"The secret of big speakers"

[Cosmik]

Linkwitz on bass reflex:

In a vented box you add a second resonant structure, which is formed by the compliance of Vb and the air mass in the vent. The driver&Vb resonator and the Vb&vent resonator are coupled acoustically. Energy is transferred between them and used to extend the low frequency response, but at a steeper cut-off rate. The large amount of stored energy for the vented box is indicated by the group delay. It leads to ringing in the time domain and a characteristic bass sound.

Probably 90% of all consumer and professional loudspeakers are vented. Thus this type of bass coloration has become the accepted norm and just because for a given box size the -3 dB corner can be pushed down further (increasing the group delay even more), than when a box is sealed. Bass quantity sells over quality. It's a male thing.

 

[pos]

It does rings like any other 4th order filter with the same Q would.

 

[KSTR]

Linkwitz on bass reflex:
[...]
The large amount of stored energy for the vented box is indicated by the group delay. It leads to ringing in the time domain and a characteristic bass sound
[...]

Sure, but what is your point? The ringing is there from the group delay and any system with the same group delay will also have the same ringing. The ringing is *not* an additional ill-effect of the speaker being ported.

Again, always remember: with minimum phase systems, in their frequency response magnitude is strictly tied to a corresponding phase and vice versa (and hence to group delay as it is a derivative of the phase).

 

[svart-hvitt]

Linkwitz on bass reflex:

For completeness, you should add AES Fellow John Watkinson, who is strongly against ported designs:

https://www.resolutionmag.com/wp-content/uploads/2016/02/Getting-the-bass-right.pdf

 

[Cosmik]

It does rings like any other 4th order filter with the same Q would.

But how have we got here? We have gone from wanting to make a cone move with a certain acceleration at a certain time, to a 4th order filter with Q. No one seems to think that that's a bit odd: to go from trying to reproduce a time domain waveform and ending up with a resonant system that rings.

 

[KSTR]

As a speaker designer you have to make compromises, depending on the application. Today, it is quite easy to undo the phase rotation of an eg. 4th order system (be it ported or whatever) to any desired response at zero cost and with little side effects, if that ever poses a problem (it sure does when the corner frequency is high as typically found in small bookshelf speakers). So you can have best of both worlds.

 

[oivavoi]

If we assume that we

For completeness, you should add AES Fellow John Watkinson, who is strongly against ported designs:

https://www.resolutionmag.com/wp-content/uploads/2016/02/Getting-the-bass-right.pdf

Assuming Watkinson/Linkwitz/@Cosmik are right about ported speakers. Would that criticism still apply if one uses a pair of ported speakers together with a subwoofer, with a steep crossover to the sub at a high frequency, for example 120 hz? (I know Watkinson thinks full range speakers are better, but for the sake of the argument)

 

[andreasmaaan]

But how have we got here? We have gone from wanting to make a cone move with a certain acceleration at a certain time, to a 4th order filter with Q. No one seems to think that that's a bit odd: to go from trying to reproduce a time domain waveform and ending up with a resonant system that rings.

The same could be said of sealed boxes though. Why do we introduce a new element that creates a 2nd order filter with Q?

The answer is obviously that the alternative is (arguably) worse.

 

[KSTR]

Assuming Watkinson/Linkwitz/@Cosmik are right about ported speakers. Would that criticism still apply if one uses a pair of ported speakers together with a subwoofer, with a steep crossover to the sub at a high frequency, for example 120 hz? (I know Watkinson thinks full range speakers are better, but for the sake of the argument)

That's actually the worst thing you can do, introducing excess phase rotation without need in a freq range where it is most audible. Even with a "fast" sealed sub tuned low this going to ruin the transient behaviour....and the steeper the crossover the more so...
Subs, unless crossed over linear-phase and with a perfect crossover function to begin with, are always detrimental to timing and speed. The "punch" of kick drums and plucked upright bass notes is always lagging behind their attack phase with this.

 

[oivavoi]

That's actually the worst thing you can do, introducing excess phase rotation without need in a freq range where it is most audible. Even with a "fast" sealed sub tuned low this going to ruin the transient behaviour....and the steeper the crossover the more so...
Subs, unless crossed over linear-phase and with a perfect crossover function to begin with, are always detrimental to timing and speed. The "punch" of kick drums and plucked upright bass notes is always lagging behind their attack phase with this.

Interesting, thanks! Do you have any links to articles/information about this issue? It's actually the first time I hear this claimed, but it does make sense.

But if one wants to try to supplement bookshelf speakers with a subwoofer (or subwoofers), what do you think is the best way of doing it?

 

[svart-hvitt]

Interesting, thanks! Do you have any links to articles/information about this issue? It's actually the first time I hear this claimed, but it does make sense.

But if one wants to try to supplement bookshelf speakers with a subwoofer (or subwoofers), what do you think is the best way of doing it?

Genelec crosses over at about 85 Hz, using DSP (GLM) for phase and timing. This method made it impossible to locate the sub once I tried it.

You mentioned cross at 120 Hz, which is (way) too high, I guess.

 

[KSTR]

Interesting, thanks! Do you have any links to articles/information about this issue? It's actually the first time I hear this claimed, but it does make sense.

This is just the normal allpass behaviour that should be well known: an allpass doesn't alter magnitude but introduces a non-flat phase response (an according group delay), and a standard crossover has an allpass response in the sum of the individual drivers. And the more of these allpasses you have, the higher their order and lower their frequency the more the sonic reproduction falls apart in the time domain: treble first, midrange next and bass last.

But if one wants to try to supplement bookshelf speakers with a subwoofer (or subwoofers), what do you think is the best way of doing it?

Design a linear phase crossover for it, preferably one with a very gradual transition before reaching the final steep -- or even infinite -- slopes. Sharp transitions generate a lot of ringing that can only cancel on-axis (if properly designed, if not, there's ringing even on-axis) and only to first order, eg. any distortion components don't cancel. And with lin-phase XOs the ringing has a pre-ringing and a post-ringing part, with the pre-ringing being more offensive as it is masked less.

 

[Cosmik]

A way of thinking of this..?

Supposing you made a speaker that comprised an actuator (electromagnetic plunger type thing) and this vibrated an array of tuning forks attached to a sheet of wood, ranging from low frequency to high. Now, for any sinusoid you cared to put in, you would get a sinusoidal output (let's say that the system is perfectly linear) that could be measured with a microphone. It would have a phase and an amplitude. If you didn't like the phase and amplitude relative to the signal you could use DSP to change both before you sent the signal to the transducer, thus resulting in a 'perfect' response.

In the frequency domain, with steady state waveforms, this would look like the perfect speaker system. The problem would be that each tuning fork would take time 'to get going', and when the signal went away it would take time 'to stop'. Clearly, with real music it would be hopeless, giving the fabled 'time smear' in a very big way.

Bass reflex is like a 'single tuning fork' speaker, where the tuning fork is stimulated by the back wave from the bass driver that is in the process of rolling off at its lower end, giving an extra bit of bass output centred on that tuning fork's frequency. Of course, the resonance (ringing) and selectivity of the 'tuning fork' is much less than a real, actual tuning fork. But the same principle applies: a steady state measurement shows a phase and amplitude response that can be made perfect by shifting the phase and changing the amplitude with DSP - in the frequency domain it looks as though the port is a 'free lunch'. But like a tuning fork (but much less) it takes time 'to get going' and time 'to stop', resulting in extra 'time smear' of transients - which most music is made of.

 

[andreasmaaan]

For completeness, you should add AES Fellow John Watkinson, who is strongly against ported designs:

https://www.resolutionmag.com/wp-content/uploads/2016/02/Getting-the-bass-right.pdf

I don't mean any disrespect to John Watkinson, but that article makes a lot of claims about the audibility of group delay that fly in the face of fairly strong evidence. And his opposition to ports is entirely based on their group delay characteristics, despite the strong evidence suggesting this is inaudible.

He also says things like:

"...during the onset of a sound the ear works in the time domain and uses the waveform to establish the size and location of the source. It is this aspect of perception that requires loudspeakers to be phase linear."

This position is unsupported by evidence, while there is plenty of evidence to the contrary. I don't place much weight on his view on this particular topic.

That's actually the worst thing you can do, introducing excess phase rotation without need in a freq range where it is most audible. Even with a "fast" sealed sub tuned low this going to ruin the transient behaviour....and the steeper the crossover the more so...
Subs, unless crossed over linear-phase and with a perfect crossover function to begin with, are always detrimental to timing and speed. The "punch" of kick drums and plucked upright bass notes is always lagging behind their attack phase with this.

Is there any evidence you base this opinion on?

To my knowledge, it isn't supported experimentally. Even using impulse responses as the stimulus and headphones as the test apparatus, group delays in this range of magnitude and frequency have proven thus far to be undetectable (to my knowledge) despite many studies having investigated it.

For example,the study @oivavoi posted in post #99 found that group delays of this magnitude and in this frequency range (a 4th order xover at 120Hz will generate under 6ms of group delay) were undetectable, even when the stimulus was an impulse response and even when subjects listened through headphones.

A loudspeaker in a room and music as the source is well-established to create conditions in which subjects are far less sensitive to these effects than the conditions in that study (and numerous others), so if they are undetectable on headphones using impulses, they will not be detectable in speakers playing music.

 

[svart-hvitt]

I don't mean any disrespect to John Watkinson, but that article makes a lot of claims about the audibility of group delay that fly in the face of fairly strong evidence. And his opposition to ports is entirely based on their group delay characteristics, despite the strong evidence suggesting this is inaudible.

He also says things like:

"...during the onset of a sound the ear works in the time domain and uses the waveform to establish the size and location of the source. It is this aspect of perception that requires loudspeakers to be phase linear."

This position is unsupported by evidence, while there is plenty of evidence to the contrary. I don't place much weight on his view on this particular topic.



Is there any evidence you base this opinion on?

To my knowledge, it isn't supported experimentally. Even using impulse responses as the stimulus and headphones as the test apparatus, group delays in this range of magnitude and frequency have proven thus far to be undetectable (to my knowledge) despite many studies having investigated it.

For example,the study @oivavoi posted in post #99 found that group delays of this magnitude and in this frequency range (a 4th order xover at 120Hz will generate under 6ms of group delay) were undetectable, even when the stimulus was an impulse response and even when subjects listened through headphones.

A loudspeaker in a room and music as the source is well-established to create conditions in which subjects are far less sensitive to these effects than the conditions in that study, so if they are undetectable on headphones using impulses, they will not be detectable in speakers playing music.

@andreasmaaan , I follow you and agree. That’s been my critique of JW as well. However, I like his theoretical reasoning. It’s beautiful, like @Cosmik ’s theorizing

It’s a shame nobody on ASR has heard his speakers (http://www.legendloudspeakers.com/). They should tick all of the boxes and answer all the shortcomings laid out in this JW article:

https://www.thebroadcastbridge.com/...echnology-part-15-a-catalogue-of-shortcomings

Maybe @Purité Audio should become a distributor of Legend? Legend search for more commercial partners.

 

[Ron Texas]

Genelec crosses over at about 85 Hz, using DSP (GLM) for phase and timing. This method made it impossible to locate the sub once I tried it.

You mentioned cross at 120 Hz, which is (way) too high, I guess.

I high pass my LS50's starting at 100hz and low pass the sub at around 70 hz. The room has a peak around 90 hz, so the wide gap helps deal with it. The sub disappears and the effect is aided by it's location which is out of my field of view. Ports on the LS50's are plugged. The difficult parts are room acoustics and getting setting the sub volume as the later is rather subjective.

 

[KSTR]

Is there any evidence you base this opinion on?

To my knowledge, it isn't supported experimentally. Even using impulse responses as the stimulus and headphones as the test apparatus, group delays in this range of magnitude and frequency have proven thus far to be undetectable (to my knowledge) despite many studies having investigated it.

For example,the study @oivavoi posted in post #99 found that group delays of this magnitude and in this frequency range (a 4th order xover at 120Hz will generate under 6ms of group delay) were undetectable, even when the stimulus was an impulse response and even when subjects listened through headphones.

(underline mine). That might just be reason why it was not detectable. Use music and linear-phase loudspeakers (this is mandatory, of course) and it is easily detected in proper ABX blind test (as is a change of absolute polarity, btw). All you need is a few allpass kernels you convolve you test signals/music with and compare that to the original (convolved with a dummy kernel so as to avoid hints from any processing artifacts). It also is a matter of training, I found that skilled listeners are readily able to detect phase shifts that, say, our 4th order crossover @ 120Hz introduces. I did a lot of these experiments both for myself as for the companies I worked for, so I have lots evidence, sadly I can't disclose. And mind you, this isn't Q&D hobby stuff, I'm working on these topics seriously for more than a decade professionally. If you look harder you might find papers and reports of experiments that give more insight than what AES etc have to offer.

I'd be happy to upload a few allpass-processed audio samples if there is any interest, and normally even with standard non-linearphase loudspeaker one can detect the differences but of course lin-phase speakers are best. The other option is you measure the phase response of your multiway, preferably 3-way speakers (or obtain the info otherwise) and use a tool like Rephase to create a correction kernel so you can test yourself.
Personally I have little interest to engage in endless discussions where everything boils down to "do you have any peer-reviewed papers that prove your point?" "there is no evidence in the literature" etc when it is so easy to get to your own personal evidence as it is in this case. Remember, this isn't useless snake oil discussions like cable directionality and such. This is real, rewarding experiments where really something is happening (well, some people are actually "phase-deaf" even after a lot of training).

 

[KSTR]

Is there any evidence you base this opinion on?

To my knowledge, it isn't supported experimentally. Even using impulse responses as the stimulus and headphones as the test apparatus, group delays in this range of magnitude and frequency have proven thus far to be undetectable (to my knowledge) despite many studies having investigated it.

For example,the study @oivavoi posted in post #99 found that group delays of this magnitude and in this frequency range (a 4th order xover at 120Hz will generate under 6ms of group delay) were undetectable, even when the stimulus was an impulse response and even when subjects listened through headphones.

(underline mine). That might just be reason why it was not detectable. Use music and linear-phase loudspeakers (this is mandatory, of course) and it is easily detected in proper ABX blind test (as is a change of absolute polarity, btw). All you need is a few allpass kernels you convolve you test signals/music with and compare that to the original (convolved with a dummy kernel so as to avoid hints from any processing artifacts). It also is a matter of training, I found that skilled listeners are readily able to detect phase shifts that, say, our 4th order crossover @ 120Hz introduces. I did a lot of these experiments both for myself as for the companies I worked for, so I have lots evidence, sadly I can't disclose. And mind you, this isn't Q&D hobby stuff, I'm working on these topics seriously for more than a decade professionally. If you look harder you might find papers and reports of experiments that give more insight than what AES etc have to offer.

I'd be happy to upload a few allpass-processed audio samples if there is any interest, and normally even with standard non-linearphase loudspeaker one can detect the differences but of course lin-phase speakers are best. The other option is you measure the phase response of your multiway, preferably 3-way speakers (or obtain the info otherwise) and use a tool like Rephase to create a correction kernel so you can test yourself.
Personally I have little interest to engage in endless discussions where everything boils down to "do you have any peer-reviewed papers that prove your point?" "there is no evidence in the literature" etc when it is so easy to get to your own personal evidence as it is in this case. Remember, this isn't useless snake oil discussions like cable directionality and such. This is real, rewarding experiments where really something is happening (well, some people are actually "phase-deaf" even after a lot of training).

 

[andreasmaaan]

@KSTR yeh I'd be interested in your files, thanks. And any studies you can point to (peer-reviewed or not).

I've looked quite hard and have not come across a decent study that found thresholds significantly dissimilar from those found in the study we've been discussing here.

I also haven't encountered a study in which there was not a significant difference in audibility between headphones/impulses and speakers/music, in exactly the opposite direction to the one you suggest is the case.

I'm not interested in engaging in endless discussions here either btw, and I am open to your claims. But not without evidence

 

[March Audio]

(.
Personally I have little interest to engage in endless discussions where everything boils down to "do you have any peer-reviewed papers that prove your point?" "there is no evidence in the literature" etc when it is so easy to get to your own personal evidence as it is in this case. Remember, this isn't useless snake oil discussions like cable directionality and such. This is real, rewarding experiments where really something is happening (well, some people are actually "phase-deaf" even after a lot of training).

That's All well and good but without it you are just a guy on a forum saying stuff. You are making claims that appear to contradict the studies that have been done, so respectfully you really do have to come up with a bit more than just saying "it is so"