Stereo Bass using subwoofers

[gnarly]

I agree it's related to their processing, but DSP does not necessarily cause excess group delay (ignoring latency since it's pure delay). In fact, for simply extending the LF response of a sealed system like this, it seems to me that it would be much easier to not cause lots of excess group delay.


Perhaps I'm missing something, but I don't see why "Huge IIR" filters would be needed in this case. Since this is a sealed box sub, it should exhibit at 2nd order highpass response, which can be trivially corrected to any desired corner frequency and Q using a minimum phase digital biquad filter. As both the sub and filter are minimum phase, the result must also be minimum phase.

I think what Levimax was saying is that to get 20Hz response from a 13" driver in a small box, a good deal of low end boost will be needed whether from shelving, PEQs, Linkwitz transform, or a combo...the IIR stew JJ speaks of. ..

And then to keep excursion from destroying the dang thing with all the boost that made its way below 20Hz, a high-pass filter will be needed.

The electrical high-pass filter is going to need a relatively sharp knee (to allow 20Hz to squeeze in at the edge of its attainable excursion, but go no lower).
So the sealed sub is going to end up with an acoustic high-pass response quite higher than 2nd order.

The more a sub manufacture tries to cram response down into a lower bottom corner, the sharper the high-pass knee will be. No way around it.
And the more the group delay will jump up.
Makes for good marketing specs at least ....aargh... Lol.

 

[bmc0]

to get 20Hz response from a 13" driver in a small box, a good deal of low end boost will be needed

Yes, and if done correctly the resulting group delay will be identical to any other minimum phase system with the same magnitude response (e.g. a larger sealed sub with a native response that is the same as the equalized system).

And then to keep excursion from destroying the dang thing with all the boost that made its way below 20Hz, a high-pass filter will be needed. [...] So the sealed sub is going to end up with an acoustic high-pass response quite higher than 2nd order.

That is accounted for (down to 10Hz, at least) in the minimum phase group delay I calculated, assuming the measurement I used is accurate. Erin's measurement shows a slope of approximately 20dB/oct—between 3rd and 4th order. I don't see a clear indication of a sharp subsonic filter down to 10Hz, but before I posted the minimum phase group delay plot I did in fact try some minimum phase highpass filters (up to 11th order butterworth) around 10Hz. I was not able reproduce the group delay shown without significantly changing the magnitude response in the 10-20Hz range.

Minimum phase filters do not cause excess group delay. A sharp minimum phase highpass will of course have significant group delay around its cutoff frequency, but its excess group delay is zero.

 

[j_j]

I agree it's related to their processing, but DSP does not necessarily cause excess group delay (ignoring latency since it's pure delay). In fact, for simply extending the LF response of a sealed system like this, it seems to me that it would be much easier to not cause lots of excess group delay.

Not if it's IIR filters.

Perhaps I'm missing something, but I don't see why "Huge IIR" filters would be needed in this case. Since this is a sealed box sub, it should exhibit a 2nd order highpass response, which can be trivially corrected to any desired corner frequency and Q using a minimum phase digital biquad filter. As both the sub and filter are minimum phase, the result must also be minimum phase.

Lots of peakiness is going to create a lot of group delay. The length of the impulse response is going to matter. Convolving two does not always work out like straight up poe/zero cancelling since there has to be another zero to avoid attempts at DC response.

 

[bmc0]

Not if it's IIR filters.

If the IIR filter is minimum phase, the excess group delay is zero.

Lots of peakiness is going to create a lot of group delay.

Where is this peakiness coming from? Apologies, but I don't understand what you're getting at.

A small example showing equalization of an ideal 2nd order system:

Q is high at 1.5, representative of a box which is way "too small". Group delay of the above (apologies for different colors):

The excess group delay is zero in all cases since the original system and the filters are all minimum phase.

 

[gnarly]

That is accounted for (down to 10Hz, at least) in the minimum phase group delay I calculated, assuming the measurement I used is accurate. Erin's measurement shows a slope of approximately 20dB/oct—between 3rd and 4th order. I don't see a clear indication of a sharp subsonic filter down to 10Hz, but before I posted the minimum phase group delay plot I did in fact try some minimum phase highpass filters (up to 11th order butterworth) around 10Hz. I was not able reproduce the group delay shown without significantly changing the magnitude response in the 10-20Hz range.

Hi, I'm not understanding why you calculated group delay from Erin's frequency response graphs, and not just use the Group delay he also measured direct w Klippel?
Was it because you suspect a big part of the Klippel Group delay shown below, is comprised of excess group delay?

Otherwise, I'm not sure how excess group delay entered the thread or why?
At any rate my bet for the huge group delay for the SB-3000 remains on a steep electrical IIR hpf. Can't know w/o meas, but seems like a good bet.

Here's a 20Hz Cheby 60dB I just put into a processor and measured transfer. Blue group delay, on left.
As I'm sure you know, it's not hard to get crazy group delay with crazy steep filters.

 

[j_j]

If the IIR filter is minimum phase, the excess group delay is zero.

What's this "excess group delay" thing? There is no "excess group delay", there could be "excess phase" but not in a pure IIR. So nobody's claiming that.

Where is this peakiness coming from? Apologies, but I don't understand what you're getting at.

Does the box go down to DC? No, it does not. There is, therefore, a huge change in response between DC and 20Hz. That's pretty (*&(&* peaky, now isn't it?

You have to consider that.

Now, in addition, in order to avoid stuffing DC into the woofer, you'll need at least another second-order HP filter to prevent excessive excursion, and that even if all you do is go to "constant excursion", as opposed to "remove DC and very low frequency excursion", which will require more than a 2nd order HP filter.

Pole-zero cancellation works, but there are both poles and zeros here that can't be cancelled.

 

[bmc0]

What's this "excess group delay" thing? There is no "excess group delay", there could be "excess phase" but not in a pure IIR.

It's the difference between the group delay of some system and that of a minimum phase system having the same magnitude response.

Does the box go down to DC? No, it does not. [...]

It seems there's been an error in communication. I fully agree with what you've stated here.
The reason I brought up "excess group delay" (or excess phase, if you prefer) is that the measured group delay for the subwoofer in question seems to be far beyond what is expected for a minimum phase system with the same magnitude response.

Hi, I'm not understanding why you calculated group delay from Erin's frequency response graphs [...] Was it because you suspect a big part of the Klippel Group delay shown below, is comprised of excess group delay?

Yes, exactly.

As I'm sure you know, it's not hard to get crazy group delay with crazy steep filters.

Sure, but look at the group delay an octave above the cutoff—it's only a little over 20ms. Erin's measurements don't seem to show any steep highpass down to 10Hz, so why is the group delay around 80ms at 20Hz relative to 100Hz rather than 15ms as predicted from the magnitude response? Assuming the measured group delay is correct, either the magnitude response is inaccurate below 20Hz or there's quite a lot of excess phase.

 

[nagster]

This is just anecdotal experience with stereo/mono subs. No instantaneous switching or double blind.
This is a browser translation, so it may be a strange sentence.

The basic configuration is 2.2. Between placing one sub next to each stereo main speaker and sending a stereo signal, and leaving the main speakers as they are and combining the two subs in a location of your choice to create a physical mono (signal remains stereo), I 100% preferred the former.

I've tried it for a long time now, whenever I had the chance, in various spaces from a small study to 800 square meters, with various music. The LPF also varied from 40Hz to 110Hz.
I felt that the difference was greater the larger the space.

I've also tried an electronic mono mix (monosub output, etc.) several times, but it always ended up being the most boring sound, so I haven't tried it for a while.

Combining the two subs in one place gave it more punch.
The sense of depth was lost. Other subjective aspects were also inferior.
On the other hand, it might be better for reggae, dub, or old hip hop. It might be different for movies. I don't know.

I got a deep space, a large space, a sense of realism, and immersion with the stereo sub placement.
The mono sub placement sounded flat. I was bored. The punch was also often a nuisance.
No, maybe it's just a lack of skill on my part.

In any case, there will be unevenness in the bass depending on the listening point, but for me, the naturalness of the space is more important.
Anyway, for now it's clear that I prefer stereo placement subs driven by stereo signals. I don't know which is more faithful or which is correct.

 

[Chrispy]

This is just anecdotal experience with stereo/mono subs. No instantaneous switching or double blind.
This is a browser translation, so it may be a strange sentence.

The basic configuration is 2.2. Between placing one sub next to each stereo main speaker and sending a stereo signal, and leaving the main speakers as they are and combining the two subs in a location of your choice to create a physical mono (signal remains stereo), I 100% preferred the former.

I've tried it for a long time now, whenever I had the chance, in various spaces from a small study to 800 square meters, with various music. The LPF also varied from 40Hz to 110Hz.
I felt that the difference was greater the larger the space.

I've also tried an electronic mono mix (monosub output, etc.) several times, but it always ended up being the most boring sound, so I haven't tried it for a while.

Combining the two subs in one place gave it more punch.
The sense of depth was lost. Other subjective aspects were also inferior.
On the other hand, it might be better for reggae, dub, or old hip hop. It might be different for movies. I don't know.

I got a deep space, a large space, a sense of realism, and immersion with the stereo sub placement.
The mono sub placement sounded flat. I was bored. The punch was also often a nuisance.
No, maybe it's just a lack of skill on my part.

In any case, there will be unevenness in the bass depending on the listening point, but for me, the naturalness of the space is more important.
Anyway, for now it's clear that I prefer stereo placement subs driven by stereo signals. I don't know which is more faithful or which is correct.

Difference in what way as space got larger? Curious, 800 sq m with what particular dimensions? That's a huge space and not likely home reproductive use relatable.

 

[levimax]

It's the difference between the group delay of some system and that of a minimum phase system having the same magnitude response.

My understanding is IIR filters are not usually minimum phase, and while they can be linearized with a FIR filter, at these low frequencies it takes a lot of taps / processing power which is not available in a subs DSP.

 

[j_j]

It's the difference between the group delay of some system and that of a minimum phase system having the same magnitude response.

That's called "excess phase", sometimes. Symmetric FIR's, for instance, have that, although they also represent pure delay.
We are talking about "group delay" which is not part of excess phase. You can only talk about group delay as a function of amplitude response if there is no excess phase in the system.

Maybe you were taught with different terminology. Dunno. I'm old, I use old-fashioned terms.

 

[j_j]

My understanding is IIR filters are not usually minimum phase, and while they can be linearized with a FIR filter, at these low frequencies it takes a lot of taps / processing power which is not available in a subs DSP.

IIR filters ***MUST*** be minimum phase as far as their poles, or they are unstable. Minimum phase, among other things, means that all poles are inside the unit circle, and strictly speaking, so are non-radius-1 zeros. (zeros can be on the unit circle, btw, but strictly ON it or inside it)

Most FIR filters are not "minimum phase" because exactly half of the zeros (they only have zeros) are OUTSIDE the unit circle. For every zero or zero pair inside the unit circle, there is an identical zero/pair at 1/<inside pair>. Literally exactly that. That makes them constant delay, but it doesn't HAVE to be like that.

Now, IIR filters can not have "linear phase", which means phi=2*pi*f*t, where f is frequency and t is the time delay. FIR filters that are symmetric or antisymmetric have pure delay characteristics, but are not minimum phase.

There's a thread here somewhere where I explained that. If I knew where it was, I'd say so.

Can't find it. Maybe I'll write it.



So, I did.

https://www.audiosciencereview.com/...onstant-delay-aka-linear-phase-whazzit.61227/ Those of you not into math, brace yourselves!

 

[nagster]

Difference in what way as space got larger?

Even when the space became larger, the type of change remained almost the same. The amount of change felt large.

Curious, 800 sq m with what particular dimensions? That's a huge space and not likely home reproductive use relatable.

It's a small multi-purpose hall that hosts plays, live shows and lectures. I don't think anyone lives there.
The stage is about 20m wide. The seating extends on both sides and is about 30m deep. The ceiling height is about 8m to 12m.
Line arrays were flown on both sides of the stage, and there were subs directly below them, but we also listened to it when the wiring was extended and gathered in the center.

 

[OCA]

I am pretty sure it is group delay, see my measurements below with red being my DIY subwoofer (big sealed enclosure) and the green line being my SVS SB-3000 (small sealed DSP). Below that is a 3rd party measurement for the SVS sub that is close to what I got. Obviously I am having some trouble integrating these 4 subs (2 of each) together due to the group delay but I don't want to derail this thread with that rather I am curious about how GD could potentially affect AE if at all. I also am thinking that the main driving force for subs these days is "smaller at all costs" and it seems to me that one of these costs is sound quality even if they do manage to get the "marketing specs." to look good.

View attachment 432964



View attachment 432965

Allpass filters are your best friends when you need to combine sealed & ported subs, I don't see why they wouldn't work with open baffles.

 

[j_j]

https://www.audiosciencereview.com/...onstant-delay-aka-linear-phase-whazzit.61227/ has a discussion of what minimum phase, constant delay (i.e. linear phase) are, and some filtering discussion, just to define the actual terminology.

 

[bmc0]

Maybe you were taught with different terminology.

I admit that I don't know to what extent it's an accepted term, but it is used in the Room EQ Wizard measurement software. See this section of the documentation.

You can only talk about group delay as a function of amplitude response if there is no excess phase in the system.

And that's why I stated that the particular subwoofer in question seems to have lots of excess phase (or excess group delay). I traced the magnitude response given in Erin's review (which I believe was measured with his Klippel NFS) and used that data to generate the following minimum phase group delay:

The measured group delay is much higher:

Implying considerable excess phase.

 

[j_j]

I admit that I don't know to what extent it's an accepted term, but it is used in the Room EQ Wizard measurement software. See this section of the documentation.


And that's why I stated that the particular subwoofer in question seems to have lots of excess phase (or excess group delay). I traced the magnitude response given in Erin's review (which I believe was measured with his Klippel NFS) and used that data to generate the following minimum phase group delay:
View attachment 433132
The measured group delay is much higher:
View attachment 433134
Implying considerable excess phase.

Well, if the person who holds the data can do this, do a linear fit on the phase of the system in the range of interest. THEN subtract out that linear fit. What remains is the non--delay part. The linear fit part is pure delay. See the thread I just posted in general audio discussions for the terms, etc. That can be minimum phase or maximum phase, or some of each, but with the time delay sorted out.

 

[OCA]

I admit that I don't know to what extent it's an accepted term, but it is used in the Room EQ Wizard measurement software. See this section of the documentation.


And that's why I stated that the particular subwoofer in question seems to have lots of excess phase (or excess group delay). I traced the magnitude response given in Erin's review (which I believe was measured with his Klippel NFS) and used that data to generate the following minimum phase group delay:
View attachment 433132
The measured group delay is much higher:
View attachment 433134
Implying considerable excess phase.

What John means there is excess phase group delay after removing minimum phase part of the response. It may as well mean "uncorrectable" part of the group delay with minimum phase (i.e. IIR) filters.

 

[bmc0]

do a linear fit on the phase of the system in the range of interest. THEN subtract out that linear fit. What remains is the non--delay part. The linear fit part is pure delay.

I don't think that pure delay is the main problem here—otherwise the measured group delay would be high everywhere. Comparing the two group delay plots in my last post, the difference at 100Hz is only 5ms or so, but at 20Hz it is almost 70ms.

What John means there is excess phase group delay after removing minimum phase part of the response. It may as well mean in-correctable part of the group delay with minimum phase (i.e. IIR) filters.

A subwoofer (sans DSP) should be minimum phase at LF when measured anechoically, so should have little or no excess phase after any pure time delay is accounted for.

 

[OCA]

A subwoofer (sans DSP) should be minimum phase at LF when measured anechoically, so should have little or no excess phase after any pure time delay is accounted for.

True as with the wavelengths involved, the enclosure can be considered a point source but that property is nearly useless for all practical purposes anywhere outside an anechoic chamber.