Hi there,
do any of you have references pointing to crossovers with other slopes than 6 dB/oct and multiples thereof? I have a fullrange driver that decreases ca. 4 dB/oct on axis and would like to try a simple passive x-o to a tweeter. (Later on I plan to do this cleanly with active x-o and more standard slopes).
do any of you have references pointing to crossovers with other slopes than 6 dB/oct and multiples thereof? I have a fullrange driver that decreases ca. 4 dB/oct on axis and would like to try a simple passive x-o to a tweeter. (Later on I plan to do this cleanly with active x-o and more standard slopes).
I don't think you're going to find an easy answer for this. But I don't think it's impossible either because there's always a digital dsp solution. It may be impossible in a passive solution though...
The 6db/oct magic number comes from the slope of the impedance rise of a cap/coil with respect to frequency.
Impedance wise, Caps are X = 1/jwc while coils are X = jwL. Since j and w are effectively not controllable, I think that in order to have a slope in betwen multiples of 6db/oct you would need a device that decreases capacitance with frequency or increases inductance with frequency or vice versa.
--
Danny
The 6db/oct magic number comes from the slope of the impedance rise of a cap/coil with respect to frequency.
Impedance wise, Caps are X = 1/jwc while coils are X = jwL. Since j and w are effectively not controllable, I think that in order to have a slope in betwen multiples of 6db/oct you would need a device that decreases capacitance with frequency or increases inductance with frequency or vice versa.
--
Danny
Years ago for a white noise to pink noise converter I built a
3dB/octave filter effective over the range 20Hz to 20KHz.
Its easily done with a series resistor and then a number of
parallel RC sections to ground, though much easier to work
out using a simulator rather than pen and paper.
I think I needed six RC sections for less than 0.5dB ripple.
Don't neglect BSC.
sreten.
3dB/octave filter effective over the range 20Hz to 20KHz.
Its easily done with a series resistor and then a number of
parallel RC sections to ground, though much easier to work
out using a simulator rather than pen and paper.
I think I needed six RC sections for less than 0.5dB ripple.
Don't neglect BSC.
sreten.
Filters are built up of components whose impedances either go +6dB/oct (inductors) or -6dB/oct (capacitors). This means that any filter ultimately has a slope of n*6dB/oct, where n is an integer. In a transition region, the actual slope can be anything, and the pinking filter that sreten described is an extreme example on how to extend this transition region to the entire audible spectrum.
Simplified, the n is the order of the filter, and often synonymous with the number of Ls and Cs in the filter. This is why n is an integer. So, saying that a filter has a slope of eg 8 dB/oct might be true, but probably only for a small frequency region, and thus pretty useless IMO. Knowing that it is a second order filter would make me feel much better since it tells me something about the design. If I am interested in the details about the response, I'd like a response curve instead.
To return to your problem, it seems as if your speaker might need a dose of a "transition region" in a range. As sreten says, BSC should also be taken inte account. Maybe it is the BSC you see? To say anything about the details on this we would need to know more about the conditions for your measurements.
HTH
Simplified, the n is the order of the filter, and often synonymous with the number of Ls and Cs in the filter. This is why n is an integer. So, saying that a filter has a slope of eg 8 dB/oct might be true, but probably only for a small frequency region, and thus pretty useless IMO. Knowing that it is a second order filter would make me feel much better since it tells me something about the design. If I am interested in the details about the response, I'd like a response curve instead.
To return to your problem, it seems as if your speaker might need a dose of a "transition region" in a range. As sreten says, BSC should also be taken inte account. Maybe it is the BSC you see? To say anything about the details on this we would need to know more about the conditions for your measurements.
HTH
Actually in a way it is BSC.
The driver is a Diatone fullrange in dipole arrangement (open baffle). The baffle is 14" wide. The measurement was on-axis, in-room using Linkwitz's burst signals and his peak detector (DIY).
What I figure is that, in the lower range and above the baffle rolloff, I get a boost from random addition of front and rear waves. As F goes up, dispersion narrows, so that I am getting more and more the "real" monopole on-axis response because the phases don't add anymore.
So I just bought tweeters to improve on the radiation characteristics of the system. The Diatone by itself is great on-axis and images really well, however due to the narrowing dispersion it is certainly not power flat and that shows as a certain dullness off axis, and a lack of perceived dynamic punch. I thought of helping him out with a tweeter.
Now, if I do a classic crossover, I somehow negate the fullrange advantage, and in order to get a clean transition iI'd have to do it in a range where the Diatone's dispersion is still clean (say, 2 kHz). I am looking for a way to just add the lacking off axis radiation above 3-4 kHz while keeping the Diatone working fullrange. So that would mean, slowly phasing in the tweeter.
Another option I thought of is to simply add the tweeter off axis. Either top firing, or rear firing. That would ensure to make its output diffuse - if not achieving true off axis fill-in, it would help the power spectrum, and if I do it above 4 kHz, the effects on imaging and the lacking time alignment should lead to increased "spaciousness" without hurting coherence and imaging. It may just work by setting the correct x-o point, and to lower the tweeter level until we're out of the important frequency range, say, fix it until 10 or 12 kHz and let it fall where it may above that.
Or, I could sum the channels out of phase at line level, that gives me the reverberant component only, and use the tweeter to radiate just that (I have the tri-amp setup prepared already).
Or, I could add the tweeter on axis and put sort of an inverted phase plug in front of it to remove the on axis component.
Am I making my life too hard ?
The driver is a Diatone fullrange in dipole arrangement (open baffle). The baffle is 14" wide. The measurement was on-axis, in-room using Linkwitz's burst signals and his peak detector (DIY).
What I figure is that, in the lower range and above the baffle rolloff, I get a boost from random addition of front and rear waves. As F goes up, dispersion narrows, so that I am getting more and more the "real" monopole on-axis response because the phases don't add anymore.
So I just bought tweeters to improve on the radiation characteristics of the system. The Diatone by itself is great on-axis and images really well, however due to the narrowing dispersion it is certainly not power flat and that shows as a certain dullness off axis, and a lack of perceived dynamic punch. I thought of helping him out with a tweeter.
Now, if I do a classic crossover, I somehow negate the fullrange advantage, and in order to get a clean transition iI'd have to do it in a range where the Diatone's dispersion is still clean (say, 2 kHz). I am looking for a way to just add the lacking off axis radiation above 3-4 kHz while keeping the Diatone working fullrange. So that would mean, slowly phasing in the tweeter.
Another option I thought of is to simply add the tweeter off axis. Either top firing, or rear firing. That would ensure to make its output diffuse - if not achieving true off axis fill-in, it would help the power spectrum, and if I do it above 4 kHz, the effects on imaging and the lacking time alignment should lead to increased "spaciousness" without hurting coherence and imaging. It may just work by setting the correct x-o point, and to lower the tweeter level until we're out of the important frequency range, say, fix it until 10 or 12 kHz and let it fall where it may above that.
Or, I could sum the channels out of phase at line level, that gives me the reverberant component only, and use the tweeter to radiate just that (I have the tri-amp setup prepared already).
Or, I could add the tweeter on axis and put sort of an inverted phase plug in front of it to remove the on axis component.
Am I making my life too hard ?
Hi MBK,
don't forget you have not only the BSC to keep in mind, but also -as you've already mentioned yourself- the radiation behaviour of your FR driver at the crossover frequency. Even at 2kHz, you will encounter a jump in the off-axis response due to highly different driver sizes (beaming of the FR, while the tweeter is happily radiating into 2pi w/ its own baffle step features...). So, even if on-axis everything looks neat, you might get into trouble off-axis. At what frequency the power response will jerk is actually dependent on the dipole response plus the 2 pi (or so) radiation of the tweeter. (As a matter of fact, you most certainly will get into trouble off-axis, in terms of what you measure - but this might nonetheless sound good.)
Also, watch out for a weird vertical interference pattern you might introduce by using filter slopes which are too flat (b/c the overlap will extend over a much too large frequency range); also the acoustical axis might change. This is especially critical if you go to really high frequencies. In turn, you can't go too low without compromising the tweeter (and the FR concept, of course). In the end, you might very well lose (much) more than you gain.
You could also try and play around with little, cone-shaped diffusors or acoustical lenses in front of your driver. But this will decrease on-axis SPL (and probably flatness of the response, too).
Cheers,
bk.
don't forget you have not only the BSC to keep in mind, but also -as you've already mentioned yourself- the radiation behaviour of your FR driver at the crossover frequency. Even at 2kHz, you will encounter a jump in the off-axis response due to highly different driver sizes (beaming of the FR, while the tweeter is happily radiating into 2pi w/ its own baffle step features...). So, even if on-axis everything looks neat, you might get into trouble off-axis. At what frequency the power response will jerk is actually dependent on the dipole response plus the 2 pi (or so) radiation of the tweeter. (As a matter of fact, you most certainly will get into trouble off-axis, in terms of what you measure - but this might nonetheless sound good.)
Also, watch out for a weird vertical interference pattern you might introduce by using filter slopes which are too flat (b/c the overlap will extend over a much too large frequency range); also the acoustical axis might change. This is especially critical if you go to really high frequencies. In turn, you can't go too low without compromising the tweeter (and the FR concept, of course). In the end, you might very well lose (much) more than you gain.
You could also try and play around with little, cone-shaped diffusors or acoustical lenses in front of your driver. But this will decrease on-axis SPL (and probably flatness of the response, too).
Cheers,
bk.
Hallo Brummkreisel,
actually the tweeter should be out of the baffle step problem range is crossed over > 2kHz. The dispersion needs to be addressed though. The Diatone is a 6.5 " driver and should be OK (omnidirectional) up to close to 2 kHz and due to the dipole cancellation effects off axis, even higher. I think it would be perfectly workable to use a steep x-o at 2-3kHz, however it would be nice if I had the best of both worlds, fullrange dipole with filled-in off axis response...
actually the tweeter should be out of the baffle step problem range is crossed over > 2kHz. The dispersion needs to be addressed though. The Diatone is a 6.5 " driver and should be OK (omnidirectional) up to close to 2 kHz and due to the dipole cancellation effects off axis, even higher. I think it would be perfectly workable to use a steep x-o at 2-3kHz, however it would be nice if I had the best of both worlds, fullrange dipole with filled-in off axis response...
In some situations it should also be possible to use asymmetric crossovers to achieve intermediate slopes over a certain frequency range. I've thought of doing this in situations where a 1st order crossover is desirable, but you want to push the response down a bit more later on to avoid a resonance 4 octaves later.
The math would require some work, maybe best done empirically/by simulation.
The math would require some work, maybe best done empirically/by simulation.
Hi MBK,
yes, you're right, the tweeter's edge diffraction features have to be dealt with (that's what I meant with baffle step features).
Some general remarks:
You will not obtain a filled off-axis response by adding a tweeter, because your Diatone acts as a dipole. Obviously, adding a 4pi or 2pi-radiating tweeter will not maintain this dipole pattern at high frequencies. You would need to use a dipole tweeter as well (some B&G stuff, or AMT, should work well). This is what your'e talking about, right?
I don't believe that your 6.5" driver really behaves omnidirectional to up to 2 kHz, since from theory, one expects the driver's directivity to become <2pi at the wavelength corresponding to its diaphragm circumference. In your case, a 6.5" ~ 165mm = lambda@700Hz or so. Going to twice that value, you're still well below 2kHz, but now already suffering from driver directivity. (Don't forget that, by using shallow filter slopes, the overlap region is larger, and the 6.5" will "bleed" into higher frequencies.)
However, in real life there is a dependence on diaphragm geometry and stiffness of when extactly the diver starts to beam. Henc, you should measure the directivity of your driver on an infinite baffle (groound plane or so), so that it is not masked by dipole cancellation e ffects, to get an idea of where you off-axis response will suffer from a more or less deep dip - and then you will have to decide how much dB you're willing to accept at, say, 60 degrees off-axis. This is the nasty part...
Then, you might consider using an even-order filter to avoid a shift of the acoustic axis inherent to odd-order filters (you don't want to use a d'Appolito configuration after all, do you?). If your tweeter can handle 12dB/oct at, say, 2 kHz, and is well-behaved in terms of group delay etc, you can use an active, phase-subtracting crossover (said to be transient perfect), so that you won't lose the nice, FR driver's sstep response - if you consider this to be a spec to be met, that is. (However, it'll be very difficult to find a tweeter that can do.) Note that in any case, you should let the filters do the job and not the drivers roll-off (off-axis response!).
Cheers,
bk.
yes, you're right, the tweeter's edge diffraction features have to be dealt with (that's what I meant with baffle step features).
Some general remarks:
You will not obtain a filled off-axis response by adding a tweeter, because your Diatone acts as a dipole. Obviously, adding a 4pi or 2pi-radiating tweeter will not maintain this dipole pattern at high frequencies. You would need to use a dipole tweeter as well (some B&G stuff, or AMT, should work well). This is what your'e talking about, right?
I don't believe that your 6.5" driver really behaves omnidirectional to up to 2 kHz, since from theory, one expects the driver's directivity to become <2pi at the wavelength corresponding to its diaphragm circumference. In your case, a 6.5" ~ 165mm = lambda@700Hz or so. Going to twice that value, you're still well below 2kHz, but now already suffering from driver directivity. (Don't forget that, by using shallow filter slopes, the overlap region is larger, and the 6.5" will "bleed" into higher frequencies.)
However, in real life there is a dependence on diaphragm geometry and stiffness of when extactly the diver starts to beam. Henc, you should measure the directivity of your driver on an infinite baffle (groound plane or so), so that it is not masked by dipole cancellation e ffects, to get an idea of where you off-axis response will suffer from a more or less deep dip - and then you will have to decide how much dB you're willing to accept at, say, 60 degrees off-axis. This is the nasty part...
Then, you might consider using an even-order filter to avoid a shift of the acoustic axis inherent to odd-order filters (you don't want to use a d'Appolito configuration after all, do you?). If your tweeter can handle 12dB/oct at, say, 2 kHz, and is well-behaved in terms of group delay etc, you can use an active, phase-subtracting crossover (said to be transient perfect), so that you won't lose the nice, FR driver's sstep response - if you consider this to be a spec to be met, that is. (However, it'll be very difficult to find a tweeter that can do.) Note that in any case, you should let the filters do the job and not the drivers roll-off (off-axis response!).
Cheers,
bk.
BK,
good points. What I mean with extended useability of the Diatone in dispersion is actually a mix of driver directivity and dipole cancellation. True, the driver becomes directional below 2 kHz. According to factory graphs, directivity of the Diatone becomes noticeable at 60 deg around 1000 Hz. At 30 deg , it is much better: just 1-2 dB down up til 4.5 kHz... (the 60 deg response is more than 10 dB down at that point). BUT, according to theory, the dipole operation means that the driver is already having a directional radiation pattern in the low end. That means, first the dipole leads to a stable directivity at the low end, then the driver beaming starts , and at some point only the natural beaming exceeds the typical dipole pattern. You are right, there is no way to do this but to measure off axis response, preferably of the combo, on a test baffle.
Siegfried Linkwitz from memory also noted for his designs that directivity in the high end of a dipole wooder is rather beneficial to smooth out off axis response changes. He uses 8" drivers crossed to monopole tweeters around 1.4 kHz and manages to successfully integrate the dispersion. I don't know exactly how that works, because as you noted the tweeter starts out as an omnidirectional monopole...
Anyway, I'd like to stay clear of the 2-4 kHz region for x-o because that's the most sensitive part of human hearing. Ideally I'd choose 4 kHz, the Fletcher Munson curve goes up sharply at that point AND, unlike the 2 kHz region, we are far away from phase sensitivity of human directional hearing. In other words at 4 kHz I can be more sloppy on the time alignment. Now, the measured in-room response on axis on dipole baffle, of the Diatone alone, also starts decreasing at ca. 4-5 dB/oct around 4-5 kHz (and comes crashing down substantially >10kHz). This is what gave me the idea of a gentle fill-in of the tweeter, at 4-5 kHz, no worries about time alignment, fullrange levels off by itself, and tweeter can be protected from over-excursion with a second HP at say, 1 kHz or so. The tweeter is a Vifa TG27-35, Fs 650 Hz, so only the excursion limit is of relevance here.
good points. What I mean with extended useability of the Diatone in dispersion is actually a mix of driver directivity and dipole cancellation. True, the driver becomes directional below 2 kHz. According to factory graphs, directivity of the Diatone becomes noticeable at 60 deg around 1000 Hz. At 30 deg , it is much better: just 1-2 dB down up til 4.5 kHz... (the 60 deg response is more than 10 dB down at that point). BUT, according to theory, the dipole operation means that the driver is already having a directional radiation pattern in the low end. That means, first the dipole leads to a stable directivity at the low end, then the driver beaming starts , and at some point only the natural beaming exceeds the typical dipole pattern. You are right, there is no way to do this but to measure off axis response, preferably of the combo, on a test baffle.
Siegfried Linkwitz from memory also noted for his designs that directivity in the high end of a dipole wooder is rather beneficial to smooth out off axis response changes. He uses 8" drivers crossed to monopole tweeters around 1.4 kHz and manages to successfully integrate the dispersion. I don't know exactly how that works, because as you noted the tweeter starts out as an omnidirectional monopole...
Anyway, I'd like to stay clear of the 2-4 kHz region for x-o because that's the most sensitive part of human hearing. Ideally I'd choose 4 kHz, the Fletcher Munson curve goes up sharply at that point AND, unlike the 2 kHz region, we are far away from phase sensitivity of human directional hearing. In other words at 4 kHz I can be more sloppy on the time alignment. Now, the measured in-room response on axis on dipole baffle, of the Diatone alone, also starts decreasing at ca. 4-5 dB/oct around 4-5 kHz (and comes crashing down substantially >10kHz). This is what gave me the idea of a gentle fill-in of the tweeter, at 4-5 kHz, no worries about time alignment, fullrange levels off by itself, and tweeter can be protected from over-excursion with a second HP at say, 1 kHz or so. The tweeter is a Vifa TG27-35, Fs 650 Hz, so only the excursion limit is of relevance here.
I think You`re spot on with this!
I have tried this several times with good success and IMO this is the best way to add a tweeter to a fullrange without compromising the fullrange approach.
The "conventional" method of adding a tweeter (on front baffle and with crossovers) to a fullrange IMO only mess up things more than doing any good (well, You may have better top end but the rest suffer).
There must be at least some misunderstandings here...
Normal, analog filters are always n*6dB down on the slope. The difference betwen the various classes, Bessel - Butterworth etc, merely describes the behaviour around the "knee" and in the stop-band, - i.e. before you reach the asymtotic slope corresponding to the filter order.
The various classes, as most of you know, are accomplished by "skewing" the values of the LC product.
This is also used in the 3-way "fill-in" filters, where ths LP and HP are both "skewed" 2nd order, and the filler BP is first order.
It is an intersting thought, though, if DSP filters can be set to any slope value ?? Some thinking needs to be done.......
Normal, analog filters are always n*6dB down on the slope. The difference betwen the various classes, Bessel - Butterworth etc, merely describes the behaviour around the "knee" and in the stop-band, - i.e. before you reach the asymtotic slope corresponding to the filter order.
The various classes, as most of you know, are accomplished by "skewing" the values of the LC product.
This is also used in the 3-way "fill-in" filters, where ths LP and HP are both "skewed" 2nd order, and the filler BP is first order.
It is an intersting thought, though, if DSP filters can be set to any slope value ?? Some thinking needs to be done.......
Hmm, tried the tweeter in a primitive setup just with a cap. It's very crude, because the sensitivity is higher by ca 3 dB on the tweeter than the Diatone, and offset both vertical and depthwise, is large. Anyway, the point was to help diffuse dispersion. Doesn't seem to work really.
Crossed low, the power spectrum seems more "right", but it doesn't sound realistic. Crossed high, it's a subtle improvement, but too little to matter really. I guess I have to do this properly, with measurements and active x-o and time alignment and steep slopes etc...
In detail:
Rear firing with 4.7 uF cap, nominal -3dB point around 4-5k: awful. Very non - integrated sound although the spectral balance seems about right.
Front firing, 10 uF cap, nominal -3 dB at around 2.5 k: better integration, but still sticks out like a sore thumb.
(it's actually perfecty listenable, like your usual HT setup in the store, and just as mis-aligned)
Front firing, 2.2 uF cap, nominal -3 dB at (gasp) 14 k: best. just adds a little sparkle and doesn't stick out. But then again, kinda pointless.
OK conclusion see above, it has to be done like a conventional X-O or not at all.
Crossed low, the power spectrum seems more "right", but it doesn't sound realistic. Crossed high, it's a subtle improvement, but too little to matter really. I guess I have to do this properly, with measurements and active x-o and time alignment and steep slopes etc...
In detail:
Rear firing with 4.7 uF cap, nominal -3dB point around 4-5k: awful. Very non - integrated sound although the spectral balance seems about right.
Front firing, 10 uF cap, nominal -3 dB at around 2.5 k: better integration, but still sticks out like a sore thumb.
(it's actually perfecty listenable, like your usual HT setup in the store, and just as mis-aligned)
Front firing, 2.2 uF cap, nominal -3 dB at (gasp) 14 k: best. just adds a little sparkle and doesn't stick out. But then again, kinda pointless.
OK conclusion see above, it has to be done like a conventional X-O or not at all.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.