I have seen a lot of people suggesting 4th order L-R for Xover. While I don't have an issue with this suggestion, I found that the number of Xover parts gets quite large if you use 4th order --> and if you translate this to Mundorf, Hovland or Crescendo, you are getting a huge bill.
Who is using 2nd order Xovers? Any good results? I have seen some commercial products, but not so sure about Xover order.
Who is using 2nd order Xovers? Any good results? I have seen some commercial products, but not so sure about Xover order.
I'm not sure you should just be asking for a comparison between one crossover design and another, without including what kind of speaker you want to build. Crossover topology is dictated by speaker drivers, enclosure, design goals etc.
Also, it would be good to tell those in the know here if you have any loudspeaker measuring equipment/software and/or crossover design software.
Also, it would be good to tell those in the know here if you have any loudspeaker measuring equipment/software and/or crossover design software.
My personal choice is the first order slope. It is the only naturally transient perfect response, but there is a transient perfect second order scheme out there, it requires a little EQ.
The slope is of course a combo of the drivers natural slope and the crossover. The crossover will probably look like a first order on paper but may be completely different. First order is demanding of the drivers. It works for me though.
The slope is of course a combo of the drivers natural slope and the crossover. The crossover will probably look like a first order on paper but may be completely different. First order is demanding of the drivers. It works for me though.
Hi indm,
I thought that single pole (first order) filters were automatically Butterworth.
If this is the case then these do not have excellent transient ability. Although second only to Bessel.
I think you need Bessel to achieve that excellence.
I'm guessing here, but I think you cannot achiveve Bessel with a single pole filter.
I thought that single pole (first order) filters were automatically Butterworth.
If this is the case then these do not have excellent transient ability. Although second only to Bessel.
I think you need Bessel to achieve that excellence.
I'm guessing here, but I think you cannot achiveve Bessel with a single pole filter.
I agree Andrew, you can't achieve bessel with a single pole - in theory. I stretch it in practice. With the text book first order, the drivers are down 3dB at the crossover point. I often design a crossover by simulating the tweeter with a capacitor dominating the crossover region, and working the mid/woofer to match phase, and 6dB down points.
I don't think this is standard, but I like it.
P.S., isn't as easy as it sounds
I don't think this is standard, but I like it.
P.S., isn't as easy as it sounds
it doens't make too much sense to call 1st order filters by butterworth/bessel/ect... its just not that complex.
filters like bessel and butterworth are actually defined fairly generically mathmatically. its up to the designer to chose what "cutoff" is. for some apps you might choose -3dB, for others -1dB or -10dB.
1st order is great if you can get away with it. but speaker locations and other issues with the speakers themselves always seem to pop up.
filters like bessel and butterworth are actually defined fairly generically mathmatically. its up to the designer to chose what "cutoff" is. for some apps you might choose -3dB, for others -1dB or -10dB.
1st order is great if you can get away with it. but speaker locations and other issues with the speakers themselves always seem to pop up.
1st order is indeed the only "natural" crossover that is transient-perfect per se. But in practice you will end up with a 3rd order (at least !) electro-acoustical transfer function. There are higher-order symmetrical and asymmetrical crossovers possible which are transient perfect - but they have to be active and they have some disadvantages like large overlap, humps (though summing flat) and transitions that are even flatter than those of 1st order crossovers.
If you want the lowest THD, IMD and the least lobing for given drivers you are better off with higher-order crossovers. But they are a no-no if you want transient perfect response. Now you'd have to pick your poison !
Regards
Charles
If you want the lowest THD, IMD and the least lobing for given drivers you are better off with higher-order crossovers. But they are a no-no if you want transient perfect response. Now you'd have to pick your poison !
Regards
Charles
caenot said:
Hi,
Acoustic 4th order L-R crossovers do not always use as many parts as you think.
e.g. http://www.zaphaudio.com/audio-speaker17.html
/sreten.
the slope used in the crossovers make an audible difference according to blind tests (and measurements), so it matters. based on my readings the ideal would be 1st order, but since that's impractical and usually calls for worse sacrifices, I would recommend extremely steep cauer-elliptical crossovers. the steeper the better. reasoning is that while steeper crossovers create higher linear distortion compared to slow crossovers, it's for a far shorter part of the frequency. so it's much better overall.
Cotdt
--- I would recommend extremely steep cauer-elliptical crossovers. the steeper the better. reasoning is that while steeper crossovers create higher linear distortion compared to slow crossovers, it's for a far shorter part of the frequency. so it's much better overall.---
An example :
http://www.diyaudio.com/forums/showthread.php?s=&threadid=6655
--- I would recommend extremely steep cauer-elliptical crossovers. the steeper the better. reasoning is that while steeper crossovers create higher linear distortion compared to slow crossovers, it's for a far shorter part of the frequency. so it's much better overall.---
An example :
http://www.diyaudio.com/forums/showthread.php?s=&threadid=6655
forr said:
As far as I remember elliptical filters have only 6dB/oct for odd orders and 12dB/oct. for even orders, which may be enough for many engineering purposes is OK. , but not always for speaker crossover, for hearing is approximately logatythmic process (Weber-Fechner principle). At least 6dB may be insufficient for pink-noise-like music signals. What do you think?
I almost always use 1st order ELECTRICAL on the tweeter highpass, which typically provides 1st order acoustic for the 1st octave below fc, asymptoping to 3rd order acoustic further away. The bass mid filter is then designed to produce a complementary response (you need to play with software like LspCAD) . This method I think is the best compromise between true 1st order filters and the steeper designs.
I don't like steep filters - they will & must produce ringing in the off-axis response.
I don't like steep filters - they will & must produce ringing in the off-axis response.
notes:
1.) steep filters are not always easy to tune. basically you are more likely to get a sharp notch or peak in the summed response due to the component tolerances.
2.) filters using stopband zeros, such as elliptic, will have very steep effective slopes, but the final slope will be less. you can get an effective slope of -60dB/oct and better then -30dB in the stop band, but the final slope might only be -12dB/oct (or even -0dB/oct). BUT if you have a signal with -100dB of rejection at all frequencies in the stopband, it is still effectively blocked even though the slope of the filter doens't reject some frequencies more then others.
the final slope is -(p-z)*6dB/oct for a lowpass. for normal filters this is just -p*6. for other filters, you can have multiple stopband zeros. the more zeros, the more attenutation you get near the cutoff frequency, but the lower the final slope will be.
If your interested, look into hourglass filters.
1.) steep filters are not always easy to tune. basically you are more likely to get a sharp notch or peak in the summed response due to the component tolerances.
2.) filters using stopband zeros, such as elliptic, will have very steep effective slopes, but the final slope will be less. you can get an effective slope of -60dB/oct and better then -30dB in the stop band, but the final slope might only be -12dB/oct (or even -0dB/oct). BUT if you have a signal with -100dB of rejection at all frequencies in the stopband, it is still effectively blocked even though the slope of the filter doens't reject some frequencies more then others.
the final slope is -(p-z)*6dB/oct for a lowpass. for normal filters this is just -p*6. for other filters, you can have multiple stopband zeros. the more zeros, the more attenutation you get near the cutoff frequency, but the lower the final slope will be.
If your interested, look into hourglass filters.
Regarding the subject of stopband vs asymptotic slopes there are two camps anyway: Those who think that steep stopband attenuation is of prime importance (definitely not the guys who prefer transient-perfect behaviour) and those who think that the asymptotic behaviour is more important.
Both have their advantages and disadvantages.
It is even possible to build transient-perfect crossovers with zeroes in the stopband but these can only be built in active fashion. For comparison I have added the behaviour of an ordinary 1st order crossover (purple and yellow).
Regards
Charles
Attachments
TheChris
---filters using stopband zeros, such as elliptic, will have very steep effective slopes, but the final slope will be less. you can get an effective slope of -60dB/oct and better then -30dB in the stop band, but the final slope might only be -12dB/oct (or even -0dB/oct).---
The final slope might, electrically, only be -12dB/oct.
Using the Hardman crossover described in my previous post : acoustically, it may be more than that. Setting the notch of the hi-pass nearly the resonance of the unit will add a 12 db/o slope for an ultimate slope of 24 dB/o. The interest of such a crossover relies in the deep rejection of the energy storing zones of resonance and breakup.
Setting the Hardman's crossover frequency at 1500 Hz, attenuations are approximately :
-10 dB at 1400 and 1700 Hz
-20 dB at 1100 and 2100 Hz
-30 dB at 900 and 2400 Hz
notches at 800 and 2800 Hz,
Any signal rejected by more than 30 dB in the stop band should not very audible.
I think the group delay is about the same as the Linkwitz-Riley.
---filters using stopband zeros, such as elliptic, will have very steep effective slopes, but the final slope will be less. you can get an effective slope of -60dB/oct and better then -30dB in the stop band, but the final slope might only be -12dB/oct (or even -0dB/oct).---
The final slope might, electrically, only be -12dB/oct.
Using the Hardman crossover described in my previous post : acoustically, it may be more than that. Setting the notch of the hi-pass nearly the resonance of the unit will add a 12 db/o slope for an ultimate slope of 24 dB/o. The interest of such a crossover relies in the deep rejection of the energy storing zones of resonance and breakup.
Setting the Hardman's crossover frequency at 1500 Hz, attenuations are approximately :
-10 dB at 1400 and 1700 Hz
-20 dB at 1100 and 2100 Hz
-30 dB at 900 and 2400 Hz
notches at 800 and 2800 Hz,
Any signal rejected by more than 30 dB in the stop band should not very audible.
I think the group delay is about the same as the Linkwitz-Riley.
Just to go back to an earlier post. I may be wrong here but I think that in the case of 1st order filters a Butterworth is a Bessel isn't it?. It only becomes an issue when the order is greater than one. But don't shout if I've got this wrong as I only did two years of engineering. (Where I went the engineers were given a hard time. You could acquire a T shirt that said. "Last year I coundn't even spell ENGINEAR and now I are one"
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.