Hi forr,
I symply constate that "official result of this scentific tests" always contraddictc listening perception of audiophile lovers.
The absurdity of this old question is evident : can a doctor affermate
that I have not pain on my head because he have made to me a perfect X-ray and the result is negative ?
about the audibility oh phase: what do they want hear with a earphone?
again, if only a few person can hear a difference, why considerate the others that do not hear? It is a paradox... can you maybe take 50
"casual" subjects on the street and to demand they can hear the difference from to say CD and SACD?? Are we crazy?
I total disagree about methodology of 99/100 of these "scientific" test.
Le Cleach is right IMHO: only a few subjects are able to detect...
life is not democratic.
Just my personal opinion, of course.
Cheers,
Inertial
I symply constate that "official result of this scentific tests" always contraddictc listening perception of audiophile lovers.
The absurdity of this old question is evident : can a doctor affermate
that I have not pain on my head because he have made to me a perfect X-ray and the result is negative ?
about the audibility oh phase: what do they want hear with a earphone?
again, if only a few person can hear a difference, why considerate the others that do not hear? It is a paradox... can you maybe take 50
"casual" subjects on the street and to demand they can hear the difference from to say CD and SACD?? Are we crazy?
I total disagree about methodology of 99/100 of these "scientific" test.
Le Cleach is right IMHO: only a few subjects are able to detect...
life is not democratic.
Just my personal opinion, of course.
Cheers,
Inertial
A few notes on x-over-regions and the audibility of phase errors...
From my experience you will have a hard time anyway reckognizing the (lack of) intrinsic phase shift of conventional x-overs. There are quite a few other effects masking the possibly audible information lying within an optimum time behaviour. The most evil fogging of spatial information to me is any kind of *stored energy issues* of which there are plenty within most (99%?) of common speaker designs.
Since any stored energy will introduce badly smeared delays to differnt frequencies - often in the magnitude of generic x-over phaseshifts or larger - one will usually not be able to reckognize the benefits of any "linear phase" system, because frequencies still do *not* arrive time aligned. This even happening on systems which appear & measure quite "time aligned" over the whole audio range and already produce a "somewhat" rectangle. Not enough, I'd like to say.
Some of the "usual suspect" causes for stored energy/delays I've always been looking into are:
- misalignment of driver's distances to listener
- driver ringing @ cone breakup (including phase breakups)
- driver ringing @ resonant freq. (tweeters & midranges)
- intrinsic ringing of any steep sloped filter circuits
- short term back reflections from cabinet (driver air flow)
- edge diffraction effects
- x-over components vibrations
- lack of rugged/stiff driver mounting
- parasitic cabinet wall resonances
So you might say "Hey, these are 20/30/50dB down from the signal!". Yes they are, but I must ask: aren't any spatial informations quite "down" compared to the full signal too? So just in case you want to experience details at, say 30dB below level within a given signal, you might want an "error level" way below these signals. Just in case...
Now the above issues are not any "must care" critera (but these actually some of my "must look at"s), and finally stored energies are far from beeing the only thingies worth considering. So your mileage will vary as will your preferences and weightings. Nevertheles I'll want to repeat: without having these (and other effects) in check, a discussion about audibility & benefits of "phase accuracy" seems quite futile.
Now the initial thread title is "Width of crossover region" and that's where I see a tight link to the above said. When thinking of a really low "error level" (not only with stored energy, but in general...), at the same time targeting to lowest possible order crossovers for your desired "phase acurrate" design, you will run into "crossover regions" covering *at least* a decade above and below the x-o-frequency.
So one of my ways of dealing with x-overs was to completely give up the term "crossover region" long ago. Any driver I look at, I will initially "implement" from about ~10Hz to ~40kHz. Then I will weight the multiple effects introduced by each driver at any frequency. (e.g. Dome tweeter at 30Hz? Cone excursion is worth a look... Dome tweeter at resonance? Top priority! Sub-woofer? Controlled performance at 5-10kHz is highly preferred.)
Now in the end it's all about listening. I might be one of the few who had the chance to mangle phase accurate multi-way systems on such a strict design level, and what can I say? The listening experience is outraging immersive. The speakers themselves completely dissapear to the ear. The just sit there as if serving not any pupose at all. Provided a decent recording (plus matching electronics) there is virtually not any sound coming out of the speakers themselves. Very irritating to the common listener. Spatial localisation especially in depth field is as frightening accurate as it can get (30-50m and more! but *only* if it's actually on the recording!). Quite some people regarding these speakers as "bad" simply because "something was missing" (hehe... missing with the speaker? or rather with the listener? Ahemm, BTW: see my signature for this too...)
Again, no one has to put this much effort into his work, but in case discussing, please do not obfuscate the (IMHO) massive importance of speaker phase / delay issues due to the fact that you are lacking knowledge of construction and/or experience with the related effects thereof.
Just my two microseconds, thanx for reading, you are now free to beat me up.
regards, redunzelizer
From my experience you will have a hard time anyway reckognizing the (lack of) intrinsic phase shift of conventional x-overs. There are quite a few other effects masking the possibly audible information lying within an optimum time behaviour. The most evil fogging of spatial information to me is any kind of *stored energy issues* of which there are plenty within most (99%?) of common speaker designs.
Since any stored energy will introduce badly smeared delays to differnt frequencies - often in the magnitude of generic x-over phaseshifts or larger - one will usually not be able to reckognize the benefits of any "linear phase" system, because frequencies still do *not* arrive time aligned. This even happening on systems which appear & measure quite "time aligned" over the whole audio range and already produce a "somewhat" rectangle. Not enough, I'd like to say.
Some of the "usual suspect" causes for stored energy/delays I've always been looking into are:
- misalignment of driver's distances to listener
- driver ringing @ cone breakup (including phase breakups)
- driver ringing @ resonant freq. (tweeters & midranges)
- intrinsic ringing of any steep sloped filter circuits
- short term back reflections from cabinet (driver air flow)
- edge diffraction effects
- x-over components vibrations
- lack of rugged/stiff driver mounting
- parasitic cabinet wall resonances
So you might say "Hey, these are 20/30/50dB down from the signal!". Yes they are, but I must ask: aren't any spatial informations quite "down" compared to the full signal too? So just in case you want to experience details at, say 30dB below level within a given signal, you might want an "error level" way below these signals. Just in case...
Now the above issues are not any "must care" critera (but these actually some of my "must look at"s), and finally stored energies are far from beeing the only thingies worth considering. So your mileage will vary as will your preferences and weightings. Nevertheles I'll want to repeat: without having these (and other effects) in check, a discussion about audibility & benefits of "phase accuracy" seems quite futile.
Now the initial thread title is "Width of crossover region" and that's where I see a tight link to the above said. When thinking of a really low "error level" (not only with stored energy, but in general...), at the same time targeting to lowest possible order crossovers for your desired "phase acurrate" design, you will run into "crossover regions" covering *at least* a decade above and below the x-o-frequency.
So one of my ways of dealing with x-overs was to completely give up the term "crossover region" long ago. Any driver I look at, I will initially "implement" from about ~10Hz to ~40kHz. Then I will weight the multiple effects introduced by each driver at any frequency. (e.g. Dome tweeter at 30Hz? Cone excursion is worth a look... Dome tweeter at resonance? Top priority! Sub-woofer? Controlled performance at 5-10kHz is highly preferred.)
Now in the end it's all about listening. I might be one of the few who had the chance to mangle phase accurate multi-way systems on such a strict design level, and what can I say? The listening experience is outraging immersive. The speakers themselves completely dissapear to the ear. The just sit there as if serving not any pupose at all. Provided a decent recording (plus matching electronics) there is virtually not any sound coming out of the speakers themselves. Very irritating to the common listener. Spatial localisation especially in depth field is as frightening accurate as it can get (30-50m and more! but *only* if it's actually on the recording!). Quite some people regarding these speakers as "bad" simply because "something was missing" (hehe... missing with the speaker? or rather with the listener? Ahemm, BTW: see my signature for this too...)
Again, no one has to put this much effort into his work, but in case discussing, please do not obfuscate the (IMHO) massive importance of speaker phase / delay issues due to the fact that you are lacking knowledge of construction and/or experience with the related effects thereof.
Just my two microseconds, thanx for reading, you are now free to beat me up.
regards, redunzelizer
I once read about a test where they came to the conclusion that there is a small audible effect on rectangulars that were passed through all-pass filters. Now it makes me wonder how scientific a comparison between rectangulars (i.e. a static signal) via headphones and music via speakers really is ???????
I belong to the camp who thinks that transient response matters. But I am also aware that this is not the only one and not the most important parameter.
Regards
Charles
I belong to the camp who thinks that transient response matters. But I am also aware that this is not the only one and not the most important parameter.
Regards
Charles
Everyone has to form their own opinions regarding the compromises they're prepared to make when designing a loudspeaker.
I personally believe a true high-fidelity speaker should be as phase accurate as possible.
http://phasecoherent.com/index.htm
I personally believe a true high-fidelity speaker should be as phase accurate as possible.
http://phasecoherent.com/index.htm
I have to ask for you understanding that I'm not able to give you every explicite detail, since I'm somewhat under non-disclusure. Nevertheless I will try to explain some of the basic principles involved with such a "strict" design.
The basic description of that system is:
Passive 4-way, point-symmetric electrodynamic transducer with
1x Tweeter: 28mm dome
2x Midrange: 54mm dome
2x Midbass: 17cm, high excursion
2x Woofer: 25cm, high excursion
Tweeters & Midranges with 6++mm custom machined front plates, mounted for closest possible distance. Plates designed for both extreme stiffness and controlled on/off-axis responce. Mid & tweeter plates not only screwed onto cabinet but also onto (into) each other. Machined basket enforcement rings for midbass & woofers. Additional mounting holes drilled into basket for midbass & woofers. All drivers mounted twice, both at front plate/basket, as well as at the magnet. (Yes, even the tweeter.)
Compound cabinet walls (stiff case plus heavy damping material) with lots of cross braces. Sand-filled sandwich was in consideration too, but with cabinet resonance character mainly defined by design & bracing, and second only by material, you may chose for your convenience. Then, weight! The more the better. Think in "tenths of tons"...
Driver center alignment in correlation to "sweet spot" listening position can be done by a front panel with angled level-stepping, sending/diffusing possible reflections of said steps into "uncritical" directions, away from listener and walls/floor (stealth airplane design anyone?). Gradually bent sides of cabinet to virtually eliminate any edge diffraction effects. Front panel stepping at the same time providing a thickness of 7-10cm (3-4") front plate for midbass and woofer (sorry, it happened just by accident, hehe...)
Now for something showing most important in any case: Conical rear cutouts for midbass & woofers! (that one being a wide baffle design there was room enough anyway...). Do "send away" any output from the driver's rear! Care to annihilate that output within the cabinet before it hits the cones again from the back after some time. Also, try experimenting with additional "deflection plates" within to further increase the "damping travel path" of these delayed energies. Will show even greater positive effects. Or get a dipole...
In short: stiff and damped case, most heavy, extremely rigid mechanics, controlled dispersion & diffraction.
Having (hopefully) done all that mechanical homework (which, btw, would apply the very same for any neat & budget fullrange speaker! Just look at zaphaudio's tricky W3-871 design! http://www.zaphaudio.com/archives.html ) we're back on topic: another few remarks on an x-over design for such a system:
Note that all of this stuff has been done before, but rarely alltogether within one design. Valuable thoughts & calculus can be found all over the web & in countless forii. (Although basically it would be searching needles in the haystack...)
What to care for with any such crossover in terms of "stored energy"?
Behaviour of midbass, midrange & tweeter at resonant frequencies has to be mastered. Electrical damping (Qe) of drivers limited by a conventional low-pass filter will more or less suffer badly. (Ring! Ring... midrange calling! tweeter on the other line too? Stress! Hectics! Transient panic! Damn, where is my original signal?) Additionally most existing drivers, even with driving them at zero source impedance, are somewhat underdamped already at their resonance frq. You will always have to do something about it.
So, in this case, applying some kinds of passive(!) Linkwitz-Riley transforms, thus moving "effective Fr", changing the actual acoustic Qts *within* x-over environment to suitable values, has shown mandatory. Yes, provided capable drivers, this can certainly be done within any reasonable ranges.
IMO, it's way cheaper to assign some new "acoustical effective" Thiele/Small params to a given (otherwise excellent) driver, instead of waiting for that other "dream of a driver" getting never built. Active systems can do many (but not all!) wonders for you here too, however, I'll stick to my passive versions for some reasons.
Then there are resonant peaks to be "mirrored out". Given a suitable driver, you will have to model networks compensating these in *both* amplitude and phase, thus omitting other certain kinds of ringing and coming yet another bit closer to your "acoustical design target" functions that way.
Of course there is the usual other stuff to handle: additional phase shifting wherever required, providing the actual crossover functions, chosing the appropiate quality for each part, seperate x-over enclosure, rugged parts mounting again and again, etc., etc., whatever will fit your intentions.
All these heretic x-over efforts? What for?
When having turned a former "loud"speaker into a fairly non-resonant "tranducer", not only the overall transfer function will start to behave quite like desired, but suddenly phase *relations* of any involved drivers will start to show extremely important. (10Hz to 40kHz for all drivers, remember?). For a given theoretical design (spreadsheet galore...) compared to the actual acoustical driver output (B&K, MLSSA, Osci...), the phase errors should stay well below a few degrees (a really tough one, believe me) and no more than 10-20deg within "lesser-critical" ranges.
It's "as easy as that"...
regards & greetings from within the haystack,
redunzelizer
Inertial
I can name, at least, three renowned scientists/engineers/researches who are very skilled in sound recording. So they know a lot about what is the sound and what affects it through the whole record/reproduction chain. I have the greatest respect for the writings of these renowned people whose theorical and practical knowledge as well as extended experience are shared by only very few audiophiles.
Phase_Accurate
--- I belong to the camp who thinks that transient response matters. But I am also aware that this is not the only one and not the most important parameter.---
The camp you belong to justifies your pseudo !
There is a general agreement that a linear frequency response must be attained before any attempt to get a good transient response from a speaker.
I can name, at least, three renowned scientists/engineers/researches who are very skilled in sound recording. So they know a lot about what is the sound and what affects it through the whole record/reproduction chain. I have the greatest respect for the writings of these renowned people whose theorical and practical knowledge as well as extended experience are shared by only very few audiophiles.
Phase_Accurate
--- I belong to the camp who thinks that transient response matters. But I am also aware that this is not the only one and not the most important parameter.---
The camp you belong to justifies your pseudo !
There is a general agreement that a linear frequency response must be attained before any attempt to get a good transient response from a speaker.
forr said:
Please take a look at the according time/frequency dependencies. In case you are seriously heading for a (near) transient-perfect response, linear frequency response will show up by itself. You just can't avoid that happening anyway!
In other words, I personally do *only* have to care for linear frequency response in case I am intentionaly building a conventional speaker with too few drivers and/or (then naturally occuring) all-pass phase shifts. Which is happening in times and is fun as well. But otherwise...
Care has to be taken not to mix things up. While a linear frequency response is necessary for good transient performance - a linear frequency response alone does not guarantee a good transient response.
The reason for this is that also non minimum-phase systems can have perfect linear frequency response (like LR 4 transfer functions).
Otherwise I must admit that your design seems to be a very careful and competent one.
Regards
Charles
phase_accurate said:
You are absolutely right. I might have explained my views a bit too irritating:
redunzelizer said:
I just wanted to express that within these "strict" designs I have moved to designing & working mainly in the time-domain. Due to the dependecies of frequency & time domain with any true "linear phase systems", I have come to the conclusion that with a "transient-perfect response, linear frequency response will show up by itself" But only with these kind of systems, of course.
I must assume that be both mean the very same thing...
regards
Crossovers and response deviations
A few months ago, I experienced what Kelticwizard suggested in post #4 :
--- 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.---
But I have done it entirely with analog systems, sine signals of different frequencies feeding :
- two similar speakers through a twin channel amplifier on which I slowly vary the volume control on one channel
- one speaker through a mono-amplifier which input received a mix of two same sinewave, one set at different levels.
- same as above but with headphones.
This was done at comfortable levels to avoid well known loudness effects. All three arrangements gave similar results. I encourage anybody to duplicate one of them.
I wanted to check the just noticeable differences of levels which can be detected. This should decide which crossovers to use in an uncompromised system.
The result of my experiences are in total disagreement with 12 dB differences said to be just noticeable. Whatever the frequency, I noticed differences vanish only when the weak signal is at a level about -30 dB under the main signal.
That was quite a surprise, but this is not at all contradictory with what we know of psycho-acoustics. There is a general agreement that just noticeable differences of levels are about 0.3 dB on sine waves and 0.5 to 1.0 dB on natural sounds like speech, music...
When feeding the same sinewave signals at different levels in two loudspeakers, A and B, set aside, the total sound pressure is the addition of the sound pressures of each output :
Ptot = pA + pB
If sound pressure pB generated by B is at –30 dB below sound pressure pA generated by A , we can write, scaling pA to 1 :
Level A = 0 dB SPL à pressure pA = 1
Level B = -30 dB SPL à pressure pB = 0.03122
Level Tot = 20 log(1+0.03122) = +0.27 dB
The result is then a change of +0.3 dB of sound pressure
Some more calculations with other values :
Level A = 0 dB SPL à pressure pA = 1
Level B = -29 dB SPL à pressure pB = 0.035
Level Tot = 20 log(1+0.035) = +0.3 dB
Level A = 0 dB SPL à pressure pA = 1
Level B = -24.5 dB SPL à pressure pB = 0.054
Level Tot = 20 log(1+0.059) = +0.5 dB
Level A = 0 dB SPL à pressure pA = 1
Level B = -18.3 dB SPL à pressure pB = 0.122
Level Tot = 20 log(1+0.122) = +1 dB
Level A = 0 dB SPL à pressure pA = 1
Level B = -11.7 dB SPL à pressure pB = 0.122
Level Tot = 20 log(1+0.259) = +2 dB
Level A = 0 dB SPL à pressure pA = 1
Level B = -7.7 dB SPL à pressure pB = 0.412
Level Tot = 20 log(1+0.412) = +3 dB
Level A = 0 dB SPL à pressure pA = 1
Level B = 0 dB SPL à pressure pB = 1
Level Tot = 20 log(1+1) = +6 dB
This little maths work (anything wrong ?) may help to determine the adequate frequencies and slopes of a crossover, taking in account the hi-passed response of the medium or tweeter (which, in most cases, behaves as a second order filter).
Using an electronic CAD software where the low-pass and the high-pass filters of a crossover are input, the response, if the drivers were perfect, can be displayed. Adding a second order hi-pass filter, of same frequency and Q resonance as the medium or tweeter used, in the hi-pass section of the crossover gives a more realistic response of the whole.
Now, the choice of the crossover is just a matter of tolerance allowed by the designer in the deviations of linearity in the frequency response. I think crossovers should not introduce deviations greater than 0.5 dB in un-compromised systems or 1 dB for good systems. Deviations of more than 1 dB introduced by a crossover should be compensated elsewhere.
Some time ago, a discussion on a french forum concluded that the resonance of speakers should be attenuated by 30 dB. I agree with that now.
A few months ago, I experienced what Kelticwizard suggested in post #4 :
--- 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.---
But I have done it entirely with analog systems, sine signals of different frequencies feeding :
- two similar speakers through a twin channel amplifier on which I slowly vary the volume control on one channel
- one speaker through a mono-amplifier which input received a mix of two same sinewave, one set at different levels.
- same as above but with headphones.
This was done at comfortable levels to avoid well known loudness effects. All three arrangements gave similar results. I encourage anybody to duplicate one of them.
I wanted to check the just noticeable differences of levels which can be detected. This should decide which crossovers to use in an uncompromised system.
The result of my experiences are in total disagreement with 12 dB differences said to be just noticeable. Whatever the frequency, I noticed differences vanish only when the weak signal is at a level about -30 dB under the main signal.
That was quite a surprise, but this is not at all contradictory with what we know of psycho-acoustics. There is a general agreement that just noticeable differences of levels are about 0.3 dB on sine waves and 0.5 to 1.0 dB on natural sounds like speech, music...
When feeding the same sinewave signals at different levels in two loudspeakers, A and B, set aside, the total sound pressure is the addition of the sound pressures of each output :
Ptot = pA + pB
If sound pressure pB generated by B is at –30 dB below sound pressure pA generated by A , we can write, scaling pA to 1 :
Level A = 0 dB SPL à pressure pA = 1
Level B = -30 dB SPL à pressure pB = 0.03122
Level Tot = 20 log(1+0.03122) = +0.27 dB
The result is then a change of +0.3 dB of sound pressure
Some more calculations with other values :
Level A = 0 dB SPL à pressure pA = 1
Level B = -29 dB SPL à pressure pB = 0.035
Level Tot = 20 log(1+0.035) = +0.3 dB
Level A = 0 dB SPL à pressure pA = 1
Level B = -24.5 dB SPL à pressure pB = 0.054
Level Tot = 20 log(1+0.059) = +0.5 dB
Level A = 0 dB SPL à pressure pA = 1
Level B = -18.3 dB SPL à pressure pB = 0.122
Level Tot = 20 log(1+0.122) = +1 dB
Level A = 0 dB SPL à pressure pA = 1
Level B = -11.7 dB SPL à pressure pB = 0.122
Level Tot = 20 log(1+0.259) = +2 dB
Level A = 0 dB SPL à pressure pA = 1
Level B = -7.7 dB SPL à pressure pB = 0.412
Level Tot = 20 log(1+0.412) = +3 dB
Level A = 0 dB SPL à pressure pA = 1
Level B = 0 dB SPL à pressure pB = 1
Level Tot = 20 log(1+1) = +6 dB
This little maths work (anything wrong ?) may help to determine the adequate frequencies and slopes of a crossover, taking in account the hi-passed response of the medium or tweeter (which, in most cases, behaves as a second order filter).
Using an electronic CAD software where the low-pass and the high-pass filters of a crossover are input, the response, if the drivers were perfect, can be displayed. Adding a second order hi-pass filter, of same frequency and Q resonance as the medium or tweeter used, in the hi-pass section of the crossover gives a more realistic response of the whole.
Now, the choice of the crossover is just a matter of tolerance allowed by the designer in the deviations of linearity in the frequency response. I think crossovers should not introduce deviations greater than 0.5 dB in un-compromised systems or 1 dB for good systems. Deviations of more than 1 dB introduced by a crossover should be compensated elsewhere.
Some time ago, a discussion on a french forum concluded that the resonance of speakers should be attenuated by 30 dB. I agree with that now.
Hi Phase_Accurate,
---If you are talking about the fundamental resonance of a driver then I strongly disagree but I fully agree when you are talking about cone break-up modes and the like.---
I thought of both. However, most of the time, the fundamental resonance of tweeter and medium behaves is standard : it's a second order high-pass filter of which simulation is easy.
On the contrary, cone break up is very different between drivers of the same size and its influence has to be seen in a less general manner than the fundamental resonance. I think it could be fairly simulated using notch peaking filters.
The only text I know which deals with out of band response unlinearities is the famous JAES paper by Marshall Leach called something like "The neglected factor in crossover design". I am very surprised that I never found other references for this problem. All responses I got on the subject ware vague, the advice being invariably : resonances must sufficiently attenuated. How much is never said. It appears to me that it could be easily simulated with a CAD program.
Here are the first results using a first order crossover set at a frequency of 2000 Hz, which can be easily scaled. Each line indicates the resonance of the high pass driver (of which response is equivalent to a second order high pass filter, chosen Q values are 1.0, 0.707 and 0.5). Then the maximal deviations of frequency response and phase are indicated.
XOVER : ORDER 1, 2000 Hz (6 dB/o, Butterworth)
DRIVER : HI-PASS, ORDER 2 (12 dB/o, Q = 1)
¨¨¨¨¨¨¨¨¨¨¨¨¨¨¨¨¨¨¨¨¨¨¨¨¨¨¨¨¨¨¨¨¨¨¨¨
2000 Hz --> -34. dB, -90° +90°
1414 Hz --> -15. dB, -45° +40°
1000 Hz --> -9.0 dB, -32° +23°
0707 Hz --> -6.0 dB, -22° +14°
0500 Hz --> -4.0 dB, -16° +09°, resonance at –12 dB
0350 Hz --> -2.5 dB, -11° +06°, resonance at –15 dB
0250 Hz --> -2.0 dB, -08° +04°, resonance at –18 dB
0125 Hz --> -1.0 dB, -03° +02°, resonance at –24 dB
0064 Hz --> -0.5 dB, +02° -09°, resonance at –30 dB
XOVER : ORDER 1, 2000 Hz (6 dB/o, Butterworth)
DRIVER : HI-PASS ORDER 2, Q = 0.7 (12 dB/o, Butterworth)
2000 Hz --> -18. dB, -53° +47°
1414 Hz --> -12. dB, -37° +31°
1000 Hz --> -8.0 dB, -27° +22°
0707 Hz --> -5.0 dB, -20° +15°
0500 Hz --> -3.5 dB, -14° +10°, resonance at –12 dB
0353 Hz --> -2.5 dB, -10° +07°, resonance at –15 dB
0250 Hz --> -1.9 dB, -07° +05°, resonance at –18 dB
0125 Hz --> -1.0 dB -04° +02°, resonance at –24 dB
0064 Hz --> -0.5 dB, -01° +07°, resonance at –30 dB
XOVER : ORDER 1, 2000 Hz (6 dB/o Butterworth)
DRIVER : HI-PASS, ORDER 2, Q = 0.5 (12 dB/o)
2000 Hz --> -15. dB, -45° +43°
1414 Hz --> -10. dB, -34° +20°
1000 Hz --> -8.0 dB, -26° +23°
0707 Hz --> -5.5 dB, -20° +17°
0500 Hz --> -4.2 dB, -14° +12°,resonance at –12 dB
0353 Hz --> -2.9 dB, -10° +09°,resonance at –15 dB
0250 Hz --> -2.3 dB, -07° +06°,resonance at –18 dB
0125 Hz --> -1.1 dB, +03° -03°,resonance at –24 dB
0064 Hz --> -0.6 dB, +02° -10°,resonance at –30 dB
For evaluations, I choose the following criteria :
Uncompormised quality : deviation less than 0.5 dB.
High quality : deviation less than 1 dB.
Average quality : deviation less than 3 dB.
We see that they are met for attenuations of respectively :
–30 dB, -24 dB and –18 dB (if deviations of about 2 dB are chosen for this last one).
This is not a demonstration that speakers can’t be designed using first order filters. It must be seen as an investigation of the influence of treble or medium drivers on the frequency response. Surprisingly to me, the phase response is always very good in all usable above cases, which means than when the crossover is set at more than two octaves above the driver resonance.
Speakers can certainly be designed using first order filters but, for ultimate quality, it probably implies some correction that higher slope crossovers do not need.
---If you are talking about the fundamental resonance of a driver then I strongly disagree but I fully agree when you are talking about cone break-up modes and the like.---
I thought of both. However, most of the time, the fundamental resonance of tweeter and medium behaves is standard : it's a second order high-pass filter of which simulation is easy.
On the contrary, cone break up is very different between drivers of the same size and its influence has to be seen in a less general manner than the fundamental resonance. I think it could be fairly simulated using notch peaking filters.
The only text I know which deals with out of band response unlinearities is the famous JAES paper by Marshall Leach called something like "The neglected factor in crossover design". I am very surprised that I never found other references for this problem. All responses I got on the subject ware vague, the advice being invariably : resonances must sufficiently attenuated. How much is never said. It appears to me that it could be easily simulated with a CAD program.
Here are the first results using a first order crossover set at a frequency of 2000 Hz, which can be easily scaled. Each line indicates the resonance of the high pass driver (of which response is equivalent to a second order high pass filter, chosen Q values are 1.0, 0.707 and 0.5). Then the maximal deviations of frequency response and phase are indicated.
XOVER : ORDER 1, 2000 Hz (6 dB/o, Butterworth)
DRIVER : HI-PASS, ORDER 2 (12 dB/o, Q = 1)
¨¨¨¨¨¨¨¨¨¨¨¨¨¨¨¨¨¨¨¨¨¨¨¨¨¨¨¨¨¨¨¨¨¨¨¨
2000 Hz --> -34. dB, -90° +90°
1414 Hz --> -15. dB, -45° +40°
1000 Hz --> -9.0 dB, -32° +23°
0707 Hz --> -6.0 dB, -22° +14°
0500 Hz --> -4.0 dB, -16° +09°, resonance at –12 dB
0350 Hz --> -2.5 dB, -11° +06°, resonance at –15 dB
0250 Hz --> -2.0 dB, -08° +04°, resonance at –18 dB
0125 Hz --> -1.0 dB, -03° +02°, resonance at –24 dB
0064 Hz --> -0.5 dB, +02° -09°, resonance at –30 dB
XOVER : ORDER 1, 2000 Hz (6 dB/o, Butterworth)
DRIVER : HI-PASS ORDER 2, Q = 0.7 (12 dB/o, Butterworth)
2000 Hz --> -18. dB, -53° +47°
1414 Hz --> -12. dB, -37° +31°
1000 Hz --> -8.0 dB, -27° +22°
0707 Hz --> -5.0 dB, -20° +15°
0500 Hz --> -3.5 dB, -14° +10°, resonance at –12 dB
0353 Hz --> -2.5 dB, -10° +07°, resonance at –15 dB
0250 Hz --> -1.9 dB, -07° +05°, resonance at –18 dB
0125 Hz --> -1.0 dB -04° +02°, resonance at –24 dB
0064 Hz --> -0.5 dB, -01° +07°, resonance at –30 dB
XOVER : ORDER 1, 2000 Hz (6 dB/o Butterworth)
DRIVER : HI-PASS, ORDER 2, Q = 0.5 (12 dB/o)
2000 Hz --> -15. dB, -45° +43°
1414 Hz --> -10. dB, -34° +20°
1000 Hz --> -8.0 dB, -26° +23°
0707 Hz --> -5.5 dB, -20° +17°
0500 Hz --> -4.2 dB, -14° +12°,resonance at –12 dB
0353 Hz --> -2.9 dB, -10° +09°,resonance at –15 dB
0250 Hz --> -2.3 dB, -07° +06°,resonance at –18 dB
0125 Hz --> -1.1 dB, +03° -03°,resonance at –24 dB
0064 Hz --> -0.6 dB, +02° -10°,resonance at –30 dB
For evaluations, I choose the following criteria :
Uncompormised quality : deviation less than 0.5 dB.
High quality : deviation less than 1 dB.
Average quality : deviation less than 3 dB.
We see that they are met for attenuations of respectively :
–30 dB, -24 dB and –18 dB (if deviations of about 2 dB are chosen for this last one).
This is not a demonstration that speakers can’t be designed using first order filters. It must be seen as an investigation of the influence of treble or medium drivers on the frequency response. Surprisingly to me, the phase response is always very good in all usable above cases, which means than when the crossover is set at more than two octaves above the driver resonance.
Speakers can certainly be designed using first order filters but, for ultimate quality, it probably implies some correction that higher slope crossovers do not need.
I'm very skeptical about those flatness criteria because the effects of room acoustics and off-axis radiation are being neglected all the time. Room acoustics are the main source of trouble and probably the main reason that lead you to believing that you could perceive a driver attenuated 30dB over one at 0dB. I recommend doing that test outdoors with no walls or ceiling, with both drivers sitting together in the floor and listening a few meters away. You can also try playing a 1Khz tone in any room and check how the perceived volume changes as you move, it's quite easy to get fluctuations in excess of +/-6dB by moving your head just 15cm.
Eva,
The design of speakers involves the mastering of how to get a wanted frequency response in anechoic conditions, on and off axis, even if you will never access to such conditions.
At this stage of the procedure, you cannot take account of room acoustics, bar on a statistically average basis.
The work I communicated above concerns only one aspect of design, often neglected, as said by you know who.
The design of speakers involves the mastering of how to get a wanted frequency response in anechoic conditions, on and off axis, even if you will never access to such conditions.
At this stage of the procedure, you cannot take account of room acoustics, bar on a statistically average basis.
The work I communicated above concerns only one aspect of design, often neglected, as said by you know who.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.