Phase-alignment based method of designing multi-way speakers

This isn't quite so. My observations are that when you flatten driver's resonance peak electrically (or mechanically) excursion is minimized so is THD in this region as THD comes from resonant character and not the opposite way.
THD always goes up with increasing excursion - since lower frequencies require more excursion (4 times as much every time you halve the frequency) distortion will go up as frequency goes down, however in addition to this there is another effect as you transition from the mass controlled region above resonance to the compliance controlled region below resonance.

Significantly above the mechanical resonance the linearity of the driver (assuming piston motion of the cone/dome) is only limited by the motor - any non-linearity in the suspension (of which there is a lot in a tweeter) don't matter and don't cause distortion because its the moving mass that controls the response to the applied force.

Below resonance the suspension compliance is what controls the response to the applied force, so any suspension non-linearities will be directly reflected as distortion in the output - hence distortion rises much more rapidly below resonance.

Electrically flattening the drivers response at resonance only reduces distortion by the amount that excursion is reduced, the fundamental rapid increase in distortion below resonance will still be there, so its still a bad idea to operate a small driver like a tweeter at and below its resonance IMHO.

Same can be said about impedance. Electrical correction of impedance peak is matching LRC values to electrical equivalent circuit of driver's mechanical properties to counter-measure its resonance. When it is done f response roll-off gets less steep as Q changes its value, impedance is flattened, THD is lowered and phase curve gets straightened. Actually that's the first and most important point of the method.
A distinction needs to be drawn between using an RLC compensator to flatten the impedance curve, and using an RLC compensator to flatten the acoustic response of the driver - the two rarely occur at the same time.

If you use an RLC compensator to flatten the impedance curve then all you're doing is providing a flatter "easier" load for a passive crossover to drive to prevent additional peaking in the response due to the interaction between the crossover and the drivers impedance, especially if the mechanical resonance occurs part way into the filters stop band where its series impedance is high.

Depending on the driver this may still leave a large peak in the acoustic response, eg if the driver's Qt is greater than ~0.7, which might be the case with a midrange driver in a small enclosure or a tweeter, there will still be a peak in the acoustic response even with a flattened impedance curve.

In this case you would want to adjust the RLC compensator to actually flatten the acoustic response, (or more specifically, match the target acoustic response) rather than aim for a flat impedance curve. If it's a parallel connected RLC network that would mean a dip in the impedance at resonance as seen by the crossover instead of the original peak.

You might do this for example, in the case of a large midrange driver in a small enclosure, which might have a cut-off frequency which is plenty low enough for the desired crossover frequency, but may have a Qt of well over 1 in that size enclosure.

What matters is the acoustic response not the impedance curve, so an RLC compensator - if needed, should be used to correct any resonant peak in the acoustic response, not the impedance curve...the impedance curve is just one of the many ingredients that helps you get the final acoustic response. Sometimes it makes sense to flatten it, other times its unnecessary or non-flat is more convenient to get the desired acoustic response.
 
Last edited:
diyAudio Moderator
Joined 2008
Paid Member
Agreed.

Where a tweeter is connected directly to a voltage source (normal amp), the short circuit represented by the source impedance will damp the circuit completely resulting in a constant voltage/frequency characteristic from the source. Under these circumstances we will see the kind of response from the tweeter as suggested by the manufacturer.

It will be the added crossover components that will serve to reduce damping. An added RLC circuit merely returns this damping, and it must be selective as the source is now non-zero.

Electrical correction of impedance peak is matching LRC values to electrical equivalent circuit of driver's mechanical properties to counter-measure its resonance. When it is done f response roll-off gets less steep as Q changes its value,
 
...What matters is the acoustic response not the impedance curve, so an RLC compensator - if needed, should be used to correct any resonant peak in the acoustic response, not the impedance curve...the impedance curve is just one of the many ingredients that helps you get the final acoustic response. Sometimes it makes sense to flatten it, other times its unnecessary or non-flat is more convenient to get the desired acoustic response.
Ah - this is what I was getting at - that certainly makes sense. :)
 
Significantly above the mechanical resonance the linearity of the driver (assuming piston motion of the cone/dome) is only limited by the motor - any non-linearity in the suspension (of which there is a lot in a tweeter) don't matter and don't cause distortion because its the moving mass that controls the response to the applied force.

Thanks for reminder about THD associated risks here, still they are also very dependent of power level applied and achieved linearity of the particular driver that is proportional to its construction excellence. But isn't it the HF driver's diameter that defines F at which it transitions to pistonic motion and not the Fr?

What matters is the acoustic response not the impedance curve, so an RLC compensator - if needed, should be used to correct any resonant peak in the acoustic response, not the impedance curve...the impedance curve is just one of the many ingredients that helps you get the final acoustic response.

I'ts like saying let us don't care about driver's resonance and associated impedance "bump", we can correct it when filters are already applied. It is satisfactory approach for one speaker setup. We can also transfer it to multi-way speakers and achieve quite satisfactory results, but is it the best possible for a given pair of speakers? What happens when phase is electrically shifted and then summed and passive filters are in-between? I'd expect more uneven acoustic response compared to phase-alignment method. The questions of implementing phase-alignment method are:

a) to what degree there will be improvement in f response (if any);
b) what we will gain additionally (for example better directivity, more linear Z - easy load for SET amps);
c) what we will lose (THD?) and to what degree.

In this case you would want to adjust the RLC compensator to actually flatten the acoustic response, (or more specifically, match the target acoustic response) rather than aim for a flat impedance curve. If it's a parallel connected RLC network that would mean a dip in the impedance at resonance as seen by the crossover instead of the original peak.

We can apply RLC in parallel to driver to damp resonance first (and achieve flat Z), then apply filters and then another RLC in parallel to whole speaker to correct for sharp peaks of the total output in the same region if they exist. In latter case it will not impact the relative phase shift between drivers (whether it was or wasn't corrected two steps before). As you correctly pointed out the negative effect would be impedance dip as it will "overcorrect" the impedance.

To avoid usage of second RLC filter we may revise our driver choice instead. Sharp response peaks may be the sign of other problems caused by particular design. Use parallel RLC filter in series with the whole speaker to compensate for rather wide "bumps". The latter means losses, so I'd go for re-selecting speakers.

By flattening Z of tweeter in the first place we will also flatten peaks in its response near Fr. Any sharp peaks then would rather be expected from LF driver so it should be carefully chosen to have smooth roll-off towards HF.
 
Agreed.

Where a tweeter is connected directly to a voltage source (normal amp), the short circuit represented by the source impedance will damp the circuit completely resulting in a constant voltage/frequency characteristic from the source. Under these circumstances we will see the kind of response from the tweeter as suggested by the manufacturer.

It will be the added crossover components that will serve to reduce damping. An added RLC circuit merely returns this damping, and it must be selective as the source is now non-zero.

Plus output from crossover will be dependent on Z curve of the speaker as Z value at given determines also response from filter.

AllenB, thanks for joining in. I felt like I'm starting to lose the ground here.
 
Last edited:
Thanks for reminder about THD associated risks here, still they are also very dependent of power level applied and achieved linearity of the particular driver that is proportional to its construction excellence.
Of course, but the way you worded your post suggested that you believe that correcting the impedance peak of the driver is what was reducing distortion, however its only doing so by reducing the output at that frequency to get rid of the peak in the frequency response - you would see the exact same distortion reduction if you applied a line level parametric EQ to do the correction. (Although only in the case of an active crossover where you can isolate the individual driver)

Since we presumably want a speaker with a flat frequency response without peaks then we will be correcting any peaks one way or another, and all methods will cause the same distortion reduction near the resonance frequency of the driver because all will reduce the amplitude by the same amount.

But isn't it the HF driver's diameter that defines F at which it transitions to pistonic motion and not the Fr?
You've apparently misunderstood me, what I mean is above the resonance frequency of the driver but below the onset of cone breakup (eg within the piston operation range of the driver) the distortion is almost entirely caused by motor non-linearity with no contribution from the suspension.

Below resonance the suspension contributes to distortion and above the cone breakup region the linearity of the cone material itself (since it is bending) contributes to distortion.

So above the fundamental mechanical resonance and below cone breakup onset is generally the most linear region for a driver.
I'ts like saying let us don't care about driver's resonance and associated impedance "bump", we can correct it when filters are already applied.
But we do care about the drivers fundamental resonance - but only what shows up in the acoustic response. What we don't care about is whether the impedance curve is flat or not. We hear the acoustic response, we don't hear the impedance curve.

The impedance curve is an internal intricacy within the "black box" that is the completed speaker design. Sometimes the crossover design might be more straight forward and convenient if we add components that flatten the impedance seen by the rest of the crossover.

However sometimes trying to flatten the impedance curve can lead to a more complex design with more components to achieve almost exactly the same result that could be achieved with fewer components by simply altering some of the other component values and "working with" the impedance curve instead of trying to flatten it.

It all depends on the situation. The suggestion that flattening the impedance curve is always the right thing to do is simply wrong. What matters at the end of the day is the acoustic response that comes out of the speaker - how you achieve it doesn't matter. Anything that you do in the network to achieve the target acoustic amplitude and phase response from each driver is fair game, provided the amplifier can drive the load.

It is satisfactory approach for one speaker setup. We can also transfer it to multi-way speakers and achieve quite satisfactory results, but is it the best possible for a given pair of speakers? What happens when phase is electrically shifted and then summed and passive filters are in-between?
Sorry but I can't interpret your question here, phase is electronically shifted by what ? If you have passive crossovers nothing you do at the input can alter the relative phasing of the individual drivers, and the summing occurs in the air, so you question doesn't make sense.
We can apply RLC in parallel to driver to damp resonance first (and achieve flat Z),
AllenB already answered this point - the only resonance that the RLC compensator is "damping" when aiming for a flat impedance curve is the increased Qt caused by the series impedance of the crossover components around the resonance frequency. If you adjust the RLC to achieve a flat impedance curve you are simply returning to the "natural" damping of the driver when driven directly by a voltage source.

That may or may not result in a flat acoustic frequency response. If the drivers natural Qt (or Qtc in a closed box like a midrange enclosure) is well above 0.7 then you can still easily have a peak in the acoustic response despite a flat impedance curve. Since we hear the acoustic response not the impedance curve, getting the acoustic response flat is what matters.

What I was trying to describe is that in some instances you can use an RLC compensator to "overcompensate" for the impedance curve to correct the acoustic response - in effect it becomes like the notch part of a Linkwitz transform. The peak in the acoustic response will have a centre frequency, Q, and amplitude which with the correct values of RLC can be corrected. What happens to the impedance curve to do this doesn't really matter.


then apply filters and then another RLC in parallel to whole speaker to correct for sharp peaks of the total output in the same region if they exist. In latter case it will not impact the relative phase shift between drivers (whether it was or wasn't corrected two steps before).
A couple of points here - the first is that I wasn't referring to any other peaks elsewhere in the drivers frequency range - I was purely speaking about its fundamental resonance, which with enclosed midrange drivers can sometimes be well above Qt 0.7.

In the case of peaks elsewhere in the response which are due to a peak in the response of an individual driver your suggestion to fix it by applying correction before the crossover to somehow maintain relative phase tracking between the two drivers is very misguided.

Peaks in the response of one driver that occur in the overlap region between two drivers (such as the cone breakup resonance of a woofer just above its low pass frequency) cause a loss of relative phase tracking between the two drivers, and this cannot be fixed by applying EQ before the crossover.

You might be able to eliminate the amplitude peak as measured on one particular axis, but it will not correct either the on axis phase response, (the summed response of two drivers is not minimum phase in nearly all crossover types) nor the relative phase tracking between the two drivers, and the amplitude correction will be wrong as you go off axis as well, particularly the vertical axis.

The correct way to fix the problem is to correct the peak in the response of the individual driver in its section of the crossover, for example an RLC notch. This will improve the phase tracking between the two drivers, flatten the summed on axis response, and also improve the response off axis.

To avoid usage of second RLC filter we may revise our driver choice instead. Sharp response peaks may be the sign of other problems caused by particular design. Use parallel RLC filter in series with the whole speaker to compensate for rather wide "bumps". The latter means losses, so I'd go for re-selecting speakers.
You can revise your driver choice to one made of unobtainium if you like. :p All drivers have resonances, whether you choose to live with those resonances or try to correct them is up to you. But saying choose a driver without resonances is just silly. Nearly all practical drivers have response variations that need correcting if you're to get the very best out of them. That's a fact of life and an important part of the design process, if you neglect it you'll get a mediocre result.
By flattening Z of tweeter in the first place we will also flatten peaks in its response near Fr. Any sharp peaks then would rather be expected from LF driver so it should be carefully chosen to have smooth roll-off towards HF.
Or, if we are using drivers made in the real world, we will be correcting any peaks in the near rolloff region of the woofer in the crossover...
 
Last edited:
The only phase related quantity that I could see being of any importance is the group delay, which must be kept within reasonable limits (e.g. under about 3 ms) between 250Hz and 2500 Hz. A small number of studies (by telephone researchers) found that people could detect more group delay above these levels and within these frequency limits.
Where did you get 3ms ? While 3ms might be ok for narrow band telephone grade audio quality, the limit is significantly lower than that for high fidelity purposes.

The typically accepted values, (originally from Blauert & Laws I believe) are:

500 Hz -3.2 ms
1 kHz - 2 ms
2 kHz - 1 ms
4 kHz - 1.5 ms
8 kHz - 2 ms

So it can be as low as 1ms at 2Khz and be detected. 1ms is still an acoustic centre offset of about a foot though, which is a fairly large acoustic centre error, and any decent crossover 4th order or lower is not going to contribute anywhere near 1ms at the crossover frequency either.

Having said that, there are those (Earl Geddes) who say that the threshold of audibility of group delay decreases with increasing SPL - so those thresholds may be fine at modest levels but cause noticeable loss in perceived quality at higher SPL. (I'm still sitting on the fence over this one, but I think it bears consideration...)

Its also been my own anecdotal experience that close acoustic centre time alignment of drivers in a mid/tweeter crossover does seem to improve imaging, even if phase tracking and flat frequency response are taken care of in both aligned and non aligned cases.

I'm yet to figure out why this is.... is it elimination of the step in group delay between the two frequency bands ? (Are we more sensitive to a broadband step in group delay between the two bands than a narrow peak at the crossover frequency ?) Is it the inevitable differences in off axis and polar responses ? I don't know. It could even be my imagination. :D
 
Last edited:
diyAudio Moderator
Joined 2008
Paid Member
Plus output from crossover will be dependent on Z curve of the speaker as Z value at given determines also response from filter.
That's what I said ;) This issue is easy to explain in a difficult way.

To be clear on what it is we are damping...

A tweeter's resonance is reflected in the impedance peak, where higher impedance=less power transfer which is offset by the resonance. Furthermore, this isn't changing.

The key point is - you don't change the tweeter's impedance with crossover components. You might apply an RLC circuit across the tweeter which results in a flattened impedance. The tweeter still has the peak, and so does the RLC but together they are flat.

If we then try to use a simple filter, an impedance peak will simply interfere with it, nothing more, nothing less. There will be a peak in the voltage applied to the tweeter but the tweeter won't change. The RLC would damp that peak.

What happens when phase is electrically shifted and then summed and passive filters are in-between? I'd expect more uneven acoustic response compared to phase-alignment method.

I am not sure I can follow all of this. Can I suggest that whenever you use the term 'phase', that you include the location and the unit of interest. Eg. 'the tweeter's acoustic phase', 'the tweeter's voltage phase', or 'the impedance phase of the tweeter with its RLC filter'.
 
I'd expect more uneven acoustic response compared to phase-alignment method.

Hi,

That the problem. Its a sweeping generalisation. Your assuming zobelling
drivers and compensating tweeter impedance peaks gives you a better
starting point. They can do but a lot of the time they simply don't.

One problem with the simulating x/o sticky is flattening impedance is optional.

It all depends on the particular drivers used and the design of the speaker.

Blindly applying a "method" is not the same as working out what needs
doing and what doesn't. The method is called "phase aligned" but I
still can't work what the meaningful definition of the term is.

Take this two way design : Zaph|Audio

audio-speaker17-crossover.gif


As its 4th order L/R acoustic the drivers are wired in phase.
The phase is not minimal, it wraps through 360 degrees.
The phase wrap causes inevitable group delay.

How would you apply the "phase aligned method" to the above design
and what would it meaningfully change in the above design ? nothing AFAICT.

rgds, sreten.

Worth a look : Crossovers

Duelnd_3way_tg1.jpg


The delay is inevitable due to the phase wrap at each crossover point.
 
Last edited:
...

audio-speaker17-crossover.gif


As its 4th order L/R acoustic the drivers are wired in phase.
The phase is not minimal, it wraps through 360 degrees.
The phase wrap causes inevitable group delay.

...

Duelnd_3way_tg1.jpg


The delay is inevitable due to the phase wrap at each crossover point.

This is a point I have been struggling with in my current design.

I am using a direct radiator mid and a horn tweeter. The problem is there is about a 5 inch difference in acoustic centers, thus an approximately 0.38 ms delay.

I theorized that I could put the crossover point at the frequency whose period is 0.38 ms (2600 Hz) then at the crossover point it would be in phase, even though 360 degrees delayed, though above and below the XO freq the phase will deviate, and this is what your graph here seems to show - though I'm not sure if I'm reading it right. If I use a 4th order XO it should keep the deviation interferences to a minimum.

So that's the theory, which the more I learn of it the more flawed it sounds, though I'm not sure what alternatives I have. I don't even know the actual acoustic displacement until I can take some measurements (Holm Impulse?).

Does anybody have actual experience with this situation??
 
Eric,

With the tweeter lagging behind the mid-woofer, you're in a lucky situation. I described it in my previous post here.

You can first design the XO to meet acoustical targets of a, say, LR4 response, looking at each driver on its own. Then you can add more low passes (or all passes) on the midwoofer to have enough group delay to outweigh the 380us time delay for all of the relevant frequencies. With that much of delay and a high XO point the filter will be tough to implement, though. But it is the only way of making a true phase coherent speaker.

Remember that you must count all of the phase rotation, "tracking phase" but with a (multiple) 360deg turn is not tracking phase. Signals will be in phase for steady state sinusoids etc but everthing else will show the remaining delay.
 
Last edited:
Eric,

With the tweeter lagging behind the mid-woofer, you're in a lucky situation. I described it in my previous post here.

You can first design the XO to meet acoustical targets of a, say, LR4 response, looking at each driver on its own. Then you can add more low passes (or all passes) on the midwoofer to have enough group delay to outweigh the 380us time delay for all of the relevant frequencies. With that much of delay and a high XO point the filter will be tough to implement, though. But it is the only way of making a true phase coherent speaker.

Remember that you must count all of the phase rotation, "tracking phase" but with a (multiple) 360deg delay is not tracking phase. Signals will be in phase for steady state sinusoids etc but everthing else will show the remaining delay.

Yes, this is what I have come to understand.

As I also said, however, the deviations will only be a fraction of the actual delay, and their effects should fall rapidly on either side of the crossover point. Something else I would like to investigate would be adding half the delay, just as you describe, to put them 180 degrees out of phase, and then reversing the polarity of one of the drivers. This will cut all the deviations in half as well. This would be an easier implementation, and the results may be acceptable - but only testing will tell.
 
only testing will tell.

No, modeling could tell you the same thing without having to build anything.

You and PRTG might want to get a copy of Passive Crossover Designer (did I mention it's free?). You can model your fantasy loudspeaker crossovers and then post the results for everyone to see. It would be nice if we could move on to talking about an actual crossover design... I think the inherent limitations will be self evident.

-Charlie
 
No, modeling could tell you the same thing without having to build anything.

You and PRTG might want to get a copy of Passive Crossover Designer (did I mention it's free?). You can model your fantasy loudspeaker crossovers and then post the results for everyone to see. It would be nice if we could move on to talking about an actual crossover design... I think the inherent limitations will be self evident.

-Charlie
Thanks Charlie.

I have PCD, thank you.

The thing about PCD is that one has to input actual impedance and response curves for the 'simulation' to work. So, to get the actual impedance and response curves, it is necessary for me to collect them with the drivers in their actual cabinets, or as close as I can manage. I am actually nearing completion of my test cabinet soon where I will be able to do that. :)

After that the modeling will certainly detail the phase and response numbers, but what I meant was how it sounded and measured. One can look at all the curves and specifications one wants, but this will only get one so close to knowing how it will actually sound and measure, IMO.

This is my first speaker project, and I actually started with the impression one could just put T/S data into a spreadsheet and pop out a design, but the more I read and researched the more limited I found such a procedure to be. I have endeavored to move forward by collecting the best data I can to input to PCD so I can then have the best simulation I can, but after building that design fully expect to have to again make adjustments because of the inevitable simulation limitations and errors.

That seems to be the way it works - and there will doubtless be plenty of discussion about crossovers at that point! :D
 
Last edited:
In the case of peaks elsewhere in the response which are due to a peak in the response of an individual driver your suggestion to fix it by applying correction before the crossover to somehow maintain relative phase tracking between the two drivers is very misguided.

Peaks in the response of one driver that occur in the overlap region between two drivers (such as the cone breakup resonance of a woofer just above its low pass frequency) cause a loss of relative phase tracking between the two drivers, and this cannot be fixed by applying EQ before the crossover.

This is where you have misunderstood me :) I'm not suggesting that RLC before filter would change relative phase - quite the opposite. What I was saying is that it can work as a filter to compensate for peaks if they exist after relative phase correction by notches/Zobels are done in parallel with drivers. Most probably there won't be any left.

The correct way to fix the problem is to correct the peak in the response of the individual driver in its section of the crossover, for example an RLC notch. This will improve the phase tracking between the two drivers, flatten the summed on axis response, and also improve the response off axis.

That's exactly first half of what my proposed method is about :)

Thanks for overall clearance to a much better degree than I could do it. I believe the controversy level in this thread will be minimized further on.
 
diyAudio Moderator
Joined 2008
Paid Member
LafeEric, I have some experience with your problem. This is essentially a 6" separation, but building the tweeter horn at ear level sets the mid slightly further from you, and in my case the total separation became 4".

Since horns aren't easy to mount close to other drivers on the baffle, your response nulls may occur not very far above and below the listening axis and the phase alignment will be more critical. In this example I needed the phase to be close for this reason.
 

Attachments

  • 1.gif
    1.gif
    36.8 KB · Views: 331
diyAudio Moderator
Joined 2008
Paid Member
To be frank I can't imagine doing this without tools as the interactions are touchy and complex. To show this I've removed the LC before the woofer, which I only used because other methods were proving difficult in fixing both the acoustic response and phase at the same time.

The woofer doesn't suffer any issues that would really need a notch filter and the response is not very different to the first case, but look at the phase...
 

Attachments

  • 2.gif
    2.gif
    38.8 KB · Views: 307