Phase coherence

[JerryLove]

I was reading some mid-2000's articles from an circuit designer down in OZ, and at one point he starts discussing phase coherency (mostly visa-vi crossovers). I think he's talking about the same issues that Roy was trying to get to on our recent Green Mountain thread.

As I understand that position: Speakers overlap. At the crossover point, for example, two speakers will be producing the same sound at the same SPL.

The basic premise is that when these two identical sounds become out of phase, the subjective experience suffers. At the most extreme, two sounds out-of-phase 180-degrees will attempt to cancel each other out "suck out". At another extreme: if one is more than 0.1s delayed from the other: then they will actually hit as separate sounds.

Giving Roy the benefit of the doubt: when he was talking about people people hearing very small distance differences: it was (presumably) not about absolute phase (which people can't hear), but about relative phase.

Phase distortions come in two forms. There's the difference in distance between the acoustical centers of the two drivers involved. There's also delay introduced by the crossover (the author had a lengthy article on how low-pass and high-pass filters add 25-degree delays).

Because the delay is tied to the frequency: while it's possible to simply time-adjust for one frequency (what happens when we reverse phase on a speaker with a 180-degree out-of-phase response to the speaker it is crossed-over with), that only fixes it for one and not for another frequency.

Now his favorite solution was a digital crossover that included phase control.. but those are relatively recent... and he does discuss what seems to be Green-Mountain's solution: the use of first-order crossovers.

Now I'm speaking way above my pay-grade here: but as I understand the 20k-ft view: 1st order crossovers introduce the least phase delay, but because they are shallow slopes demand that the speakers perform several octaves beyond their crossover points (one of their largest problems).

So as I understand GM's solution: Align the voice-coils on the vertical plane, use a first-order crossover to limits phase problems from the crossover, and then try really hard to find speakers that don't fall apart (resonance, cone breakup, lobing, etc) with asked to do far more octaves than a single driver can really do.

Can anyone tell me if I've got the generalities right? Can anyone comment on relative phase and the effect on audibility?

 

[GranteedEV]

Even if you were to take away breakup, resonance etc, there's still the fact of power handling and excursion, no?

 

[jostenmeat]

Hi Jerry, I'm trying to be self effacing for the two of us here, but we both probably know enough just to be dangerous.

Firstly, may I ask you how delay is tied to frequency? (sorry, and thanks.)

1st order does require that the drivers remain relatively flat throughout a much wider passband. The xover point between tweeter and mid will be very high (since a tweeter can only go so low, but even then it still requires a wide passband nevertheless), and of course the mid must be able to really go quite high. The drivers need to match well. I have read that Dynaudio, BW, and particularly Thiel have extensively used 1st order. I am not as sure about the first two.

The biggest reason to not use 1st order, and it seems to be a doozie, is because offaxis response hurts for it. It seems with all of the recent DBT results of listener preference that offaxis response is a really important thing. However, while I can give an educated guess as to why 1st order doesn't do well here, I am not going to embarrass myself unnecessarily without the benefit of further research.

While you mention "align the voice coils on the vertical plane", in fact Thiel is known for their sloped baffles. They are sloped 15 degrees backwards, because 1st order xovers are 15 degrees out of phase; the sloped baffle now puts them perfectly in phase. It was TLS Guy who taught me about the last part, after picking his brain on 1st order. Annunaki was the first to teach me about the fundamental design differences at the beginning of my post, years ago.

 

[JerryLove]

Hi Jerry, I'm trying to be self effacing for the two of us here, but we both probably know enough just to be dangerous.

Firstly, may I ask you how delay is tied to frequency? (sorry, and thanks.)

I don't understand the math / science behind it but, looking (I think it was a 1st order crossover), the (and I might have these reversed) low-pass filter creates a wave -25 degrees out-of-phase, and the high pass +25 degrees out of phase. As the wavelength varies, the temporal difference between those two points on the wave changes (higher wavelength, means less fractions-of-a-second between peaks, and so different phase variation)

The biggest reason to not use 1st order, and it seems to be a doozie, is because offaxis response hurts for it. It seems with all of the recent DBT results of listener preference that offaxis response is a really important thing. However, while I can give an educated guess as to why 1st order doesn't do well here, I am not going to embarrass myself unnecessarily without the benefit of further research.

1st order does not seem to be popular: there are many downsides (not all of which I can hope to repeat off the top of my head) to them. All speaker design is about compromise.

While you mention "align the voice coils on the vertical plane", in fact Thiel is known for their sloped baffles. They are sloped 15 degrees backwards, because 1st order xovers are 15 degrees out of phase; the sloped baffle now puts them perfectly in phase. It was TLS Guy who taught me about the last part, after picking his brain on 1st order. Annunaki was the first to teach me about the fundamental design differences at the beginning of my post, years ago.

That's almost certainly why (you mention this above) Thiel uses slopped baffles and first-order crossovers. I'd assume they are time-aligning their speakers.

 

[jostenmeat]

I don't understand the math / science behind it but, looking (I think it was a 1st order crossover), the (and I might have these reversed) low-pass filter creates a wave -25 degrees out-of-phase, and the high pass +25 degrees out of phase. As the wavelength varies, the temporal difference between those two points on the wave changes (higher wavelength, means less fractions-of-a-second between peaks, and so different phase variation)

Oh. Firstly, did you mean 15 degrees, and not 25? Thanks for your offering on the second part.

1st order does not seem to be popular: there are many downsides (not all of which I can hope to repeat off the top of my head) to them. All speaker design is about compromise.

Yes, but even taking the compromises aside, just finding the drivers that can even be implemented as 1st order will really narrow down the field. I think this is partly why Dyn/BW can pull it off: they make their own drivers.

That's almost certainly why (you mention this above) Thiel uses slopped baffles and first-order crossovers. I'd assume they are time-aligning their speakers.

Yes, that is what I said. However, and I think even TLS was saying this (? I don't want to misquote him, so ignore that), while it looks to be a good goal on paper, this compromise may not be worth it in the grand scheme of things (I'm talking 1st order in general), because we cannot hear the difference with subtle phase shift (at least compared to something like significantly compromised offaxis response), or something like that.

 

[JerryLove]

And those are the two great questions:

1) Which compromise is perceptually more important.
2) Which compromise can be better exploited / mitigated.

Which brings me back to a question I asked in the GreenMountain thread that was never really answered: Why not use a digital, active crossover and get phase-coherency without having to give up crossover slope?

 

[jostenmeat]

And those are the two great questions:

1) Which compromise is perceptually more important.
2) Which compromise can be better exploited / mitigated.

If I am understanding current trends correctly, 1st order is not the best compromise right now.

Which brings me back to a question I asked in the GreenMountain thread that was never really answered: Why not use a digital, active crossover and get phase-coherency without having to give up crossover slope?

You'll have to ask the manufacturers themselves. Because speaker experts like TLS Guy are proponents of active speakers, and believe they will gain in popularity, eventually. Have cake, eat it.

 

[InTheIndustry]

YouTube: Phase Technology-Absolute Phase Crossover.

Phase Technology is my favorite speaker manufacturer, period, point blank. President of Phase Tech, Ken Hecht (who's father Bill invented the soft dome tweeter) is, IMO, an incredible speaker engineer and a great guy. They don't come any nicer or more patient than Ken, who also does the Phasetech videos.

I have NEVER been anything but excited to work with their products. The company & Ken's engineering ideals regarding not just his crossover design, but driver composition, cabinet design, & implementations are 1st rate.

Ken's dARTS system is our flag ship theater piece in our showroom & is the finest system I've ever heard - at any price point - when set up properly. dARTS is a 100% active, closed system that offers phase control/correction as well as many other unique features. Watch the video on that as well. All of the Phase Tech videos are kind of cool and informative if you're into audio & speaker design.

 

[walter duque]

YouTube: Phase Technology-Absolute Phase Crossover.

Phase Technology is my favorite speaker manufacturer, period, point blank. President of Phase Tech, Ken Hecht (who's father Bill invented the soft dome tweeter) is, IMO, an incredible speaker engineer and a great guy. They don't come any nicer or more patient than Ken, who also does the Phasetech videos.

I have NEVER been anything but excited to work with their products. The company & Ken's engineering ideals regarding not just his crossover design, but driver composition, cabinet design, & implementations are 1st rate.

Ken's dARTS system is our flag ship theater piece in our showroom & is the finest system I've ever heard - at any price point - when set up properly. dARTS is a 100% active, closed system that offers phase control/correction as well as many other unique features. Watch the video on that as well. All of the Phase Tech videos are kind of cool and informative if you're into audio & speaker design.

Friend of mine has the Premier Collection and drives them with a Cinepro amp. Absolutely one of the finest systems I have ever heard. Fantastic speakers.

 

[JerryLove]

BTW: This is the article I read that made me start this thread:

http://sound.westhost.com/ptd.htm

 

[Swerd]

BTW: This is the article I read that made me start this thread:

http://sound.westhost.com/ptd.htm

In your OP, some of what you said is right and some of it you've got confused. I was going to try to straighten it out, but the variations in phase and time in various odd-order or even-order crossover is a complex matter. Simple words don't cut it. You need to see phase diagrams of the sine waves of voltage and current as they change when passed through a capacitor or an inductor.

Just to refresh myself before writing this, I re-read chapter 7 in Speaker Building 201 by Ray Alden. It introduces the subject pretty well. If you are curious about this, get yourself a copy.

There are no easy answers to the question, but Ray Alden summarizes the current thinking by people such as Floyd Toole and Sigfried Linkwitz who have good evidence that listeners seem to prefer the sound of speakers with even order crossovers (2nd or 4th order) that produces a flat first arrival signal that minimizes room reflections. Whether speaker drivers are in or out of phase or time with eachother seems to matter much less.

Interestingly, when you consider a 2-way speaker with a 4th order LR type crossover, the woofer and tweeter are out of phase with eachother in the crossover range by 360°. That makes them in phase, but one cycle out of time. I know I can't tell the difference, and I wonder if others can.

All design features in speakers are some type of compromise. There are no simple answers.

 

[Rickster71]

Here's another speaker that uses phase correct alignment. http://www.sourcespeaker.com/Coherentpulse61a.html

 

[highfigh]

Hi Jerry, I'm trying to be self effacing for the two of us here, but we both probably know enough just to be dangerous.

Firstly, may I ask you how delay is tied to frequency? (sorry, and thanks.)

1st order does require that the drivers remain relatively flat throughout a much wider passband. The xover point between tweeter and mid will be very high (since a tweeter can only go so low, but even then it still requires a wide passband nevertheless), and of course the mid must be able to really go quite high. The drivers need to match well. I have read that Dynaudio, BW, and particularly Thiel have extensively used 1st order. I am not as sure about the first two.

The biggest reason to not use 1st order, and it seems to be a doozie, is because offaxis response hurts for it. It seems with all of the recent DBT results of listener preference that offaxis response is a really important thing. However, while I can give an educated guess as to why 1st order doesn't do well here, I am not going to embarrass myself unnecessarily without the benefit of further research.

While you mention "align the voice coils on the vertical plane", in fact Thiel is known for their sloped baffles. They are sloped 15 degrees backwards, because 1st order xovers are 15 degrees out of phase; the sloped baffle now puts them perfectly in phase. It was TLS Guy who taught me about the last part, after picking his brain on 1st order. Annunaki was the first to teach me about the fundamental design differences at the beginning of my post, years ago.

Phase delay is also known as phase shift.

I have never heard of speakers being 15 degrees out of phase using a first order filter. Also, 15 degrees tilt has nothing to do with being electrically out of phase by that amount. It may be a coincidence if they come back into phase with that physical tilt but it's not going to work for all frequencies.

I don't know if you have calculated wavelengths of different frequencies, so I'll include the formula here- λ=c/f, where lambda is the wavelength, c is speed of sound and f is frequency. As an example, if c=1126ft/sec (@20° in dry air), a 2500Hz wave will be about 5.4" and if phase alignment is to be accomplished by tilting the baffle, that 5.4" needs to be looked at as 360° at the frequency in question. This means the distance from the center of one voice coil to another needs to be known in order to determine how much tilt is needed.

If the listening level is low and/or the speakers are placed on the long wall and no side wall is near for first reflections, off-axis response is much less significant. This is also the case if the system is set up for a small "sweet spot".

Your comment about tweeters is important and is the main reason for selecting one with a low Fs and staying as far from that as possible- the rule of thumb I have always heard/seen is 1-1/2 octaves from Fs to F3, especially in a first order filter.

I found this interesting link:
http://www.linkwitzlab.com/crossovers.htm

 

[jostenmeat]

Thanks highfigh. I may have yet again successfully misremembered and/or misunderstood. The 15 deg is in regards to the lobing tilt with odd order xover. Here is what TLS said about a year and a half ago:

Thiel really believes in first order filters. Yes, the slope backwards is important. Odd order crossovers have 15 degrees of tilt to their lobing pattern, which with tweeter on top is downwards. The backward slope corrects for the 15 degree tilt, so the acoustic axis is towards the listener, and it corrects for the 1/4 wave length time advancement in the crossover.

Thanks for that formula, even if I have used it before. As for small sweet spots, yes I'm quite familiar with that in my stereo rig.

 

[highfigh]

Thanks highfigh. I may have yet again successfully misremembered and/or misunderstood. The 15 deg is in regards to the lobing tilt with odd order xover.

Thanks for that formula, even if I have used it before. As for small sweet spots, yes I'm quite familiar with that in my stereo rig.

That 15° is a physical issue, not electrical and that's what I was responding to. The number I have always heard and read is that a crossover filter causes 90° per "order", so a 1st order causes 90°, a 2nd order causes 180° and so on. This is why a 2nd order crossover usually has the mid-range driver wired with the polarity reversed.

 

[3db]

I'm not a speaker designer but as an electrical engineer, I was introduced to filter circuits. Before talking about phase and even order of filter , one must know what kind of filter is being used as this dictates the roll off and phase of response of the filter.

Characteristics of a Butterworth Fillter:
http://en.wikipedia.org/wiki/Butterworth_filter

Characteristics of a Chebyshev Filter:
http://en.wikipedia.org/wiki/Chebyshev_filter

Charateristics of an Elliptic Filter:
http://en.wikipedia.org/wiki/Elliptic_filter

Crossover filters from what I read here use some form of the above and there are bound to their rules. For instance a 1st order butterworth (if I remember this correctly ) slopes down at 3db per octave and their phase shift is 45 degrees). A 2nd order butterworth slopes at 6db per octave with a phase of 90 degrees I don't know without rersearching what the characteristics are for the other filters but I vaguely remember something about Chebyshev 1st order is not as steep in the 1st order but the phase shift is smaller.

 

[TLS Guy]

This is a confused thread.

For a start there is confusion with tilt due to lobing error of odd order filters and phase delay and advancement of low and high pass filters. The tilt of odd order filters is 15 degrees off the horizontal axis. If the tweeter is on top the tilt is down, if on the bottom the tilt is up. Even order filters have symmetrical lobing patterns.

Now the phase error of a first order crossover with both low and high pass filters is 90 degrees. (45 for each filter).

For second order it is 180 degrees

For third order it is 270 degrees

For fourth order it is 360 degrees which is a whole wavelength at crossover frequency.

Now the high pass filter leads and the low pass filter lags.

So to calculate the distance by which the high pass filter leads the low pass you divide the velocity of sound by the crossover frequency and then divide that by four for first order, by 2 for second, by three for third and 1 for fourth.

However in a real world design things are seldom that simple. The reason being that because of acoustic driver slopes, the high and low pass filters nearly always have different orders and turnover frequencies to get a smooth transfer. The actual crossover slopes are the addition of the acoustic slopes of the drivers and the electrical slopes of the filters. However it gets more complex than that as because the voice coils of drivers are inductive devices, they have phase shifts also. So you seldom have neat phasing errors of 90, 180, 270 and 360 degrees.

As far as phase coherent time aligned speakers are concerned, we have had two threads discussing this recently.

I'm going to link to one of my posts in a recent thread on that issue.

 

[3db]

That 15° is a physical issue, not electrical and that's what I was responding to. The number I have always heard and read is that a crossover filter causes 90° per "order", so a 1st order causes 90°, a 2nd order causes 180° and so on. This is why a 2nd order crossover usually has the mid-range driver wired with the polarity reversed.

!st order Butterworths exhibit 90 degrees instead of 45 degrees. This could be as I Haven't seen filter theory in 30 years and my memory is a bit rusty. But be careful about filter generalizations. This much I do remember..Butterworrth exhibit 90 degrees per order but this is not true for the other filter types I've mentioned.

 

[3db]

This is a confused thread.

For a start there is confusion with tilt due to lobing error of odd order filters and phase delay and advancement of low and high pass filters. The tilt of odd order filters is 15 degrees off the horizontal axis. If the tweeter is on top the tilt is down, if on the bottom the tilt is up. Even order filters have symmetrical lobing patterns.

Now the phase error of a first order crossover with both low and high pass filters is 90 degrees. (45 for each filter).

For second order it is 180 degrees

For third order it is 270 degrees

For fourth order it is 360 degrees which is a whole wavelength at crossover frequency.
.[/URL]

What you arer stating about 90 degrees per order only holds true for Butterworth filters but not for the others I have mentioned. Are you implying than that crossover networks are built on Butterworth designs?

 

[Alex2507]

There are no simple answers.

I am doomed.