up the imaging and harmonic integrity. 1st order systems don't "add to the naturalness," they simply don't destroy it the way higher order xovers do!
there was a recent review in Stereophile of an mbl speaker that uses high-order crossover. in the review John Atkinson says that the speaker is not phase-coherent due to the crossover, but phase coherence "doesn't matter" in loudspeakers anyway. what a crock! just another reason to hate that magazine so very much. (i still get it only because those bastards keep auto-renewing me, and i'm too lazy to call up their subscription department.)
1st order filters have drawbacks l as you mention, but when properly implemented they
NONSENSE!
The drivers out of band response pretty well assures that you will wind up with a higher order actual acoustic slope in real life. Go read a few books before you pass yourself off as a speaker designer.
http://www.amazon.com/exec/obidos/search-handle-form/103-2174812-1655011
I wonder why several decades of research into higher order cross
over slopes have occured if first order slopes are so slam dunk.....
LMAO,
Fred
PS Crossver design is very difficult to optimize even when you know what what you are doing and make good measurements.
NONSENSE!
The drivers out of band response pretty well assures that you will wind up with a higher order actual acoustic slope in real life. Go read a few books before you pass yourself off as a speaker designer.
http://www.amazon.com/exec/obidos/search-handle-form/103-2174812-1655011
I wonder why several decades of research into higher order cross
over slopes have occured if first order slopes are so slam dunk.....
LMAO,
Fred
PS Crossver design is very difficult to optimize even when you know what what you are doing and make good measurements.
whatever!
fred i guess i should clarify, i'm just stating my preference for the sound of 1st order designs, i'm not saying they're perfect. maybe you could refer me to a commercial speaker with a high order crossover that sounds really good, as i have not heard one.
btw the link you gave does not work for me.
cheers,
marc
p.s. wait i'm confused now... what exactly are you saying?
fred i guess i should clarify, i'm just stating my preference for the sound of 1st order designs, i'm not saying they're perfect. maybe you could refer me to a commercial speaker with a high order crossover that sounds really good, as i have not heard one.
btw the link you gave does not work for me.
cheers,
marc
p.s. wait i'm confused now... what exactly are you saying?
Re: 1st order filters have drawbacks l as you mention, but when properly implemented they
Such harsh words Fred Sounds like you've been reading too many books and not building enuff speakers.
He did say properly implemented. This means overlap on the order of two octaves. Even with this you usually make other compromises.
Some of the best commercial designs use 1st order XOs
Because speaker design in general got headed away from single full range speaker and headed down the multi-way route -- partly because of the more is better attitude. Now after some dissatisfaction with many-way speakers, there is a return to looking at full-range and simplier systems, just as the tube (particularily SE) has made a huge resurgence.
XO design is indeed very hard. One of the reasons i tend to avoid it, by choosing designs that can get away with minimal XOs, or easier to implement active designs. And you know what, some of this simplier designs let me at the music in a way much more expensive, more complex designs don't...
dave
Fred Dieckmann said:
Such harsh words Fred
Sounds like you've been reading too many books and not building enuff speakers.He did say properly implemented. This means overlap on the order of two octaves. Even with this you usually make other compromises.
Some of the best commercial designs use 1st order XOs
Because speaker design in general got headed away from single full range speaker and headed down the multi-way route -- partly because of the more is better attitude. Now after some dissatisfaction with many-way speakers, there is a return to looking at full-range and simplier systems, just as the tube (particularily SE) has made a huge resurgence.
XO design is indeed very hard. One of the reasons i tend to avoid it, by choosing designs that can get away with minimal XOs, or easier to implement active designs. And you know what, some of this simplier designs let me at the music in a way much more expensive, more complex designs don't...
dave
why several decades . . .
"I wonder why several decades of research into higher order cross
over slopes have occured if first order slopes are so slam dunk....."
You can use different x-o topologies/slopes/etc. to realize different radiation patterns and other design goals.
If a design goal is a very loud system, a higher order x-over is necessary (for protection of higher frequency drivers).
If a design goal is time/phase/transient (or very low cost) response a lower order x-over is necessary.
Matt MacBeth
P.S.
"Go read a few books before you pass yourself off as a speaker designer." - Sounds like a friendly way of saying "I'm smarter than you"!
I'm glad everyone doesn't always get jumped on by "experts" when sharing comments on this board; it would be a very lonely place.
"I wonder why several decades of research into higher order cross
over slopes have occured if first order slopes are so slam dunk....."
You can use different x-o topologies/slopes/etc. to realize different radiation patterns and other design goals.
If a design goal is a very loud system, a higher order x-over is necessary (for protection of higher frequency drivers).
If a design goal is time/phase/transient (or very low cost) response a lower order x-over is necessary.
Matt MacBeth
P.S.
"Go read a few books before you pass yourself off as a speaker designer." - Sounds like a friendly way of saying "I'm smarter than you"!
I'm glad everyone doesn't always get jumped on by "experts" when sharing comments on this board; it would be a very lonely place.
That makes sense. For the low/mid I used the Audax HM130 Z4, just because of the very smooth roll off at high frequencies. The tweeter was a Peerless WA10, but there are many better tweeters like the ScanSpeaks. F_xov was at 2 kHz. There are more higher-order filter solutions for transient correct xov’s. But they all end up at something between 6dB/oct and 12 dB/oct. I am also investigating if a class of filters that are called “Transitional Gaussian filters" can be made of use.
Because of pure aluminium ribbon tweeters are my favourite, I will see if I can make a phase coherent xov for it. Oh boy, back to the bomb-shelter
By the way, good reading about all kinds of speaker design are the JAES of the past 50 years. And everybody who wants to go really deep: this book. Reprints of the most important articles in the JAES can be found here (the top 3 ones), including the original work of Thiele and Small. Does this make you happy Fred?
Fred's Going to Blow
Marc,
I am more or less in agreement with you. If you have the right drivers, in my experience anyway. Higher order slopes usually hurt the sound ie. 24db slopes,usually suck.
Don't worry about Fred . Since a muzzle was put on him, the pressure has been building up inside him. He was probably due for a blow up sooner or later.
Jam
Marc,
I am more or less in agreement with you. If you have the right drivers, in my experience anyway. Higher order slopes usually hurt the sound ie. 24db slopes,usually suck.
Don't worry about Fred . Since a muzzle was put on him, the pressure has been building up inside him. He was probably due for a blow up sooner or later.
Jam
Crossovers
http://www.amazon.com search books by Vance Dickason
I don't know what you are calling "high" order cross overs. The Magnepan speakers sound excellent with what I suppose you might consider "high" order networks. There are any number of speakers with second and third order networks that sound great. First order networks put a lot of IM distortion in tweeter that makes many of them sound rough. The Theil CS3 and CS 3.5 come to mind when describing a first order high pass network limitations.
The order of a network is very dependent on the driver response and crossover frequency. There are many asymmetric slope crossovers that work great also. The crossover really has to be designed to work with the specific drivers which makes the design as much art as science. My active 4th order L-R blows away first order networks on my speakers by the way. The compromises in flat frequency response and increased IM distortion make first order networks rather problimatic in most cases. It is interesting that most serious articles on crossover design inevitably talk about the limitations of first order networks.
http://www.rane.com/note107.html#paper
http://www.passlabs.com/pdf/phasecrx.pdf
http://www.integracoustics.com/MUG/MUG/tweaks/index.html
"Crossover Upgrades" section
http://www.amazon.com search books by Vance Dickason
I don't know what you are calling "high" order cross overs. The Magnepan speakers sound excellent with what I suppose you might consider "high" order networks. There are any number of speakers with second and third order networks that sound great. First order networks put a lot of IM distortion in tweeter that makes many of them sound rough. The Theil CS3 and CS 3.5 come to mind when describing a first order high pass network limitations.
The order of a network is very dependent on the driver response and crossover frequency. There are many asymmetric slope crossovers that work great also. The crossover really has to be designed to work with the specific drivers which makes the design as much art as science. My active 4th order L-R blows away first order networks on my speakers by the way. The compromises in flat frequency response and increased IM distortion make first order networks rather problimatic in most cases. It is interesting that most serious articles on crossover design inevitably talk about the limitations of first order networks.
http://www.rane.com/note107.html#paper
http://www.passlabs.com/pdf/phasecrx.pdf
http://www.integracoustics.com/MUG/MUG/tweaks/index.html
"Crossover Upgrades" section
hi fred,
you're right, you have to consider the behavior of the entire system when talking about crossovers. i was referring more to the electrical filters themselves, not the combined effect with the driver behavior. i also should have been clear when i said "high-order" xovers, i was referring primarily to those above 2nd order and which are necessarily non-linear phase. phase-coherent 2nd order filters are possible to implement and sound very good. i've also heard speakers with 3rd order filters sound decent, though not as often.
hmm, so on your 4th order L-R, you don't hear any ill effects from the group delay and associated non-linear transient response? i agree the increased distortion and reduced power handling, even with wide-band drivers, is a weakness of 1st order filters, but imperfect transient response in turn is a weakness of the vast majority of higher-order xovers. at least in the passive implementations i've heard on commercial speakers they seem to detract from the music - they definitely have a sonic signature.
p.s. i don't consider any Thiel speaker to be a shining example of 1st order x-overs... to me they all sound very unmusical. i have also heard they are not true 1st order xovers, being much more complex than a simple series cap/inductor... dunno how much validity there is to that but the sound of their speakers doesn't seem to have the usual virtues of a purist 1st order design. a better example is something like, oh, the Meadowlark Swift, a very simple and inexpensive little speaker that sounded fantastic when i heard it at the Stereophile show. also the Audio Concepts Sapphire, which has been shown by measurement to have minimal phase shift and very clean impulse response.
you're right, you have to consider the behavior of the entire system when talking about crossovers. i was referring more to the electrical filters themselves, not the combined effect with the driver behavior. i also should have been clear when i said "high-order" xovers, i was referring primarily to those above 2nd order and which are necessarily non-linear phase. phase-coherent 2nd order filters are possible to implement and sound very good. i've also heard speakers with 3rd order filters sound decent, though not as often.
hmm, so on your 4th order L-R, you don't hear any ill effects from the group delay and associated non-linear transient response? i agree the increased distortion and reduced power handling, even with wide-band drivers, is a weakness of 1st order filters, but imperfect transient response in turn is a weakness of the vast majority of higher-order xovers. at least in the passive implementations i've heard on commercial speakers they seem to detract from the music - they definitely have a sonic signature.
p.s. i don't consider any Thiel speaker to be a shining example of 1st order x-overs... to me they all sound very unmusical. i have also heard they are not true 1st order xovers, being much more complex than a simple series cap/inductor... dunno how much validity there is to that but the sound of their speakers doesn't seem to have the usual virtues of a purist 1st order design. a better example is something like, oh, the Meadowlark Swift, a very simple and inexpensive little speaker that sounded fantastic when i heard it at the Stereophile show. also the Audio Concepts Sapphire, which has been shown by measurement to have minimal phase shift and very clean impulse response.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.