Never Put Subwoofers In Corners... Even with DSP and Multi-Sub Setups?

[Chromatischism]

Multiple subs are far more effective at smoothing and do something no dsp can do, smooth over a much larger area of a room. Two subs and dsp after careful setup gets you huge benefits.

Yep. I follow ideal placement strategies as much as I can. Red line is 2 subs, center of front and back walls, pre-EQ1. This is considered the #1 placement for dual subs in rectangular rooms. However, EQ is still needed for the system to reach its potential:

You can not do this with bass traps. They are simply not effective below about 80 Hz, not even the bigger ones.

1. delays were not manually optimized pre-EQ because that is taken care of by SubEQ – otherwise the red line would look a little better.

 

[Chromatischism]

There was a pretty good article in Audiohlics fifteen years ago

They have more recent articles, too.

https://www.audioholics.com/subwoofer-setup/multiple-subwoofer-setup-calibration-1

 

[QMuse]

"Partial cancellation" among subwoofers is highly desirable. Partial cancellation and partial reinforcement are mechanisms by which multiple subs result in smoother in-room response. IF you had NO partial cancellation then adding more subs WOULD NOT IMPROVE the summed in-room response. The peaks would be just as tall, and the dips would be just as deep. Fortunately that is not the case.

Imo there is nothing of value to be gained by trying to phase-align multiple subs. By the time you hear the low frequencies, the room's effects are dominant. From a perceptual standpoint, in home-audio-sized listening rooms, there is no such thing as direct sound (which is completely different from the perceptual standpoint further up the spectrum). Fix the low-frequency in-room frequency response and you will have fixed what matters.

By way of anecdotal evidence (I don't have Harman-class resources), I have many customers who report subectively and (when available) measurably improved in-room bass from deliberately introducing significant phase differences into their distributed multi-sub systems.

There is a crossover between sub(s) and mains. Unless you get the sub(s) phase to match mains phase around the XO freqeuncy they will never sum correctly. Phase aligning between subs and mains needs to be done the same way you're doing phase aligning between woofer and tweeter around their XO point and that has nothing to do with room modes and cancelling the dips in subs summed response outside the XO region. Phase aligning between drivers (in this case between sub woofer and main woofer) is simply a procedure that needs to be done around any XO point you create.
As you are speaker designer I'm sure we can agree on that.

Now, what happens below subs and mains XO point is another thing. Phase mismatch between subs below XO region will cause additional dips in LF summed response and that has nothing to do with room modes. The same is happening in the scenario with 2 main speakers although in that scenario there is no XO around 80Hz or so. Phase mismatch in that scenario is usually caused by differences in positioning of the left and right speaker which cause phase shifts at LP. That can be corrected to some extent with phase linearization between left and right speaker in the LF region taking care that pre-ringning is not induced due to excessive phase correction.

 

[RichB]

I don't remember. I looked into group delays of vented box systems years ago and concluded that the calculated group delay of my system was perceptually benign, but don't remember the specifics of where the perceptual thresholds are. And I'm not aware of how that would be any different for multiple subs versus a single sub.



If you're talking about arrival time differences, 20 milliseconds of relative delay corresponds to a path length difference of about 22 feet, which is imo a much greater path length difference than we're likely to have in a home audio listening room.

If I recall, the mini-DSP running Dirac has about 40 milliseconds of delay. That is way out of bounds for the gamer community and gets perilously long for AV sync. I wonder if ASR can measure this @amirm comes up with a PEQ or standard REQ test.

- Rich

 

[patate91]

There is a crossover between sub(s) and mains. Unless you get the sub(s) phase to match mains phase around the XO freqeuncy they will never sum correctly. Phase aligning between subs and mains needs to be done the same way you're doing phase aligning between woofer and tweeter around their XO point and that has nothing to do with room modes and cancelling the dips in subs summed response outside the XO region. Phase aligning between drivers (in this case between sub woofer and main woofer) is simply a procedure that needs to be done around any XO point you create.
As you are speaker designer I'm sure we can agree on that.

Now, what happens below subs and mains XO point is another thing. Phase mismatch between subs below XO region will cause additional dips in LF summed response and that has nothing to do with room modes. The same is happening in the scenario with 2 main speakers although in that scenario there is no XO around 80Hz or so. Phase mismatch in that scenario is usually caused by differences in positioning of the left and right speaker which cause phase shifts at LP. That can be corrected to some extent with phase linearization between left and right speaker in the LF region taking care that pre-ringning is not induced due to excessive phase correction.

I guess the Dunning Kruger effect plays a role here, but time and phase aligning multiple subs is not an easy task for people who never made it or have limited knowledge.

I know it gives a lot of benefits but it needs to be setup correctly, if not things will ggetworst.

 

[QMuse]

I guess the Dunning Kruger effect plays a role here, but time and phase aligning multiple subs is not an easy task for people who never made it or have limited knowledge.

I know it gives a lot of benefits but it needs to be setup correctly, if not things will ggetworst.

Time/phase aligning (which is the same thing) is indeed a quite complex task, and most people overestimate their capabilities to do it due to the lack of knowledge. Subs that have DSP instead of passive XO, thus introducing app. 8-15ms of additional internal delay made things even more complicated as you can't align them based purely on sub's distance to the LP.

Btw, level aligning subs to mains is far less complicated, yet from what I have seen and read here most of the folks don't do that right either.

 

[patate91]

Time/phase aligning (which is the same thing) is indeed a quite complex task, and most people overestimate their capabilities to do it due to the lack of knowledge. Subs that have DSP instead of passive XO, thus introducing app. 8-15ms of additional internal delay made things even more complicated as you can't align them based purely on sub's distance to the LP.

Btw, level aligning subs to mains is far less complicated, yet from what I have seen and read here most of the folks don't do that right either.

Don't forget that D-K effect has 2 effect, the one you cited and the other one his : Expert can underestimate their abilities. For an expert or someone who has some experience find certain things to do or understand easy. But for non expert it is not easy.

 

[Hipper]

You can not do this with bass traps. They are simply not effective below about 80 Hz, not even the bigger ones.

This is simply not true.

In my room, 3.86m x 4.2m x 2.4m high, simply installing two GIK Soffit Traps, one in each of the front corners, yielded the following results:

Green with no treatment, purple with the two bass traps. Not startling improvement but a little better. More impressive is the reduction in decay times:

Ignore the 40Hz issue which is road traffic noise.

And here is the room with lots of bass traps, including Soffit Traps in all four corners and all ceiling-wall corners:

Yes, both are with no smoothing. Not perfect by any means but much smoother. Using EQ I was able to complete the job.

I'm not saying subs can't do this. I've no experience of them and find the idea very interesting.

But can they also help with decay times? With all the treatment, the decay times came down from 400ms overall but with some low frequencies well over 600ms, to around 200ms right across the spectrum. I think it's that that makes the biggest contribution to the better sound.

Edit: this last question was answered by the links from March Audio's post 19, page 1, comparing the waterfall plot for one sub from the first link with that of four subs with DSP in the second link.

 

[Chromatischism]

I'm seeing a little bit of evidence below 80 Hz in your first measurement, but nothing that's outside of what I see from measurement variation. Your second measurement though shows a big difference. You definitely tapped into some issues with the additional corners.

Do you have a post-EQ graph?

RE: decay times, those are mostly down due to the reduction in amplitude, and will see a similar result from multi-sub + EQ. Do bass traps reduce decay even if amplitude isn't affected? That I don't know. Bass traps work to make the room appear larger, like adding stuffing inside a speaker. A room acoustician would have to answer that.

I have (currently) four GIK Tri-traps, and went with them because after measuring my space, I couldn't fit the soffit traps. I didn't see any changes below 80 Hz with those.

 

[Duke]

There is a crossover between sub(s) and mains. Unless you get the sub(s) phase to match mains phase around the XO freqeuncy they will never sum correctly.

So we need to back up and define what "sums correctly" would look like in the context of human hearing perception at low frequencies, and the answer is highly counter-intuitive. Briefly, "sums correctly" in this context looks like "smooth frequency response."

The in-room sound field at low frequencies is complex, with loudspeaker outputs and reflections and reflections of reflections combining in and out of phase, and mostly in between, in ways which change with location throughout the room. Fortunately there is a nice neat way to get an excellent grasp of the net effect: Look at the in-room steady-state frequency response.

Floyd Toole indicates that speakers + room = a "minimum phase system" at low frequencies, which in turn implies that what we perceive is predicted by the in-room frequency response. From page 200 of the first edition of his book:

"At subwoofer frequencies the behavior of room resonances is essentially minimum phase (e.g., Craven and Gerzon, 1992; Genereux, 1992; Rubak and Johansen, 2000), especially for those with amplitude rising above the average spectrum level. This suggests that what we hear can substantially be predicted by steady-state frequency-response measurements if the measurements have adequate frequency resolution to reveal the true nature of the resonances." [emphasis Duke's]

Let's look at the bass region as a whole first, and then come back to the crossover region. The two are of course related, but understanding the former facilitates understanding the latter.

Recall that the ear has poor time-domain resolution at low frequencies, but heightened sensitivity in the SPL domain. (The latter is indicated by the way equal loudness curves bunch up south of 100 Hz. A 5 dB change in SPL at 40 Hz is perceptually comparable to a 10 dB change in SPL at 1 kHz. Arguably, this is huge.)

So what matters most to the ears is what's happening in the SPL domain at low frequencies, which in turn is dominated by room interaction. Thus when we have optimized the room interaction for smooth frequency response, we have solved the problem that matters most to the ears.

The phase of the individual subwoofers can play a role, but it's probably not the role you are thinking. Let me introduce a term that describes a highly desirable but highly counter-intuitive property of an in-room low-frequency sound field: Decorrelation.

What we want is for the in-room bass energy to be decorrelated; that is, for it to NOT be all neatly in-phase. When the in-room bass energy is highly correlated, we have huge modal peaks and dips. When it is highly decorrelated, we have smooth bass. Room size plays a role: The larger the room, the more room-interaction peaks and dips we have (i.e. the greater the decorrelation), and the closer together they are and therefore the more perceptually benign they are. This is why large rooms usually have more natural-sounding bass than small rooms.

The ear will sum peaks and dips which are close enough together, rather than hearing them separately. This is what happens at shorter wavelengths (higher frequencies) in normal home audio listening rooms: A measurement of the in-room response includes what looks like "grass" because of all the reflection-induced peaks and dips, but we don't hear the individual blades of grass. Instead we hear a continuum (with some weighting due to arrival times, which is beyond the scope here). Our home audio rooms are far too small to achieve "grass" at low frequencies, but a distributed multi-sub system can be a worthwhile step in that direction because the summation of multiple dissimilar peak-and-dip patterns results in more peaks and dips, as well as smaller peaks and dips. In other words, a distributed multi-sub system is a way to get a small room to behave sort of like a larger room in the bass region.

And the mechanism by which a distributed multi-sub system does this in a small room does this is, decorrelation. It accomplishes decorrelation by spreading the bass sources around in the physical domain (locations). This spacing facilitates the outputs of the subs summing in semi-random phase throughout the room.

We can further increase decorrelation of the in-room bass energy by dialing in phase differences between the different subwoofers, which in turn can result in further smoothing of the in-room response.

At frequencies below the room's modal region (in the "pressure zone") the bass wavelengths are long enough for the outputs of the distributed subs to sum essentially in-phase, which typically results in a rise in the in-room response, which in turn can sound boomy. By introducing a significant phase variation between the different subs, we can extend semi-random phase summation to below the modal region as well, which can prevent that undesirable boominess.

Hopefully by now we have shown that the in-room frequency response is what matters most to the ears, and that decorrelation is a key to achieving smooth in-room frequency response (and I have nothing against also using EQ).

Now let's turn to the crossover between subs and mains. Typically that's at 80 Hz or less, so it's well within the region where what matters most is the in-room frequency response. I have nothing against getting a nice neat time-and-phase coherent transition in the crossover region, UNLESS doing so compromises what really matters: The frequency response.

On the other hand if we prioritize getting the in-room frequency response right, including in the crossover region between subs and mains, then we are solving the problem that matters:

"At subwoofer frequencies... what we hear can substantially be predicted by steady-state frequency-response measurements"

Phase aligning between drivers (in this case between sub woofer and main woofer) is simply a procedure that needs to be done around any XO point you create.
As you are speaker designer I'm sure we can agree on that.

If you had said, "Identifying and prioritizing the issues of greatest perceptual consequence (in this case between sub woofer and main woofer) is simply a procedure that needs to be done around any XO point you create", THEN I could agree with you.

 

[Duke]

If I recall, the mini-DSP running Dirac has about 40 milliseconds of delay. That is way out of bound for the gamer community and gets perilously long for AV sync. I wonder if ASR can measure this @amirm comes up with a PEQ or standard REQ test.

The latency issue you describe implies that one may not be able to simply apply DSP to the subwoofers; the main speakers may well require a compensating time delay as well.

And here is the issue I see, from my perspective on the commercial side of the fence: Virtually everyone I have dealt with who uses DSP has gone through multiple units, initially proclaiming that each unit is "totally transparent" and then becoming dissatisfied and moving on to the latest & greatest, which for a while will also be "totally transparent". (I have been a dealer for two different DSP manufacturers, but the return rate from customers who initially loved the unit and then became dissatisfied within a few weeks was too high for me.) As a subwoofer manufacturer, I don't want to always be playing catch-up to the "latest & greatest", which may change every few months. So I'd rather not need DSP in the signal path for my product to function effectively. If someone wants to add their own DSP unit because it makes a worthwhile net improvement, of course that is fine with me.

 

[QMuse]

So we need to back up and define what "sums correctly" would look like in the context of human hearing perception at low frequencies, and the answer is highly counter-intuitive. Briefly, "sums correctly" in this context looks like "smooth frequency response."

The in-room sound field at low frequencies is complex, with loudspeaker outputs and reflections and reflections of reflections combining in and out of phase, and mostly in between, in ways which change with location throughout the room. Fortunately there is a nice neat way to get an excellent grasp of the net effect: Look at the in-room steady-state frequency response.

Floyd Toole indicates that speakers + room = a "minimum phase system" at low frequencies, which in turn implies that what we perceive is predicted by the in-room frequency response. From page 200 of the first edition of his book:

"At subwoofer frequencies the behavior of room resonances is essentially minimum phase (e.g., Craven and Gerzon, 1992; Genereux, 1992; Rubak and Johansen, 2000), especially for those with amplitude rising above the average spectrum level. This suggests that what we hear can substantially be predicted by steady-state frequency-response measurements if the measurements have adequate frequency resolution to reveal the true nature of the resonances." [emphasis Duke's]

Let's look at the bass region as a whole first, and then come back to the crossover region. The two are of course related, but understanding the former facilitates understanding the latter.

Recall that the ear has poor time-domain resolution at low frequencies, but heightened sensitivity in the SPL domain. (The latter is indicated by the way equal loudness curves bunch up south of 100 Hz. A 5 dB change in SPL at 40 Hz is perceptually comparable to a 10 dB change in SPL at 1 kHz. Arguably, this is huge.)

So what matters most to the ears is what's happening in the SPL domain at low frequencies, which in turn is dominated by room interaction. Thus when we have optimized the room interaction for smooth frequency response, we have solved the problem that matters most to the ears.

The phase of the individual subwoofers can play a role, but it's probably not the role you are thinking. Let me introduce a term that describes a highly desirable but highly counter-intuitive property of an in-room low-frequency sound field: Decorrelation.

What we want is for the in-room bass energy to be decorrelated; that is, for it to NOT be all neatly in-phase. When the in-room bass energy is highly correlated, we have huge modal peaks and dips. When it is highly decorrelated, we have smooth bass. Room size plays a role: The larger the room, the more room-interaction peaks and dips we have (i.e. the greater the decorrelation), and the closer together they are and therefore the more perceptually benign they are. This is why large rooms usually have more natural-sounding bass than small rooms.

The ear will sum peaks and dips which are close enough together, rather than hearing them separately. This is what happens at shorter wavelengths (higher frequencies) in normal home audio listening rooms: A measurement of the in-room response includes what looks like "grass" because of all the reflection-induced peaks and dips, but we don't hear the individual blades of grass. Instead we hear a continuum (with some weighting due to arrival times, which is beyond the scope here). Our home audio rooms are far too small to achieve "grass" at low frequencies, but a distributed multi-sub system can be a worthwhile step in that direction because the summation of multiple dissimilar peak-and-dip patterns results in more peaks and dips, as well as smaller peaks and dips. In other words, a distributed multi-sub system is a way to get a small room to behave sort of like a larger room in the bass region.

And the mechanism by which a distributed multi-sub system does this in a small room does this is, decorrelation. It accomplishes decorrelation by spreading the bass sources around in the physical domain (locations). This spacing facilitates the outputs of the subs summing in semi-random phase throughout the room.

We can further increase decorrelation of the in-room bass energy by dialing in phase differences between the different subwoofers, which in turn can result in further smoothing of the in-room response.

At frequencies below the room's modal region (in the "pressure zone") the bass wavelengths are long enough for the outputs of the distributed subs to sum essentially in-phase, which typically results in a rise in the in-room response, which in turn can sound boomy. By introducing a significant phase variation between the different subs, we can extend semi-random phase summation to below the modal region as well, which can prevent that undesirable boominess.

Hopefully by now we have shown that the in-room frequency response is what matters most to the ears, and that decorrelation is a key to achieving smooth in-room frequency response (and I have nothing against also using EQ).

Now let's turn to the crossover between subs and mains. Typically that's at 80 Hz or less, so it's well within the region where what matters most is the in-room frequency response. I have nothing against getting a nice neat time-and-phase coherent transition in the crossover region, UNLESS doing so compromises what really matters: The frequency response.

On the other hand if we prioritize getting the in-room frequency response right, including in the crossover region between subs and mains, then we are solving the problem that matters:

"At subwoofer frequencies... what we hear can substantially be predicted by steady-state frequency-response measurements"



If you had said, "Identifying and prioritizing the issues of greatest perceptual consequence (in this case between sub woofer and main woofer) is simply a procedure that needs to be done around any XO point you create", THEN I could agree with you.

Ok, plz let me check if I got this right..

Let's assume you have 2 mains and 4 subs. You put those 4 subs around the room (following some best practice guidances) and dial random phase shift on their control boards so none of them is in phase with each other nor with mains, hoping for the smoothest summed response? Because "decorrelation" will take care of the rest?

Do you really think that is how modern bass management systems are supposed to work?

 

[Chromatischism]

I don't want to speak for him, but I don't think his response precludes there being an "optimal" phase/delay for optimal decorrelation. I think that is what systems like multi-sub optimizer and sound field management calculate.

 

[Duke]

Let's assume you have 2 mains and 4 subs. You put those 4 subs around the room (following some best practice guidances) and dial random phase shift on their control boards so none of them is in phase with each other nor with mains, hoping for the smoothest summed response?

That's not what I do, nor what I advocate, but I can see how what I wrote could have come across like that. Thanks for giving me a chance to clarify.

In my commercial multi-sub system, normally four subs are driven by a single channel of amplification. My system is designed for set-up by people who do not use measurements, though good measurements of course make it easier and faster and give better results. It is aimed at a price point and does not include DSP. The amplifier's phase control is used to fine-tune the blend between woofers and mains. Decorrelation can be increased by reversing the polarity of one of the subs.

Because "decorrelation" will take care of the rest?

"Decorrelation" is when the in-room reflections do not stack up atop one another to create huge peaks and dips. "Correlation" is when they do. You can try to fix correlation with EQ but you're going to need a lot of EQ. To maximize correlation, use a single large subwoofer and place it in a corner. This will give you the deepest loudest bass, but also the worst frequency response.

Do you really think that is how modern bass management systems are supposed to work?

No, I would expect a modern bass management system to be vastly more sophisticated and capable. But routinely I have customers measuring +/- 3 dB across the bass region, which is arguably pretty good. This often with zero bass trapping and often without use of the single band of parametric EQ included in my amplifier. (I have NOTHING against bass trapping or EQ or soundfield optimization, but my point is that a distributed multi-sub system can be a fairly effective solution on its own because the concept is based on sound acoustic and psychoacoustic principles.)

I don't want to speak for him, but I don't think his response precludes there being an "optimal" phase/delay for optimal decorrelation. I think that is what systems like multi-sub optimizer and sound field management calculate.

That's a very good way of putting it. I don't claim that what I do is the optimum; I claim cost-effectiveness in my price ballpark.

 

[Chromatischism]

@Duke

My personal experience (only 1 data point) tells me I am plenty sensitive to small changes to levels in bass frequencies, as you suggest. However, how do you explain the Fletcher-Munson curve which suggests otherwise?

Is it just a matter of being insensitive to absolute levels, while being sensitive to relative level changes?

https://mixingsound.files.wordpress.com/2016/05/fletcher-munson-curve.png?w=529

 

[Duke]

@Duke

My personal experience (only 1 data point) tells me I am plenty sensitive to small changes to levels in bass frequencies, as you suggest. However, how do you explain the Fletcher-Munson curve which suggests otherwise?

Is it just a matter of being insensitive to absolute levels, while being sensitive to relative level changes, correct? So we are sensitive to frequency.

https://mixingsound.files.wordpress.com/2016/05/fletcher-munson-curve.png?w=529

Yes, we are extra-sensitive to relative level changes. This is predicted by the way the equal-loudness curves bunch up south of 100 Hz.

At 60-80 dB, the vertical spacing between equal-loudness curves at 50 Hz is about HALF as much as the vertical spacing up a 1 kHz. A loudness (perceptual) increase of 20 phons corresponds to a 20 dB increase in SPL at 1 kHz, but a loudness increase of 20 phons corresponds to only about 10 dB increase in SPL at 50 Hz.

That can be kinda had to follow, so let me put it another way: A 10 dB change in SPL at 50 Hz sounds like as much of a difference as a 20 dB change at 1 kHz. And a 3 dB change at 50 Hz sounds like as much of a difference as a 6 dB change at 1 kHz. So a little change in SPL in the bass region goes a long way, perceptually.

To make matters worse, subwoofer amps tend to have teeny tiny knobs, making it that much harder to make fine level adjustments.

 

[Chris A]

However, how do you explain the Fletcher-Munson curve which suggests otherwise?

Equal loudness curves at low frequencies actually indicate much more sensitivity to relative SPL than frequencies above ~100 Hz. This is one reason why the typical practice of attenuating deep bass by "mastering guys" is such a chore to undo (demaster, that is). During demastering, if you're off by only 1-2 dB from the bass levels that were there during the recording (before mixing and mastering), you'll hear it most acutely.

The clue to what is occurring is that you must first get the deep bass levels back to a threshold of hearing level (usually quite high SPLs relative to higher midbass frequencies), then the equal loudness curves are so close to each other that you can easily blow it if you boost or attenuate the bass at any frequency between ~30 Hz--about the lowest frequency that you can hear in-room before tactile senses take over from the ear's hearing system--and 100 Hz. It takes a few trials at 83 dB (the magic mastering SPL level discussed elsewhere) to restore the bass levels correctly. If you demaster at lower or higher average SPL, your tracks will either have too much or too little bass response when played back at the reference 83 dB level at the listening position(s).

Chris

 

[Chromatischism]

That can be kinda had to follow, so let me put it another way: A 10 dB change in SPL at 50 Hz sounds like as much of a difference as a 20 dB change at 1 kHz. And a 3 dB change at 50 Hz sounds like as much of a difference as 6 dB change at 1 kHz. So a little change in SPL in the bass region goes a long way, perceptually.

We may be interpreting the equal loudness curves differently. My reading of the research was that the level in phons on the curves corresponds to how much is required to achieve equal loudness. As in, tones played at the levels indicated in those curves will all sound the same loudness to us. That means we are most sensitive between 2-4 kHz, and least sensitive as we approach 20 Hz. However the steep nature of the curve and the logarithmic spacing suggest modest relative changes at one narrow frequency compared to adjacent frequencies would be very audible. That has been my experience as well.

 

[Duke]

Tones played at the levels indicated in those curves will all sound the same loudness to us. That means we are most sensitive between 2-4 kHz, and least sensitive as we approach 20 Hz. However the steep nature of the curve and the logarithmic spacing suggest modest relative changes at one narrow frequency compared to adjacent frequencies would be very audible. That has been my experience as well.

^^This is my understanding.^^

Sorry that my wording was awkward; I'm not firing on all cylinders today, and probably should have waited for another day to post here.

 

[Chromatischism]

Thinking more about this, and from my own experience, I tweak my sub integration almost endlessly...and once I'm closing in on the right settings, meaning the crossover has been decided and the overall slope is dialed in, I listen for hours with my test tracks, making sub level changes in the Denon AVR on the order of 0.5 at a time. This is real OCD-level stuff. You've seen my bass response above. Small differences are audible and you can tell when you're getting close to the right balance. I either eventually decide on sub levels or fall asleep in exhaustion, whichever comes first