Revelation: Using multiple subs to cancel room modes

[noah katz]

I thought I was pretty well versed on room modes, but this was a revelation to me.

I learned about this in the Amps/Receivers forum of all places; thanks to M Code for the link to this Harman subwoofer white paper

http://www.harmaninternational.com/wp/index.jsp?articleId=1003

The revelation was that all odd-order axial modes (1st, 3rd, 5th, etc.) of a pair of opposite room boundaries can be cancelled by placing a sub at each one, two opposite walls, for example.

The attached picture of a room’s first mode’s amplitude and phase shows how this works. The key is that while the amplitude at a given distance from each opposite wall is the same, the phase is opposite, so two subs on opposite walls cancel each other’s odd-order modes.

The picture showing the second mode shows that the even-order modes won’t be cancelled because the phases are the same near the opposite walls. Other placements can cancel even-order modes but they’re not too practical (IIRC).

Besides cancellation, another strategy is to preventing a mode from being activated at all by place a sub at a mode’s null point. For odd-order modes this is a boundary’s midpoint, i.e., in the middle of a wall between the corners.

Both strategies can be combined by placing pairs of subs at the midpoint of opposite walls.

This is impractical for floor-ceiling modes, but assuming all listeners’ ears are at the same height, this can be handled with EQ, which BTW is assumed to be used in all the cases studied in the paper.

Dozens of combinations of quantity and location of subs were modeled and tested. The best, meaning smoothest response and least variation in response over a typical listening area away from the walls, was 4 subs, each at mid-wall position.

Very nearly as good and best bang-for-the-buck configuration was two subs at the midpoints of opposite walls.

A bit behind these two were 4 corner subs, and 4 subs consisting of a pair at intermediate positions on opposite walls.

All other combinations were much worse.

Other points of interest:

While a very large number of subs can eliminate all modes, even 50 in a normal sized room is way too few to do so.
Output level per sub (referred to as LF factor in the paper) does not increase linearly with increasing number of subs; once again a pair of subs is the sweet spot.

[DMF]

It makes sense that placing two subs along opposite walls cancels the odd-order modes. But it seems to make as much sense that they would aggravate the even-order modes. Odd-order has a null on the centerline, which is the primary sweetspot. Even order has a peak on the centerline. I wonder if the EQ is lopping off the peaks and net result just has the even-order nulls?

Doesn't seem like a bad compromise. ..

What about the other two axes, one wonders?

[KBK]

I just use custom acoustics packages. My friend who does this professionally, is now going to make his products available to the public, for the first time..and he's been using these patented 'tube traps' for over 15 years. His are called 'Pressure Zone Absorbers' and are much more effective/efficient than ASC tube traps. He's done about 50 films (film sets) now, besides everything else he's done. :)

[QQQ]

I use top secret egg shell cartons (wasn't this thead about subs ;)?).

Thanks for the thread Noah.

[DMF]

Gee, you guys really know how to make positive contributions to a thread.

Maybe this is old news to you, but what about the others here?

[QQQ]

That was exactly what my tongue in cheek post was pointing out. As usual KBK tried to use a thread as an avenue to brag about something completely unrelated to the thread. I'll try to make my comment less obtuse next time so as not to confuse anyone.

Sorry Noah, again, thanks for the thread.

Tom Holman has also been promoting the use of subs along opposite side walls with 10.2 for some time.

[rynberg]

Originally posted by DMF
It makes sense that placing two subs along opposite walls cancels the odd-order modes. But it seems to make as much sense that they would aggravate the even-order modes.

Ah, but in most rooms, the second mode of the width will be ABOVE the subwoofer crossover.....

[bpape]

The drawback to this is that neither sub then sits at the end of the length dimension to reinforce the deepest mode of the room (assuming length is the longest dimension). While generally speaking, supporting a room mode may not be desirable, in many cases, the sub(s) can use all the help they can get down in the sub 30Hz region.

Not only are we now not reinforcing this, but with the relationship of length to width in many rooms, we will be coming very close to reinforcing (with both subs sitting at the end of the width dimension) the 2nd mode of the length (as it is generally pretty close to the 1st width mode).

Using the terribly technical crawl around method ;) I usually find that best starting point is 1 sub on the front wall - say 1/3 from the right front corner and maybe a foot out from the wall. The other is usually 1/7 of the length from the front wall on the left side and a foot out from the wall. If the room is very short, you may need to move the left sub a bit farther into the room.

Yeah. There are times when that isn't THE best. BUT it rarely sounds bad. It doesn't short or overly reinforce much of anything and makes a good starting point to work from.

[noah katz]

Youre welcome Q, and don't worry, I understood you perfectly :)

"I wonder if the EQ is lopping off the peaks and net result just has the even-order nulls?"

TI think that's the case.

"The drawback to this is that neither sub then sits at the end of the length dimension to reinforce the deepest mode of the room (assuming length is the longest dimension). While generally speaking, supporting a room mode may not be desirable, in many cases, the sub(s) can use all the help they can get down in the sub 30Hz region."

A premise of the paper was that in these days of relatively affordable super capable sub drivers, sufficient output was available and emphasis could be placed on smooth response.

[bpape]

Agreed. However, a lot of folks considering the 2 subs option are those with:

Limited budgets
Limited floorspace
etc.

A lot of the questions on the board as about 2 little subs or 1 big sub. Few are as fanatical as some of us in terms of $$$ outlay for 2 subs that will really do 20Hz. I guess even fewer will pay $$$$$$ for one beastie!

How many people actually have the room to put 2 true 20Hz subs half way down BOTH of their sidewalls? Many barely have room for a walkway. I'm not disagreeing with the premise, just qustioning the reality of implementation within the average HT room.

[noah katz]

"How many people actually have the room to put 2 true 20Hz subs half way down BOTH of their sidewalls?"

Good point. As luck would have it, I have one so placed already, and could have the other one pretty close to the opposite wall midpoint by building it into my entertainment center when I re-do it.

[bpape]

Nice. Let us know how it turns out.

[noah katz]

It'll be awhile, I've got several projects ahead of that.

I was hoping Mark Seaton, Mysphyt, et al would pop in here. Oh well.

[noah katz]

Something else just ocurred to me, cancelling/not exciting the odd-order modes, in particular the first, would make bass traps work better, because they wouldn't have to deal w/the lowest frequency mode.

[bpape]

??? You're not going to deal with the room fundamentals with this arrangement. Best case would be the 3rd harmonic and you'll still have issues down that low unrelated to room modes in most cases - just caused by positioning and the other room dimensions.

[noah katz]

"You're not going to deal with the room fundamentals with this arrangement."

Why not?

The odd harmonics include the fundamental (1st mode), and it's cancelled for the opposite walls they're, and not excited by being halfway between the side walls.

[bpape]

Don't think so. You're sitting at the end of the width and @ 1/3 of the length and 2/3 of the length. You'll still have a pressure buildup of the fundamental at the ends of the length. You're also still at the same end of the height dimension.

[noah katz]

"You're sitting at the end of the width and @ 1/3 of the length and 2/3 of the length."

Huh? Re end of the width, unless you're disputing the white paper's results, that cancels the odd harmonics, including the 1st = fundamental. Midpoint = 1/2 an2 1/2, not 1/3 and 2/3, and according to the paper that prevents the odd harmonics for the width from being excited.

BTW these refer to the subs' placement, not the listeners.

[bpape]

My bad on the 1/3 vs. 1/2. Sorry.

Yeah. I totally agree that in theory it would deal with the odd harmonics of the length. I'm just not clear how placing subs at half way of the length but both on the floor and both against a side wall will do anything to deal with any of the modes of the width and height dimensions.

[noah katz]

No problem.

Correct, nothing is done for the floor/ceiling modes.

Not quite clear on what you're not clear on, but have you looked at the paper?

[SVonhof]

Most of this converstation has gone way over my head, but one thing I have to say... Not all rooms are rectangular. I wish I had a rectangular room. But, the builder of our tract home said that was the way it was designed, that was the way they would build it.

So, we have a room that is about 15.5' across the front and 13.5' across the back and 23' front to back. The switch from 15.5 to 13.5 happens about midway on the right side when looking at the screen. How do you figure that out?

I have two subs. One is on the right front screen wall, about a foot away from the right wall and about 10" out from the front wall. The other sub is currently about 2 feet from the back wall, centered along the 13.5' length. I was thinking of moving it to the front, next to the other one, but I just am not sure. I will try it after my recovery from surgery, but that will be a few weeks. The option to buy more subs is probably not a good one, since my wife just would not go for it...

[noah katz]

Here's a guy who tried it:

http://f20.parsimony.net/forum36475/messages/4708.htm

"Written by John Smolley at 30 Dec 2004 04:12:01:

Thought you guys might be interested to know that today, in prep for the IB plunge (literally n2 my crawl space), took my Infinity subs and tried them in one of the Harmon papers preferred locations--opposing each other mid wall. I should noter that the paper specifically cites center location front and back walls. For my set-up tried them centered on L and R walls. Locations that based on the "expert" advice I've heard, likely never have tried in 20 years of solitary confinement.

Results were impressive to say the least-and let me say this too--I tried the speakers with their attached towers in a variety of locations before, and with the included mic and test disk, was able to get some obective data. Data was sampled at about 1/10 octave resolution from 95 to 20Hz--best case before (w/o EQ) +/- 5.5 db but closer to a 8dB notch around 60--subjectively suck-out, loss of drum impact, but room shaking lows. Even the singl filter parametric EQ built into the system offered only modest improvement.

With Harmon white paper suggested placements, +/- 2dB MAYBE 2.5dB with two gentle troughs. Had I not the 2 subs to play with and the paper, would have never guessed. Also, below 70 Hz no directional cues noted--maybe some nearby objects rattling a bit more energetically, but with mains in place, this somewhat unorthodox placement will likely be unnoticed.
Guess I'm just suggesting this paper is well worth reading, esp if you are in the prelim stages of design. While it may rely heavily on computer modeling, the results I got seem to make a strong case for paying attention. Something bugs me about the front and behind suggestion, which in my room would have been close to stealth setup--I reckon if i added some delay to the rear I might overcome this prejudice, but the L & R wall is so good, even w/o EQ why bother?
Cheers to all,
John S. "

[Randy Mathis]

I've been debating this myself.

I have been wondering if I should do...

...an IB with 4 15" Dayton IB's at the front right of the screen...

...or...

...two sonosubs with a 12" shiva Mark IV each, tuned to about 18hz with one at the front right/left of the room and the other at the opposite.

Right not i'm hearing that the shiva's are not available but I can't start until Summer anyway.

Any suggestions?

The sonosubs would be cheaper.

I just wonder if they would play as deep.

I have a horrible RCA 12" sub right now. Nothing under 22hz and rattles and buzzes a bit above that.



Edit: I'm thinking that the two sonosubs is the way to go. That way I could eventually add two more for best results. The IB would be nice so as to save floor space.

[Mark Seaton]

Hi Noah,

Sorry about not jumping in here earlier, I've been fighting off a cold and been keeping fairly busy. I have posted the links to the three different Harman white papers and commented quite extensively in previous posts here. I have given this one some more thought in the past few months though.

An interesting alternate means of thinking about room modes is as multiple image sources rather than modes fitting into the dimensions of the room. This is in fact what is happening, where losses at the reflecting boundaries attenuate the image. While the usefulness of this view depends on your speaker knowledge, the room-subwoofer system can then be analyzed as a listener sitting within or near an array of subwoofers.

This is something we have to deal with all the time in the pro world. For example, when you have sources close enough with respect to the wavelength, the sources behave more in unison. This would imply that a placement resulting in more densely packed images would result in smoother response. Interestingly enough, the study's results follow this. For example, when using only 4 subwoofers, placement at the midpoint of each wall results in the closest image possible from each wall.

As I gave this more thought, I was continually frustrated with the realization that the rear wall images are a nightmare to work around, as the room depth is almost always long enough to make for very problematic response in the range of interest. I am working on convincing myself that the best solution would be a very lossy rear wall below 100Hz with a few subwoofers across the front wall.

Another interesting observation is that given the criteria of the white paper's model and the resulting images of the 4 subwoofers, I would expect a somewhat similar result from a single subwoofer placed dead center of the room. It makes an interesting arguement for an IB manifold into the attic or false ceiling.

Cheers,

[noah katz]

Hi Mark,

Ah, good, there you are :) Hope you're feeling better.

I'm not grasping the image concept. What exactly constitutes each image? What's lacking in the amplitude/position way of looking at it?

"I am working on convincing myself that the best solution would be a very lossy rear wall below 100Hz with a few subwoofers across the front wall."

That would be great, but would you do that, other than make the room behind the back wall into a bass trap?

Thanks

[noah katz]

How about this half-baked idea, kind of like sonic holography, where speakers cancel the undesired output from each other:

Rather than, or in addition to, EQ'ing each sub, how about using each sub to act as an active bass trap for the other? The advantage would be smoother response everywhere in the room because the modes would be prevented, albeit only partially, from developing in the first place.

The room dimensions and speaker locations would make this deterministic, but the variable of room RT and flexibility of the boundaries would probably mean this would take DSP and a measuring mike to be workable.

[Mark Seaton]

Originally posted by noah katz
I'm not grasping the image concept. What exactly constitutes each image? What's lacking in the amplitude/position way of looking at it?

There's really no difference in the result of a fully complete model as they are analogous methods. My point is only to offer another perspective on the situation. The same analogy can be used when modeling speakers where a single speaker on the ground can be modeled as a pair in free space, or a single speaker in a corner can be modeled as 8 speakers in free space or 4 in half space. While some have software models which include tangential and oblique modes, when considering solutions and in graphic representaitons it is rare to see people looking at more than two axies. The full picture of modal interaction can make for difficult understanding and makes for an even more difficult time of defining the basic requirements of a solution. We end up with some interesting concepts with ratios of possible relative placements for subs and listeners, yet all are quite conditional.

The image concept is quite simple. Each boundary is an acoustic mirror. If you mirror a subwoofer's location about each surface, and then mirror those images about the walls again, and again, you form an array of subwoofers in each axis (X, Y & Z). The images at greater distances become less significant, to a point you can be fairly comfortable with the trends found in modeling a finite number of images in each axis. A lossy boundary will reduce the intensity of an image, which can simplify things further. Of course it should be noted that flexible boundaries can sometimes have other odd effects on the time/phase relation. I believe this was noted in some of the Harman papers which cite this as a point of failing for the models.

This approach is more interesting when you consider much of the work being done in the pro audio world. We already have programs which will model interactions of multiple speakers quite readily. One difference here is that we are not in a far field situation where the polar pattern of the array matters, but rather the interactions in a specific near-field area. Again, this parallels much of the current work with line arrays and other steerable arrays. Some of these products have worked out software which actually optomize the near field radiation of the array per your input parameters, rather than you having to experiment and see what comes out.

Looking at an array of speakers we also can consider simple known concepts of what happens with respect to source(subwoofer) spacing with respect to wavelength. I had done a little modeling of this effect on a couple of situations and it did appear there was correlation with modal predictions. Again, looking at the problem in this context, I believe it no coincidence that the recommended configurations from the Harman paper result in more closely spaced arrays.

Originally posted by noah katz
How about this half-baked idea, kind of like sonic holography, where speakers cancel the undesired output from each other:

Rather than, or in addition to, EQ'ing each sub, how about using each sub to act as an active bass trap for the other? The advantage would be smoother response everywhere in the room because the modes would be prevented, albeit only partially, from developing in the first place.

The room dimensions and speaker locations would make this deterministic, but the variable of room RT and flexibility of the boundaries would probably mean this would take DSP and a measuring mike to be workable.

Many room designers will use bass traps in dealing with problematic room modes. A speaker could certainly be configured to act in a similar manner, but with a bit more flexibility. A couple years ago I was thinking about doing something similar using the open format, matrix DSP products in the pro-market (i.e. 2x2, 4x4 and 8x8 signal processors). My thought was that we might be able to use a pair of higher power subwoofers to simultaneously act as a subwoofer and a bass trap or counteracting source. In these DSP products it is very simple to split an input signal, delay one part of the signal with a wide variety of available modifications, and add it back into the original signal. I had created some configuration files in the DSP software, but never had the time to start looking into the process and pheasabiltiy of executing and calibrating such a system.

One of the major issues at hand is in clearly defining the problem (as it would apply to a realizable, dedicated home theater), as well as what the ideal solution would be.

Very thick absorption and bass traps a few feet deep could be possible along a single wall of a room, as I know some designers who already use 8-12" thickness of acoustical devices behind the fabric covered walls.

[noah katz]

Oh right, I'm familiar with the image concept for room boundary reinforcement. But for modes I don't get the benefit of this model; how do you visualize the effect of frequency and boundary flexibility on the images.

I see you're way ahead of me on the active bass trap idea. It seems like there's a lot of promise there.

One question is how much excursion it would take; first order one would guess it shouldn't take more than the sub that produced it, but I wonder if it needs more to compensate for the fact that the wave has now spread far and wide.

Perhaps MultEQ will make all this unnecessary, though, if the claims of time domain correction bear out.

Thanks

[DMF]

Well, I tried it tonight. And it bloody works!

I had one near mid-long wall and the other near the opposing front corner. I knew the placement wasn't good, but the room is still in progress. But tonight I got a wild hair and moved the one from the corner to the side across from the other (not quite mid-wall). Cranked up 'Fight Club'. HUGE improvement in the impact, and some in the extension. I'll have to take a plot soon.

Meanwhile, wife is having a bit of difficulty digesting Fight Club. :cool:

[noah katz]

"Well, I tried it tonight. And it bloody works!"

Great! Keep us posted.