Underdamped response = missing sound quality?

[dfuller]

@Heinrich I don't know what's up with the fuss. I'm doing optics engineering so I already know how damping plays out in electronics and optics, but not speakers.

Well, figure a speaker's cabinet to driver interaction is a resonant HPF. The higher the Q, the more of a hump and more ringing you're going to have. In this case it would be Qtc, the Q of the whole system (box included). Note that this is only sort of related to its overall slope, you can have lower Q ported systems (which are still 4th order) and higher Q sealed systems (which are still 2nd order).

Ideally, you don't go higher than Qtc=0.707, above that you start to get response boosts around the tuning frequency from filter ring (which of course impedes transient response). Of course doing EQ to bring it back to ideal also improves GD since this is a minimum phase system.

The audibility of this is variable, but usually not unless pathological.

As far as port alignments go, by far the most common is the QB3 - Quasi-Butterworth, 3rd order. Maximal linear extension, not perfect but generally fine (read: any ringing is probably inaudible assuming the port tuning isn't totally knackered) transient response, smaller box size.

SBB4 (Super Boom Box, 4th order) is also fairly common and gives better transient response (read: less ring) than QB3 but less linear extension for a given tuning and requires a much larger box.

Graph from AudioJudgement

There are other non-flat (peaking) alignments like BB4 (Boom-Box 4th order), SQB3 (Super Quasi-Butterworth 3rd order) that show up from time to time - they give better bass extension at the expensive of transient response. I am under the impression that Neumann tends to use a corrected SQB3 alignment (but I can't state this for a fact unless Markus or someone chimes in).

 

[bachatero]

Ok, so now we know that speakers are minimum phase systems with group delay and that an underdamped response is just fine because critical or over damping negatively affects frequency response.

This technically answers my original question about underdamped response but it doesn't feel satisfactory with how much of a change plugging the sub's port made to transient response. If "transient response" is just group delay, why do sealed sub makers almost never advertise this feature if it's table-stakes?

Well, no.
Your signal seems to be 10 cycles on, 10 cycles off. So it repeats after 0,5s. Therefore in the frequency domain it consists of amplitudes at 2Hz, 4Hz, 6Hz ...... ∞. All these frequencies are necessary for realising the signal (more or less), otherwise you would have a continuous sine without any changes (like start and stop) and certainly no transients.

This is true, but this isn't very helpful in our case because almost no content has a constant bass signal. Most bass lines look exactly like a repeating tone and in movies, explosions are filled with transients.

 

[terryforsythe]

I‘m not an engineer, but heard about science.

But you tell engineers the following:

I could throw so much at you, all from real science, your ears would ring. The discussion of frequency, phase, time etc needs at least a confident knowledge of basic math, calculus and such.

We are well versed in "basic math, calculus and such." That is much of the engineering curriculum.

it doent‘t make sense to talk about „a frequency“ to begin at a point in time.

That statement does not make sense.

As you know for sure, there‘s the Fourier series.

Yes. Much time studying linear systems is dedicated to the Fourier series, and converting back and forth between time domain and frequency domain representations of signals.

The discussion of frequency, phase, time etc needs at least a confident knowledge of basic math, calculus and such. I‘ve not seen any of that, sigh, in this thread.

Please enlighten us and provide the "basic math, calculus and such" that we should be discussing.

For obvious reasons I cannot give you a full picture of the meaning of all this.

Then please stop with this nonsense.

 

[dfuller]

Ok, so now we know that speakers are minimum phase systems with group delay and that an underdamped response is just fine because critical or over damping negatively affects frequency response.

I don't know if I'd say "underdamped is fine" - it isn't ideal, better to get closer to the "magic number" of 0.707 if possible as that gives maximal linear extension with minimal harm to the transient behavior. There are of course other compromises, cabinet size and bass extension being part of it.

This technically answers my original question about underdamped response but it doesn't feel satisfactory with how much of a change plugging the sub's port made to transient response. If "transient response" is just group delay, why do sealed sub makers almost never advertise this feature if it's table-stakes?

Plugging the port also massively affects frequency response. Ported alignments have much deeper linear bass extension as a general rule, sans EQ.

This is the same speaker (the Neumi BS5P) ported and sealed (measurements courtesy of EAC) . The response difference is pretty stark - 20hz difference in F3 vs 200hz, ~80hz for the sealed alignment and ~60hz for ported.

 

[Cbdb2]

View attachment 436207

Thats an impossible signal. It has infinite BW.

 

[Heinrich]

Then please stop with this nonsense.

Sure. I was under the impression you ignored the basics, but now I‘m convinced it‘s my fault. Please go ahead.

 

[terryforsythe]

Plugging the port also massively affects frequency response. Ported alignments have much deeper linear bass extension as a general rule, sans EQ.

This is the same speaker (the Neumi BS5P) ported and sealed (measurements courtesy of EAC) . The response difference is pretty stark - 20hz difference in F3 vs 200hz, ~80hz for the sealed alignment and ~60hz for ported.

For speakers without a subwoofer or DSP, you definitely get deeper bass extension using bass reflex. Sometimes, though, the group delay introduced by bass reflex in a speaker is less than optimal when crossing over to a subwoofer. For that, I have had better results using an acoustic suspension speaker, or plugging the port if the speaker already is ported. As I previously noted, one other person reported the opposite experience.

For a subwoofer, a ported design results in less excursion at the tuning frequency and deeper bass response without DSP, both of which are advantagious. But, greater group delay is introduced. Some people say it is audible, some people say it is not and I have seen strudies to back that up. I have not compared the two. My subwoofer is acoustic suspension and has no problem reaching down below 20Hz with a flat response, but it is large and is backed by a lot of power (18"/1200W).

 

[dfuller]

Sometimes, though, the group delay introduced by bass reflex in a speaker is less than optimal when crossing over to a subwoofer.

It's certainly generally easier to do a crossover without adding in a helmholtz resonator to it!

 

[MAB]

Ok, so now we know that speakers are minimum phase systems with group delay and that an underdamped response is just fine because critical or over damping negatively affects frequency response.

It was explained on page 1.
I can try to elaborate.

The stimuli you are applying aren't pure tones. They are periodic 40Hz tone bursts, can be modeled many ways. The pulses in time domain are represented by a series of harmonics in the frequency domain. For illustration:

You modulate a 40Hz sinusoidal with a pulse, in your example d=0.5 (50% duty cycle). This implies an entire series of discrete frequencies and amplitudes.

The thing is, the speaker's response to the 40Hz tone burst you apply is going to be dramatically affected by the rolloff of your speakers, which is in the mid 40Hz range.

And by stuffing the port, you change the rolloff characteristics of your woofer, right in the region where reproduction of the leading and trailing edges of the transient are critically important. Which is why @voodooless said in post 8 that you first need to EQ the sealed system to the same FR as the ported system before seeing if there were any residual time domain changes. Which is the same thing as saying you can likely get exactly the same sound with EQ, and retain the efficiency benefits and distortion benefits of the ported design.

Note this is also why simply matching volume isn't the same as matching response. If you were truly using a pure 40Hz tone, you are correct, matching SPL would be sufficient. @thewas pointed this out also early on. The phenomena you see is artifact of the bursts, and the high-pass filter changes. If you EQ to the same FR, you will see the same time domain behavior!

The solution to the above is doable, I recall either Maple or Maxima had it explicitly as an example in their training documentation, I will try to find since I think it is a good read, with great pictures.

You can test this out, even in Audacity for example.
Below is the same tone burst you used, with the bass response of ported and sealed frequency response using filters in Audacity! I got the bass response from the ASR review. And simulated the effect of sealing the ports, we can argue if I got the exact new corner freq and slope of your speakers with stuffed ports spot-on, but it doesn't change the result too much.

You can see the so-called sealed response has less ringing. And, as you point out, this effect doesn't change even if we match the volumes. It's not that sealed has some time-domain advantage over ported. It just has different frequency response. Also, I am guessing the effect you observed is more apparent around 40Hz. This makes sense since stuffing the ports make the biggest difference in that same region. If you had a speaker with an 80Hz f3, tone bursts using ~80Hz would show the largest sensitivity to changes in FR due to stuffing the port.

I am just modeling the change in bass response due to the changes in the woofer rolloff, no other interactions. For your situation, need to take into account the desk which is often a bump around 160Hz, and your room. You can apply all different equalizations to the above waveform in Audacity or other tools, and see how these additionally affect the response, and create ringing. For sure the bass response of your speakers plays the largest role in your example.

 

[Holmz]

It was explained on page 1.
I can try to elaborate.

The stimuli you are applying aren't pure tones. They are periodic 40Hz tone bursts, can be modeled many ways. The pulses in time domain are represented by a series of harmonics in the frequency domain. For illustration:
View attachment 436394
You modulate a 40Hz sinusoidal with a pulse, in your example d=0.5 (50% duty cycle). This implies an entire series of discrete frequencies and amplitudes.

The thing is, the speaker's response to the 40Hz tone burst you apply is going to be dramatically affected by the rolloff of your speakers, which is in the mid 40Hz range.
View attachment 436398
And by stuffing the port, you change the rolloff characteristics of your woofer, right in the region where reproduction of the leading and trailing edges of the transient are critically important. Which is why @voodooless said in post 8 that you first need to EQ the sealed system to the same FR as the ported system before seeing if there were any residual time domain changes. Which is the same thing as saying you can likely get exactly the same sound with EQ, and retain the efficiency benefits and distortion benefits of the ported design.

Note this is also why simply matching volume isn't the same as matching response. If you were truly using a pure 40Hz tone, you are correct, matching SPL would be sufficient. @thewas pointed this out also early on. The phenomena you see is artifact of the bursts, and the high-pass filter changes. If you EQ to the same FR, you will see the same time domain behavior!

The solution to the above is doable, I recall either Maple or Maxima had it explicitly as an example in their training documentation, I will try to find since I think it is a good read, with great pictures.

You can test this out, even in Audacity for example.
Below is the same tone burst you used, with the bass response of ported and sealed frequency response using filters in Audacity! I got the bass response from the ASR review. And simulated the effect of sealing the ports, we can argue if I got the exact new corner freq and slope of your speakers with stuffed ports spot-on, but it doesn't change the result too much.

View attachment 436407
You can see the so-called sealed response has less ringing. And, as you point out, this effect doesn't change even if we match the volumes. It's not that sealed has some time-domain advantage over ported. It just has different frequency response. …

Well it has both.
One can wax on about the FFTs going both ways… but the lower graph looks more like the top one.
The change in the FR is not the causal mechanism of the poor time domain performance, it is a consequence of it.

Everyone seems to just want to wave a magic wand over everything and make it frequency response uber alles.

 

[dfuller]

Everyone seems to just want to wave a magic wand over everything and make it frequency response uber alles.

Magnitude response on its own is damn near useless, but magnitude and phase? Now we're getting somewhere.

 

[MAB]

Well it has both.
One can wax on about the FFTs going both ways… but the lower graph looks more like the top one.

That was my point. I think it got missed. And further more, if you EQ the sealed alignment back to the same FR as ported, you have the same phase response. Unless you have a poorly performing port with lots of nonlinearities.

The change in the FR is not the causal mechanism of the poor time domain performance, it is a consequence of it.

Both need to be considered. It was the entire point. You need to EQ to same FR to assess correctly time domain performance. Or at least know the FR of each configuration.

Everyone seems to just want to wave a magic wand over everything and make it frequency response uber alles.

I was being careful to avoid this, and not really sure why you comment like this. For instance I wrote:
"that you first need to EQ the sealed system to the same FR as the ported system before seeing if there were any residual time domain changes."
I hope you read that carefully. To evaluate the time domain issues as were claimed in this thread, you also need to know the frequency response of the two alignments, or better yet that they are matched. That was the entire point of the above example I did. It may be there are some non-linearities in the KH150 port that sealing it up address. But without also including the frequency responses, the problem is underspecified.

Agree that it is incomplete to sweep everything into the frequency response corner. As was just stated, magnitude and phase. Which is why people were asking over and over if the frequency response was matched.

 

[Heinrich]

That was my point. I think it got missed. And further more, if you EQ the sealed alignment back to the same FR as ported, you have the same phase response. Unless you have a poorly performing port with lots of nonlinearities.

Both need to be considered. It was the entire point. You need to EQ to FR to asses correctly time domain performance. Or at least know it.

I was being careful to avoid this, and not really sure why you comment like this. For instance I wrote:
"that you first need to EQ the sealed system to the same FR as the ported system before seeing if there were any residual time domain changes."
I hope you read that carefully. To evaluate the time domain issues as were claimed in this thread, you also need to know the frequency response of the two alignments, or better yet that they are matched. That was the entire point of the above example I did. It may be there are some non-linearities in the KH150 port that sealing it up address. But without also including the frequency responses, the problem is underspecified.

Agree that it is incomplete to sweep everything into the frequency response corner. As was just stated, magnitude and phase. Which is why people were asking over and over if the frequency response was matched.

Topics like this come up every other week. The advertising wants us to believe that audio is special, in that standard analysis doesn‘t apply. Thanks for the well executed explanation, especially considering that the audience is engineeres, as far as I‘ve learned.

Reiterated, watch the room’s echoes for the full picture.

 

[Heinrich]

O/k, a few notes.
Do we know the tuning of the speaker in question, is it the Neumann or the subwoofer?
Both may have an additional (level dependent?) subsonic filter applied, in order to prevent damage. Or there‘s a limiter at work, frequency dependent, filtered. Don‘t know how that would be described in a linear theory.

Filters show group delay, the higher the order the more, the lower the edge frequency the more. Sealed translates to second order, ported to forth order. We want the lowest reasonable edge frequency.

Physiologically group delay is considered quite harmless, except for outliers like the Klipsch horn. Room reverberation renders the speakers‘ contribution insubstantial.

Fun question: Would it be possible to transform a ported design‘s output to that of the lower order sealed speaker‘s output? Get from fourth order to second order, phase andlogically group delay included? Fun answer: Yes, just adjust the frequency response of the ported speaker, using a linear equalizer, to that of a sealed. Since all contributions, speaker, equalizer are „minimum phase“, phase follows frequency response exactly.

Practical? Could be, with a limiter and a low tuning frequency of the port. Patent? Nohhh …

 

[bachatero]

You'd think that I could just easily change the sub's sealed response to match that of ported, but I did so and it added too much distortion down low to sound good, so that's a non-starter.

 

[voodooless]

You'd think that I could just easily change the sub's sealed response to match that of ported, but I did so and it added too much distortion down low to sound good, so that's a non-starter.

Do it the other way around

 

[Heinrich]

You'd think that I could just easily change the sub's sealed response to match that of ported, but I did so and it added too much distortion down low to sound good, so that's a non-starter.

You may think I‘m not as friendly as I ought to be. But that is because of the language barrier alone, please take my apologies.
In regard to your problem you formulated a suspicion. The bass would be delayed, pumping air and so forth, masking milliseconds later some other content. But you were not sure, if that was audible or objectionable.

Actually the case isn‘t that simple.

My question wasn‘t meant as a proposal. But anyway, you may have heard about the „Linkwitz transform“, or a „biquad“. It is used to shift the so called ‚pole‘ of a transfer function to a more digsirable place, correcting the Q also. With sealed speakers that is a single pole, with ported there are two, mathematically multiplied. Why not shifting that two instead of only one? It would be very easy. The consequences for the mechanical design are manageable.

Fun question, because we, the tough guys, do it already, and very much more successfully when correcting the bass for well controlled in-room response. If there‘s something more to be desired it is always the mechanical limits, e/g excursion, distortion etc. Today (!) that is all well understood, but cannot be exemplified in all sufficient detail on an passer-by internet board. I would love somebody to try anyway.

 

[bachatero]

Do it the other way around

But if I do it this way, I'd have to lose out on bass response or increase distortion, both of which are non-starters.

 

[voodooless]

But if I do it this way, I'd have to lose out on bass response or increase distortion, both of which are non-starters.

That why bass-reflex is such a nice principle if implemented correctly

 

[Heinrich]

But if I do it this way, I'd have to lose out on bass response or increase distortion, both of which are non-starters.

Seems I made it to the ‚ignore list‘. Fair enough. What „we“ are talking about in harmonious loops is the very foundation of bass alignment, well understood since the 60s of the last century (Thiele/Small), but tainted by myths and misconceptions fueled by the advertising. It appears less promising to me to pick one buzz word, or the other, and trying to figure it all out from there. Better to take a thorough read of the vast literature before to get a full picture of the landscape.

That why bass-reflex is such a nice principle if implemented correctly

Ja, but with today‘s possibilities, especially arbitrary, pretty cost effective equalization in the digital domain (24bits and more) we address even room resonance with a smile on our faces. Just tune to 30Hz and the rest will follow.