Speaker time alignment, does it matter?

[echopraxia]

If time alignment of left and right speakers is known to matter greatly, why does there seem to be the implication that time alignment between drivers doesn't matter much (or am I misreading that)? It seems hard to believe that it wouldn't matter the same if not more between drivers than between speakers.

 

[RayDunzl]

If time alignment of left and right speakers is known to matter greatly, why does there seem to be the implication that time alignment between drivers doesn't matter much (or am I misreading that)? It seems hard to believe that it wouldn't matter the same if not more between drivers than between speakers.

Timing difference left/right come to your left/right ear at different times - all the sound is delayed.

The typical speaker has the tweeter and woofer physically aligned vertically. Vertical time difference does not excite that (ITD) mechanism - the left/right still comes at the same time. Other mechanisms assist with vertical location sensing.

Maybe turn one speaker on its side while the other is still vertical, see what happens.

---

Snap your fingers while moving your hand horizontally around your head, then vertically, front to back. I find much more "difference" in the lateral direction.

 

[Matias]

I posted this in a speaker review thread, and it got buried, so perhaps it was best to give it a thread of its own.

As I had mentioned, when I was getting into audio, companies and designers like Snell, Fried, and Thiel, amongst other notables, were the preeminent builders. As such, I recall them having a focus on time alignment of drivers. With many of today's powered models using DSP, I would imagine that this is a non-factor. However, does it make sense to ask Amir to measure things like step response, phase response (same? https://bksv.com/media/doc/17-198.pdf), and acoustic distance/source in the basic battery of speaker testing just to provide an additional perspective?

Also, I'm seeing lots of CSD's being posted that appear to have ringing or long settling times in the lower frequencies but no one is discussing them (or I missed it). Is the CSD plot being provided considered inaccurate, or are problems like this no longer considered a problem, audibly? https://www.audiosciencereview.com/...er-csd-waterfall-audio-measurement-png.67181/

Help a newb out!

IMO, yes it does matter. That is one of the reasons I love my Thiel CS3.7 speakers.
Some active speakers have time alignment switchable on/off because it is easy to implement in DSP. I once heard a demo of Goldmund speakers and it was clearly noticeable. Anyone who experienced such a demonstration should confirm this.

Now of course frequency response on axis and off axis, cabinet ressonances, frequency extension and port noises also matter. Along with subjective qualities like soundstage, timbre, physical design and finishing, price and value, etc. So time alignment is 1 variable among many to weight when choosing a speaker.

 

[polmuaddib]

If your speaker baffle is not sloped like Thiel, Avalon etc... then you can raise the front part a little bit, make them sloped and listen. Simple experiment. If you hear a change in soundstage (messed up since it isn't designed for that) then time alignment might matter. Not really proof, but merits more research...

 

[echopraxia]

Timing difference left/right come to your left/right ear at different times - all the sound is delayed.

Vertical difference does not excite that mechanism - the left/right still comes at the same time. Other mechanisms assist with vertical location sensing.

Maybe turn one speaker on its side while the other is still vertical, see what happens.

---

Snap your fingers while moving your hand horizontally around your head, then vertically, front to back. I find much more "difference" in the lateral direction.

Of course we are more attuned to spatial difference horizontally than vertically, but I am speaking of time delays in different frequencies now (independent of their position in space). One could imagine adding a time delay to certain frequencies on a coaxial speaker: At some point, the delay must become audible, so the question is at what point does that occur?

I tried doing the same experiment while covering one ear, and I still could perceive a shift in the character of the sound. Of course, not a very scientific test since my other ear was probably still hearing something.

Does anyone know of any free software that would let me experiment with time delays of a subset of frequencies, to see how obvious it is? I'm curious now.

 

[RayDunzl]

Does anyone know of any free software that would let me experiment with time delays of a subset of frequencies, to see how obvious it is? I'm curious now.

@pkane is known for creating such toys.

 

[House de Kris]

I did a similar experiment as sergeauckland, but my results seem to be different. My tweeters can be moved front-to-back by about three inches. The crossovers between tweeter and midrange was at about 1.7kHz. As is my typical style, I listened first then measured. I used a six cycle tone burst at the crossover frequency, limited by a hamming window to reduce out of band energy. I could hear differences with just a quarter inch, or less, of movement of the tweeters. Later I used a microphone and scope and verified they were indeed in time alignment. In fact, using measuring gear I was unable to improve on the location I found by ear. From this I came to the belief that it is important.

Similar is a story of James Lansing in, I believe the 40s, when he was making theater speakers. He used flamenco dance recordings and moved the large tweeter horns astern and aft on the bass bins to make the foot stomping sound right (as in one cohesive stomp instead of fragmented and disjointed).

 

[MakeMineVinyl]

A Linkwitz-Riley crossover requires time alignment between drivers, either physically or with digital delay on one of the drivers.

 

[Another Bob]

Proper crossover design takes into account the difference in arrival time (if any) between the different drivers. If you move a driver without changing the crossover or the listening axis, the acoustic outputs will no longer sum as intended, and the frequency response will change. This is more audible than the actual arrival time difference. See the dips and peaks in the vertical off-axis response of multi-way speakers? Same thing.

 

[mhardy6647]

I did a similar experiment as sergeauckland, but my results seem to be different. My tweeters can be moved front-to-back by about three inches. The crossovers between tweeter and midrange was at about 1.7kHz. As is my typical style, I listened first then measured. I used a six cycle tone burst at the crossover frequency, limited by a hamming window to reduce out of band energy. I could hear differences with just a quarter inch, or less, of movement of the tweeters. Later I used a microphone and scope and verified they were indeed in time alignment. In fact, using measuring gear I was unable to improve on the location I found by ear. From this I came to the belief that it is important.

Similar is a story of James Lansing in, I believe the 40s, when he was making theater speakers. He used flamenco dance recordings and moved the large tweeter horns astern and aft on the bass bins to make the foot stomping sound right (as in one cohesive stomp instead of fragmented and disjointed).

This is the 'Eleanor Powell double tap' story --

https://books.google.com/books?id=t...r powell tap dancing western electric&f=false

'originally' quoted by me at
https://hifihaven.org/index.php?thr...r-powell-tap-dancing-problem.3323/#post-61269

 

[KaiserSoze]

I did a similar experiment as sergeauckland, but my results seem to be different. My tweeters can be moved front-to-back by about three inches. The crossovers between tweeter and midrange was at about 1.7kHz. As is my typical style, I listened first then measured. I used a six cycle tone burst at the crossover frequency, limited by a hamming window to reduce out of band energy. I could hear differences with just a quarter inch, or less, of movement of the tweeters. Later I used a microphone and scope and verified they were indeed in time alignment. In fact, using measuring gear I was unable to improve on the location I found by ear. From this I came to the belief that it is important.

Similar is a story of James Lansing in, I believe the 40s, when he was making theater speakers. He used flamenco dance recordings and moved the large tweeter horns astern and aft on the bass bins to make the foot stomping sound right (as in one cohesive stomp instead of fragmented and disjointed).

This is one of those things where an effect is easy to produce in well-controlled conditions using a test tone, but the question is then whether you would actually notice it when listening to music. If you put the two drivers a half-wavelength apart (4" for 1.7 kHz) you'll have full cancellation if you are equidistant from both drivers, and this would likely be detectable for particular music choices. But you heard a difference with 1/32 of a wavelength (.25"/8" = 1/32), and I'm not sure whether it would be possible to explain that without doing some math that I'm not inclined to do. It would be much easier to explain if you were listening very close and at a listening position in the vertical plane where distance to the tweeter diaphragm differed from the distance to the back of the midrange cone by something closer to 4".

 

[KaiserSoze]

A Linkwitz-Riley crossover requires time alignment between drivers, either physically or with digital delay on one of the drivers.

Does that mean that in order to get the same effect with two drivers that aren't time aligned you use asymmetric slopes?

 

[BenB]

As long as we're all telling stories, I'll tell one of mine:
When I was in college, I used DSP to improve the frequency and phase response of my speakers. (A pair of ported DIY MTM speakers with a passive crossover @ 2.2 kHz). I flattened the frequency response across the passband (about 40 Hz to 20 kHz), and made them phase linear. I verified the adjustments caused beautiful impulse responses (nearly perfect diracs). It made an obvious audible improvement. Later, I built another pair of speakers with a passive crossover @2kHz. These ones were nearly perfectly flat in the frequency domain with the passive crossover. I added digital correction to them, and verified the beautiful impulse responses... but this time there was no audible improvement. There wasn't much of any magnitude correction to do, and the phase correction just didn't matter. I loaded my correction impulse into foobar so I could turn it on and off instantly. It didn't make a difference. I moved my listening position close to the speakers in order to get more direct sound and less reflected sound. Nothing doing. Try as I might, I couldn't tell the difference, even with my young ears (at the time).
I was pretty disappointed since I had originally convinced myself that my DSP knowledge gave me the secrets to audio nirvana (this was back in the very early 2000s). I have a job in DSP that I've been doing for almost 20 years now, and I know very well how one might leverage DSP to "improve" the sound of a speaker... and I design passive crossovers.
Later I proved to myself (using foobar and ABX testing) that speakers with steep crossovers in the 200 Hz to 1 kHz range actually do make a (barely) discernible impact to the sound due to phase "distortion", but didn't find that there was any clear preference. Now phase isn't really something I concern myself with... magnitude and dispersion on the other hand...

 

[MakeMineVinyl]

Does that mean that in order to get the same effect with two drivers that aren't time aligned you use asymmetric slopes?

I don't know the answer to that question.

 

[GelbeMusik]

Does that mean that in order to get the same effect with two drivers that aren't time aligned you use asymmetric slopes?

I'll give the German again. Obviously many folks don't know what the buzzword "time alignment" is actually all about. You cannot understand the topic from the advertising, or with a premature focus on audio or even x-overs in particular. In oder to learn about time / frequency / wave etc one would take best advantage from re-reading the referring college script.

Only so much. If companies emphasize some time alignment effort, this is most often half the truth ( and the other way round ). In order to make a crossover work, time alignment is a precondition. So every xo which works is logically time aligned! Otherwise the drivers would not integrate, which would be seen in the frequency response--easily as big holes and hills.

What they usually refer to is a special time alignment, namely all drivers radiating kind-of in phase. This is especially not good. It narrows the vertical dispersion with, as can be seen as standard, vertically arranged drivers. This again spoils the balance in frequency space. That is why so called "time aligned" speakers measured so badly in the past.

What can be done today is to correct the phase of the grand total output of all individual drivers together digitally. This makes the group delay--whatever that is--constant over much or all of the frequency band. But, except for few of the coaxial speakers, this holds only for a single point in space. At all other points the constant group delay is not. So far it seems to be a more idealistic than reasonable effort.

Second there is very strong evidence, that the human hearing is not capable to detect, or more so is annoyed by non constant group delay within bounds. To hit or overrun those bounds is quite a hard task to accomplish. Regular designs are safe. Typical, and therefore I think the ancient Klipsch experiment, it was common in non-digital times with long horns. But then the frequency response was heavily compromised too.

After this German epic, or sermon if You will, it is all in the college script I mentioned.

 

[napilopez]

Wow, I just tried adding a 5ms delay to one of my Genelecs and the soundstage changes HUGELY. What was pinpoint imaging of a voice in front of me became smeared wider, and shifted to the side. It's hard to describe, but it felt like the voice shifted in position to my right (I delayed my left speaker), but without the same precise pinpoint imaging that it had before -- sounds almost like an echo or larger/wider soundstage, weirdly enough.

I tried with 1ms, and it's also extremely easy to perceive the difference. So what does this mean when you say "anything <7ms is treated as a single sound"? I am also confused by this statement of what it's trying to imply -- are you saying time domain is important, or is not?

The precedence/haas effect in action! If you have Toole's book, I suggest you read/re-read section 7.6.4 and 7.6.5 (at least in the latest edition). It basically explains everything you heard.

The time domain stuff has an audible effect, but it's not always clear its better or worse. It's also much easier to spot in direct comparisons because it breaks the adaptaptation our brains constantly automatically make for delayed sounds.

The reason wide directivity speakers sound 'bigger' is precisely because the delayed reflected sounds are combining with the primary sound to create image shift/spreading like you are describing. In your case it's more dramatic because you are tweaking the delay for speakers at the same SPL, but nonetheless that's the principle at work. If you were in a non-reflective environment, (the original Haas study was done on a roof) the sound would likely shift towards the right completely.

If time alignment of left and right speakers is known to matter greatly, why does there seem to be the implication that time alignment between drivers doesn't matter much (or am I misreading that)? It seems hard to believe that it wouldn't matter the same if not more between drivers than between speakers.

This is a very different scenario though. The precedence effect has to do with localization, so as you described, delaying one speaker shifts the image towards the other speaker. It's a very obvious effect. But crucially, the precedence effect does not mask the other aspects of the delayed sounds. They still contribute to timbre and whathaveyou. So you hear everything, just it gets shifted towards the position of the original sound.

 

[GelbeMusik]

If time alignment of left and right speakers is known to matter greatly, why does there seem to be the implication that time alignment between drivers doesn't matter much (or am I misreading that)? It seems hard to believe that it wouldn't matter the same if not more between drivers than between speakers.

This is an interesting question. It goes back to the ( quite faulty ) foundations of stereo as a concept. It can be shown, that the human hearing reacts strongly to inter-aural time differences. That is the difference in time, at which some sound arrives at the left and right ear. The nearly same effect can be achieved by exposing the ears, again left and right to sound of different volume. These differences, time and volume, are perceived as an estimate on the direction, from which the sound comes in. Some physical phenomenon is used by the senses as to construct a picture of the surrounding.

This trained ability of the hearing is used with stereo, but very imperfectly. The often told theory behind it is hilarious nonsense. It assumes a person who is hearing straight forward and never moves its head. The audiophile in this delightfull stereo situation cannot easily be distinguished from a manequin. Actually, some folks tries to follow the rules

O/k, what stereo does is to induce and exploit those mentioned inter-aural differences in the studio as to mimic directional "information" by treating the "signal". But because the theory never holds, the sound engineer / mixer will better test listen to the recording. Only once she is satisfied with her settings, after many correctiions for good, it is published.

Stereo never ever reproduces the original sound field!

Back to Your question. The quite artifical situation of stereo listening is shown to not react to phase lag, excessive group delay, time misalignment, if they are identical in both speakers used. That is a matter of fact. It holds within some uncertain bounds, sure. But the limits are never hit with regular quality designs. Take it for granted, that You will not experience any drawbacks from phase eq. egd eq. tma etc whatever they name it. As long as it is the same for both speakers. It seems the hearing doesn't use the related physical phenomenon for creating a picture of the outside world.

Why is it all so complicated? Because stereo addictives from audiophilia often cite scientific investigations, which are not: not focussed on or even related to stereo. Group delay of so and so is perceptable blah blah and so loudspeaker have to ... (put in some hilariously excessive requirement). Such conclusions are plain wrong. They lack "the middle piece" in their derivations. They secretly assume that stereo has to reproduce the "signal", which is the original sound field. They say, because it is not the original "signal" the "information" is lost. Plain wrong, stereo re-invents the "information" on direction by introducing artificial add-ons to some excerpt of the "signal".

And that can be done and is only done as a difference between the two channels of the program material. As long as the intrinsic phase/volume ondulations of each speaker of a re-producing stereo pair are kept the same, these are irrelevant.

 

[KaiserSoze]

I'll give the German again. Obviously many folks don't know what the buzzword "time alignment" is actually all about. You cannot understand the topic from the advertising, or with a premature focus on audio or even x-overs in particular. In oder to learn about time / frequency / wave etc one would take best advantage from re-reading the referring college script.

Only so much. If companies emphasize some time alignment effort, this is most often half the truth ( and the other way round ). In order to make a crossover work, time alignment is a precondition. So every xo which works is logically time aligned! Otherwise the drivers would not integrate, which would be seen in the frequency response--easily as big holes and hills.

What they usually refer to is a special time alignment, namely all drivers radiating kind-of in phase. This is especially not good. It narrows the vertical dispersion with, as can be seen as standard, vertically arranged drivers. This again spoils the balance in frequency space. That is why so called "time aligned" speakers measured so badly in the past.

What can be done today is to correct the phase of the grand total output of all individual drivers together digitally. This makes the group delay--whatever that is--constant over much or all of the frequency band. But, except for few of the coaxial speakers, this holds only for a single point in space. At all other points the constant group delay is not. So far it seems to be a more idealistic than reasonable effort.

Second there is very strong evidence, that the human hearing is not capable to detect, or more so is annoyed by non constant group delay within bounds. To hit or overrun those bounds is quite a hard task to accomplish. Regular designs are safe. Typical, and therefore I think the ancient Klipsch experiment, it was common in non-digital times with long horns. But then the frequency response was heavily compromised too.

After this German epic, or sermon if You will, it is all in the college script I mentioned.

Okay, to be truthful I wasn't able to follow all that, even though I read it four or five times.

My understanding is that what makes the L-R crossover slopes special is that the two drivers remain phase-coherent throughout the overlap region notwithstanding the continuous shift in phase that occurs in each of the two outputs generated by the crossover. I am entirely certain that phase-coherency is the true and correct property that L-R crossovers preserve. The reason I am sure of this is that the desired effect is manifestly for the two wavefronts to be mutually phase-coherent such that there will be no destructive interference and no cancellation when the two wavefronts merge into a single wavefront. It could not possibly be more manifest that phase-coherency is the property preserved by the L-R crossover, and as such, it is similarly manifest that phase-coherency is also the property on which they depend, i.e., the crossover cannot preserve a property that does not exist to begin with. I have not said anything there that I would have thought less than obvious to anyone who is likely to read this. Similarly, were I to point out that time-alignment is neither a necessary condition nor a sufficient condition for assurance of phase-coherency prior to the effect of the L-R crossover, I would be stating something that I would regard as virtually self-evident. It isn't a sufficient condition, because, for one example, the inductance of the coil introduces a frequency-dependent phase shift. Neither is time-alignment a necessary condition, because the phase offset that occurs when the two drivers are not equidistant from the listening position can be corrected via a phase offset introduced either electrically or mechanically.

But you wrote:

In order to make a crossover work, time alignment is a precondition. So every xo which works is logically time aligned!

I would probably have been less bothered by that had you not written it right after you wrote this:

Obviously many folks don't know what the buzzword "time alignment" is actually all about. You cannot understand the topic from the advertising, or with a premature focus on audio or even x-overs in particular. In oder to learn about time / frequency / wave etc one would take best advantage from re-reading the referring college script.

Your statement about what makes a crossover work implicitly defines "time alignment" to have the meaning that is rightfully the meaning of "phase coherency". Then you characterized phase-coherency as a special sort of time alignment:

What they usually refer to is a special time alignment, namely all drivers radiating kind-of in phase. This is especially not good. It narrows the vertical dispersion with, as can be seen as standard, vertically arranged drivers. This again spoils the balance in frequency space. That is why so called "time aligned" speakers measured so badly in the past.

And this confused me even more. How could it be "especially not good" for the two drivers to be phase-coherent given that, as concerns L-R crossovers, the reason that time alignment ostensibly matters is that time-alignment is ostensibly necessary in order that the two wavefronts will be coherent? And pardon me if I misinterpret what you are meaning to say, but I believe that you are saying that if the two drivers are phase-coherent, that this will squash the main lobe in the vertical plane. Why would phase-coherency do that? The main lobe in the vertical polar plane is defined by the locus of the nulls where the two wavefronts are 180 degrees out of phase. Why would phase coherency of the two wavefronts induce destructive interference at polar angles (measured from horizontal) that are smaller than the polar angles at which complete destructive interference will occur if the two wavefronts aren't phase-coherent?

I read the rest of it, several times. But I've already written more than I meant to and more than I should have.

 

[JungleXray]

Proper crossover design takes into account the difference in arrival time (if any) between the different drivers. If you move a driver without changing the crossover or the listening axis, the acoustic outputs will no longer sum as intended, and the frequency response will change. This is more audible than the actual arrival time difference. See the dips and peaks in the vertical off-axis response of multi-way speakers? Same thing.

So just so I’m clear on this, a properly designed passive crossover counld be designed with time alignment in mind? Not with a stepped or sloping baffle?

 

[mk05]

Why speaker time alignment doesn't work - by PS Audio