Certainly not in my case. Impedance correction is necessary in some cases (and as Bear says can be detrimental, it certainly was in my case with my midwoofers). Not using it can be to your advantage if you have an optimiser (or are willing to try lots of variations). The rising impedance can be used to simplify the crossover complexity when it is taken into account.
Absolutely! it has given me a whole new outlook, I was obsessed with getting a flat frequency response probably at the detriment of other things that were more important!
Tony.
I have not read this entire thread but simply googled as a member and found it. Raising it from the dead a little I know but it seems to me that high end speakers are often marked by excessive sibilance. In a hall the highs are incredibly muted naturally from what I have read. Most large halls roll of around 10k drastically.... if my memory serves. They also (alot of them) shout like crazy. I guess I like my Mahler more on the back row. I know I am going to get yelled at but to some small degree a speaker is like a well "plated" meal... it seems to taste better - I think. But I am no audiophile admittedly.
Since the thread has come back from the dead, I might as well post an update. Lynn's comments about girl and guitar speakers gave me the incentive to redo my crossovers, and the speakers are MUCH better now.
I originally had 2nd order acoustic (bessel) at 3Khz, but changed to 4th order acoustic (bessel) at 2.8Khz. The silibance is all but gone, and they cope much better now with dense source material. I still have one CD that is not good (but the problem is not silibance) but I've not yet tracked down whether this is source, amp or speakers.
As well as the steeper slope, another area that is much better is the phase tracking through the crossover region, whether this has helped with the silibance problem or not though I can't say.
Tony.
I originally had 2nd order acoustic (bessel) at 3Khz, but changed to 4th order acoustic (bessel) at 2.8Khz. The silibance is all but gone, and they cope much better now with dense source material. I still have one CD that is not good (but the problem is not silibance) but I've not yet tracked down whether this is source, amp or speakers.
As well as the steeper slope, another area that is much better is the phase tracking through the crossover region, whether this has helped with the silibance problem or not though I can't say.
Tony.
high a.wayne, You will have to be more specific 
Do you mean check the transfer function of each of the LP and HP of the crossover into say an 8 ohm dummy load? I haven't done that. I did simulate, and also did acoustic and impedance measurements after I built it, which matched quite well the simulated results.
Tony.
Do you mean check the transfer function of each of the LP and HP of the crossover into say an 8 ohm dummy load? I haven't done that. I did simulate, and also did acoustic and impedance measurements after I built it, which matched quite well the simulated results. Tony.
Seems to me that "sibilance" often can be thought of as a type of dynamic intermodulation distortion. That would explain how increasing the slope and raising the frequency a little on wintermute's tweeters would alleviate the sibilance problem in his speakers... speculating of course.
_-_-bear
_-_-bear
There's been a little bit of OT discussion (my fault
) about the causes of sibilance in another thread in the last couple of days, starting at my post #1011:http://www.diyaudio.com/forums/lounge/200865-sound-quality-vs-measurements-21.html#post2863455
There wasn't really any consensus on the cause but my feeling is still that its primarily due to resonances in the sibilance (6-8Khz approx) region, as revealed in a CSD plot, and the way that a high Q resonance will ring over a long period of time. (Several ms in some cases) Such resonances may or may not be obvious or even visible for that matter in the frequency response plot, especially if two drivers are crossing over nearby and the resonance is in the stop band of the drivers filter.
In your case your original crossover would have only had the mid/woofer 12dB down by 6Khz, so if there are any significant high Q cone breakup resonances in that 6-8Khz region in your mid/woofer, they would have still been high enough in amplitude to hear "beneath" the output of the tweeter, and potentially add sibilance.
I call this a buried resonance, I'm not sure what the correct term might be. A resonance that's some way down the roll-off slope of the crossover (on the order of 12dB or more) will have very little effect on the summed frequency response, (any deviations from flat will be very minor) yet you will be able to measure the time domain effects of this resonance with a CSD plot. (and hear it)
What this looks like on a CSD is a quick initial decay in response (due to the tweeter which is the dominant output at that frequency, and presumably fairly clean, if its a good tweeter) followed by a lower level but long lingering resonance that originates from the mid/woofer.
By going from a 2nd order to 4th order crossover you're reducing any potential mid/woofer cone breakup resonances in the sibilance region from 12dB down to at least 24dB down, (a bit more in fact since you lowered the crossover frequency as well) which would make a big difference.
As a general rule of thumb I like to keep any out-of-band high Q breakup resonances of drivers at least 20-24dB below the nominal summed output level - I find 12dB down can be quite audible. If I can't achieve that with crossover slope and frequency then I'd apply a notch filter to the resonance as well, even if its already some way down the slope.
Failure to take care of "buried" resonances like this with metal cone mid/woofer designs I think is one of the major reasons I don't like the sound of many such designs - even if its 12-15dB down, if its not notched out the breakup resonance can still be audible especially at higher SPL's.
So the question is, if you measure the CSD of the unfiltered response of your mid/woofer, do you see any significant resonances around 6-8Khz ? If so, do you still see them in the CSD in the summed response ?
I guess you no longer have your original crossover to do a comparative CSD measurement of the summed response with 2nd and 4th order networks ?
Keeping those out of band buried resonances as low as possible in amplitude is one good reason for preferring higher slopes like 4th order, certainly for mid/tweeter crossovers where cone breakup is a big issue.
A side benefit of the steeper slope is a big reduction in excursion and low frequency drive to the tweeter which in itself will give significant quality improvements. (Although I'm not sure if I would classify these as sibilance improvements)
Last edited:
4th order electrical is like having too much NFB,
I should have clarified, wintermute was talking about going from 2nd order acoustic to 4th order acoustic, which is what I was talking about as well.
I can't agree about 4th order filters (acoustic or electrical) killing the music though. Harder to get right than lower order, yes, but if done right I think the result is definitely superior, with the out of band resonance issues being one reason why.
Another major one is reduced overlap region which leads to much less destructive interference off axis, which I find tends to improve imaging and off axis performance.
Each to their own though
@DB,
True, somewhat and mainly for monopoles, The reduction of overlap also leads to a lack of coherency between driver, the speaker never sounds as one, i always hear the acoustic centers of the different drivers.
Non domestic use, where listening distances are much greater and power handling is at a premium , yes, i would absolutely prefer 24 db slopes
(electrical) not so for domestic or near field listening and definitely not on dipoles...
True, somewhat and mainly for monopoles, The reduction of overlap also leads to a lack of coherency between driver, the speaker never sounds as one, i always hear the acoustic centers of the different drivers.
Non domestic use, where listening distances are much greater and power handling is at a premium , yes, i would absolutely prefer 24 db slopes
(electrical) not so for domestic or near field listening and definitely not on dipoles...
4th order electrical is like having too much NFB,
all the best speakers i've ever heard used 4th acoustical order slopes though i'd love to hear some high slope FIR stuff.
I've heard some good 2nd order slopes for lower spl speakers but no audible advantage to my ears as far as driver integration.
All the 1st and 3rd order speakers i've heard have tended towards messy and unfocused. I'd love to hear the geddes stuff but i suspect 4th order would improve those too.
I doubt i'd dislike 6th order acoustic assuming that's the result of 4th order electrical. I'd imagine it's a law of diminishing returns though.
The key to high order slopes, is starting with transparent drivers. since IMO rigid mid drivers are more transparent and so while they likewise need more crossover attenuation they also telegraph it less than poly etc.
Last edited:
Interesting comment, as that's more or less opposite to what I find.
In my experience lack of coherency of two drivers, eg being able to identify two separate sound sources is usually due to poor or non existant phase tracking in the crossover region.
If you have a phase tracking design where the relative driver phase tracks close to zero degrees over the full overlap region, and the amplitude response is also accurately flat the two should sound very coherent and not distinguishable as a separate source. It doesn't even require the drivers to be very close together either, although it helps.
For example I have a test baffle with an 8" dual cone full range driver and a ribbon tweeter - crossed over at 4Khz at 18dB/oct, however because the tweeter acoustic centre is around 20mm forward of the full range driver they are very close to in-phase tracking over most of the overlap region.
They're also quite far apart, about 250mm from centre to centre, which is about 2.9 wavelengths at the crossover frequency. (Due to their physical size 220mm centre to centre is about the closest they could ever be, which is still 2.5 wavelengths)
On paper 2.9 wavelengths apart crossed over at 4Khz sounds like a complete nightmare from the perspective of driver integration, and yet at normal listening distances (2 metres plus) coherency between the drivers is outstanding, with no hint of sounds coming from separate drivers. The image source height is exactly at the tweeter. Imaging and coherency is also very good out to about +/- 15 degrees on the vertical axis. Why when they're so far apart ? I don't know.
Sure, if you get within a metre or so you can hear the two drivers as separate sound sources, but that's understandable given how far apart they are.
Consider also how our perception of source location works. Binaural hearing and a wide range of frequencies (and phase at midrange frequencies) are involved in locating a sound source on the horizontal left-right plane, so the sounds from both mid/woofer and tweeter contribute to horizontal location. This is why its important for the drivers to be vertically stacked so that both drivers give the same cue for horizontal location.
However vertical height is perceived almost entirely through high treble - in particular around 8Khz which is the major cue for vertical elevation. In a design with the tweeter crossed at 4Khz, and fairly steeply, this 8Khz cue will be coming entirely from the tweeter, thus the apparent height localizes to the tweeter, (assuming the tweeter is flat through the 8Khz region) and the overall source location of the complete sound appears to be the tweeter.
I've done experiments where I've kept the tweeter height the same and moved the mid/woofer even lower and further apart and its possible to increase the distance to about 300mm centre to centre before localisation and coherency starts to fall apart at a normal listening distance, which is an amazingly large distance for such a high crossover frequency. (I wouldn't recommend it of course, but it was an interesting experiment)
If anything I think steeper filters help driver integration, provided phase tracking is accurate, particularly when the driver spacing is greater than ideal. Anything beyond about 24dB/oct is into the diminishing returns area though I think, and you then start to get an abrupt power response transition if the directivity of the drivers don't match at the crossover frequency, whereas the lower slopes will give a more gradual transition with less of a power response hole. (A good reason not to use digital brick wall filters)
Last edited:
such is the case when DI does not match... it's simple really.
In IE a 2nd order crossover, say a 5" mid is becoming directive near 3khz. the 2nd order crossover ensures that despite the directivity, there is still meaningful output from a wide-dispersion tweeter, contributing to the sound power IE from 2khz - 3khz and the vice versa applies from 3khz to 4khz with the mid's contribution to total output reducing overall sound power.
Replace that with high order slopes - in essense proverbial 'brick wall filter' for example, and only the individual drivers' directivity contributes to sound power. If DI is mismatched, you get an abrupt transition...2-3khz is ALL the mid and 3-4khz is ALL the tweeter... Not a big issue if DI is matched though... it just rarely is.
Hi Simon, I've never had any luck trying to get CSD plots. I do have impulse measurements of the original crossover and the current so I could possibly import them into something to do CSD plots.
The woofer has some breakup around 6.5K and some more at higher frequencies but it is a long way down db wise.
I think that Lynn's premise was that if the slope is too shallow (at too low a frequency) that too much bass energy gets to the tweeter which then causes excessive distortion, I guess that this could be causing modulation of the dome which causes some odd IM type distortion. This was what I suspect was happening.
Attached graph shows dark blue, raw response of midwoofers. green response with 12db/octave rolloff (basically just notch filters) and black, the response with the 24db / octave rolloff. Similar notch filters but with added 2nd order electrical filter.
Tony.
The woofer has some breakup around 6.5K and some more at higher frequencies but it is a long way down db wise.
I think that Lynn's premise was that if the slope is too shallow (at too low a frequency) that too much bass energy gets to the tweeter which then causes excessive distortion, I guess that this could be causing modulation of the dome which causes some odd IM type distortion. This was what I suspect was happening.
Attached graph shows dark blue, raw response of midwoofers. green response with 12db/octave rolloff (basically just notch filters) and black, the response with the 24db / octave rolloff. Similar notch filters but with added 2nd order electrical filter.
Tony.
Attachments
@DB,
When the xover and the acoustic centers are spot on you cannot localize the drivers even when as close as 280mm You can with steep slopes, you may not be sensitive to this as others are and when last comparing studio monitors for nearfield listening, it was immediately apparent when using monitors with high order slopes, you can hear it immediately, (1M listening distance)
Myself favor odd order slopes, baffle offset and careful attn (treatment) to early reflections, best for coherency, imaging and dynamics IMO and from experience, no 24 db x-over speaker i have ever heard in 35 yrs has even come close.
For point source speakers, as close as possible for mid/twt works for me ....
When the xover and the acoustic centers are spot on you cannot localize the drivers even when as close as 280mm You can with steep slopes, you may not be sensitive to this as others are and when last comparing studio monitors for nearfield listening, it was immediately apparent when using monitors with high order slopes, you can hear it immediately, (1M listening distance)
Myself favor odd order slopes, baffle offset and careful attn (treatment) to early reflections, best for coherency, imaging and dynamics IMO and from experience, no 24 db x-over speaker i have ever heard in 35 yrs has even come close.
For point source speakers, as close as possible for mid/twt works for me ....
The solution in my mind comes from demos at Magico:
1. Use very flat tweeter.
2. Tilt it upwards starting at 2 kHz or so, so you are +2 dB by 20 kHz
3. Use Mundorf supreme caps.
Honestly I do not like this sound! It is not natural sounding, and I do not like those caps, but they add a Walt Disney like quality to the very top which many like.
Best,
E
1. Use very flat tweeter.
2. Tilt it upwards starting at 2 kHz or so, so you are +2 dB by 20 kHz
3. Use Mundorf supreme caps.
Honestly I do not like this sound! It is not natural sounding, and I do not like those caps, but they add a Walt Disney like quality to the very top which many like.
Best,
E
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.