I assume you brought a stereo pair... If so, there might be another reason why that crossover worked that good. Look at the comb pattern at the ear(s) from the cross talk, i.e. left speaker arriving at right ear and vice versa. That dip usually is around the ~1k-2-k mark. Having good power response seals the deal.
Toole's measurements (right) comparing center speaker with phantom center measured at the ear position. Left graph is a trace of the phantom measurement.
the first part .. well i would not call it "fact" but you can try your self for real and experience the phenomena.
the secund part is one way to achieve a goal.large overlap and low order filtering is one way. i think this part would be closer to opinion. a general advice would be that most of the time a 2nd order filter is more than enough. sortha rule of thumb.
I am surprised that no-one has really addressed the real issues here, though Gnobuddy touched on it on the last page, and the word "ringing" has been mentioned several times. In my opinion intermodulation (or any kind of nonlinear distortion) is a complete red herring.
The problem with a brickwall filter is the Gibbs Phenomenon. This is mathematically identical to diffraction of acoustic or electromagnetic waves around a sharp corner, and occurs whenever there is a sharp cutoff in a transfer function (such as that in a high-order speaker crossover). So when you split a signal into high- and low-pass parts with a brickwall filter, there is severe ringing in the time domain in both parts of the signal.
Now, if the filter is working perfectly, when you electronically sum the two parts again you get the original signal, with or without some phase issues depending on the implementation of the filter - the ringing on one side of the crossover point will cancel with identical "anti-ringing " on the other. On the other hand, acoustic summation never works properly. In an anechoic chamber you might get some approximation to correct recombination on a line extending right out away from a point on the speaker between the drivers, but anywhere away from this line there are different phase shifts between the two signals and you will get a mess. At one point in space you might get cancellation of the ringing at one frequency but not another, while at a nearby point the ringing that cancelled ten centimetres away might be in phase and actually be reinforced.
In a real listening environment you won't even get correct summation on axis, as the sound from each driver will get to you after multiple reflections. With real drivers, the situation gets even worse, as you have to factor in non-ideal phase responses and different off-axis behaviour for drivers with different dimensions.
I've never heard a system with this kind of filter, but I would expect the result to sound completely weird, with big response anomalies around the crossover frequency changing rapidly as you move your head. Stereo imaging will be awful, and listening to any actual music will be very fatiguing.
Alex
The problem with a brickwall filter is the Gibbs Phenomenon. This is mathematically identical to diffraction of acoustic or electromagnetic waves around a sharp corner, and occurs whenever there is a sharp cutoff in a transfer function (such as that in a high-order speaker crossover). So when you split a signal into high- and low-pass parts with a brickwall filter, there is severe ringing in the time domain in both parts of the signal.
Now, if the filter is working perfectly, when you electronically sum the two parts again you get the original signal, with or without some phase issues depending on the implementation of the filter - the ringing on one side of the crossover point will cancel with identical "anti-ringing " on the other. On the other hand, acoustic summation never works properly. In an anechoic chamber you might get some approximation to correct recombination on a line extending right out away from a point on the speaker between the drivers, but anywhere away from this line there are different phase shifts between the two signals and you will get a mess. At one point in space you might get cancellation of the ringing at one frequency but not another, while at a nearby point the ringing that cancelled ten centimetres away might be in phase and actually be reinforced.
In a real listening environment you won't even get correct summation on axis, as the sound from each driver will get to you after multiple reflections. With real drivers, the situation gets even worse, as you have to factor in non-ideal phase responses and different off-axis behaviour for drivers with different dimensions.
I've never heard a system with this kind of filter, but I would expect the result to sound completely weird, with big response anomalies around the crossover frequency changing rapidly as you move your head. Stereo imaging will be awful, and listening to any actual music will be very fatiguing.
Alex
i did not digg it up but a test was conducted with a digitally separated musical material.
so supposedly it had no ringing but the results where STILL verry well belov expectation and it made the listeners have headaches.
like real head aches.
so supposedly it had no ringing but the results where STILL verry well belov expectation and it made the listeners have headaches.
like real head aches.
I just don't get all these comments.
The DEQX is considered by many to be an excellent sounding crossover with many posts and threads on this forum related to it, and many positive reviews on line.
In the DEQX world a 60db/oct filter isn’t steep. The default suggested slopes are 96db/oct and can be as high as 300db/oct.
So where are all these headaches, horrific artifacts and pre-ringing which makes it's use offensive?
The DEQX is considered by many to be an excellent sounding crossover with many posts and threads on this forum related to it, and many positive reviews on line.
In the DEQX world a 60db/oct filter isn’t steep. The default suggested slopes are 96db/oct and can be as high as 300db/oct.
So where are all these headaches, horrific artifacts and pre-ringing which makes it's use offensive?
I think there is a little time/frequency domain confusion here. An abrupt step indeed does ring but that violates Nyquist, just try it and see for yourselves there are several free tools that support overlap and add FFT filtering and you can have a 1 bin wide transition between high and low frequencies. I just tried it in CoolEdit with a short Taiko drum piece and nothing obviously horrible. Not to say one way or the other that this is actually a good thing to do but I think the actual artifacts are being exaggerated.
I've used the DEQX for years now. No headaches. I generally use filters of 60 or so dB/oct.
There's a learning curve. But the most important (and hardest) thing is to find a place to take measurements with a nice long window that doesn't include reflections. In-room measurements for mid/high are fine with a shorter window, but it's nice to have corrections going as low in frequency as possible. Once good measurements are obtained, you can play with crossover parameters til your heart's content.
Sheldon
There's a learning curve. But the most important (and hardest) thing is to find a place to take measurements with a nice long window that doesn't include reflections. In-room measurements for mid/high are fine with a shorter window, but it's nice to have corrections going as low in frequency as possible. Once good measurements are obtained, you can play with crossover parameters til your heart's content.
Sheldon
That's because there was no reason, it's just a statistical accident.
We are not going to agree, but I respect your right to believe something very different from what I believe.
Peace, and good music to you!
-Gnobuddy
Wouldn't a coaxial driver minimize the effects of a sudden transition from one frequency band to another?
Yes, sort of. That presumes that the acoustic center of the coaxial driver remains constant across all relevant frequencies. The shift may be small enough to be inconsequential.
http://www.newaudio.it/Tannoy/AUTOGRAF MINI/0705045_c[1].jpg
this hot thing from tannoy was the all capitals hit when it was available, and yess it makes a huge lot of difference.
it would avoid a lot of problems you would have with a brickwall xover, but it would not solve everything.
first, non-digital way is simply impossible as it would mean infinate delay as pointed out by others.
the other thing is still the ringing witch may be canceld, or may not be.
in theory it should be, in real life no chanse.
one of the more noteworthy things would be to mention is material, and construction.
even in this case, the tweeter is different from the other driver. has different partial resonations. so even then the transmission would not be smooth, it would be rather drastic.
the resonatin the cone does builds up mechanical ringing, at with a brick wall filter, it would be the most obvious as it cuts in and out.
not uite what i personally would like to hear ever.
this hot thing from tannoy was the all capitals hit when it was available, and yess it makes a huge lot of difference.
it would avoid a lot of problems you would have with a brickwall xover, but it would not solve everything.
first, non-digital way is simply impossible as it would mean infinate delay as pointed out by others.
the other thing is still the ringing witch may be canceld, or may not be.
in theory it should be, in real life no chanse.
one of the more noteworthy things would be to mention is material, and construction.
even in this case, the tweeter is different from the other driver. has different partial resonations. so even then the transmission would not be smooth, it would be rather drastic.
the resonatin the cone does builds up mechanical ringing, at with a brick wall filter, it would be the most obvious as it cuts in and out.
not uite what i personally would like to hear ever.
This makes no sense. Resonances or cone break-up in band are an inherent property of a given driver. Getting rid of them means not reaching those frequencies for the driver. This is precisely a major benefit of steep digitally derived filters. For passive crossovers these resonances are dealt with by adding a corresponding notch filter - not always an easy job as it has to be accounted for in the calculation for the entire filter.
And by the way, I'm talking steep relative to passive crossovers, not infinitely steep (if infinitely steep is the definition of "brick wall" for this thread, then discussing theory is fine, but the discussion of practical applications is rather pointless).
Sheldon
Last edited:
See the attached thread referencing Joseph Audio.
http://www.diyaudio.com/forums/solid-state/5663-infinite-slope-crossover.html
http://www.diyaudio.com/forums/solid-state/5663-infinite-slope-crossover.html
then let me put it thisway. the cone acts sortha like a pendulum, it retains some momentum.
if you cut off the input, it will still have momentum. no crossover can deal with that part.
it isin't anything major, but a verry sharp crossover , let it be a not infinate but verry verry sharp, will get you one kind of driver inherited "sounding" , and swtiching to an other speaker will give a nother kind of. with our verry sharp filter, we just simply allow this to shine. and if the material passes the crossover point freqvently, then its going to give you one hell of an annoying sound.
witch may not be interesting in a public adress setup, i can imagine it would be masked by the crowd and stuff.
if you cut off the input, it will still have momentum. no crossover can deal with that part.
it isin't anything major, but a verry sharp crossover , let it be a not infinate but verry verry sharp, will get you one kind of driver inherited "sounding" , and swtiching to an other speaker will give a nother kind of. with our verry sharp filter, we just simply allow this to shine. and if the material passes the crossover point freqvently, then its going to give you one hell of an annoying sound.
witch may not be interesting in a public adress setup, i can imagine it would be masked by the crowd and stuff.
I don't want to intrude in this very interesting discussion,but don't forget that one of the greatest contribution to a speaker sound is the room...
And finally, whatever we try to get the perfect speaker or not, some of the best sounding ones to the greater number of listeners are usually not the perfect (measurements wise) ones.
In fact most speakers are imperfect, not just because the technology is, but by design! The main purpose is not perfect sound (remembered the cd marketing!), but to please humans, and we the humans are very imperfect creatures, that came a long way from evolution. We are not yet measuring instruments, but still searching in the jungle what are the beasts that can eat us 😀
And finally, whatever we try to get the perfect speaker or not, some of the best sounding ones to the greater number of listeners are usually not the perfect (measurements wise) ones.
In fact most speakers are imperfect, not just because the technology is, but by design! The main purpose is not perfect sound (remembered the cd marketing!), but to please humans, and we the humans are very imperfect creatures, that came a long way from evolution. We are not yet measuring instruments, but still searching in the jungle what are the beasts that can eat us 😀
I always feel a speaker with gentler filter sounds more neutral than one with steeper filter. I had a chance to test many different filters in the same design. In my case, 18dB sounds better than 24dB with time and phase aligned speaker.
If I would design speaker from scratch, I would choose units and enclosure design that allow filter as gentle as possible. I now consider steep filter only to be used to hide the defect of the unit or speaker design...
And the best filter would be no filter.
If I would design speaker from scratch, I would choose units and enclosure design that allow filter as gentle as possible. I now consider steep filter only to be used to hide the defect of the unit or speaker design...
And the best filter would be no filter.
Do you know exactly how YG achieves this?
I listened to a YG speaker last year and it's probably the most more coherent passive speaker which uses high Xover that I've ever heard.
Hi there,
what a usual or common crossover, having just say 2-4th order (acoustically ...) in high- and lowpass filters, does for us in most of the recent 2-3 way designs is
to smooth out changes in radiation patterns from the "lower" (bass-/midrange) driver to the "higher" driver (tweeter), especially when looking at horizontal directivity.
For the vast majority of loudspeakers still having unmatched radiation patterns of drivers (and their kind of mounting/baffling) at the crossover frequency "blending" both driver's patterns at crossover is the only way to avoid a sharp (mostly down-) step in the Directivity Index (DI) at the XO frequency of the speaker (but see below, for the vertical dispersion).
As we know e.g. from research done by Dr. Floyd Toole at the NRC (Canada) and at Harman (US), having a "smooth" (mostly slowly rising with frequency) DI over frequency is an important factor in the listener's estimation and ranking of a loudspeaker due to perceived quality ...
Putting other possible issues such as ringing, practicality etc., that may come with very steep filters, aside for a while, the above said is a requirement having high priority and has to be solved first. As i pointed out, such directivity issues are in no way adressed or mitigated by using a very steep crossover but are worsened by using them.
Having drivers instead with well matched DI, mounted close enough compared to wavelength (vertically) not to cause a too obvious narrowing of the vertical dispersion at XO, we are more free to chose our crossover due to this respect. But now also the - only apparent - need for very steep ('brickwall') filter has gone, after having the underlying and relevant issues solved.
So my personal answer to the question
"Why not use brickwall filters ?" would be simply
"Because it does not solve relevant issues due to listener's preference and/or technical quality of a loudspeaker."
what a usual or common crossover, having just say 2-4th order (acoustically ...) in high- and lowpass filters, does for us in most of the recent 2-3 way designs is
to smooth out changes in radiation patterns from the "lower" (bass-/midrange) driver to the "higher" driver (tweeter), especially when looking at horizontal directivity.
For the vast majority of loudspeakers still having unmatched radiation patterns of drivers (and their kind of mounting/baffling) at the crossover frequency "blending" both driver's patterns at crossover is the only way to avoid a sharp (mostly down-) step in the Directivity Index (DI) at the XO frequency of the speaker (but see below, for the vertical dispersion).
As we know e.g. from research done by Dr. Floyd Toole at the NRC (Canada) and at Harman (US), having a "smooth" (mostly slowly rising with frequency) DI over frequency is an important factor in the listener's estimation and ranking of a loudspeaker due to perceived quality ...
Putting other possible issues such as ringing, practicality etc., that may come with very steep filters, aside for a while, the above said is a requirement having high priority and has to be solved first. As i pointed out, such directivity issues are in no way adressed or mitigated by using a very steep crossover but are worsened by using them.
Having drivers instead with well matched DI, mounted close enough compared to wavelength (vertically) not to cause a too obvious narrowing of the vertical dispersion at XO, we are more free to chose our crossover due to this respect. But now also the - only apparent - need for very steep ('brickwall') filter has gone, after having the underlying and relevant issues solved.
So my personal answer to the question
"Why not use brickwall filters ?" would be simply
"Because it does not solve relevant issues due to listener's preference and/or technical quality of a loudspeaker."
Last edited:
I've been advocating active steep slope (not infinite slope) filters around here for some time. My speakers now are two ways with a slope in the 250db/ octave range.
I use a PC for the source and a program called equalizer APO to control and equalize the drivers individually. Soundcard => amps => drivers. This allows me to experiment a lot with different slopes.
If there are any disadvantages to steep slopes I have a hard time finding them. I can hear lobing pretty easily and steep slopes and low crossovers are the only way to get rid of it. So the advantages outweigh the disadvantages at least in an active set up. Acoustic interference is much more noticeable than steep slopes, but I wouldn't go steeper than needed.
I think I can hear some ringing if the slope is too steep. But here's the thing, just start with a more shallow slope and steepen it at certain intervals. Its the high q nature of the slope itself that causes the ringing.
Obviously driver spacing and dispersion matter. In some ways the slope depends on the speaker. But with a steep slope you can hand off from a woofer to a tweeter in the 1000hz range when both are basically omnidirectional. This is very important for driver integration.
There are a couple other things to consider. Because one driver is handing off to another so quickly you want to match dispersion obviously but also the tonal qualities of the drivers. Stay away from break up modes. Also possibly the qts of the drivers could come into play. A very damped woofer with an underdamped tweeter. Most would say this isn't audible and doesn't matter except at fs, but I suspect it might.
I've wondered about this. I thought it was a law of physics that sound travels through a medium at distinct frequencies?
Yes this is an issue. Also matching levels exactly. All in all steep slopes are much more suited to active systems with processing power. But the end result can sound almost as integrated as a full range driver.
I use a PC for the source and a program called equalizer APO to control and equalize the drivers individually. Soundcard => amps => drivers. This allows me to experiment a lot with different slopes.
If there are any disadvantages to steep slopes I have a hard time finding them. I can hear lobing pretty easily and steep slopes and low crossovers are the only way to get rid of it. So the advantages outweigh the disadvantages at least in an active set up. Acoustic interference is much more noticeable than steep slopes, but I wouldn't go steeper than needed.
I think I can hear some ringing if the slope is too steep. But here's the thing, just start with a more shallow slope and steepen it at certain intervals. Its the high q nature of the slope itself that causes the ringing.
Obviously driver spacing and dispersion matter. In some ways the slope depends on the speaker. But with a steep slope you can hand off from a woofer to a tweeter in the 1000hz range when both are basically omnidirectional. This is very important for driver integration.
There are a couple other things to consider. Because one driver is handing off to another so quickly you want to match dispersion obviously but also the tonal qualities of the drivers. Stay away from break up modes. Also possibly the qts of the drivers could come into play. A very damped woofer with an underdamped tweeter. Most would say this isn't audible and doesn't matter except at fs, but I suspect it might.
I've wondered about this. I thought it was a law of physics that sound travels through a medium at distinct frequencies?
Yes this is an issue. Also matching levels exactly. All in all steep slopes are much more suited to active systems with processing power. But the end result can sound almost as integrated as a full range driver.
Last edited:
- Status
- Not open for further replies.