[Roy Johnson]

JohnR- You're right. I oversimplified. Actually the two drivers are 90 degrees out of phase on every tone (at least that's the goal), so the summation at every frequency has a constant relationship- there is no relative phase shift between the two drivers. The group delay is constant.

Scotty- My apologies for not responding sooner. You are correct about Ashley's and Kaminsky's research- I should have mentioned it. If you examine the graph of the frequency response of their circuits, you'd see
a) there is not more than a couple of dB difference in ultimate attenuation these circuits offer compared to a parallel circuit- insignificant to the drivers at those frequencies so far away from the crossover point, as I mentioned before.
b) these circuits do run the drivers harder by a couple of dB in the overlap region than does a parallel circuit, which can be significant to the drivers' distortion levels and power handling. And
c) with the relaxed constraints, the final frequency response would not be flat, by several dB. Could this be the what gave the impression of "better dynamic coupling"? Maybe, maybe not. Life is short.

When you overlay a graph of the series' slopes onto those of a parallel circuit's, why isn't there much attenuation difference seen? Because the parallel circuits start rolling off much farther from the crossover point, e.g., a series circuit works the two drivers each nearly a couple dB louder for a good part of an octave on each side of the crossover point. This means ~60% more power sent to each driver, right in the region where we would really prefer to push them less. Only below/above the crossover point does a brief region of 12dB/octave rolloff exist for either side of the filter, as you said.

This is for a two-way speaker. If you make a three-way speaker with a series first-order network, the woofer circuit does indeed add a 2nd-order rolloff to the bottom of the mid, but (in my experience) too far below the actual crossover point to help the mid very much.

Thanks for the link to Fried's thoughts. Appreciated that. Do note his remarks on parallel networks are primarily on circuits of 2nd-order and higher. But, never say never- I will re-examine the sonic benefits of the series circuit. However, Scotty, the parts are not really in shunt.

An easier way to see what is happening is to look at Kaminsky's transfer function that led to that graph (previous page). You will see that the Voltage going to either driver is a function of both the capacitor and the inductor (hence the quasi-12dB/oct rolloff), and of both drivers' impedances (the least linear parameters).

In other words, for the series circuit, a non-linearity in one driver's impedance (any deviation from being a "resistive" load for the crossover) affects both drivers.

A driver deviates from being a resistive load in several important ways:

A front-suspension resonance in the woofer.
The woofer cone "going soft" before the crossover point (happens in most models of woofers and mids- seen as a decrease in or a flattening-off of a raw driver's inductive impedance rise w/frequency, as low as 300Hz!!).
The unavoidable, eventual woofer-cone breakup above the crossover point.
The imperfection of the capacitor used in either driver's Zobel network,
Improper design of the Zobel circuits for each driver (a pervasive problem, in my experience).
Thermally-caused impedance changes in either driver as power levels change.
Dynamic impedance changes from a lack of dustcap-pressure venting behind the woofer (at more than 1/8" of stroke in my experience),
From magnetic-field non-linearites created/encountered as the woofer or the tweeter strokes.
From the low-frequency resonance of the tweeter.
From the non-linearities of the tweeter's suspension compliance on very small signals and very large signals.

Throw in the imperfection of the main capacitor for the tweeter and of the inductor for the woofer, as those would also be part of the impedance non-linearity... (we've all heard poor capacitiors. Ever heard a large-wire inductor electrically ring in the midrange circuit? Ecch!)[/list:u]With a series crossover, a non-linearity in one driver, one circuit element, affects both drivers. A parallel circuit does not have this crosstalk.


Since a series circuit can use wider-tolerances for the capacitor and inductor (usually the capacitor), then from experience, this saves <15 minutes for testing and matching, per stereo pair, for a manufacturer ordering caps in bulk. It does make sense for the home constructor, to avoid buying a lot of extra caps from which to match.

If the speaker manufacturer is using high quality, good-sounding film capacitors (which still vary in tolerance), there's little to no savings in the cost for bulk purchases, from specifiying looser tolerances beyond the +/- 5% usually seen in each batch of caps. There is a significant savings if those were certain imported film caps or any electrolytic caps (still used in most higher-order circuits, for woofer/mid, by very respected companies).



Thanks for such a good forum, to all. I have only been here a few times, and couldn't ask for more considered responses from all parties. Makes the sharing of experiences and theories far more effective, which reminds me there is always more to learn. Reminds me also what I can and cannot talk about, since there's only so much a designer can share, which I would like to touch on:

Within these limits, I believe it's most useful for my comments to point a direction to further someone's research, to clarify confusion and misconception, and finally, to do no damage that keeps speaker design a psuedo-science. Which would be making un-substantiated claims like "cones draining away energy" from a cabinet. Ones to which Physics says, "No- that is not the explanation for the change in sound."
What I cannot say in my capacity as a professional designer, consistently amounts to my teaching of those who have not done the research. I would love to share all my info with a home constructor, but cannot risk it. So I point, and explain. Hope that's ok.

A manufacturer can brag about a new titanium-based technology, but cannot show WHY publicly, via graphs (so Stereophile tries to), or many times privately to the marketing dept, because if you look with an experienced eye to the proper graph, you will see many times that this is not an audible advance, or it is not the main reason the new model sounds better.

A consumer believes the designer has the answers or does not. From asking the right questions or not. No matter what, it's up to you to listen, and have a close look at a reviewer's choice of music, listening habits, past experience, and room layout if you are to learn something from reviews.

It is easiest to believe a designer knows what he is doing if you see an actual physics-based explanation of the pros and cons- the WHY behind each pro and con, which is seldom done. Anything less defines self-promotion, and frankly, a waste of your time.

It's been a long week. My apologies if I wasn't succinct.

Best,
Roy Johnson
Green Mountain Audio

[TheChairGuy]

Welcome to AC, Roy, your contributions are most helpful  

   TCG

[rblnr]

This thread has been tremendously educational.  Thanks all.

[Duke]

Hello Roy,

I've thoroughly enjoyed your posts here and elsewhere.  Pardon me if I now display how little I've actually grasped...

From what I understand, a 4th order crossover has 360 degrees of phase shift at the crossover frequency - in other words, the woofer's response will lag one wavelength behind the tweeter's.  

But, what if you physically moved the tweeter so that its voice coil is now one wavelength behind the woofer's at the crossover frequency?  Factoring in that 360 degree phase shift, would the two drivers now be "in phase" at the crossover point?  Would this improve the otherwise dismal   step response of the system?

I'm sure there is something fundamental that I'm missing here...

[Roy Johnson]

Thanks guys- I'm glad this didn't put you to sleep!

Duke, everything you say is accurate- regarding the 360 degree difference at the crossover point, and how offsetting two drivers by that one wave's length improves the impulse response.

The drivers are "in phase" only at that one crossover frequency. The time delay at other frequencies varies. It is not a single fixed amount, equal to one wave period (the period of the crossover frequency). So the offset is only helpful at the crossover frequency. At other frequencies, the drivers are not 360 degrees out of phase, so the offset doesn't fully help.

The resulting impulse response is still not a compact up/down pulse. It is a smooth oscillation of say, the tweeter pulse (spike) arriving first up, then down, then smoothly turning into the mid's pulse arriving, up then down, then the woofer pulse arriving, up then down. Those three impulses "fall/revolve/evolve into" smoothly each other, so to speak, since they are "in phase".

Without that one-wavelength offset, the impulse response looks like three individual spikes.
With the setback, it looks like three rolling waves- a narrow one followed by a wider one, and then by a very wide one. The overall sound is smoother with the offset, but still dis-jointed.
BTW, the heights of those three "pulses" (spikes) should also be equal... seldom seen in the graphs in Stereophile! That would mean the woofer's pulse is as loud as the mid's pulse, as loud as the tweeter's pulse. If they are unequal, the designer manipulated the steady-state tone balance to hide some of the artifacts of the phase shift, as I have mentioned.

The incoherency at your ear is audible, under double blind conditions, regardless of the recording, the music, the tone range. And that "phasiness" bothers women more quickly than men, especially when turning up the volume.

A coherent source arrives with more dynamic snap, has better focus from from to rear (which is the "direction" of the time delays), and leaves with a faster, cleaner decay. The music is not as "noisy".
It is more independent of the speakers' locations and less dependent on the recording quality, your electronics and your wires.

I think you probably have a good grasp Duke- but you have to know that the math gets worse. There is an article I wrote for Andrew Marshall's Audio Ideas Guide (Spring 1997 issue??) that lays out much of the math, so you could actually see the how the time-delay numbers are calculated, but this issue of his magazine is not online. I've re-read it, and yikes- what a pain to wade through... it's a lot to digest. What was discussed here however, combined with the info on Audiogon regarding "time coherence" pretty much lays it out- the pros and the cons.

BTW, I think it's important to mention again, that the concept of a first-order speaker exhibiting "comb filtering" is true, for very specific frequencies, at a very specific point in space.
Fortunately, music doesn't have pure single tones, so we don't hear comb filtering in a first-order speaker. What we do hear is some of the depth of the soundstage collapse when we stand up, because we've spoiled the woofer-mid-tweeter timing, which means echoes are smearing over. We also hear sibilants and transients start to smear. And yet those time delays caused by you standing up and wandering around, are far less at every frequency than what a higher-order design delivers on any axis. This is not my opinion- these time delays are mathematically predictable and verifiable. It is claimed, "You can't hear the time delays of the 4th-order design, and there are test tones that demonstrate you can't!" My response would be, "Try listening to music to easily hear the effects of time-delay distortion."

I hope this is helpful. Those who know the math will see that I have left out many variables, by assuming each driver has a dead-flat response, before the crossover was applied (perfect drivers!!) and w/o cabinet reflections and w/o "floor and wall-behind" bass-boost. And with the listener at a particular location.

Even with those variables, I do think what's been said here is most of the reason for the differences heard between a first-order design (or a no-order design, like a full-range electrostat), and a high-order design.

Best to all,
Roy
Green Mountain Audio

[Val]

I would like to comment that a first-order crossover is just one form of the many unavoidable compromises in speaker design and implementation. I know it can work well, as demonstrated by the superb sound of the Vandersteen 5A, for example. But I also agree with the previous post regarding the successful implementation of Modaferri's infinite slope by Joseph Audio. For another example of a well-implemented high-order crossover, take a look at the impedance and phase responses of the PMC OB1 in the latest issue of Audio Ideas Guide. Both measurements are as flat as I have seen in my many years of audiophilia, and Andrew Marshall made the connection between those measurements and the wonderful sound of the speaker. The OB1, as all PMC designs, has a 4th-order Linkwitz-Riley crossover.

[Roy Johnson]

Good points Val.
It's important to note that the impedance and phase responses you mention are only flat (linear) because of many added parts in the crossovers, and do not represent, even remotely, anything about the acoustic phase response- contrary to what Andrew Marshall states. That flat impedance curve is only what the amplifier sees, and does not reflect the still-varying phase angle between voltage and current the drivers see. And now the sound is less clear because of the haze the extra crossover parts and circuit-board traces add.

Is the implementation ok, as you say? Yes, those are two fine examples of high-order design. In my experience with those designs, the phase shift in them is clearly audible. But I also know what an absence of acoustic phase shift sounds like.

If one does not have an intimate experience with low phase-shift speakers, intimate meaning a lot of different music heard thru them for many hours, from Jazz to bluegrass, to opera, folk, rock, symphonic, then the phase error can escape your attention at first listen.

However, to that listener, the phase shifts become easily audible when the music played encompasses the crossover point(s). Most demo music selected by the salesman or by the factory though, does not span the crossover point, but has sounds laying well away from it to either side- like high treble with a mellow voice, not the mid-treble with higher-voice, for a tweeter crossover. Even then, with instruments and sounds not "requiring" the two drivers be in synch, the sound from the tweeter is clearly in "a different room" than the sound from the mid. Why? Because of the different time delays affect depth.

Also, in most of those designs, if not all of them, you can hear the treble sound come from the height of the tweeter. Why? Because of the timing difference compared to the sound from the mid. Among the ways we localize sound is by first-arrival, and the tweeter in those designs usually arrives first.

To find the faults of any speaker the most quickly, always play music that requires output from the driversaround the crossover point.

Best,
Roy

[JohnR]

Hi, since this is the Lab, I would appreciate it if we could keep the discussion primarily technical. Thanks

[Aether Audio]

http://www.audiocircle.com/index.php?topic=10046

[Roy Johnson]

SP Pres- I hadn't seen your writings before. Thanks Bob. You cover many important points in your posts on that thread, and do so very well. I disagree with several things, but that's not for here.

JohnR- Thank you for intervening. If you mean subjectivity is not in keeping with the Lab pages' orientation, then I'd agree that keeps the threads focused on technical matters. I'll try hard to respect that.

Please allow me three short sentences which I believe have merit as a background to the Lab forum:
We all look to objectivity to find and to hold truth, and to make better tools.
Because we associate objectivity with truth, it is easily invoked to invalidate experiential results.
Yet experiences should be a goad to finding new and better methods of objective measurement!

Two links: Subjective vs. Objective Research, Changing  Minds
Thanks John.

The goal of this Lab forum of course, is to disseminate knowledge, so we help to create better designers and better fidelity. Knowledge is developed from what is considered objective, and from real-world experiences which are reported objectively (a necessary condition).

I considered well how that last post might be received John, because of your guidelines. I posted it as I believe the phase-response issue is mired primarily because the lab tests to determine if phase shift was audible were fundamentally flawed:

Although the clicks and squarewaves employed met the mathematical standards for "complex signals", and those signals were easily repeated during the tests, and by any other researcher, my view is that these signals were not complicated enough to mimic music or any familiar sound, nor familiar enough to the headphone-wearing listeners.[/list:u]Thus, I objectively reported how, subjectively, music-listening reveals much to most any listener I have encountered, if we use it wisely.
And that would be by selecting the type of music in accordance with physics- asking if that proposed musical passage can reveal the existence (audibility) of the particular fault physics says is there. This can only be answered by asking what should we likely hear happen to that proposed passage, or to that note, or melody, or harmony, or rhythm, or timbre, or dynamic contrast, or image specificity, or depth of the image, or to the position of the image.

An experienced listener, a scientific listener, a trained listener, would then listen for "that". Convinced for himself, he would employ blind tests on himself and others to allay his subjectivity.

Physics shows the most severe phase-response foul-ups in a speaker happen around the crossover points. Hence my suggestion to play music that requires output from the drivers around/across the crossover frequency.

I will try harder to keep subjective comments, mine and those reported to me, out of this thread. Thanks for allowing me to respond.

Best,
Roy

PS: a neat reference: http://inventors.about.com/