[David Ellis]

I have learned a bit about this over the past few years and several crossover designs.  I feel comfortable in this conclusion and have ran it past a credible folks that agreed.  The conclusion is this:

"Slanted baffles generally make life easier on the crossover designer"

While there are several issues to address in any crossover, the initial two issues pertinent to this discussion are phase and amplitude.  Ideally, the phase of the drivers And the amplitude of the drivers are such that they will product a dead-flat response in the crossover region.  This necessitates proper/even phase response from the drivers.  If the phase is not proper/even, there will be response cancellations.   With a good crossover, the drivers will be perfectly in-phase at the crossover point.  They will also be completely out of phase when the leads of the tweeter are reversed.  This is called the "reverse null".

The result WILL be a null at/around the crossover frequency.  The 1801 demonstrates this nicely:



It should be obvious the 1801 doesn't need a slanted baffle for optimum performance.  This is true.  However, it doesn't necessarily mean this is completely optimal for all situations with all drivers.  

First, the acoustic center of the tweeter is generally the base of the dome where the dome meets the voice coil.  The acoustic center of the woofer is generally the base of the cone where the cone meets the voice coil.  The is an approximation, but will be very close in nearly all situations.  The exception to this might be those subwoofer drivers with flat cones.  I cannot comment on this.    Given the relative position of the acoustic center of the drivers it is obvious that 2 drivers on a flat baffle will not be acoustically aligned.  The tweeter waveform starts before the woofer wave form.  This is important.  The woofer phase response is DELAYED.

Mitigating this is accomlished via 2 methods.  One - the baffle certainly can be tilted backward.  Two, the crossover can "fix" the phase delay.

In an ideal world, the crossover could advance the phase of the woofer so that it will catch-up to the tweeter.  Or, in an ideal world, the tweeter will delay the phase of the tweeter.  In the real world, neither of these happen.  This is because.

1.  Inductance delays phase.  There are inductors in series behind the woofer.

2.  Capacitance advances phase.  There are capacitors in series behind the tweeter.

Crossovers actually make matters worse for the relative acoustic response of the drivers until... the phase of the woofer is so delayed and the phase of the tweeter is so advanced that it actually aligns.  At this point the the phase of the tweeter is a full wave-form BEHIND the woofer.  The beauty is the wave form is aligned!  There is no cancellation.

There are a few speakers that have an extreme slant to the baffle (off axis problems here)  that provide a completely aligned waveform.   This is often called "time-aligned", but some marketing departments have abused this term.  The key to identifying these time-aligned speakers is via observation from the side.  If the woofer dust cap is in-front of the tweeter dome, the speaker might actually be time-aligned.  If the woofer is NOT ahead of the tweeter, and it's called "time aligned"    be suspect of the marketing department.

I many cases a slight slant to the baffle will make life easier on the crossover designer, or better for the impedance curve.  This allows the crossover designer to give more attention to amplitude and less attention to phase.  A slanted baffle might allow fewer crossover parts, or a higher impedance.  

For the 1801, there is indeed some degree of compromise herein.  There is a slight dip in the impedance curve around 3khz to accomodate the phase situation.  This impedance dip is necessary to accomodate the phase.  This is indeed a compromise.  On the other side of the compromise is the off axis response of the drivers for a time-aligned speaker.  For the 1801 I chose a flat baffle and phase alignment + a bit of impedance ripple.  

http://www.ellisaudio.com/1801imp.jpg

Most good speakers are phase aligned (like the 1801).

Oh, while I am on the "soapbox", I must comment thus, anybody who builds/advertises a "no compromise" loudspeaker (or any piece of hifi gear) is an idiot.     There are always compromises.  

Hopefully my comments are fairly clear.  Whew, that's some typin'

[ak645]

Hi Dave,

 Thanks for the great post! Very informative.I notice some speaker builders make the claim that their speakers are "phase correct" because they use first order crossovers.But if they don't somehow time align the drivers this cannot possibly be true.Am I understanding your post correctly?

 Andy

[mac]

With a good crossover, the drivers will be perfectly in-phase at the crossover point. They will also be completely out of phase when the leads of the tweeter are reversed. This is called the "reverse null".

This is only true when you're talking about even-order LR crossovers.

[David Ellis]

Very informative.I notice some speaker builders make the claim that their speakers are "phase correct" because they use first order crossovers.But if they don't somehow time align the drivers this cannot possibly be true.Am I understanding your post correctly?

Fair question,

The simple answer to your primary question:  A speaker doesn't have to be 1st order to be "phase correct".

Other issues...there is a vast foray of terminology used to describe loudspeaker phase relationship among drivers.  So much of this is terribly convoluted by marketing departments.  As I (Dave) describe matters, "Time-Aligned" is where both the drivers phase matches perfectly.  This requires the driver acoustic centers to be significantly offest with the tweeter behind the woofer.  "Phase aligned" is where the signal from the tweeter is 1 wave behind the woofer.  This is much more typical, and true with well designed crossovers on flat baffle speakers.  There is yet another group (not addressed yet), where the commercial speaker is NOT phase correct in any fashion.  In this case the crossover designer simply ignored phase.     I don't know how prevalent this is, but I am certain it happens.

[David Ellis]

First, thanks for taking the time to read my post.  I was sincerely hoping someone would Q.C. the above post.  It's on the edge of my knowledge.

This is only true when you're talking about even-order LR crossovers

I must admit some degree of historical ignorance here.  While I do understand the basics of the various mathematical crossover typologies, I thought all of them paid attention to phase.   Is it really true all other crossover typologies have no electronic phase attention?     If so... well... that's just dumb.  However, maybe the era the other typologies were generated, getting impedance and amplitude correct was good enough.

Also, when I sit to design a crossover, I honestly don't consider any of the traditional typologies.  This is because driver's are all imperfect.  Dave's philosophy is:

1.  Avoid cone and motor resonance

2.  Get the phase and ampltude correct

3.  Use few components as possible.

4.  Maintain tolerable impedance.

The process goes as such, measure, model, tweak listen, measure, model, tweak listen, measure, model, tweak listen,...  I am sure you understand.

There is a partial caviat that I experienced. A friend in Great Falls managed to accomplish a mathematical crossover for the Vifa P17 and SEAS27TFFC with extremely good sucess.  He was an EE, and got a bit lucky too.  All that was needed after he was finished was reversing the tweeter leads!  The crossover had a few bumps, but was very good.  These drivers are very easy to work with.

Thanks again for reading Mac.  If true, I am very surprised that L.R. is the only typology that considers phase.  Please confirm.

[mac]

I am very surprised that L.R. is the only typology that considers phase. Please confirm.

My point was that only (even-order) LR crossovers will exhibit a "perfect null" when reversing the electrical polarity of the tweeter.

References:

Linkwitz, Seigfried H, "Active Crossover Networks for Noncoincident Drivers",
J. Audio Eng. Soc., Vol.24, No. 1, pp. 2-8 (Jan/Feb 1976)

Linkwitz, Seigfried H, "Passive Crossover Networks for Noncoincident Drivers",
J. Audio Eng. Soc., Vol.26, No. 3, pp. 149-150 (Mar 1978

http://www.rane.com/pdf/note107.pdf

[David Ellis]

My point was that only (even-order) LR crossovers will exhibit a "perfect null"

So... what is an imperfect null, and what is the impact?

I will admit some degree of, well, focus in this regard.  I really don't care what traditional model my crossover conforms with.  This is because I believe discussion of conformity to typology XXX or YYY or ZZZ has little/no utility.  I believe accomodating drivers in the above noted fasion is infintely more significant.  

In the good-old days of paper cones, cone tweeters, and no/poor measurement equipment, these models were paramount.  Today, things have changed.

[mac]

So... what is an imperfect null, and what is the impact?

I probably should have said "deep, symmetrical" instead of "perfect".  An odd-order crossover or an even-order one with an incorrect delay (or measurement axis that's off from the design axis) will yield less of a deep, symmetrical null than the one you have pictured.  A first-order crossover will not even produce a discernable null.

All you're doing with your reverse null test is confirming that the Linkwitz-Riley crossover works as advertised.  

[David Ellis]

A first-order crossover will not even produce a discernable null.

Hm, this is strange, but... the drivers I have used all need a 2nd order acoustic crossover or more.  Hence, I haven't experimented with a completed 1st order crossover.  However, I'd think that at some point the two drivers should be completely out of phase and product a deep cancellation.   I'll experiment with this.

I probably should have said "deep, symmetrical" instead of "perfect". An odd-order crossover or an even-order one with an incorrect delay (or measurement axis that's off from the design axis) will yield less of a deep, symmetrical null than the one you have pictured.

Herein, I wish to express some personal bias with those who get wrapped-up in the TERMINOLOGY of the traditional mathematical crossover models.  As I re-read the discussion above between Mac and I there is little/no substance.  The only matter present is clarification of the language regarding terminology.  There is only confusion for those outside the realm of crossover semantics.  This is unfortunate IMO.

In an effort clarify matters, I believe this is an effective summary of the assertions.

Dave's (my) assertion - All good crossovers will exhibit good amplitude, phase and impedance response .

Mac's assertion - There a group of terminology describing this phenomena.

Dave's response - Yep, I know there is a copious littany of terminology.  This terminology has little/no utility today.  It only makes matters more complicated for the layperson.