Time aligned speakers - do they make sense?

[Holmz]

With non-coincident drivers, any speaker can only be phase- or time-aligned at one particular axis, regardless of whether the alignment is done electronically or physically. This is why the vertical off-axis frequency response looks so bad. I consider a speaker to be phase-aligned on the axis where the frequency response through the crossover region is smoothest. As mentioned above, it is time-aligned when the step response is optimized. I have time-aligned speakers using electronic crossovers, and I think there is subtle, but noticeable improvement, which I notice mostly as improved imaging.

I’ll take an on-axis perfect step response over not being correct anywhere.

 

[HarmonicTHD]

With all due respect to Toole...I'm totally unconvinced of his conclusions regarding phase audibility.
Certainly the uniform standard-type speakers that show up in the book, didn't have the capability to demonstrate flat phase across the board, my guess doubly so when a sub was added.

Is this your personal opinion or do you have evidence or other sources?

The studies of

Dr. Floyd Toole, Dr. Wolfgang Klippel, Andrew Jones, and James Croft​

which „thewas“ linked above, all say pretty much the opposite to your statement.

 

[Mnyb]

Hopefully it can be done digitally nowadays.

the “old way” of using first order passive filters has so many other drawbacks imho, that i’m not sure if it was worth doing ? Very demanding on the drivers .

I’ve had a pair of small theil speakers and they where good , but they where well made in other aspects to.

 

[eddy555]

well only first order filters (min phase) coupled with offset acoustic centers can give you a truly time coherent Right triangle like looking step response ...

 

[KSTR]

well only first order filters (min phase) coupled with offset acoustic centers can give you a truly time coherent Right triangle like looking step response ...

I sense a lot of confusion here.
Time coherency is not the commonly used technical wording and it is ill defined.

Linear phase crossover is the correct nomenclature.
Linear phase crossovers are possible with other targets than first order acoustic(!) slopes as well. First order is just widely known and popular.
With aligned acoustic centers, though, the higher order solutions suffer from significant cancelling around XO frequency as phase is more than 120 degrees apart (with first order we have 90 degree phase offset).

With the tweeter set back further than acoustic center alignment would require and with the help of some allpass networks it is possible to create linear phase crossovers of fourth order which don't suffer from excessive phase difference or even destructive summing.

In general, IME the most important aspect of an XO for "timing precision" is exact phase tracking between ways, not resultant total phase of the sum.
The typical phase offset target is zero which gives a Linkwitz-Riley crossover in case of an isolated single XO frequency.
This phase offset shall be maintained around XO point until there is more then 40dB level difference between ways.
Other phase offset are viable targets as well, I often use a 60 degree target. 90 degree target gives the know odd-order Butterworth XO (again, for an isolated single XO frequency).

As for sound, making a good speaker with tight phase offset tracking linear phase is the icing on the cake, because the tight phase tracking already makes for a good speaker even though the step response doesn't look too pretty.
Fixing a crippled speaker (with bad phase tracking) with phase pre-correction does help but will never achieve the quality of the well-aligned speaker even without phase pre-correction.
I always design XO from the phase offset target because that's what important. Magnitude target then follows automatically and is just dialed in globally.

Once you have a fixed phase tracking, it is a piece of cake to turn the speaker into a linear phase system by corrective phase EQ of the source signal (so, globally on the signal feed).

 

[gnarly]

Is this your personal opinion or do you have evidence or other sources?

The studies of

Dr. Floyd Toole, Dr. Wolfgang Klippel, Andrew Jones, and James Croft​

which „thewas“ linked above, all say pretty much the opposite to your statement.

It's my personal opinion, from my own tests/evaluations, and reading every paper/study I can find on phase audibility.

I've read the AudioExpress article, and give the author credit for admitting his starting-point biases.
I've read most of the references quoted in that article.

I am 100% convinced that no one truly knows the extent of phase audibility, because sufficient testing has yet to be accomplished.

All past studies that I've found have a fatal flaw, in that they compare various degrees of phase rotation or group delay to each other.
The comparison needs to be ZERO phase rotation/group delay (linear phase), vs various degrees or rotation.
Full-range loudspeaker, linear phase reproduction has not been historically possible until rather recently for such tests.
And It will take some very careful statistical work, using such loudspeakers, to truly isolate phase audibility.

Headphone studies of all-pass vs not, prove nothing imo, as headphones fail to impart the impact of bass, and how it is perceived not only by the ears, but also via tactile body sensations..

The prosound community is clearly more onboard with the idea that phase needs more study and is very important to SQ.
Like the AudioExpress article starts off, linear phase is showing up in prosound studio monitors. Meyer's Bluehorn is an example.
Martins MLA, and EAW's Anya, are example of large scale line arrays that use a great deal of phase control on individual drivers for the purpose of nearly incredible beam steering.
Anybody gonna tell me the entire immersive audio kick doesn't depend on exact timing...?
Or that acoustic noise cancellation doesn't defend on timing down to the phase level?

Like said, I don't think anyone knows how audible phase is....the good science has yet to be done.
But for me, until good science proves otherwise, I say how can phase not matter?

i say all we really have is the frequency domain, and the time domain.
And even the frequency domain is rooted in time.
How can folks dismiss the time domain claiming we can't hear it?
Is our inability to hear the time domain just part of the recording circle-of-confusion? Will it become more apparent when recordings are mastered on linear-phase studio monitors?
And then replayed on linear-phase home speakers?
Who knows?

 

[gnarly]

Linear phase crossover is the correct nomenclature.
Linear phase crossovers are possible with other targets than first order acoustic(!) slopes as well. First order is just widely known and popular.

Indeed, the graphs I've posted in this thread of the 4-way MEH, are using 16th order xovers, but with zero phase rotation being linear phase.

(Would be 1440 degrees of phase rotation if IIR....which would clearly be an enormous non starter LOL)

 

[skoch]

It could be important but it's more about phase than time. If the waves from the tweeter and midrange (or woofer) are out-of-phase at the crossover frequency they will cancel and you'll get a dip in the response.

...Just checking with an online calculator, 5kHz has a wavelength of 2.7 inches. So if the midrange is 2.7 inches behind the tweeter, the tweeter will lag by exactly one wavelength and they are back in-phase. If there is a 1.35 inch difference that's a half-wavelength and they will be out-of-phase.

Exactly one-half wavelength is 180 degrees so that could easily be corrected by reversing the connections to one of the drivers. Also, crossovers often introduce a phase-lag in the low-pass and a phase-lead in the high-pass so that should be taken into account.

...You'd also have to calculate the effective radiation plane from the drivers, which are usually a cone or dome so it's not so obvious. (The manufacture should know this.)

At lower frequencies (longer wavelengths) it's not as critical since there will be smaller phase differences.

But at lower frequencies you can hear differences in phase more. An out of phase crossover region at 100hz between 2 drivers can be heard very clearly, but at 2khz i don't think anybody in the world can hear a phase difference. Even a 180 degrees difference.

 

[ernestcarl]

But at lower frequencies you can hear differences in phase more. An out of phase crossover region at 100hz between 2 drivers can be heard very clearly, but at 2khz i don't think anybody in the world can hear a phase difference. Even a 180 degrees difference.

Why don't you test it for yourself with Foobar2000's ABX plugin. It shouldn't be too hard to do. Obviously, the conditions would not be the same as other people's setup, but the conclusion should be applicable to you, at least.

 

[ernestcarl]

how can phase not matter?

My takeaway is that it does matter... but it seems to me it's much more a question of how important it really is (i.e. achieving "zero phase") over other things/challenges that also need to be addressed.

 

[levimax]

Using 16th order / brickwall filters to improve driver alignment and phase brings up the question of which is more audible phase issues or filter ringing neither of which are well studied as far as I can tell.

 

[MakeMineVinyl]

My takeaway is that it does matter... but it seems to me it's much more a question of how important it really is (i.e. achieving "zero phase") over other things/challenges that also need to be addressed.

Dialing in good time coherence is easier done with DSP than improving the SINAD of of your DAC/amplifier/preamp. If someone has the capability, I'd certainly do it.

 

[ernestcarl]

Dialing in good time coherence is easier done with DSP than improving the SINAD of of your DAC/amplifier/preamp. If someone has the capability, I'd certainly do it.

I do it with all my speakers that are connected to a PC and stereo/MCH DAC. It seems rather a waste not to when the capability is already there...

 

[valerianf]

Time alignment of the tweeter is very important in the crossover midrange-tweeter section:

Time Alignment Part Three: Delays and Crossovers for Tweeters and Mids - Audiofrog

In part two, we discussed comb filtering and how it affects both frequency response and imaging. Comb filtering is caused when two sources of the same sound are different distances […]

www.audiofrog.com

 

[KSTR]

Indeed, the graphs I've posted in this thread of the 4-way MEH, are using 16th order xovers, but with zero phase rotation being linear phase.

(Would be 1440 degrees of phase rotation if IIR....which would clearly be an enormous non starter LOL)

Maybe a misunderstanding here, I was talking about linear phase analog (or even passive) higher order crossovers... the sum is linear phase, the individual ways are not (and cannot be) linear phase.

 

[gnarly]

Using 16th order / brickwall filters to improve driver alignment and phase brings up the question of which is more audible phase issues or filter ringing neither of which are well studied as far as I can tell.

Glad you brought this up....
provides a good chance to help dispel yet another oft repeated internet myth.....

A distinction MUST be made between linear phase xovers, and one sided linear phase high pass or lowpass filters.

Fully complementary linear phase xovers do not have pre=ring potential. Whereas linear phase high pass of low pass filters used alone do.
Iow, xovers are cool.
System high pass or low pass, or adjusting drivers' phase independently of frequency response, are not cool.

Here is a graph to illustrate. It's the same 4-way MEH I've been posting in this thread.

The green trace has a Butterworth 18dB/oct minimum phase high pass filter, @100Hz.
The blue trace has a 96dB/oct linear phase high pass filter, @100Hz.

Note the traces when preceeding time 0.
Linear phase hpf (blue) shows huge potential for pre-ring: That's the dive below zero.
Minimum phase hpf (green), none..

Now, both of the the traces include three 96dB/oct linear phase xovers at 300, 750, and 6300Hz.
Where is the pre-ring in the green trace? Not any.
It's only the system hpf, in this case the 100Hz hpf, that can cause pre-ring.

Now, if the linear phase (blue) is combined with a sub having a 100Hz linear phase low pass filter, the step response will change to look just like green. Symmetric cancellation of pre-ring poetential will occur. (and is what i do in practice when adding sub)

Fully complementary linear phase xovers, not matter how steep or shallow, cancel each others sides pre-ring potential.

If DIY folks fully grasped this, i imagine they'd be jumping up and down in joy, realizing the ease of active xover implementations now available
(without the oft repeated internet myth regarding pre-ring.)

In fairness and full discloser, a case can be made that any lack of fully complementary acoustic xover, can result in pre-ring, because the two sides don't completely cancel each other acoustically. But this is only possible in the xovers' summation regions and where lobing potential is high.
In practice, I've found using high order/or brick wall, narrows the xover range so well, this potential issue becomes a bit of a joke.

I honestly don't know if lower order linear phase xovers raise the potential, as I get inferior polar response with them compared to steep, and don't bother with low order anymore...

Hope this all made sense....

 

[gnarly]

Maybe a misunderstanding here, I was talking about linear phase analog (or even passive) higher order crossovers... the sum is linear phase, the individual ways are not (and cannot be) linear phase.

Thx for clarification.

Curious why you see a distinction between linear phase analog (or passive) and digital. Seems to me the method of implementation doesn't matter at all, other than the odds of actually achieving linear phase.
When I've looked at analog implementations using banks of all-pass filters, it appears anything beyond low order to be a pipe dream.

In my mind, it is fair to characterize individual drivers, as being able to achieve linear phase.
The key is linear phase within their intended summation range, which for me is to -30 to -40dB.

If out of band minimum phase flattening is first applied to raw driver's response, and then steep linear phase xovers are added,
it is pretty easy to keep phase linear all the way to/thru the -30 to-40dB frequency points.
So effectively, the driver becomes linear phase in my mind. (Admittedly with the need for summation with another lin phase driver as per my just prior post.)

 

[levimax]

Glad you brought this up....
provides a good chance to help dispel yet another oft repeated internet myth.....

A distinction MUST be made between linear phase xovers, and one sided linear phase high pass or lowpass filters.

Fully complementary linear phase xovers do not have pre=ring potential. Whereas linear phase high pass of low pass filters used alone do.
Iow, xovers are cool.
System high pass or low pass, or adjusting drivers' phase independently of frequency response, are not cool.

Here is a graph to illustrate. It's the same 4-way MEH I've been posting in this thread.

The green trace has a Butterworth 18dB/oct minimum phase high pass filter, @100Hz.
The blue trace has a 96dB/oct linear phase high pass filter, @100Hz.

Note the traces when preceeding time 0.
Linear phase hpf (blue) shows huge potential for pre-ring: That's the dive below zero.
Minimum phase hpf (green), none..

Now, both of the the traces include three 96dB/oct linear phase xovers at 300, 750, and 6300Hz.
Where is the pre-ring in the green trace? Not any.
It's only the system hpf, in this case the 100Hz hpf, that can cause pre-ring.

Now, if the linear phase (blue) is combined with a sub having a 100Hz linear phase low pass filter, the step response will change to look just like green. Symmetric cancellation of pre-ring poetential will occur. (and is what i do in practice when adding sub)

Fully complementary linear phase xovers, not matter how steep or shallow, cancel each others sides pre-ring potential.

If DIY folks fully grasped this, i imagine they'd be jumping up and down in joy, realizing the ease of active xover implementations now available
(without the oft repeated internet myth regarding pre-ring.)

In fairness and full discloser, a case can be made that any lack of fully complementary acoustic xover, can result in pre-ring, because the two sides don't completely cancel each other acoustically. But this is only possible in the xovers' summation regions and where lobing potential is high.
In practice, I've found using high order/or brick wall, narrows the xover range so well, this potential issue becomes a bit of a joke.

I honestly don't know if lower order linear phase xovers raise the potential, as I get inferior polar response with them compared to steep, and don't bother with low order anymore...

Hope this all made sense....

View attachment 206986

Very interesting, I learned a lot, thank you. I am still a little fuzzy on some details. After reading your post my understanding is that linear phase crossovers don't ring because the ringing is cancelled out by the two filters at the crossover point. So here a few questions. How about minimum phase crossover filters and post ringing? Seems like they would be more prone to ringing? It is popular with DIY digital crossovers to not use the exact same slope or even the exact crossover point in order to "blend" the drivers together from flatter FR but with what you are saying this would be a bad idea from the "ringing" standpoint. Is that right? Do you have any insight into how audible "ringing" is for either crossovers or independent high and low pass filters? Thanks

 

[MakeMineVinyl]

Time alignment? Who needs time alignment?

 

[fineMen]

Time alignment of the tweeter is very important in the crossover midrange-tweeter section:

Time Alignment Part Three: Delays and Crossovers for Tweeters and Mids - Audiofrog

In part two, we discussed comb filtering and how it affects both frequency response and imaging. Comb filtering is caused when two sources of the same sound are different distances […]

www.audiofrog.com

If You see a slanted front baffle, and the manufacturer swaggers about "time blah", that for sure tells, that the cross-over isn't well done.
Well done is a correct phase alignment, with a main lobe ( https://en.wikipedia.org/wiki/Acous...bination thereof) is not practically possible. ) aiming at the listener, and the lobe being wide especially towards the ceiling.