Forr:
The crossover point is the frequency at which both drivers are outputting equal SPL. The crossover region is traditionally considered the number of octaves between the -12 dB point for each driver. So if you have a 12 dB/octave crossover where each driver is 6 dB down at the crossover region, the entire crossover region will be about an octave.
This is because once one driver is outputting 12 dB less than the other in a speaker system, it is considered as not being able to be heard. Just for an experiment, you can try this yourself with your computer speakers. . Download a freeware tone generator from David Taylor's website here: www.satsignal.net => Audio Tools. Choose any convenient frequency-say 500 Hz. Fill in 500 Hz in both boxes where is says sweep frequency. Set the right speaker at 0 dB, the left speaker at 12 dB down. Then check and uncheck the left speaker to see if you can detect when it is playing and when it is not. It will be hard to tell a difference.
However, if you move the left speaker up to -6 dB, then check and uncheck the left speaker, the difference will be detectable.
PS: I understand that SY has conducted experiments which show that people can perceive a subtle difference in music when one driver is playing more than 12 dB lower than other. However, these differences are apparently extremely subtle, so the general reference level between points for the crossover region is still generally considered -12 dB, unless otherwise specified.
The crossover point is the frequency at which both drivers are outputting equal SPL. The crossover region is traditionally considered the number of octaves between the -12 dB point for each driver. So if you have a 12 dB/octave crossover where each driver is 6 dB down at the crossover region, the entire crossover region will be about an octave.
This is because once one driver is outputting 12 dB less than the other in a speaker system, it is considered as not being able to be heard. Just for an experiment, you can try this yourself with your computer speakers. . Download a freeware tone generator from David Taylor's website here: www.satsignal.net => Audio Tools. Choose any convenient frequency-say 500 Hz. Fill in 500 Hz in both boxes where is says sweep frequency. Set the right speaker at 0 dB, the left speaker at 12 dB down. Then check and uncheck the left speaker to see if you can detect when it is playing and when it is not. It will be hard to tell a difference.
However, if you move the left speaker up to -6 dB, then check and uncheck the left speaker, the difference will be detectable.
PS: I understand that SY has conducted experiments which show that people can perceive a subtle difference in music when one driver is playing more than 12 dB lower than other. However, these differences are apparently extremely subtle, so the general reference level between points for the crossover region is still generally considered -12 dB, unless otherwise specified.
Forr: You are correct, the greater the attenuation rate, the narrower the crossover region. So a 6 dB/octave network should have about a three octave crossover region, a 24 dB/octave crossover closer to a half an octave for the crossover region. I am not sure of the exact figures, and I believe the size of the crossover region might vary somewhat even among crossovers of the same order.
Hi Kelticwizard,
Many thinks for the link and for this simple idea using a signal generator. I wonder why I have not had it myself or had never seen mentionned anywhere. I think this is an experience which evrybody involved in making crossovers should make. I think the width of the crossover region is an important notion.
Using David Taylor's signal generator was not sufficiently conclusive for me because of a little perturbance when changing the level of one channel. I will repeat it but using my hardware generator. However I already am in rough accordance with -12 dB point for the limit of the crossover region.
Maybe you have some references for this tradition considering the number of octaves between the -12 dB point for each driver, I was unaware of Sy's experiments. Because of the curves of equal loudness contours, the operating frequency may affect the results.
Looking at the Vance Dickason's book, I roughly estimated the following crossover regions for a central frequency of 1000 Hz and different slopes. Types of filters Butterworth, Bessel, LR, etc... change the width a bit.
1st order Butterworth, 6 dB/o : 250-4000 Hz ... = 4 octaves
2nd order Butterworth, 12 dB/o : 500-2000 Hz ... = 2 octaves
2nd order Linkwitz-Riley 12 dB/o : 560-1700 Hz < 2 octaves
3rd order Butterworth, 18 dB/o : 640-1560 Hz = 1 1/3 octaves
4th order Butterworth, 24 dB/o : 700-1400 Hz = 1 octave
4th order Linkwitz-Riley, 24 dB/o : 740-1350 Hz < 1 octave
Not mentionned by Dickason, Elliptic crossover (Hardman) :
800 - 1200 Hz ... = 2/3 octave
Regards ~~~~~ Forr
§§§
Many thinks for the link and for this simple idea using a signal generator. I wonder why I have not had it myself or had never seen mentionned anywhere. I think this is an experience which evrybody involved in making crossovers should make. I think the width of the crossover region is an important notion.
Using David Taylor's signal generator was not sufficiently conclusive for me because of a little perturbance when changing the level of one channel. I will repeat it but using my hardware generator. However I already am in rough accordance with -12 dB point for the limit of the crossover region.
Maybe you have some references for this tradition considering the number of octaves between the -12 dB point for each driver, I was unaware of Sy's experiments. Because of the curves of equal loudness contours, the operating frequency may affect the results.
Looking at the Vance Dickason's book, I roughly estimated the following crossover regions for a central frequency of 1000 Hz and different slopes. Types of filters Butterworth, Bessel, LR, etc... change the width a bit.
1st order Butterworth, 6 dB/o : 250-4000 Hz ... = 4 octaves
2nd order Butterworth, 12 dB/o : 500-2000 Hz ... = 2 octaves
2nd order Linkwitz-Riley 12 dB/o : 560-1700 Hz < 2 octaves
3rd order Butterworth, 18 dB/o : 640-1560 Hz = 1 1/3 octaves
4th order Butterworth, 24 dB/o : 700-1400 Hz = 1 octave
4th order Linkwitz-Riley, 24 dB/o : 740-1350 Hz < 1 octave
Not mentionned by Dickason, Elliptic crossover (Hardman) :
800 - 1200 Hz ... = 2/3 octave
Regards ~~~~~ Forr
§§§
forr said:
I agree. I gave the Taylor link with some misgivings, because it does have that little disturbance when turning the low channel on and off. However, I thought it served as a quick and easy method to roughly illustrate the point. I applaud your decision to take things further and use a hardware signal generator without that disturbance.
forr said:
Hoo boy, I would have to look that one up. My basic references for my loudspeaker knowledge are Abraham B. Cohen's HiFi Loudspeakers and Enclosures, David Weems' Building, Designing and Testing Loudspeakers, and the AES Loudspeaker Anthology Vol 1 & 2. I never have gotten around to getting Dickason's book, I know I should. I can't point at this moment to any one specific sentence or passage in any book or article, but over the course of reading these sources the consensus seems to have emerged that the generally recognized definition of "crossover region", unless otherwise specificed, seems to be the distance in octaves between the -12 dB down points.
forr said:
SY's quote, and a discussion of many of these same points, is contained in this thread, especially the first two pages of it. Incidentally, in that thread, I too gave the crossover region for a 6 dB/octave filter as four octaves. I have reconsidered since then. I figure that because the crossover point is 3 dB down in a first order network, and it takes sometime for a first order filter to drop 3 dB, that a first order crossover would not be two octaves on either side of the crossover point. But I haven't tested this, so i might be wrong.
One more thing, on a related topic. If you consider third order crossovers as higher order, I have a paper by Bang and Olufson engineers on using a one octave filler driver to achieve a square wave response. Thisis for second and third order crossovers. I can send it to you if yoiu Email me.
While this does not affect the crossover region per slope exactly, it has everything to do with making the crossover region more desirable, since phase shift is absent in the crossover region with these crossovers.
Here is a thread on it.
And here is john k's spreadsheet for calculating the values for the second order version of this crossover.
Here is a pic of the step response, on axis. Off-axis, apparently it does not make a square wave.
While this does not affect the crossover region per slope exactly, it has everything to do with making the crossover region more desirable, since phase shift is absent in the crossover region with these crossovers.
Here is a thread on it.
And here is john k's spreadsheet for calculating the values for the second order version of this crossover.
Here is a pic of the step response, on axis. Off-axis, apparently it does not make a square wave.
Attachments
The branches will of course have phase-shift but their sum is without phase-shift.
No speaker will ever do this for more than one point in space apart form a single-driver point-source.
There are many possibilities to achieve a transient-perfect response that don't need an additional driver like the filler driver concept does.
Regards
Charles
Hi Kelticwizard
---One more thing, on a related topic. If you consider third order crossovers as higher order, I have a paper by Bang and Olufson engineers on using a one octave filler driver to achieve a square wave response. Thisis for second and third order crossovers. I can send it to you if yoiu Email me.---
Thanks, but I already know Eric Baekgaard's works for Bang & Olufsen. I am even currently listening to a pair of Beovox S45 using a filler driver !
I am aware of a recent active version of this filler driver where a direct comparison between the filler solution and a conventionnal crossover is possible.
Hare are current attemps towards the obtention of good square waves with 18 dB/o crossovers :
http://freerider.dyndns.org/anlage/LeCleach1.zip
http://freerider.dyndns.org/anlage/LeCleach2.zip
I found the Dickason's book very good, especially for newbies. I very much like the discussion about the filters, their slope, phase and group delays. Slight criticisms : it lacks a chapter on active filters (my french edition is the translation of the fifth one) and I would have prefer passive crossovers formulas written with "Pi" than with already calculated constants.
regards ~~~ Forr
§§§
---One more thing, on a related topic. If you consider third order crossovers as higher order, I have a paper by Bang and Olufson engineers on using a one octave filler driver to achieve a square wave response. Thisis for second and third order crossovers. I can send it to you if yoiu Email me.---
Thanks, but I already know Eric Baekgaard's works for Bang & Olufsen. I am even currently listening to a pair of Beovox S45 using a filler driver !
I am aware of a recent active version of this filler driver where a direct comparison between the filler solution and a conventionnal crossover is possible.
Hare are current attemps towards the obtention of good square waves with 18 dB/o crossovers :
http://freerider.dyndns.org/anlage/LeCleach1.zip
http://freerider.dyndns.org/anlage/LeCleach2.zip
I found the Dickason's book very good, especially for newbies. I very much like the discussion about the filters, their slope, phase and group delays. Slight criticisms : it lacks a chapter on active filters (my french edition is the translation of the fifth one) and I would have prefer passive crossovers formulas written with "Pi" than with already calculated constants.
regards ~~~ Forr
§§§
PHASE ACCURATE
---There are many possibilities to achieve a transient-perfect response that don't need an additional driver like the filler driver concept does.---
---Nice work there by Mr. Le Cleach, though his approach is not transient-perfect but transient-improved.---
Any reference for transient-perfect crossovers ?
---There are many possibilities to achieve a transient-perfect response that don't need an additional driver like the filler driver concept does.---
---Nice work there by Mr. Le Cleach, though his approach is not transient-perfect but transient-improved.---
Any reference for transient-perfect crossovers ?
O.K. I must admit that a really transient-perfect speaker doesn't actually exist. But there are transient-improved topologies that come close enough to be called like that.
1.) Single-driver fullrange.
2.) 2nd order with overlap and delayed tweeter
3.) 2nd and third order with overlap and EQing.
4.) The combination of points 2 and 3 like it is used in some active studio monitors (Meyer, K+H, Relec)
5.) Transient-perfect constant-voltage crossovers. Only the first order version can be done passively (i.e. the ordinary 1st order crossover). The other ones have to be built as active subtractive topologies.
6.) Digital FIR filters
2, 3 and 4 could even be done passively but one is better off with active solutions.
Regards
Charles
1.) Single-driver fullrange.
2.) 2nd order with overlap and delayed tweeter
3.) 2nd and third order with overlap and EQing.
4.) The combination of points 2 and 3 like it is used in some active studio monitors (Meyer, K+H, Relec)
5.) Transient-perfect constant-voltage crossovers. Only the first order version can be done passively (i.e. the ordinary 1st order crossover). The other ones have to be built as active subtractive topologies.
6.) Digital FIR filters
2, 3 and 4 could even be done passively but one is better off with active solutions.
Regards
Charles
Hi Phase_Accurate,
The question, then, is : is a transient perfect response important ?
Most renowned authors and studies agree that it is detectable on some selected signals having special waveforms but not on programs (music...).
Nevertheless, if it is thought it must be an aim for ultimate fidelity, two remarks :
- the achievement of a transient perfect reponse must not be detrimental to the forward frequency response which is recognised as as the primary quality of the louspeaker (see Toole's studies among professionasl and critics).
- most improved transient crossover imply slow slopes. A hi-passed loudspeaker has a constant (12 dB/o) or increasing (6 dB/o) excursion down to its resonance, so it is prone to intermodulation.
Because of the transient nature of sound, easily associated with square waves, good figures for these last ones may have more impact than a linear frequency response on many customers or amateurs. For long, I was in this error despite the fact that my favourite speakers were not at all designed with consideration towards the transient response.
I am not sure to ever adopt the Le Cleac'h's solution, however I think it is a good compromise (easy to test with numeric crossovers such as the Berhinger DCX ) and worth to mention. Curiously, I should have think that people who have tried it and whom I am aware of would have show more enthousiasm.
The question, then, is : is a transient perfect response important ?
Most renowned authors and studies agree that it is detectable on some selected signals having special waveforms but not on programs (music...).
Nevertheless, if it is thought it must be an aim for ultimate fidelity, two remarks :
- the achievement of a transient perfect reponse must not be detrimental to the forward frequency response which is recognised as as the primary quality of the louspeaker (see Toole's studies among professionasl and critics).
- most improved transient crossover imply slow slopes. A hi-passed loudspeaker has a constant (12 dB/o) or increasing (6 dB/o) excursion down to its resonance, so it is prone to intermodulation.
Because of the transient nature of sound, easily associated with square waves, good figures for these last ones may have more impact than a linear frequency response on many customers or amateurs. For long, I was in this error despite the fact that my favourite speakers were not at all designed with consideration towards the transient response.
I am not sure to ever adopt the Le Cleac'h's solution, however I think it is a good compromise (easy to test with numeric crossovers such as the Berhinger DCX ) and worth to mention. Curiously, I should have think that people who have tried it and whom I am aware of would have show more enthousiasm.
If I can exprime only my personal opinion,
Some renowned researchers, at his time , was able to "demonstrate":
- all amplifiers sound the same
- all cables sound the same
- A/D + D/A zero differences
- Mp3 was "sufficient"
- audiophile lovers have only mirages and imaginations...
Then I am not surprise if they claim the inaudibility of "transient perfect speakers".
I think we must experimentate with our head and our ears...
Good weekend at alls,
Inertial
Some renowned researchers, at his time , was able to "demonstrate":
- all amplifiers sound the same
- all cables sound the same
- A/D + D/A zero differences
- Mp3 was "sufficient"
- audiophile lovers have only mirages and imaginations...
Then I am not surprise if they claim the inaudibility of "transient perfect speakers".
I think we must experimentate with our head and our ears...
Good weekend at alls,
Inertial
Even Le Cleach's admits that only few people are able to detect transient perfect responses. I would'nt deny the authorority of authors who have study the subject and who tell that this response is detectable with some signals and not on music. But there is one more question, difficult to answer. When listening to these special test signals, which are the prefered renditions : good transient or not ? When studying the audibility of group delays at very low frequencies, Fincham found that there was a detectable difference between small and large group delays, but he did'nt mention any preference among the testing listeners. A preference is not necessarily what one may think. It must not be forgotten that distorsions (group delay is a kind a distorsion) are not always interpreted as less good sounding.
Phase_Accurate
---99% of all multiway speakers aren't transient-perfect it is not that astonishing that most designers regard it as unimportant---
Tests on the audibility of phase or group delay have made with earphones to eradict the uncertainty of transients rendered by traditional loudpseakers.
Inertial
You seem to imply that science and audiophily are contradictory. Statistically, the percentage of audiophiles among researchers in acoustic engineering must be higher than the percentage of researchers in acoustic engineering among audiophiles. Otherwise I may misunderstand the real meaning of the expression "audiophile".
---99% of all multiway speakers aren't transient-perfect it is not that astonishing that most designers regard it as unimportant---
Tests on the audibility of phase or group delay have made with earphones to eradict the uncertainty of transients rendered by traditional loudpseakers.
Inertial
You seem to imply that science and audiophily are contradictory. Statistically, the percentage of audiophiles among researchers in acoustic engineering must be higher than the percentage of researchers in acoustic engineering among audiophiles. Otherwise I may misunderstand the real meaning of the expression "audiophile".
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.