Multiple Subwoofer Placement

[keithyates]

Hi, Nyal,

Your 2nd question first. We've talked about doing an AES paper. Just kicking it around. In the meantime, hey, there's work to do!

Which brings me to your 1st question. The process took 30-40 man-hours in the early days and is now down to about 20. The optimization routine we wrote -- and which "talks" to the CFD engine while it's beavering away through the night -- is the key to keeping the engineering time at a manageable level.

BTW, we don't "count" computer time, we just let 'er rip all day & night, but FWIW it takes a dedicated multicore multiprocessor 64-bit workstation to run the CFD simulations, with 32GB RAM a bare minimum. Processing time is sensitively dependent on the number of possible subwoofer locations we include in the study, and what the max allowable number of subs is. On the first job we did, I marked out 70 sub locations. Put a mic at each of the 8 head locations, moved the sub to location #1 and captured the impulse response at each seat. Moved the sub to location #2 and repeated until we'd gotten it all (70 sub locations x 8 mic locations = 560 IRs). Then modeled the same thing in software and compared them, overlaid the curves. ("Holy mackerel!" I screamed when I first saw the prediction curves overlaid on their measured counterparts.) Anyway, took the workstation a couple days to get through it.

Not there quite yet, but at the current rate of demand, in another 5 or 6 months the workstation will become the choke-point for this thing we call BOSS ("Bass Optimization & Specification Service"). In the meantime I have enough acoustical engineers with CFD chops to spread the work load; they just dial into the workstation from wherever they are and get it in the queue. (I've been at the CFD thing for 12 years now, and have made that specialty one of the things I look for in new hires.)

 

[Jeff Hedback]

That is awesome (truly meaning full of awe) Keith.

I use what I'd have to call the "little 'ol me" version of that concept using BEM modeling via ABEC. It does account for ~ any shaped room, speaker location(s), listener location(s), surface impedance and surface treatments ALL with limitations. Even with the limitations, it has been extremely helpful and accurate in both predictive design and diagnostic analysis situations.

 

[microstrip]

In the middle of the appreciations of Keith great posts, my question is embarrassing ...

What is the best place for a 15" sub-woofer speaker to be used in an IB configuration in the front wall? The front wall is 13 ' x 8 '. And in a similar size back wall ?

 

[Jeff Hedback]

Hello microstrip,

That's really not quite enough information to say "try this". However, I'll share my thoughts on how to start given your speakers. If most of what you enjoy is classical, place the sub 1'-2' right of midpoint between L & R speakers. If most of what you enjoy is produced mixes, shade the sub in same manner but right of center.

An old trick, but a great one, is to place the sub in your seat and crawl around until you hear the smoothest response. You'll want to pick a good track with smooth low end and play through the sub only and then full system to confirm.

Now that you have "location" settled, you'll want to focus on:
- volume
- crossover
- phase
- EQ
- in that order

 

[keithyates]

In the middle of the appreciations of Keith great posts, my question is embarrassing ...

What is the best place for a 15" sub-woofer speaker to be used in an IB configuration in the front wall? The front wall is 13 ' x 8 '. And in a similar size back wall ?

I'd start off with a couple of basic questions:

First, are you looking to optimize LF sound quality in only 1 spot in the room (presumably where you park your noggin), or are you looking to optimize (decrease spatial variance) through a broader region, say, at 2, 3 or 4 seats?

Secondly, is your room shape simply an extruded rectangle (a.k.a. parallelipiped, shoebox)? If not, e.g., if your room has an alcove or L-shape layout or slanted/cathedral/soffited ceiling or raised seating platform or an opening into another room or a window, my experience suggests taking rule-of-thumb type advice cum grano salis.

 

[microstrip]

I'd start off with a couple of basic questions:

First, are you looking to optimize LF sound quality in only 1 spot in the room (presumably where you park your noggin), or are you looking to optimize (decrease spatial variance) through a broader region, say, at 2, 3 or 4 seats?

Secondly, is your room shape simply an extruded rectangle (a.k.a. parallelepiped, shoebox)? If not, e.g., if your room has an alcove or L-shape layout or slanted/cathedral/soffited ceiling or raised seating platform or an opening into another room or a window, my experience suggests taking rule-of-thumb type advice cum grano salis.

Thanks for your suggestions. My room is a perfect parallelepiped, the speakers are the large Soundlabs and, as the front wall is adjacent to an empty space, I can use it for an IB subwoofer as well as the back wall. The room is parallelepiped like (13'x 8' x 30'), one third of it is used as my office, the main space as listening room. The walls are 2 feet stone, happily the front one is only 1 feet thick. I can have the hole for the speaker done any where in this wall or the similar back one, but can not consider experimental holes!
The speakers are placed 7 feet way from the front wall and I listen at 12 feet from the speakers. Most of time only one or two people seat at this distance.

 

[andy_c]

I'll be moving into my first home in a few months, so I can finally get some subwoofers! I've been looking into the whole Geddes multi-sub approach recently, and I thought I'd post a couple of things I recently found.

The page that gets referenced most often for how to do the setup and measurements for the Geddes approach is this one by Markus Mehlau. On that page, he repeats some instructions for how to do the measurements, originally given to him by Earl Geddes in this post on diyaudio.com in 2008. I've repeated part of it below.

Set your spectrum analyzer to a very low bandwidth but not less than 200 Hz and fairly long averaging time. This will yield a very long average of the sound signal. Take your mic and move it through large spatial positions near but arround the prefered listening position. Be carefull as small bumps of the cable can generate large erroneous signals into the mic. The sweeping has to be smooth. When the analyzer has completed its run you will have a plot of the frequency and spatial averaged low frequency sound field.

This raises a number of questions. What constitutes a "fairly long averaging time"? How quickly or slowly should the mic be moved when doing the measurement? How is the repeatability of the measurement given the human factor of waving the mic around in a somewhat uncontrolled manner?

It turns out I'm not the only one with such questions. There's a thread over at AVS forum, started in 2009, in which Roger Dressler asks Earl some similar questions. Earl's answer surprised me. It turns out he is no longer using the mic sweeping technique! Yet Markus' page is considered a de facto standard for how to do the measurements. Here's the pertinent quote from Earl.

I should explain that I have given up on the microphone sweeping technique because of too many problems. I now just use several fixed points in space and average the spectrums over those. I try and choose about 6 points that would be "typical" listener locations, but preferably not symmetrical as this would bias the results. The exact locations are not that important, but vary the height, width and distance back from the speakers in a semi random fashion, keeping inside of the "football".

I don't recall seeing that in any other thread beside the one at AVS, so I thought I'd bring it up here just in case others have not seen it. I've heard that REW has an averaging function, so I'll probably use that. I can imagine it would be useful to see each individual sweep as well, and not just the average.

The other concern I have is related to Earl's recommendation to not use a high-pass filter on the mains. My mains are of vented-box type, so I'd like to protect them from excessive cone excursion below the box tuning frequency. I may be misinterpreting this, but what looks to be a frequency response plot of one of the Geddes speakers is here, showing them to be -6dB at 80Hz. It's not clear which Geddes speaker it is, because if you look up the Summa, Abbey, Nathan etc., they all show the same plot, yet they have different woofers. Since the speaker's LF cutoff is so high, what he's saying makes more sense to me. He keeps harping on the large overlap between the mains and the subs, namely that the low-pass cutoff of the subs ends up being higher than the high-pass cutoff of the mains. But I wonder if he's overstating that, given the 80Hz cutoff frequency of the Geddes speaker in that graph. Just from an energy standpoint, it seems counterintuitive to me that it can be done with so much overlap without a bass hump in the overlap region. One cannot simply turn down just the woofer of the mains by itself to compensate for having multiple sources of bass in the same way one can adjust the level of each individual sub. Although he does say in this post that he would be fine with a 50 Hz high-pass. However, if those frequency response plots of the Geddes speakers are correct, and they are what I think they are, it seems to me that one could roughly duplicate that response with an 80Hz HPF when used with mains that are capable of going much deeper than 80Hz.

 

[amirm]

I think a few things are clear if I can call them that . One, this is work in progress as one clearly sees from him changing his process and view.

Second, the process is more random than predictable. Namely, instead of having specific placement for subs, he wants a number of them, including the mains acting as one above 50 Hz. By having all of these playing at once, there is likely more filling effect.

I am not a fan of his stance on No EQ. I think Earl often mixes what is the cheapest way to do things with what is the best way to do things. He concedes the EQ point later in that thread. To EQ, you need to have a high-resolution EQ and measurement. I have yet to see Earl's own room measurements but I bet it would show that it is not as flat as it could be. Resonances can be lowered with EQ as another powerful tool.

Not sure I am answering any of your questions but thought I summarize things a bit .

 

[Mark Seaton]

I'll be moving into my first home in a few months, so I can finally get some subwoofers! I've been looking into the whole Geddes multi-sub approach recently, and I thought I'd post a couple of things I recently found.

Along the lines of Amir's post above, I'd point out that Earl's approach is more so a continued effort to implement/execute what he has mathematically derived to be a catch-all approach...

IMO the approach makes some assumptions or dismisses priorities that are not of concern to Earl, but are to many. Some simple examples are that Earl is not very concerned with maximum headroom, where he assumes his speakers and subwoofers are more than enough. Earl also has little interest in extension & output below 20-25hz.

Give these assumptions and priorities consideration in how they will work out for you. I do believe Dr. Geddes to have a wealth of knowledge and skills, and often glean insight or perspective from his postings. I suggest many to learn from his approach and determine for themselves how well that will or will not work in meeting their expectations within their own system.

I don't recall seeing that in any other thread beside the one at AVS, so I thought I'd bring it up here just in case others have not seen it. I've heard that REW has an averaging function, so I'll probably use that. I can imagine it would be useful to see each individual sweep as well, and not just the average.

Averaging is very useful, but it is dangerous to not first give a look at the individual responses for trends in deviation. Straight averaging also assumes you place the same importance on every location. In a well behaved, and carefully planned room this can work just fine, but when real-world compromises come into play, we often have decisions to make. A great example is the situation of a rear row of seats backed against a rear wall... Do you leave the main row of seats with rather anemic bass if there is a 6-10dB rise in the lower regions at the rear row? Automagic systems and blind averaging do this without your input. Of course if you press Earl on this situation he would likely explain that he would never place one row at the back wall. Good info to know in applying a given approach to your room!

Averaging functions are available in the Room EQ Wizard, XTZ, and the Dayton OmniMic systems as a few examples.

The other concern I have is related to Earl's recommendation to not use a high-pass filter on the mains. My mains are of vented-box type, so I'd like to protect them from excessive cone excursion below the box tuning frequency. I may be misinterpreting this, but what looks to be a frequency response plot of one of the Geddes speakers is here, showing them to be -6dB at 80Hz. It's not clear which Geddes speaker it is, because if you look up the Summa, Abbey, Nathan etc., they all show the same plot, yet they have different woofers. Since the speaker's LF cutoff is so high, what he's saying makes more sense to me. He keeps harping on the large overlap between the mains and the subs, namely that the low-pass cutoff of the subs ends up being higher than the high-pass cutoff of the mains. But I wonder if he's overstating that, given the 80Hz cutoff frequency of the Geddes speaker in that graph. Just from an energy standpoint, it seems counterintuitive to me that it can be done with so much overlap without a bass hump in the overlap region. One cannot simply turn down just the woofer of the mains by itself to compensate for having multiple sources of bass in the same way one can adjust the level of each individual sub. Although he does say in this post that he would be fine with a 50 Hz high-pass. However, if those frequency response plots of the Geddes speakers are correct, and they are what I think they are, it seems to me that one could roughly duplicate that response with an 80Hz HPF when used with mains that are capable of going much deeper than 80Hz.

Earl's speakers basically are high passed, with a 2nd order acoustic filter being the sealed box for the woofer. Your vented loudspeakers should definitely be set to small, and through measurements you should look for the best integration with your subwoofers. There are plenty of cases where a speaker with anechoic extension to 30Hz will have a huge hole in the response at the listening area below 80Hz. The identical speaker can have significantly different observed extension when placed in a left/right vs. center location. This is no fault of the loudspeaker, but rather the interaction with the room.

I would argue that the most important starting point when reading any catch-all approach will be to determine the assumptions made, and how your room matches or deviates from those assumptions.

 

[Gedlee]

Hi Andy

One thing to keep in mind is that in the LF modal region the room dominates everything and as such there can be no more standard approachs than there are standard rooms - which means none. Everything is optional to achieve what works. There are certainly some "rules", like multiple subs makes this process easier and with better results, but nothing is absolute.

Contrary to statements made here, I do use EQ, but only in the modal region, never above 500 Hz since that would change the direct field and the direct field needs to be optimized in the loudspeaker not in the room.

Overlapping the subs with the mains is desirable, but not essential - remember, its whatever works, but I have found it far easier to get a flat response with overlap than without. And as people who have tried this comment, the audible integration of the subs to the mains sounds better that way. If there is a peak in the response from this overlap, then tame it down with EQ. In my designs I try and get about a gradual 3-6 dB boost at 40-50 Hz over what it is at say 200 Hz. This is to compensate for LF subjective effects in small rooms that requires more bass to sound nuetral. The overlap tends to help achieve this objective.

When using ported subs that are not tuned very low or use "smaller" woofers, a HP filter is always going to be required - one other reason why I never use ported mains. But using "small" on a surround setup is way too high, you just need to limit the cone motion below the ported box resonance so that it does not get excessive. A large speaker in a closed box is never going to have this problem.

As to "time averaging", all this means is that in the modal region we must be looking at the steady state since this is all that makes any sense. So the time of the measurement just has to be long enough that we have a steady state measurement.

And to Marks point - yes, nothing is going to cure a lack of sufficient output at LFs, so I always assume that this is not an issue. Of course, with multiple subs it's a lot less of a problem than it is with fewer subs. Below 20 Hz? Are you serious? Sounds more like bragging rights than reality.

 

[andy_c]

Hi Earl - good to see you back.

One thing to keep in mind is that in the LF modal region the room dominates everything and as such there can be no more standard approachs than there are standard rooms - which means none. Everything is optional to achieve what works. There are certainly some "rules", like multiple subs makes this process easier and with better results, but nothing is absolute.

Yes, I probably shouldn't have used the word "standard" to describe the techniques on Markus' page - maybe "guideline" would have been better. Anyway, what I'm trying to do is to leverage your experience and knowledge, so if there are some techniques you've abandoned due to too many problems, I'd want to go directly to your latest technique. One thing I've noticed with your approach is that, with the exception of Markus' site, the information is split up among many forum posts. So if there's a technique that supersedes what's on Markus' page, such as the abandonment of the mic sweeping technique, I'd like to use the best practice and also make note of the change so others might benefit.

Contrary to statements made here, I do use EQ, but only in the modal region, never above 500 Hz since that would change the direct field and the direct field needs to be optimized in the loudspeaker not in the room.

Yes indeed. I did notice this quote of yours from another thread that others might find interesting:

I have come to the conclusion that there are three regions in a small room. Above modal (the Schroeder frequency), "modal" and "mode". Modal is where there are several modes, but seperated enough as to be discrete. "Mode" is below "modal", where one can litterally have as few as one or two modes in the entire bandwidth of the source. In this case using multiple subs doesn't work and the sub litterally has to be fit to the room. I now realize that I have seen this in every room that I have done LFs in, including my own. The first mode in my room is relatively undamped, very discreate, very high gain and independent of the source location. A 1/2 octave or octave above this and everything changes. - multiple subs is the solution and EQ is almost not required. But in the "Mode" region, EQ is the only solution.

Now, regarding the overlap issue...

Overlapping the subs with the mains is desirable, but not essential - remember, its whatever works, but I have found it far easier to get a flat response with overlap than without. And as people who have tried this comment, the audible integration of the subs to the mains sounds better that way. If there is a peak in the response from this overlap, then tame it down with EQ. In my designs I try and get about a gradual 3-6 dB boost at 40-50 Hz over what it is at say 200 Hz. This is to compensate for LF subjective effects in small rooms that requires more bass to sound nuetral. The overlap tends to help achieve this objective.

Okay, this is starting to make more sense to me. Since your desired response has a hump, the overlap may help you achieve that response. But I am wondering one thing. How does one know that this LF subjective effect is a property of small rooms in general, and not of, say, having a bunch of recordings that are somewhat lean in this area (i.e. what Floyd and Sean call the "circle of confusion")?

When using ported subs that are not tuned very low or use "smaller" woofers, a HP filter is always going to be required - one other reason why I never use ported mains. But using "small" on a surround setup is way too high...

And yet a THX-compliant pre/pro or receiver will set a crossover frequency of 80Hz in this scenario. If I look at the response of one of your speakers (shown below), I see an 80Hz cutoff.

Aren't these two scenarios more or less equivalent though? That is, taking mains that have deep response and applying an 80Hz HPF, and using mains having the response above (-6dB at 80Hz).

Another thing - and I'm just thinking out loud here - is that I don't see a way in HT pre/pros or receivers to have them apply an HPF to the mains while leaving the sub out unfiltered (to allow the variable LPFs of the DCX2496 to be used on the subs). Since I have a music-only setup, I think there will be an analog DIY preamp in my future with the option of an HPF for the mains and unfiltered sub out.

I'm going on a house-hunting trip in another state, so I won't be able to respond until next week.

 

[Gedlee]

Aren't these two scenarios more or less equivalent though? That is, taking mains that have deep response and applying an 80Hz HPF, and using mains having the response above (-6dB at 80Hz).

Another thing - and I'm just thinking out loud here - is that I don't see a way in HT pre/pros or receivers to have them apply an HPF to the mains while leaving the sub out unfiltered (to allow the variable LPFs of the DCX2496 to be used on the subs). Since I have a music-only setup, I think there will be an analog DIY preamp in my future with the option of an HPF for the mains and unfiltered sub out.

Depends on what the characteristics of HP filter is and how close it is to the HP of the mains. If the electrical HP is second order and the HP of the mains is more than an octave below this, then they are roughly equivalent. Thats not usually the case however. Using closed box mains is just so much easier.

I don't think that any receivers do what you suggest either, but I don't do that and don't need to do that so its not an issue for me. I found out a long time ago that using multiple subs means that one wants the mains to be closed box as everything else just gets complicated. My Summas were originally ported, but then I sealed the ports on all of them - a much better solution.

 

[mirekti]

So I am digging this thread from the graveyard as I have a quite specific setup.
At the moment I am using P3ESRs as my main speakers which are HPFed @100Hz at the line level. Under these speakers I am using two Rythmik F8 subs which are fed by full line level signal through miniDSP.
In a way I made a full range speakers out of P3s which have capability of DSPing bottom two octaves.

If I wanted to apply this method, would I need to buy extra three subwoofers in addition to those two that I already own?
Or I could add only one or two for this to work?
In theory, I could put P3s back on the stands, feed them full range, and reuse current subwoofers to place one in the corner, the other somewhere else and purchase a third one for a full setup.

I also found this -> MSO https://www.andyc.diy-audio-engineering.org/mso/html/tut_intro.html so given the time when this thread was started, would it be better to keep everything as it is, add one subwoofer and try with MSO or decouple P3s from subs and go with the Geddes approach?

 

[sbnx]

Hello Mirekti,

I just completed a Geddes setup with full range speakers and 4 subs. 2 of the subs are Rhythmic F18's. The result was really good. You can not tell that subs are playing at all except that the bass is fuller and it gives a nice added sense of spaciousness to the sound. I am going to experiment a little more with sub placement etc to see if I can improve even more.

I looked up the measurements of your speakers. They have smooth bass rolloff and are back to flat at 80Hz (after the classic British midbass bump). This should be a really easy Geddes approach. I would say that since you already own 3 subs then decouple them from the mains to see what happens. If you don't like it you can always go back.

One note is that you are going to need a way to put in the Gain/EQ/delay settings necessary to get smooth (flat) bass. I use a Xilica XP4080 for this and it works really well. My preamp has all outputs active so it is easy for me to run a set of XLR's from the preamp to the input of the Xilica. Then 4 of the Xilica outputs go to the subs. (I do not run the mains through the Xilica) One nice feature is that it is super easy to mute the subs. So if I want to listen with or without them it is as simple as clicking "Mute". One other thing is that you could theoretically have a couple of target curves for the bass. It is very easy to select between them. So if you are in a rock mood you could select a curve with more bass and perhaps a bump from 60 to 120 Hz. For jazz you could select a flat curve etc. Dr. Geddes used a Behringer 2496 for his contoller. It is cheaper than the Xilica but some folks on this forum reported issues with the Behringer so I avoided it.

I will also add that there is a bit of a learning curve to this. Don't expect just to place the 3 subs in random locations and get it to work. There are others on this forum that run multi subs and they will likely chime in here as well.

Good Luck,
Todd

 

[DaveyF]

I am running two REL 8” subs in my very small dedicated room. To say that getting the ‘blend’ correct with two subs is a lot more difficult than one, would be a major understatement!
It took me about two months of experimenting to get the subs to disappear into the mix...so much that they are now invisible as to the ‘blend’.
Worth the effort, but it can be a little frustrating to keep on missing the mark...but once you hit it...it’s pretty obvious.