Phase Reversal in 2nd Order Crossovers

I've come across several sources that indicate that there is a phase reversal with 2nd order (12db) crossovers.

I've seen several 2-way designs that recommend wiring the woofer backwards (connect positive of amp to negative of woofer). And in 3-way designs, that the midrange be wired backwards.

I'm wondering this this 'phase reversal' is a true inverting of the signal, or if we are dealing with some strange phase shift between voltage (signal) and current?

While I plan to try it, I'm wondering if I open my three way cabinets and send mid-band frequencies out of the amp (mid-bass, mid-mid, and mid-high), then compare the signal coming out of the amp with the signal coming out of each section of the crossover, will I literally see the XO signal inverted?

In short, is wiring the speaker backwards because of a true and literal inversion of the signal as it arrives at the speaker component, or are we talking about a shift between signal and current that require the speaker component to be wired backwards.

And more important, is this all a load of baloney, or is it a serious approach to internal speaker wiring?

Enquiring minds want to know.

Steve/BlueWizard
 
A first order low-pass introduces 0-90 (45 at F3) degrees of phase lag, high-pass 90-0 degrees of phase lead (45 at F3, it's repeating waveform so there's no time travel involved), and the combination has the outputs 90 degrees out of phase at all frequencies. Together the two sum flat.

If you cascade two together for a LR2 filter you get 0-180 degrees of phase lag (90 at F6) from the low-pass, 180-0 of phase lead from the high-pass (90 at F6), and the outputs are 180 degrees out of phase at all frequencies. Together the two cancel for no output (assuming you're equidistant from both driver's acoustic centers along direct and all reflected paths). Invert one and they sum flat.

Inverting the woofer for a two-way or midrange for a three-way leaves high-frequency transients with the correct acoustic polarityand yields a flat sum.
 
PeteMcK,

Sound.WestHost was one of the site that lead me to ask the question. I've already acknowledged that either and only one of the woofer, midrange, or tweeter should be wire in reverse.

But are we seeing a somewhat intangible shift between voltage and current, or are we literally see an inverted signal? If the signal is literally inverted, I should be able to observe it as so on an Oscilloscope; comparing amp output to individual crossover outputs.

Now, based on what Drew Eckhardt said, -

"If you cascade two together for a LR2 filter ...the outputs are 180 degrees out of phase at all frequencies."

Implying that the signal is literally inverted.

But then he adds -

"...0-180 degrees of phase lag (90 at F6) from the low-pass, 180-0 of phase lead from the high-pass...

First, I admit I am somewhat out of my element here, but that implies to me, Low being Lagging and High being Leading, that relative to each other and at any one specific frequency, the lows and highs are one frequency cycle apart. That would be a phase shift, but not necessarily an inversion of the signal. Either that or I don't understand the concept of 'phase'. Keep in mind that I'm looking at this from a strictly voltage perspective.

Again, is this an observable phenomenon, or is it only a calculated one.

He finishes with -

"Inverting the woofer for a two-way or midrange for a three-way leaves high-frequency transients with the correct acoustic polarity and yields a flat sum."

Again, trying to learn here, is the goal of switching the wiring on one speaker to get all the speaker either Leading or Lagging. In a sense, and again, I'm guessing, by reversing the wire leads on a speaker, are you trying to give a lagging speaker the impression of being a leading speaker?

Given that everything is alway out of phase at all frequencies, the question has to relate to the leading and lagging aspect. So, is this current leading or lagging voltage, meaning the effect is out of phase, but the signal to the speakers is not literally and observably inverted?

Do you see what I'm getting at?

'Phase' can be applied in one of two ways. Two signals (meaning two voltages or waveforms) can be out of phase by 180degrees. One goes one volt positive the other goes one volt negative, the result is Zero volts. That is a simple inverted signal.

The other 'phase' is related to the presents of inductance and capacitance, and in which case, both signals(amp and XO), that is both voltages, will be in phase, but the current, and therefore the work being done, will be out of phase.

But here is my problem, in the center of the effective frequency band of either the woofer, midrange, or tweeter, isn't the crossover essentially out of the circuit. At mid-band aren't the effects of the crossover negligible?

Since you are making references to leading and lagging, I have to assume capacitance and inductance, which is exactly what a crossover is, and to voltage relative to current, but again, mid-band aren't the XO components essentially out of the circuit?

I'm just trying to educate myself, or more accurately, I'm trying to get you to educate me.

My simplest and most basic question is, using an oscilloscope to measure only signal voltages, should I be able to observe this phase shift? (Assuming a single mid-band frequency.)

I'm interested in understanding beyond that, but that is my most basic question.

Steve/BlueWizard
 
The last time I tried an XO that showed correct phase by inverting one driver, the sound came out funny. In one polarity, the cymals sounded good but the drums were muffled; in the other polarity, the drums sounded correct and the cymbals sounded muffled.
 
Effects of crossover is NOT negligible, even in the middle of the passband.
And yes, the phase shift can be observed by measuring signal voltage.

Most crossover network only ensure the amplitude to sum flat, but in the expense of phase distortion.
Johnk's old site have some transient response graphs that show how xover phase shift affect pulse shape.

1st order
An externally hosted image should be here but it was not working when we last tested it.


LR2
An externally hosted image should be here but it was not working when we last tested it.


3rd order Butterworth
An externally hosted image should be here but it was not working when we last tested it.


LR4
An externally hosted image should be here but it was not working when we last tested it.


http://www.geocities.com/kreskovs/TimeAligned2.html
 
BlueWizard said:
I've come across several sources that indicate that there is a phase reversal with 2nd order (12db) crossovers.

..........


And more important, is this all a load of baloney, or is it a serious approach to internal speaker wiring?

Enquiring minds want to know.

Steve/BlueWizard


Hi,

Generally the latter. What matters is the acoustic not electrical
response of the drivers. Genuine 2nd order acoustic crossovers
are few and far between so the rule cannot be generally applied.
(talking about standard bass/mid+tweeter 2 way above)

Builders of active crossovers like pretending acoustic issues do not exist .....

:)/sreten.
 
This subject is difficult because it pertains to the phase relationship between the amp and speaker. Each amp has its’ own way of interrelating with a given speaker.

My experiments with the matrix quad systems years ago gave me an understanding that may be of benefit here. I will add here that the quad systems were disappointing because the tolerances of the matrix and all-pass filters were sloppy.

The out-of-phase problem with speakers is only for the bass. It doesn’t make a lot of difference in the hi-freq.

The all-pass filter in quad matrix generates quadrature: a sin-cosine relationship. 0-degrees and 90-degrees.

The 90-deg lag has no effect below 200Hz, but it has a tremendous affect on the treble.

A 180-deg lag effects the bass, but not the treble.

So long as the bass speakers are in phase they will be OK. So long as the treble speakers are not 45-135 degrees (or thereabouts) out-of-phase they will be OK.

If one tweeter lags the other by 90-degrees the sound will appear to come from the tweeter with phase lead. This is part of the function of the matrix system. I have tested this and it has a very great effect.

By the way, there is no such thing as actual phase lead in the sense that the phase is shifted to lead the signal. This is impossible. Phase lead is actually 270-deg lag. In a repeating signal such as a sin wave it is perceived as phase lead, but in audio where things are always changing it is not necessarily phase lead.

If you happen to have a graph of the output of an all-pass filter you will notice that the sin (0-deg) input and the sin output are not the same. There is a repeating +/-45 degree variance through the audio spectrum of 20-20Khz.

I built a high definition (close tolerance components) SQ decoder years ago and there is a switch whereby I can compare the sound of the original signal to the output of the matrix decoder (front channels only).

The stereo groove of the SQ record are the front channels of the matrix, except the +/-45 degree variance is in the matrix decoder output.

The 45-degree variance at various frequencies is hardly discernible in a comparison between stereo groove and front decoder channels.

The point here is that so long as both speakers have the same phase characteristic the 0 to 180 degree shift due to the crossover may not be all that important.

But then again, if the same frequency is coming from the tweeter and midrange (in the same speaker box) and they are not in phase then a muddy sound may result, not sure. They are close to one another so there is no directional issue.
 
The problem with this whole subject is that the speaker system is a reactive load. The ideal load for a power amp would be a purely resistive load where the current and voltage are always in phase.

The speaker is also a generator. When the suspension system pulls the cone back to the mid-point it generates a signal that is seen by the amplifier output terminal. Thus, the feedback loop picks up this signal.

The feedback loop of the amp must deal with this complex reactive relationship in general.

It is my present interpretation of the graphs that they depict the manner in which the amp is trying to correct for reactance in the load.

The positioning of the speakers relative to one another (within the speaker box) is surely important. The problem with interpreting impulse graphs is that they involve the amplifier feedback.

When the acoustic energy of the woofer reaches the tweeter it may cause the tweeter to generate a signal which gets into the signal going to the feedback loop. Thus, the relative positioning of the speakers is an important issue.

The problem I have with the graphs is in what they actually depict. They are showing how the amp is dealing with the load in order to match it with the input signal. The graphs may not actually have much to do with the sound coming from the cone of the speaker.
 
Mm...

I would like to recommend anyone who looks at the pulse or impulse response of crossover filters to build an allpass filter that mimics the transfer function and listen to it.

I have done this in blind tests with my students the last ten years and about 1/100th of them manage to identify the allpass filter with 99% confidence.

The impulse response may look awful for the allpass filter. Yet the difference is, if not impossible to detect, at least very small.

All you need to build an allpass filter is one opamp, 3 resistors and one capacitor per channel. Possibly the inputs should be buffered, so throw in another opamp to make sure.
 
Svante said:
Mm...

I would like to recommend anyone who looks at the pulse or impulse response of crossover filters to build an allpass filter that mimics the transfer function and listen to it.

I have done this in blind tests with my students the last ten years and about 1/100th of them manage to identify the allpass filter with 99% confidence.

The impulse response may look awful for the allpass filter. Yet the difference is, if not impossible to detect, at least very small.

All you need to build an allpass filter is one opamp, 3 resistors and one capacitor per channel. Possibly the inputs should be buffered, so throw in another opamp to make sure.

Was this test done with headphones or speakers, if you don't mind my asking?

Pete B.
 
The problem is that the focus must be on the acoustic response...

The problem is that the focus must be on the acoustic response...as sreten has already pointed out.

Originally posted by hailteflon The problem with this whole subject is that the speaker system is a reactive load. The ideal load for a power amp would be a purely resistive load where the current and voltage are always in phase.

This issue of the driver phase requirements, inverted connections or not, has to do with the acoustic output of the drivers. A 2nd order crossover, meaning a combination of a 2nd order lowpass with a 2nd order highpass, has its two sections with acoustic outputs that are 180 degrees out-of-phase throughout the crossover region if they are both connected to the amp with the same electrical polarity.

One of the two sections must have its electrical connection inverted so that the acoustic output of one driver is thus inverted so that the acoustic sum of the two sections then sums as desired for the acoustic target. That is, the sum will then be flat for the 2nd order Linkwitz-Riley. If the target were a 2nd order Butterworth, then the CORRECT connection, one inverted driver, will result in a +3db bump at crossover on the design axis due to the lobing. So it's necessary to define the specifics of any crossover, not just the order.

But again, the focus must be on the acoustic output, not the electrical aspect other than for normal or inverted connection requirements.

The speaker is also a generator. When the suspension system pulls the cone back to the mid-point it generates a signal that is seen by the amplifier output terminal. Thus, the feedback loop picks up this signal.

Outside of tubes, the output impedance is low enough that they can be considered to be ideal voltage sources for purposes of loudspeaker system design. There is little if any, effect on the gain of the signal. What does occur is that the non-zero output impedance of the amp, even though small, will add a slight reduction in the damping of the back-emf. But again, this is very small for solid state amps.

The feedback loop of the amp must deal with this complex reactive relationship in general.

The amp output impedance is the issue, not the loop gain. The complex nature is easily handled by most SS amps, but it can be a problem for a tube amp.

It is my present interpretation of the graphs that they depict the manner in which the amp is trying to correct for reactance in the load.

That interpretation is in error. The charts represent the perfect, ideal response of the acoustic output of the crossovers depicted. An amp simply makes matters worse for whatever non-ideal nature it may have, that of not being the perfect voltage source. SS amps when used within their design limits have a small impact, while tube amps will have a much larger one.

The positioning of the speakers relative to one another (within the speaker box) is surely important. The problem with interpreting impulse graphs is that they involve the amplifier feedback.

Again, the latter is just not correct. In fact, if one uses an MLS-based measurement that uses a feedback probe, all influence prior to the probe attachment point is totally eliminated. The preamp, amp, cabling, connections and all other influences are of no concern whatsoever. The cross-correlation made between the known MLS output of the software and the measured signal as it appears at the probe point corrects for any and all signal degradation. The probe cabling and feedback channel are then the only things not corrected, but high quality sound cards minimize it to the point of little concern.

When the acoustic energy of the woofer reaches the tweeter it may cause the tweeter to generate a signal which gets into the signal going to the feedback loop. Thus, the relative positioning of the speakers is an important issue.

This is, in essence, not an issue whatsoever. There will not be any tweeter "feedback". It is not modulated in any significant way. Were this so, the exceedingly tight correlation between design and followup measurements would not hold as they actually do. This is just speculation that is not borne out at all in empirical testing.

The problem I have with the graphs is in what they actually depict. They are showing how the amp is dealing with the load in order to match it with the input signal. The graphs may not actually have much to do with the sound coming from the cone of the speaker.

This is not correct. The graphs depict the ideal acoustic responses of the various crossovers. They have absolutely nothing to do with the impact of the amp, whether large or small.

Dave
 
the focus must be on the acoustic output

Let's put that one in bold letters. The acoustic output is what matters. Further, the notion of reversing the polarity of one of the drivers assumes that the acoustic centers are coplanar (generally not the case) and that any acoustic rolloffs from the drivers are at least a decade away from the crossover frequency (almost never the case).
 
The impulse response may look awful for the allpass filter. Yet the difference is, if not impossible to detect, at least very small.

Even if there might not be any audible difference I design my active crossovers for perfect step response because:

1.) A speaker should be able to reproduce the input waveform, someting that is expected from even the worst of all amplifiers for instance - so why not expect it from speakers ?

2.) Because I know how to do it !!!!! :D

I know that there are many good sounding speakers with high-order allpass crossovers and bad sounding transient-perfect designs do also exist, no need for mentioning that ! ;)

Regards

Charles
 
Ex-Moderator R.I.P.
Joined 2005
Its quite impressive with such advanced knowledge in here !

I cannot say that I understand it all, but I know fore a fact that nice acoustical measurements does not automatically lead to good sound, I have heard some of those too - and I dont think you can totally neglect the electrical side of a crossover, actually I am convinsed that its VERY important, and in these days with the focus on the acoustical side "the other side" tend to be forgotten

I would really like to know what is actually going on inside the filter itself .... electrically:clown:
 
The electrical XO is a means to an end only

tinitus said:
Its quite impressive with such advanced knowledge in here !

I cannot say that I understand it all, but I know fore a fact that nice acoustical measurements does not automatically lead to good sound, I have heard some of those too - and I dont think you can totally neglect the electrical side of a crossover, actually I am convinsed that its VERY important, and in these days with the focus on the acoustical side "the other side" tend to be forgotten


I don't think anyone was implying anything WRT measurements and quality of sound, only that a design is based on a goal and that goal, such as a 2nd order Linkwitz-Riley for example, has specific requirements. There is, of course, no quarantee that upon meeting those requirements that the design will sound good nor is it necessarily even appropriate given the drivers. The focus is on the specifics of the selected target. You can't just design for "good sound", there are principles to use as guides to that end. If the selected target is chosen poorly, that's another issue (and thread) altogether.

I would really like to know what is actually going on inside the filter itself .... electrically:clown:

This is a simple matter to determine. Pretty much any CAD software will allow one to show the transfer function of the crossover. That transfer function is that which is necessary to achieve the desired target when combined with the specified driver's electro-acoustic transfer function, i.e. it's raw SPL response. Simply put, when the two are combined the result is that the acoustic output from the XO/driver combination is the specified target.

Dave