Hi dorkus,
The main reason why 1st order filters don’t work, at least not for the tweeters:
De excursion of a dynamic speaker rises with 12 dB/oct with downward going frequency above fs of the speaker. If you filter at 6 dB/oct the excursion will rise at 6dB/oct. Imagine what will happen with the excursion limited dome of a tweeter at low frequencies. Some tweeters have fs around 400 Hz.
So any 1st order filter needs to be supported with extra filtering below the xov for tweeters. Also the power handling will be severe limited.
The main reason why 1st order filters don’t work, at least not for the tweeters:
De excursion of a dynamic speaker rises with 12 dB/oct with downward going frequency above fs of the speaker. If you filter at 6 dB/oct the excursion will rise at 6dB/oct. Imagine what will happen with the excursion limited dome of a tweeter at low frequencies. Some tweeters have fs around 400 Hz.
So any 1st order filter needs to be supported with extra filtering below the xov for tweeters. Also the power handling will be severe limited.
i was referring primarily to those above 2nd order and which are necessarily non-line
Assymetrical subtractive crossovers can have one side be a high order and still have correct phase response for the summed response. I am curious how the actual transient response of the phase coherent second order filter looks when the actual driver responses are included. Considering an crossover filter response without the contributions of the drivers response is pretty useless if you ask me. You are listening to the contributions of all the responses in an actual speaker. I think phase coherence is an over rated aspect of speaker design since there are damned few phase coherent recordings. Multiple microphones and mixing make it kind of a moot point.
Assymetrical subtractive crossovers can have one side be a high order and still have correct phase response for the summed response. I am curious how the actual transient response of the phase coherent second order filter looks when the actual driver responses are included. Considering an crossover filter response without the contributions of the drivers response is pretty useless if you ask me. You are listening to the contributions of all the responses in an actual speaker. I think phase coherence is an over rated aspect of speaker design since there are damned few phase coherent recordings. Multiple microphones and mixing make it kind of a moot point.
I think Kreskovsky's work is first rate. Getting phase
coherence from two pole filters is not trivial (at least not
to me!)
The application of these filters still encounters some gaps,
as with the 1st order designs. The loudspeaker drivers not
only contribute amplitude variations, but also phase variations,
and so the real life sum of these signals is unlikely to be
anywhere near as good.
My own experience is that you can start with this and many
other types of filters, and only after much work and
measurement do you end up optimized for particular drivers,
and you are still very unlikely to see a square wave with
your microphone.
Still, it's a nice goal. Perhaps it will sound good, too.
coherence from two pole filters is not trivial (at least not
to me!)
The application of these filters still encounters some gaps,
as with the 1st order designs. The loudspeaker drivers not
only contribute amplitude variations, but also phase variations,
and so the real life sum of these signals is unlikely to be
anywhere near as good.
My own experience is that you can start with this and many
other types of filters, and only after much work and
measurement do you end up optimized for particular drivers,
and you are still very unlikely to see a square wave with
your microphone.
Still, it's a nice goal. Perhaps it will sound good, too.
Re: i was referring primarily to those above 2nd order and which are necessarily non-line
i dunno, yes there are lots of crappy recordings out there with phase all over the place but there are a few good ones made by engineers like Tony Faulkner that are indeed phase coherent, using Blumlein or ORTF micing techniques etc. also phase coherence is still an issue even with multi-mic'ed recordings - i think phase distortion can effect the perceived harmonic integrity of acoustic instruments... a reason why i think 16/44 CD robs sound of its "natural" quality, even though the frequency response and dynamic range are technically fine. so i disagree that it is moot.
p.s. hi Mr. Pass! both you and Fred you are absolutely right of course that getting linear-phase, linear-amplitude summed response from a system is extremely difficult even after you master the electrical filter aspect of it, due to the amplitude/phase response of the drivers themselves. i think this may be how some speakers get fairly linear phase out of 2-pole filters... combined with the poles of the drivers themselves maybe you can get lucky. i think something to take into account though is the ringing of the filter itself, which is something that can NOT be compensated for with the driver or other filter element.
Fred Dieckmann said:
i dunno, yes there are lots of crappy recordings out there with phase all over the place but there are a few good ones made by engineers like Tony Faulkner that are indeed phase coherent, using Blumlein or ORTF micing techniques etc. also phase coherence is still an issue even with multi-mic'ed recordings - i think phase distortion can effect the perceived harmonic integrity of acoustic instruments... a reason why i think 16/44 CD robs sound of its "natural" quality, even though the frequency response and dynamic range are technically fine. so i disagree that it is moot.
p.s. hi Mr. Pass! both you and Fred you are absolutely right of course that getting linear-phase, linear-amplitude summed response from a system is extremely difficult even after you master the electrical filter aspect of it, due to the amplitude/phase response of the drivers themselves. i think this may be how some speakers get fairly linear phase out of 2-pole filters... combined with the poles of the drivers themselves maybe you can get lucky. i think something to take into account though is the ringing of the filter itself, which is something that can NOT be compensated for with the driver or other filter element.
Pjotr said:
actually i've seen plenty of speakers that have only the 1st order HP filter for the tweeter and sound just fine - excellent in fact. you are right however about the excursion problem and the resultant limited power handling. this is why you need a very robust tweeter with such a xover... tweeters that have sealed rear cavities, aperiodic venting etc. are designed for exactly this sort of situation i believe.
p.s. i should really clarify for everyone... when i say 1st order xover, i really mean 1st order ELECTRICAL FILTER before the driver, not overall system response. my bad.
Yes dorcus,
Thiel is using tweeters with a huge excursion capabilities, + and – 5 mm, someone told me. But the excessive IM is still left. At high levels the Thiels become rather muddy at the mid/highs. I heard them, I got the feeling to be smashed against the wall and that was not due to the amplifier.
Thiel is using tweeters with a huge excursion capabilities, + and – 5 mm, someone told me. But the excessive IM is still left. At high levels the Thiels become rather muddy at the mid/highs. I heard them, I got the feeling to be smashed against the wall and that was not due to the amplifier.
yah i get that same feeling when i listen to Thiels...
i did not get that at all from the Meadowlark speaker however, which is much smaller, uses a nice reasonable soft-dome tweeter (better behaved at high excursions than bad metal domes i think), and much cheaper to boot. i heard some flamenco guitar track being played at very high volume levels, powered by Rogue tube monoblocks, and it was incredibly clean and dynamic. there was not a huge amount of treble energy in the recording though, i don't expect it to fare as well on big orchestral music with full-tilt percussion and brass.
i did not get that at all from the Meadowlark speaker however, which is much smaller, uses a nice reasonable soft-dome tweeter (better behaved at high excursions than bad metal domes i think), and much cheaper to boot. i heard some flamenco guitar track being played at very high volume levels, powered by Rogue tube monoblocks, and it was incredibly clean and dynamic. there was not a huge amount of treble energy in the recording though, i don't expect it to fare as well on big orchestral music with full-tilt percussion and brass.
both you and Fred you are absolutely right of course that getting linear-phase
Funny how I became more right after Mr. Pass weighed in on the matter........... My real point is to focus on the tradeoffs of lower order filters. Linear phase at the expense of non flat amplitude response and increased distortion may not be worth the trade offs. Unfortunatly there is no such thing as a free lunch in speaker design as well as everything else.
I think assymetrical crossovers optimised for the drivers response may be the best compromise between filter order and amplitude flatness. Active filters based on this principle are probably were the real advances can be made. The the possible inclusion of all pass networks and equalization in active networks allows approaches that are just not practical with passive crossovers.
Loudspeaker Recipes: Book 1: Four Two-Way Systems
by Vance Dickason
http://www.pvconsultants.com/audio/frdgroup.htm
http://www.linkwitzlab.com/
http://www.passlabs.com/pdf/XVR1MAN1.PDF
Funny how I became more right after Mr. Pass weighed in on the matter........... My real point is to focus on the tradeoffs of lower order filters. Linear phase at the expense of non flat amplitude response and increased distortion may not be worth the trade offs. Unfortunatly there is no such thing as a free lunch in speaker design as well as everything else.
I think assymetrical crossovers optimised for the drivers response may be the best compromise between filter order and amplitude flatness. Active filters based on this principle are probably were the real advances can be made. The the possible inclusion of all pass networks and equalization in active networks allows approaches that are just not practical with passive crossovers.
Loudspeaker Recipes: Book 1: Four Two-Way Systems
by Vance Dickason
http://www.pvconsultants.com/audio/frdgroup.htm
http://www.linkwitzlab.com/
http://www.passlabs.com/pdf/XVR1MAN1.PDF
Re: both you and Fred you are absolutely right of course that getting linear-phase
hehe...
honestly, i was never arguing with you on that point. read our back and forth comments more closely. my initial comments were on the electrical filters themselves, not on the system as a whole, so i am at fault for not clarifying that point. high-order crossovers still have phase non-linearity which may or may not be complementary with the driver response, and there is the issue of energy storage and ringing as well. assymetrical crossovers are indeed promising, as are other designs (e.g. allpass) when implemented actively.
Fred Dieckmann said:
hehe...
honestly, i was never arguing with you on that point. read our back and forth comments more closely. my initial comments were on the electrical filters themselves, not on the system as a whole, so i am at fault for not clarifying that point. high-order crossovers still have phase non-linearity which may or may not be complementary with the driver response, and there is the issue of energy storage and ringing as well. assymetrical crossovers are indeed promising, as are other designs (e.g. allpass) when implemented actively.
i was never arguing with you on that point
"there is the issue of energy storage and ringing as well"
I give up, you are starting to sound like Grey......... BTW an all pass filter is for driver time alignment and is not a crossover filter per say. At any rate, the links are there for those who want to explore this suject in depth.
bla, bla, bla,
Fred
"there is the issue of energy storage and ringing as well"
I give up, you are starting to sound like Grey......... BTW an all pass filter is for driver time alignment and is not a crossover filter per say. At any rate, the links are there for those who want to explore this suject in depth.
bla, bla, bla,
Fred
Re: both you and Fred you are absolutely right of course that getting linear-phase
This is the compromise i most often use lately. I let the midbass roll off with its natural 2nd order roll-off and roll-in the tweeter 1st order (single cap). My current XO is at 10k, high enuff that the tweeter isn't too stressed. I have to live with a little narowing of the dispersion as the midbass reaches the top of its range.
dave
Fred Dieckmann said:
This is the compromise i most often use lately. I let the midbass roll off with its natural 2nd order roll-off and roll-in the tweeter 1st order (single cap). My current XO is at 10k, high enuff that the tweeter isn't too stressed. I have to live with a little narowing of the dispersion as the midbass reaches the top of its range.
dave
Now there are many "experts" on this, but not that many perfect speakers to prove their theories. I don`t like to try to discuss what gives the better sound without having ability to do real listening-tests as well, but here`s my opinion.
I`ve been working on a "perfect speaker project" for about twenty years,wondering where I`ll end up if ever. Not at all by following "the book" but more like picking up some good ideas and trying to include the best under one shell.
From the beginning I went for a four-way; allmost all dynamic speakers suffers from having a too low X-over for the tweeter. So I`ve added a 2" dome to take the dirty work, letting the tweeter just to do what it`s best at. That leaves me with a top section that are able to handle houndreds of watts without ever beeing close to give up, and with a relaxed open soundstage like nothing else. (-but what a job to tune..)(X-overs by 2 & 7khz)
The clue to make a really good speaker is to find some really good drivers that don`t need any kind of adjustments and let them do just what they`re good at and nothing more.
Now I`ve ended up with using only the simplest 1st order filter, some my think that`s an easy way but it isn`t. Not if you`re going for a superb result. It takes a lot of tuning and searcing for the best suited drivers while all the time searcing for the simplest possible solutions to not ruin the sound. Like ending up without any kind of absorbers to steal energy, just diffractors.(-this also does things for the bass-preformanse, but that`s another story..)
It`s not enough just to make a flat frequense-curve, the speaker must be able/optimoced to let any impulse from top to bottom slipping through with as little loss as possible to still make the music sound alive. And if you want to recreate an authentic stereophonic soundstage there can not be any components in paralell to (relatively) shortcut the signal. All this to make the speaker extremely fast. Allmost like real music..
After finishing my speakers some years ago they "speak for themselves.." There`s a lot of parameters to be considered in a speaker, a lot of things that needs to be taken care of. But whatewer, there`s nothing like a good 1st-order-speaker if it`s just done right!
I`ve been working on a "perfect speaker project" for about twenty years,wondering where I`ll end up if ever. Not at all by following "the book" but more like picking up some good ideas and trying to include the best under one shell.
From the beginning I went for a four-way; allmost all dynamic speakers suffers from having a too low X-over for the tweeter. So I`ve added a 2" dome to take the dirty work, letting the tweeter just to do what it`s best at. That leaves me with a top section that are able to handle houndreds of watts without ever beeing close to give up, and with a relaxed open soundstage like nothing else. (-but what a job to tune..)(X-overs by 2 & 7khz)
The clue to make a really good speaker is to find some really good drivers that don`t need any kind of adjustments and let them do just what they`re good at and nothing more.
Now I`ve ended up with using only the simplest 1st order filter, some my think that`s an easy way but it isn`t. Not if you`re going for a superb result. It takes a lot of tuning and searcing for the best suited drivers while all the time searcing for the simplest possible solutions to not ruin the sound. Like ending up without any kind of absorbers to steal energy, just diffractors.(-this also does things for the bass-preformanse, but that`s another story..)
It`s not enough just to make a flat frequense-curve, the speaker must be able/optimoced to let any impulse from top to bottom slipping through with as little loss as possible to still make the music sound alive. And if you want to recreate an authentic stereophonic soundstage there can not be any components in paralell to (relatively) shortcut the signal. All this to make the speaker extremely fast. Allmost like real music..
After finishing my speakers some years ago they "speak for themselves.." There`s a lot of parameters to be considered in a speaker, a lot of things that needs to be taken care of. But whatewer, there`s nothing like a good 1st-order-speaker if it`s just done right!
1. test pilot, why dont you share with us your design. just remember to wear your asbestos suit. fred is still around i have been trying to do exactly what you stated but wanted to stickto 3 drivers to avoid a filter in the all critical midrange (120Hz to 4000Hz - there are some that would argue that the midrange goes as low as 70Hz and while that maybe true i cant find a dynamic driver that can cover such a wide range with equal aplomb).
2. many posts ago Mr. Pass posted some pointers on how to design filters. it implied that one should add/change one component at a time and listen a lot.
3. any speaker is designed for a listening envoriment and also the tastes of the designer. we diyers have an advatage (over commercial designs) in that we are working with a known enviroment and known tastes (ours). Also in some ways we are able to avoid cutting corners (esp with box building).
i have been trying to do exactly what you stated but wanted to stickto 3 drivers to avoid a filter in the all critical midrange (120Hz to 4000Hz - there are some that would argue that the midrange goes as low as 70Hz and while that maybe true i cant find a dynamic driver that can cover such a wide range with equal aplomb).2. many posts ago Mr. Pass posted some pointers on how to design filters. it implied that one should add/change one component at a time and listen a lot.
3. any speaker is designed for a listening envoriment and also the tastes of the designer. we diyers have an advatage (over commercial designs) in that we are working with a known enviroment and known tastes (ours). Also in some ways we are able to avoid cutting corners (esp with box building).
Before breaking one's b@lls to pull off a "phase coherent" design, it might be worthwhile to do some controlled tests to determine whether that's something worth the tradeoffs. What literature I have suggests that, the esthetic knowledge that square waves aren't being passed put aside, the sorts of phase responses characterisitic of filters like LR4s are actually not audible.
I spent years doing speaker designs that were nominally first order. To do it right, you need to look at 5 way as a minimum, given the reality of driver rolloffs. After all, an acoustic LR4 derived from a second order filter and a second order rolloff has the same phase response as a flat driver with a fourth order filter. And after years of work getting my speakers to pass (at one point in space!) a recognizable square wave, my belief in the superiority of my approach crumbled when (after reading a paper by Stanley Lipschitz) I did some bypass testing with unity-gain allpass filters and found I couldn't distinguish the output from the input. Nor could any of my other true-believer friends.
John and I have discussed this and his strong feeling was that even if you can't hear it, phase coherence should be preserved at that one point in space out of an esthetic design goal. I've got a somewhat different view, preferring to concentrate on the demonstrably audible, but I can respect John's design goal as a "fun" thing to do. There are other approaches than first order to this goal, John's newest crossover being one, the B&O filler driver being another, the Spica asymmetrical XO with time delay being yet another. All methods only achieve phase coherence at one point.
I haven't figured out how to get both my ears into that one magic point in space, though...
Anyway, phase coherence at that one point can't hurt. But one must be VERY careful that tradeoffs of things that ARE audible (e.g., distortion, polar response) aren't being made to achieve a goal that is of questionable audibility.
I spent years doing speaker designs that were nominally first order. To do it right, you need to look at 5 way as a minimum, given the reality of driver rolloffs. After all, an acoustic LR4 derived from a second order filter and a second order rolloff has the same phase response as a flat driver with a fourth order filter. And after years of work getting my speakers to pass (at one point in space!) a recognizable square wave, my belief in the superiority of my approach crumbled when (after reading a paper by Stanley Lipschitz) I did some bypass testing with unity-gain allpass filters and found I couldn't distinguish the output from the input. Nor could any of my other true-believer friends.
John and I have discussed this and his strong feeling was that even if you can't hear it, phase coherence should be preserved at that one point in space out of an esthetic design goal. I've got a somewhat different view, preferring to concentrate on the demonstrably audible, but I can respect John's design goal as a "fun" thing to do. There are other approaches than first order to this goal, John's newest crossover being one, the B&O filler driver being another, the Spica asymmetrical XO with time delay being yet another. All methods only achieve phase coherence at one point.
I haven't figured out how to get both my ears into that one magic point in space, though...
Anyway, phase coherence at that one point can't hurt. But one must be VERY careful that tradeoffs of things that ARE audible (e.g., distortion, polar response) aren't being made to achieve a goal that is of questionable audibility.
SY said:
agreed. while i still believe phase coherence is important, i don't have an A/B listening test to back it up as you do. but designing a phase coherent speaker that sucks in other regards is silly of course. it does happen that many phase coherent speakers sound good to me though. (no, not Thiel...)
maybe it's not the phase coherence per se that makes them sound good but the concomittant design qualities... although i have a sneaking suspicion that poor phase performance tends to muck up the harmonic structure of acoustic instruments. i'd have to do a formal test to back up this assertion however. i've found phase-coherent designs do tend to have the most natural harmonic presentation, but i can't say this is true "all other things being equal."
Hi,
Some 10 years ago I worked on the physical department of the Technical University Delft. We had a group of enthusiastic colleagues building DIY speakers. Also colleagues of the acoustical department were involved. We did some listening tests in those days with different filter topologies. De filters were summed up in the electrical domain and listening was done by headphone.
Only with test signals like square waves and saw tooth’s a significant difference could be heard. With normal music no one could hear a significant difference I must admit. But one thing was remarkable, filters with steep phase slopes tended to sound worse. If a steep higher order LP filter was fed by pink noise a pitch could be heard at the cut-off frequency. This was reflected in the group delay curve. Filters that showed a peak in the group delay tended to have that “pitchy” sound at the frequency of the peak in the group delay. This evidence was confirmed by many literature.
I am building DIY speakers for 25 years now. One of my first projects was a combination of a Decca DK30 and a KEF B110. Despite all my efforts and yes I experimented a lot with active filtering, active compensation of unit response and so on, this DK30 - B110 combo is still one of my best sounding concoctions. I gave up building a speaker system that can reproduce faithfully at SPL’s of more than 90 dB for domestic use. To my opinion that is only possible with horn loaded systems. It appears to me that the only way to build a “transient correct” system regarding dispersion is the “Tannoy way” of using concentric units. But the Tannoy’s do have other flaws (yes I have some 8” units here).
I don’t know why but it seems to me a “transient correct” system is holding some secrets. Maybe it is you need the best available units for, who knows… Anyway it stays intriguing.
Here is a link to some very good articles of Lynn T. Olson, a very experienced speaker designer: http://www.unik.no/~robert/hifi/articles/index.html
Chaio
Some 10 years ago I worked on the physical department of the Technical University Delft. We had a group of enthusiastic colleagues building DIY speakers. Also colleagues of the acoustical department were involved. We did some listening tests in those days with different filter topologies. De filters were summed up in the electrical domain and listening was done by headphone.
Only with test signals like square waves and saw tooth’s a significant difference could be heard. With normal music no one could hear a significant difference I must admit. But one thing was remarkable, filters with steep phase slopes tended to sound worse. If a steep higher order LP filter was fed by pink noise a pitch could be heard at the cut-off frequency. This was reflected in the group delay curve. Filters that showed a peak in the group delay tended to have that “pitchy” sound at the frequency of the peak in the group delay. This evidence was confirmed by many literature.
I am building DIY speakers for 25 years now. One of my first projects was a combination of a Decca DK30 and a KEF B110. Despite all my efforts and yes I experimented a lot with active filtering, active compensation of unit response and so on, this DK30 - B110 combo is still one of my best sounding concoctions. I gave up building a speaker system that can reproduce faithfully at SPL’s of more than 90 dB for domestic use. To my opinion that is only possible with horn loaded systems. It appears to me that the only way to build a “transient correct” system regarding dispersion is the “Tannoy way” of using concentric units. But the Tannoy’s do have other flaws (yes I have some 8” units here).
I don’t know why but it seems to me a “transient correct” system is holding some secrets. Maybe it is you need the best available units for, who knows… Anyway it stays intriguing.
Here is a link to some very good articles of Lynn T. Olson, a very experienced speaker designer: http://www.unik.no/~robert/hifi/articles/index.html
Chaio
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.