Hi Davey,
One way is second order. Here it is, typically what must not be done :
http://hifi-france.ifrance.com/filtre1/filtre-actif.html
You will have an idea of an other fancy conception from the same author if you go to the home page looking for the tube amp.
One way is second order. Here it is, typically what must not be done :
http://hifi-france.ifrance.com/filtre1/filtre-actif.html
You will have an idea of an other fancy conception from the same author if you go to the home page looking for the tube amp.
I don't think his frequency response is very flat, though he uses quite some overlap (midrange is a wideband driver !). But most response errors will be minumum-phase and could therefore easily be corrected.
The answer I am obviously owing to all: The output voltages sum to unity. If the whole speaker (i.e. at least its forward sound pressure) should behave like that - which it has to, otherwise the whole thing wouldn't make sense - driver behaviour has to be taken into account of course.
One possible solution for a fullrange + woofer is shown here:
http://www.diyaudio.com/forums/showthread.php?postid=292070#post292070
Info on transient-perfect crossovers is scarce in general. Even when searching hard you won't find many pages on the subject on the web apart fom John's.
Regards
Charles
The answer I am obviously owing to all: The output voltages sum to unity. If the whole speaker (i.e. at least its forward sound pressure) should behave like that - which it has to, otherwise the whole thing wouldn't make sense - driver behaviour has to be taken into account of course.
One possible solution for a fullrange + woofer is shown here:
http://www.diyaudio.com/forums/showthread.php?postid=292070#post292070
Info on transient-perfect crossovers is scarce in general. Even when searching hard you won't find many pages on the subject on the web apart fom John's.
Regards
Charles
Phase-accurate
---One possible solution for a fullrange + woofer is shown here:
http://www.diyaudio.com/forums/show...2070#post292070---
Yesterday, I carefully read the famous Leach's AES paper about the "neglected factor in the design of crossovers" : there is a scheme same as yours. Ingenious.
---One possible solution for a fullrange + woofer is shown here:
http://www.diyaudio.com/forums/show...2070#post292070---
Yesterday, I carefully read the famous Leach's AES paper about the "neglected factor in the design of crossovers" : there is a scheme same as yours. Ingenious.
re post #17
phase_accurate, I follow you about polynomial partitioning. With enought time I could figure out how to do that. One of the filters could be a standard response, like L-R or Butter worth. The other would not. Or, maybe I could figure out some symetrical way to do it. But it still seems inelegant to me because the flat summation depends on component values.
If on the other hand, I have one filter, say the high pass filter, and then subtract its signal from the original, I will have both HP and LP. If I then work out the transfere functions for both, they should be the same as if I had used two filters.
But in this second subtractive case, it depends upon the component values used in the subtraction. Some inverting/noninverting schemes are very in accurate. Others better. Are there some highly accurate? I tried to explore this once. I think there might be. Its the same problem involved in receiving a differential signal and getting high CMRR, without a transformer.
But there are still other issues. Does subtractive to get flat summation really work with the loudspeakers? After all the common conventions with passive crossovers and single amping did not develop overnight. So if we go to active, subtractive crossovers, with multi-amping, does it work?
Likewise, is commercially made equipment done that way. Do most of the powered subwoofer systems being sold today use subtractive crossovers? How exactly? Or do they use two filters?
Likewise I have read that many of the high power PA systems for rock concerts are using multi-amping and active crossovers. They use a four way speaker system. The upper 3 ways are horns, and they are suspended from the ceiling. The bottom way is just a bunch of big drivers. These go under the stage. It is all multi-amped and active crossover. But is it subtractive or paired filters? And exactly how does it work and what are the design trade-offs?
phase_accurate, I follow you about polynomial partitioning. With enought time I could figure out how to do that. One of the filters could be a standard response, like L-R or Butter worth. The other would not. Or, maybe I could figure out some symetrical way to do it. But it still seems inelegant to me because the flat summation depends on component values.
If on the other hand, I have one filter, say the high pass filter, and then subtract its signal from the original, I will have both HP and LP. If I then work out the transfere functions for both, they should be the same as if I had used two filters.
But in this second subtractive case, it depends upon the component values used in the subtraction. Some inverting/noninverting schemes are very in accurate. Others better. Are there some highly accurate? I tried to explore this once. I think there might be. Its the same problem involved in receiving a differential signal and getting high CMRR, without a transformer.
But there are still other issues. Does subtractive to get flat summation really work with the loudspeakers? After all the common conventions with passive crossovers and single amping did not develop overnight. So if we go to active, subtractive crossovers, with multi-amping, does it work?
Likewise, is commercially made equipment done that way. Do most of the powered subwoofer systems being sold today use subtractive crossovers? How exactly? Or do they use two filters?
Likewise I have read that many of the high power PA systems for rock concerts are using multi-amping and active crossovers. They use a four way speaker system. The upper 3 ways are horns, and they are suspended from the ceiling. The bottom way is just a bunch of big drivers. These go under the stage. It is all multi-amped and active crossover. But is it subtractive or paired filters? And exactly how does it work and what are the design trade-offs?
Hi:
I am of the opinion that a symmetrical subtractive crossover for any given slope can be done as a combination of standard and all-pass filters. An example of a 24dB/oct. filter for crossover frequencies of 1 and 10kHz is given in the picture/simulation below.
Regards,
Milan
I am of the opinion that a symmetrical subtractive crossover for any given slope can be done as a combination of standard and all-pass filters. An example of a 24dB/oct. filter for crossover frequencies of 1 and 10kHz is given in the picture/simulation below.
Regards,
Milan
Attachments
zenmasterbrian
Most active PAs nowadays use LR crossovers (funny that those were called HiFi crap by the PA guys 20 years ago !!!) because they offer better and easier control of lobing, amplitude response and driver nonlinear distortion. Some also use digital FIR crossovers ore filters with notches like the NTM.
As already mentioned the primary goal of using constant voltage crossovers is improved transient-response. This comes at the cost of less steep slopes around the crossover frequency which brings increased IMD and lower achievable SPL for a given driver combination. This is exactly the contrary of what you want to achieve in the PA busines where you have to maximise your bang (pun intended) for the buck/truckspace/weight/.... .
For the same reasons most subwoofer combinations at home use steep filters like LR.
Regarding the accuaracy of the subtractive implementation: The components that specify the gain of the stages involved have to be very accurate. If not, the final slope of the subtractively derived branch will suffer. But the frequency-response determining components (i.e. caps !!!) need to be less accurate. If a constant-voltage crossover is built using two seperate filter branches it is vice versa: Higher demands on capacitor accuracy lower demand on gain accuracy.
Milan
The circuit that you posted is a classic. It is an elegant solution for x-overs that sum perfectly flat in the frequency domain. But it is not transient-perfect because it shows an allpass behaviour.
Regards
Charles
Most active PAs nowadays use LR crossovers (funny that those were called HiFi crap by the PA guys 20 years ago !!!) because they offer better and easier control of lobing, amplitude response and driver nonlinear distortion. Some also use digital FIR crossovers ore filters with notches like the NTM.
As already mentioned the primary goal of using constant voltage crossovers is improved transient-response. This comes at the cost of less steep slopes around the crossover frequency which brings increased IMD and lower achievable SPL for a given driver combination. This is exactly the contrary of what you want to achieve in the PA busines where you have to maximise your bang (pun intended) for the buck/truckspace/weight/.... .
For the same reasons most subwoofer combinations at home use steep filters like LR.
Regarding the accuaracy of the subtractive implementation: The components that specify the gain of the stages involved have to be very accurate. If not, the final slope of the subtractively derived branch will suffer. But the frequency-response determining components (i.e. caps !!!) need to be less accurate. If a constant-voltage crossover is built using two seperate filter branches it is vice versa: Higher demands on capacitor accuracy lower demand on gain accuracy.
Milan
The circuit that you posted is a classic. It is an elegant solution for x-overs that sum perfectly flat in the frequency domain. But it is not transient-perfect because it shows an allpass behaviour.
Regards
Charles
phase_accurate
I understand the issues with SPL limitations and lobbing when you don't use use steep crossovers.
You say these systems use LR. Are they using LR for both of the adjacent ways? Or are they using LR for one of the ways and subtraction to get the other.
If it is the latter, I assume they use subtraction for the lower freq driver.
That Milan schematic, maybe it is some combination or balanced thing where they are both derived and both subtracted from each other, or some such. So it is more complex, but basically it is subtractive, and hence flat summing.
So if I go and buy a consumer grade powered subwoofer, am I getting something that is subtractive? If so, I'd like to see something that documents all the various approaches and design issues.
These matters are undertreated in hobbyiest literature, and in most prof. literature too.
Thanks
I understand the issues with SPL limitations and lobbing when you don't use use steep crossovers.
You say these systems use LR. Are they using LR for both of the adjacent ways? Or are they using LR for one of the ways and subtraction to get the other.
If it is the latter, I assume they use subtraction for the lower freq driver.
That Milan schematic, maybe it is some combination or balanced thing where they are both derived and both subtracted from each other, or some such. So it is more complex, but basically it is subtractive, and hence flat summing.
So if I go and buy a consumer grade powered subwoofer, am I getting something that is subtractive? If so, I'd like to see something that documents all the various approaches and design issues.
These matters are undertreated in hobbyiest literature, and in most prof. literature too.
Thanks
When you use LR crossovers it only makes sense when both branches are LR of the same order. That does not necessarily mean that one uses active filters of the desired order(s) but that the combination of filter and driver response shall give a symmetrical LR crossover.
You can build subtractive LR crossovers with the circuit posted by Milan (moamps). But these make it more difficult to take driver response into consideration.
There is only one subwoofer that I know of which uses a transient-perfect subtractive approach (and I am note even fully sure about that) to be coupled to the rest of the frequency range. Keep in mind that you usually want the least amount of midrange coming out of a sub and this is quite difficult to achieve subtractively. It really only makes sense for stereo subs that are placed close to the mains.
Regards
Charles
You can build subtractive LR crossovers with the circuit posted by Milan (moamps). But these make it more difficult to take driver response into consideration.
There is only one subwoofer that I know of which uses a transient-perfect subtractive approach (and I am note even fully sure about that) to be coupled to the rest of the frequency range. Keep in mind that you usually want the least amount of midrange coming out of a sub and this is quite difficult to achieve subtractively. It really only makes sense for stereo subs that are placed close to the mains.
Regards
Charles
Any books that talk about about active and subtractive cross overs. I've never found any.
For audio, its all passives. Books on active filters, but not for this application.
Anyone work out all the math of how you could obtain symetrical filters, but with just one filter and subtraction? Anyone analyze all the tollerances? Anyone analyze a filter pair that is flat summing, but do it interms of the polynomials and the poles and zeros? Anyone have a filter pair that is both symmetric, but also flat summing.
I have found the literature lacking.
For audio, its all passives. Books on active filters, but not for this application.
Anyone work out all the math of how you could obtain symetrical filters, but with just one filter and subtraction? Anyone analyze all the tollerances? Anyone analyze a filter pair that is flat summing, but do it interms of the polynomials and the poles and zeros? Anyone have a filter pair that is both symmetric, but also flat summing.
I have found the literature lacking.
What do you want to achieve exactly ?
Do you just want linear frequency response ? In this case the easiest thing would be to use an LR4 topology. If your mains are closed-box you can design their response into the highpass function (which would then consist of a 2nd order electrical and a 2nd order electro-mechanical higpass function).
If you want to use a subtractively derived LR4 then you'd have to correct for the drivers' responses after the crossover. I.e. you would need much more components.
If you want a transient-perfect crossover for your subs (a mono sub is definitely out of the question in this case) then there is no chance to find anything in the literature (at least not to my knowledge). This would definitely be the hardest version to achieve !
Regards
Charles
Do you just want linear frequency response ? In this case the easiest thing would be to use an LR4 topology. If your mains are closed-box you can design their response into the highpass function (which would then consist of a 2nd order electrical and a 2nd order electro-mechanical higpass function).
If you want to use a subtractively derived LR4 then you'd have to correct for the drivers' responses after the crossover. I.e. you would need much more components.
If you want a transient-perfect crossover for your subs (a mono sub is definitely out of the question in this case) then there is no chance to find anything in the literature (at least not to my knowledge). This would definitely be the hardest version to achieve !
Regards
Charles
Zenmasterbrian
---I have found the literature lacking---.
Most if not all answers to your questions about crossovers and at least two articles about current amps can be found in the five AES anthologies.
---I have found the literature lacking---.
Most if not all answers to your questions about crossovers and at least two articles about current amps can be found in the five AES anthologies.
AES anthologies??
phase, mostly I want a different way to do it. I want to go multiamped and eliminate passive crossovers. I want to find the best ways to do it. I'm not keen on symetrical filters, because they are often not flat summing. They also depend on component tollerances for what kind of summing they do have. I believe there are people who have worked out all the possibilities. I'm especially interested to know what current practices are in commericial equipment.
Most systems now have active crossovers and multi-amping for subwoofers.
phase, mostly I want a different way to do it. I want to go multiamped and eliminate passive crossovers. I want to find the best ways to do it. I'm not keen on symetrical filters, because they are often not flat summing. They also depend on component tollerances for what kind of summing they do have. I believe there are people who have worked out all the possibilities. I'm especially interested to know what current practices are in commericial equipment.
Most systems now have active crossovers and multi-amping for subwoofers.
Zenmasterbrian
---I'm especially interested to know what current practices are in commericial equipment.---
Most if not all answers...
JAES Anthologies can be bought here, for example :
http://www.audioxpress.com/bksprods/bkindexltheory.htm
---I'm especially interested to know what current practices are in commericial equipment.---
Most if not all answers...
JAES Anthologies can be bought here, for example :
http://www.audioxpress.com/bksprods/bkindexltheory.htm
forr, I will look into that. I'll also spend more time with the Transconductance Amps thread mentioned earlier.
Thanks🙂
Thanks🙂
Still want to pursue this. I've also got a new thread:
diyAudio Forums > Top >Source >Digital >Open Source, Open Architecture!😀
diyAudio Forums > Top >Source >Digital >Open Source, Open Architecture!😀
http://www.national.com/an/AN/AN-346.pdf#page=6
Above is a good link. Thanks phase. I'm still exploring other links.
Above is a good link. Thanks phase. I'm still exploring other links.
John K., I like your dipole speakers, and your web site.
Some people get a dipole, with rear drivers. Sometimes out of phase, sometimes in phase.
Generally, surround sound is supposed to have dipoles for the side surrounds. But many systems are done that way.
I like you designs.
Some people get a dipole, with rear drivers. Sometimes out of phase, sometimes in phase.
Generally, surround sound is supposed to have dipoles for the side surrounds. But many systems are done that way.
I like you designs.
- Status
- Not open for further replies.