They're the exact same thing, just at a different section of the circuit. So his comparison is totally warranted.
FWIW, I've only found one approach to crossovers that has been consistently satisfactory: control the directivity of the treble unit (waveguide, coincident/Dual Concentric driver, etc.) and effect the handover where the directivity of the lower driver narrows to the pattern of the upper driver in the horizontal plane. The only interesting question left, I think, is how narrow or wide that pattern should be. But there's a whole long thread on that here already. 🙂
Other approaches can work ok, especially in heavily-treated rooms were one hears relatively more direct sound. But none of them seem to do as well as the above-outlined approach.
Given the results of the Meyer and Moran paper, I wouldn't be concerned about a DA/DA loop.
IMO, the most pressing shortcoming of DSP crossovers today is that too much boost will lead to digital clipping. Not an issue for most uses, but for extreme Linkwitz Transforms for small subs, or dipole bass compensation, it can be an issue.
There's a teeny tiny bit of truth the argument, but it's mostly irrelevant. In addition to David Smith's points, on more: take a speaker with wide impedance variations in the midrange, and put it on two amps. One has high source (aka output) impedance, one has low source impedance. They will sound different, with the response of the former adversely affected by the interaction.
Dave,
Only one that was designed not to be an easy load. A raw driver can be far worse than a well designed system with proper impedance compensation. High capacitance is the tough part. Rising inductance as you say, less so.
Hi OnAudio,
Excuse me for being a little late joining the party, but here is my two cents worth -
One issue that I am aware of that effects the quality of audio reproduction is that due to group delay, which has been documented and confirmed as being an undesirable byproduct of Xover filters. Group delay varies in a Xover as function of its crossover frequency, its Q and order. And from recent work by people like Geddes, it seems that our sensitivity to group delay is a function of frequency and also SPL.
So from this, my choice of filter order would include group delay as well as all the other many parameters that affect the overall quality of reproduction.
To show the variation in group delay between the various Xover filter orders, I’ve attached two sets of curves, one for even order and the other for odd order Xovers.
All of the responses are from LR type filters. Of interest in the odd order Xovers is that minimum group delay is significantly affected by the polarity of the tweeter.
The polarity of the tweeter in the even order Xovers is set to give a flat frequency response. One thing that stands out for me is that the 3rd order group delay with the tweeter polarity reversed is the same as that for the 2nd order Xover!
So Why not use 3rd order instead of 2nd order?
Another issue that comes to mind is that the LP and HP responses of a 2nd order LR Xover is –6.02dB down at the crossover, which gives perfectly flat on-axis acoustic summation at the crossover frequency, but only half the power. On the other hand, the summation using a 3rd order Xover gives a flat acoustic response on-axis, with only -3.01dB loss at the Xover frequency, thus providing no power loss. In my opinion, this is an issue that needs further thought.
Regards
Peter
Excuse me for being a little late joining the party, but here is my two cents worth -
One issue that I am aware of that effects the quality of audio reproduction is that due to group delay, which has been documented and confirmed as being an undesirable byproduct of Xover filters. Group delay varies in a Xover as function of its crossover frequency, its Q and order. And from recent work by people like Geddes, it seems that our sensitivity to group delay is a function of frequency and also SPL.
So from this, my choice of filter order would include group delay as well as all the other many parameters that affect the overall quality of reproduction.
To show the variation in group delay between the various Xover filter orders, I’ve attached two sets of curves, one for even order and the other for odd order Xovers.
All of the responses are from LR type filters. Of interest in the odd order Xovers is that minimum group delay is significantly affected by the polarity of the tweeter.
The polarity of the tweeter in the even order Xovers is set to give a flat frequency response. One thing that stands out for me is that the 3rd order group delay with the tweeter polarity reversed is the same as that for the 2nd order Xover!
So Why not use 3rd order instead of 2nd order?
Another issue that comes to mind is that the LP and HP responses of a 2nd order LR Xover is –6.02dB down at the crossover, which gives perfectly flat on-axis acoustic summation at the crossover frequency, but only half the power. On the other hand, the summation using a 3rd order Xover gives a flat acoustic response on-axis, with only -3.01dB loss at the Xover frequency, thus providing no power loss. In my opinion, this is an issue that needs further thought.
Regards
Peter
Thanks PLB, lets assume one picks such a crossover point such that cabinet resonance fixes the power issue. Or picking such a point to cater for 200Hz room boom 😉
lots of phase rotation there PLB, makes for tiny sound and poor coherency between drive units ...
Hi OnAudio,
"Thanks PLB, lets assume one picks such a crossover point such that cabinet resonance fixes the power issue"
Do you have any suggestions for simulation software for that? Or that can accurately model the polar response?🙄
"Or picking such a point to cater for 200Hz room boom "
A tweeter that can handle 200Hz? 🙄
Regards
Peter
"Thanks PLB, lets assume one picks such a crossover point such that cabinet resonance fixes the power issue"
Do you have any suggestions for simulation software for that? Or that can accurately model the polar response?🙄
"Or picking such a point to cater for 200Hz room boom "
A tweeter that can handle 200Hz? 🙄
Regards
Peter
I cannot accept that argument without documentary evidence from a respected authority.
Regards
Peter
Yes, at 344M/s, that's for the 4th order LR. For the 2nd and 3rd order LR filters it's 318uS = 4.3"
Regards
Peter
Correct me if I'm wrong, but how can you have an odd order LR type filter ? Since a LR filter is synthesised from two cascaded identical butterworth sections, surely that can only lead to even orders ?
Yep, because the relative phase shift between drivers through the crossover region is 90 degrees with either tweeter phasing, but the total phase rotation when the drivers are out of phase is 180 degrees, but with the in phase case the total rotation must be 360 degrees. More rapid phase rotation means higher group delay...
Depends whether you're using it in a true quadrature phase crossover (relative acoustic phases are actually 90 degrees) or whether you just have a 3rd order network but due to driver offsets etc end up having an in-phase tracking network.
Just because the slopes are 3rd order doesn't necessarily mean you end up with a 90 degree phase shift unless you take steps like acoustic centre alignment, nor that it's necessarily desirable...
One downside of an actual quadrature configuration is that the vertical lobes are asymmetric, with up to a 3dB peak in the frequency response at crossover above or below the design axis, depending on the tweeter phasing.
If the tweeter is above the woofer and in phase with it, the peak in the lobe will be above the design axis, if its out of phase it will be below. (If the tweeter is below the woofer that reverses obviously)
If your crossover region is 2-4 Khz where the ear is very sensitive and easily irritated, a 3dB peak in the response just there when you go somewhat off the vertical axis is a bad thing. Better to have a flat response on axis and a gentle droop in that area as you go off the vertical axis.
The second problem is that 90 degree phase tracking is difficult to achieve in practice, putting you in something of a "knife edge balance" situation even on axis. Any more than an extra 30 degrees phase shift and you'll start seeing cancellation between the two drivers (greater than 120 degrees) so the potential for notches in the response at frequencies where the phase tracking is less than ideal is quite great.
If you assume that a certain tolerance of phase tracking error is inevitable, lets say +/-20 degrees, and then calculate the frequency response deviation that will occur from flat with an initial phase shift of 90 degrees, with that from an initial phase shift of 0 degrees, there is MUCH less error in frequency response for the 0 degrees situation for the same amount of phase error.
Furthermore, as you go off axis you can go further symmetrically off axis before cancellation from phase tracking errors starts occurring if you started at 0 degrees than if you started at 90 degrees.
The relatively low group delay of the inverted polarity 3rd order filter you plotted also only applies if the acoustic phasing of the drivers is really 90 degrees. If it's actually in-phase due to driver offset or other phase shift then the group delay increases anyway. Any shift in the delay of either driver radically alters the shape and height of the group delay peak.
But bear in mind this only occurs if the driver spacing is half a wavelength at the crossover frequency or more. As you go below a half wavelength spacing the power response "hole" is progressively reduced. So it's an issue in a midrange to tweeter crossover certainly, but not really an issue in a woofer to midrange crossover, unless the drivers are physically a long way apart.
But as mentioned above, the drawback of this flat power response is a peak in the response at some vertical off axis angle. The extra power response has to come from somewhere, it doesn't come out of thin air.... 😀
Then you have to prove that a dip in the power response in the reverberant sound field is a greater evil (or that its even audible at all) than a large peak in the near on axis response just above or below the design axis which could easily be aimed at the listener depending on their seating height. Good luck proving that 😉
I'm not against 3rd order networks by any means - my current speakers use a 3rd order crossover, but it's not quadrature tracking (not anymore anyway) but instead tracks in-phase, and there is no doubt to me that tracking in phase is superior to tracking at 90 degree phase when you take into consideration the three dimensional sound field that a speaker produces, rather than just thinking about on axis response and "total sound power".
I used to like the concept of a 90 degree tracking odd order filter for the same reasons you espouse, but I've since learnt the error of my ways.. 😛 IMHO, in-phase driver tracking through the crossover region, no off axis response peaking, and symmetrical crossover lobes are all more important than flat power response or slightly lower group delay at the crossover frequency.
Another side benefit of a LR type filter over an odd order filter in a tweeter application is less strain on the tweeter - you're -6dB down at the crossover frequency instead of -3dB, which is a big deal for a tweeter. As far as on axis response goes, the extra 3dB of power applied to the tweeter in the odd order case is "wasted" due to the 90 degree phase shift not fully summing on the listening axis.
Last edited:
LR filter is just a filter with a Q of 0,7 which are 6dB down at xover. I don't see a reason why this would not work for odd orders.
vac
Yes, they are. They're both just signal processing.
Do you have any grounds for your extraordinary claim that there's magical about passive circuits vs. line level ones (passive or active), besides just audiophool dogma?
Its true that theoretical LR filters (acoustical or electrical) will be cascades of identical Butterworth and therefore of even order, but in practice you can create "in phase" networks with even or odd combinations. Since interunit driver delay is arbitrary you can choose whatever filter order that gets phases to overlap, adjust to have -6dB at the crossover point, and achieve response as flat as you could want. It can be near perfect in practice even if not perfect in theory.
With regard to "Q of 0.7", lets not forget that 0.7 is the Butterworth Q for second order only. All other orders will have a range of Q's ranging from higher to lower. This is a commonly repeated error because we so frequently discuss second order filters.
David
With regard to "Q of 0.7", lets not forget that 0.7 is the Butterworth Q for second order only. All other orders will have a range of Q's ranging from higher to lower. This is a commonly repeated error because we so frequently discuss second order filters.
David
I keep referring to a simple system thread where I worked from 1st to 2nd to 3rd order networks until I achieved phase overlap in an old AR system
Crossover mods for the AR4x - The Classic Speaker Pages Discussion Forums
Its all trial and error but you can pick a crossover point, start with a crossover order that might work and shape a trial crossover until the it looks appropriate with a soft (-6 dB) corner and then compare the phases of the upper and lower sections. Are they parallel or do they cross? If parallel, then how far apart? If 0 degrees apart then you are done. If 180 degrees apart then flip the polarity of one unit and you are done. If 90 degrees apart then you have a problem.
If they cross at a significant angle then try and figure which side needs more or less crossover slope. Remember these rules: adding or subtracting orders impacts the stop band and band edge but not the pass band. In a 2 way an extra order on the tweeter bends the phase below the crossover up. An extra order on the woofer bends the phase above the crossover point down. Each extra order ultimately bends the phase by 90 degrees, but only by 45 degrees at the crossover point.
In the end, its just a game to find the right combination that get the phases close enough (maybe within 30 degrees) gets the corners in the right spot and with the right shape so that the summed combination is adequately flat. I can usually make it work with reasonable crossover orders, but rarely I've had to rethink the crossover frequency to get a reasonable compromise.
Regards,
David
Having just got a second order crossover up and running on my bookshelves, I'm rather pleased with the imaging and stereo stage, which is outstanding.
The resistor on the bass capacitor really works well to adjust phase and smooth a crossover peak:
Always a work in progress, but I reckon that phase matching at crossover is very satisfactory. 😎
An externally hosted image should be here but it was not working when we last tested it.
The resistor on the bass capacitor really works well to adjust phase and smooth a crossover peak:
An externally hosted image should be here but it was not working when we last tested it.
Always a work in progress, but I reckon that phase matching at crossover is very satisfactory. 😎
- Status
- Not open for further replies.