S&V Experiment: one 15, two 12, or four 8 inch subs

[amirm]

I love it when someone spends time and energy to figure out something instead of just talking about it. Such is the case with Brent Butterworth, constructing the thee different configuration subs in the title to see which performed best: http://www.soundandvisionmag.com/listening-tests/2011/03/subwoofers-4-2-or-1

"Everybody knows that four subwoofers are better than one. But are four little subs better than a single big one or two midsizers? Only a blind test could tell.

I can tell you in one paragraph how to set up a pair of small speakers, but I could write a book about setting up subwoofers. It’s the most challenging aspect of home audio because the resonances in a room tend to stress certain bass frequencies and strangle others. The effects of those resonances change from place to place in a room, so the sound may be perfect in one seat and a mess the next chair over.

A few years ago, researchers at Harman International found that using four subwoofers in a room — in the corners or at the midpoint of each wall — delivered smoother response across multiple seats than a single subwoofer could. But discovering that four subs are better than one creates a new conundrum for the audio enthusiast. Sure, a few of us have the luxury of asking, “Should I buy one kick-ass subwoofer or four kick-ass subwoofers?” But the question most of us must ask goes more like, “Should I spend my $1,200 subwoofer budget on one kick-ass subwoofer, two pretty good midsize subwoofers, or four little bargain subs?” In other words, would you get more benefit from the deep extension and powerful dynamics of, say, a big 15-inch sub, or from the smoother response of four 8-inch subs in the corners?

Unfortunately, finding a simple answer to this question may be impossible. Say you compared Brand X’s 15-inch sub to four of its 8-inch subs. Well, what if the driver in the 15-incher kind of sucks? Or what if the engineers tuned the 15-incher for tight bass while they tuned the 8-inchers for maximum output? Performance differences like these would make the results of a comparison meaningless.

The more I thought about it, though, the more I realized that while I couldn’t find the definitive answer to this question, I might be able to discover some useful clues. To do so, I had to create as level a playing field as possible. Which meant ensuring that the subs had as similar a design as possible. Which meant I’d have to build them myself...."

Don't want to quote the whole article. Please read it. It is very worthwhile.

 

[Ron Party]

Thanks, Amir. Worthwhile read indeed.

 

[DS-21]

I liked that he made an effort to hold Vd reasonably constant, but varied the number of subwoofers to get there. I would've also held total box volume constant (rather than Q - that is to say, the 8's would've each been in a cabinet with 1/4 the internal volume of the 15's cabinet, etc.), because we all know that low-end efficiency is dominated by the cabinet size, but holding Q constant is not an unreasonable choice. I also don't know why he wouldn't EQ for as-close-as-possible to equivalent response, because then he's comparing transfer functions and not subwoofers. His method "punishes" the multisub setup for having a likely higher LF cutoff, even when they could likely go lower (depending on their throw) with EQ.

Given the self-imposed constraints, the results were exactly what anyone with experience would expect: the sub with most volume displacement lumped in the most favorable loading gave the strongest lows, and the most subs gave the smoothest upper bass.

The problem with that article, as I see it, was the rigidity of the formulation: identical subs in every setup. There are no rational grounds for all of the subwoofers in a system to be the same size, or same design. The room will "tune" them all differently anyway, and chances are in a multisub system that proper setup will dictate very different levels for each one. We can take advantage of that fact by designing a system with different output capabilities at each spot. (That is to say, using the example of Dr. Toole's setup in Sound Reproduction, one can four identical sub and run two of them at -6dB and the fourth at -12dB, OR we can design two of the four subs to be ~4dB less capable and the fourth to be 9-10dB less capable.) By using multisubs smarter, it's possible to get all four desirable "sound quality" traits (deep extension, copious output, smooth upper-bass response, and low seat-seat variation) at a lower end cost and/or with a less intrusive visual footprint.

 

[FrantzM]

Agree with DS-21.. The subs do not have to be identical...

 

[Mark Seaton]

Agree with DS-21.. The subs do not have to be identical...

That all depends on what you are testing and considering. This is the assumption (same subs fed the same signal) of Todd Welti's study/paper posted at Harman, but with a rather important additional flexibility of having EQ available for the summed response. We also have to remember that the study was ranking based on a minimum of seat to seat variation over a potential area, not the smoothest response at a few defined listening positions. Rooms with greatly differing wall construction/rigidity (ie foundation wall vs only stud/drywall) can also throw the ideal results out the window.

The article makes for an interesting example/case study, but has quite a few variables in the mix, not the least of which is the ground plane frequency response and output capability of the subwoofers.

 

[Randy Bessinger]

Isn't that why Harman developed Soundfield Management? Which to me seems very similiar to Earl Geddes sub set-up method, just with less work and more flexibility (with Harman's being more expensive).

 

[amirm]

I think the article was well-intentioned. it tried to reduce some of the variables which in reality, are almost infinite in nature. Since it agrees with other experiments and research, it is reassuring that and adds to the conclusions of other work.

 

[Mark Seaton]

Isn't that why Harman developed Soundfield Management? Which to me seems very similiar to Earl Geddes sub set-up method, just with less work and more flexibility (with Harman's being more expensive).

Hi Randy,

The two approaches don't really conflict in concept at all, really only in execution and deployment. I don't think I'd label one more or less expensive, other than the fact that you can use something like a DCX-2496 or MiniDSP to try and follow Earl's approach. I haven't yet had time to look closely enough at the Sound Field Management execution and the utilization of the various subwoofers at VLF to make a comparison, but Earl's approach tends to leave much of the heavy lifting at lower frequencies to a single subwoofer which can work for or against you in different applications.

I agree with Amir that the general correlation makes for a nice litmus test and correlation/confirmation of the general concept. A few measurements with something like the Dayton OmniMic, XTZ Room Analyzer or a REW setup would be the next step in such an experiment/adventure.

 

[andy_c]

I don't think I'd label one more or less expensive, other than the fact that you can use something like a DCX-2496 or MiniDSP to try and follow Earl's approach. I haven't yet had time to look closely enough at the Sound Field Management execution and the utilization of the various subwoofers at VLF to make a comparison, but Earl's approach tends to leave much of the heavy lifting at lower frequencies to a single subwoofer which can work for or against you in different applications.

The one thing about SFM I don't like (as implemented in the BassQ - not sure about other implementations) is the fact that the test signal it uses only energizes the subs, not the mains. In the Geddes approach, the test signal energizes both mains and subs, so it seems to me the potential is there in Earl's technique to do a better job of mains/sub integration than in SFM. OTOH, it seems that realizing this potential in the Geddes approach is harder than in SFM due to the need for iterative manual adjustment of the DCX2496.

 

[DS-21]

Earl's approach tends to leave much of the heavy lifting at lower frequencies to a single subwoofer which can work for or against you in different applications.

I don't follow you there. As I interpret his approach, it has three prongs:

1) Use a "primary" (he calls it "ULF") sub augmented by "filler" (he calls them "BroadBand") subs the in modal region, set up sequentially.
2) Subs and mains ideally using only one side of the driver's output (closed, 4th order BP, presumably acoustic lever if anyone ever actually builds one) with sufficient volume displacement and/or excursion protection to not require highpass filtering, so all of the speakers (3+ subs, mains) can sum to contribute to room pressurization below that.
3) EQ to your desired low-end cutoff.

Under Earl's approach, if, say, you have a third sub that's calibrated 8dB down broadband but you want it to contribute more at ULF because it's capable of doing so, that's a simple matter of using a shelf filter to bring up the ULF on that third sub. While I doubt Earl (given his oft-stated position on the utility of response that low) would do that himself, it still falls entirely within his approach.

That sounds like an interesting hypothesis to test. As it happens, I'm going to be setting up a multisub system with closed-box subs in the near future. I think what I'll try, just to see if it works - in this room, with these subs in these locations, at least - the way I understand it to in real life. That is to say, I'll calibrate for smooth response in the modal region and then, based on reasonable estimates of each sub's ultimate output capability, bring up the lows on the "BB" subs so that they're contributing far more at ULF than they are in the modal region. Thoughts?

 

[amirm]

The one thing about SFM I don't like (as implemented in the BassQ - not sure about other implementations) is the fact that the test signal it uses only energizes the subs, not the mains. In the Geddes approach, the test signal energizes both mains and subs, so it seems to me the potential is there in Earl's technique to do a better job of mains/sub integration than in SFM. OTOH, it seems that realizing this potential in the Geddes approach is harder than in SFM due to the need for iterative manual adjustment of the DCX2496.

I thought the reason Earl's method energizes the mains at the same time is due to desire to turn them into additional "subs." It is not meant for blending them into mains one way or the other. The overall response may still be wrong int he overlap.

 

[Phelonious Ponk]

Interesting read, thanks for the link, Amir. The thing that strikes me is, if you're a dedicated listener sitting in the sweet spot, the feedback from the listeners is almost exactly what I'd expect to hear from an evaluation of single well-placed subs of those sizes. Though while squeezing the really deep rumble out of an 8" sub might not be likely, I think with the right component and driver choices and the right design (and probably significantly more money), the 12 and 15 inchers could be more articulate.

Tim

 

[andy_c]

I thought the reason Earl's method energizes the mains at the same time is due to desire to turn them into additional "subs."

That's true, but it does not preclude other potential advantages.

It is not meant for blending them into mains one way or the other. The overall response may still be wrong int he overlap.

If you look at Markus Mehlau's page, the white noise test signal is used and the combined system response is measured up to 200 Hz. That's well above the sub/mains overlap in almost every case one could think of. So if that response is optimized, then by definition one can say the sub/mains overlap is optimized as well.

 

[terryj]

So if that response is optimized, then by definition one can say the sub/mains overlap is optimized as well.

Not sure if I fully agree andy.

One thing that earle does not recognise it 'time' in the bass region. I think his comments go something like 'there is no evidence for audibility of time in the sub region' (only saying that to try and ensure I am not misquoting him)

I have flip flopped on that one lots of times!! In the end, I think I now (again, sigh, wish I could be consistent myself!) agree with him.

BUT, leaving aside audibility, let's move on to solely FR. Which brings us back to your point.

It is easy to see (and trivial to show) that changes in timing between the subs and the mains produce (often wildly) differences in the FR.

So the upshot is that after EQ and sub placement, we have a nice looking FR that we want (that should allow for individual preference).

BUT, we can apply eq to a graph that is 'correct' (ie the result of the subs and mains being in time with each other) or one that is 'wrong', ie the subs and mains are just playing when they want so to speak.

Ok, the absolute timing of the bass signal may not be audible, but eq surely is. So we can ignore time and eq the result...but that surely is using FR domain to correct things from the time domain, not gonna work really, OR we can set the timing right and then apply the EQ to what the eq should be applied to.

And usually, the FR looks far smoother with correct timing anyway, so we are left with less eq to apply and not filling in holes made by incorrect timing (which would not exist if the time was right)

Can I say I have an awful lot of trouble getting this message across to the EQ brigade? I don't know why it is like hitting my head against a brick wall (not referring to you obviously, just the usual 'who cares about time?' response. Heck, try talking about time when referring to the mids and tweeters! Boy, don't I get the backlash then!!)

(It would not be too difficult for me to post some graphs to show this if there is any interest)

 

[andy_c]

Okay, let me modify my statement a bit.

"So if that response is optimized in the frequency domain, then by definition one can say the sub/mains overlap is optimized in the frequency domain as well".

Better? I was implicitly assuming that only the magnitude of the response in the frequency domain was being considered per Earl's previous posts on the subject. It seems SFM could have the advantage in that regard if the need to consider phase turned out to be audible.

BUT, leaving aside audibility, let's move on to solely FR. Which brings us back to your point.

It is easy to see (and trivial to show) that changes in timing between the subs and the mains produce (often wildly) differences in the FR.

Right, but since a constant delay gives a phase lag directly proportional to frequency, this difference in response shows up most strongly in the upper frequency range of the subs. Probably it's most pronounced in the frequency region where the subs and mains "cross over" (even though Earl does not like to use that phrase).

Ok, the absolute timing of the bass signal may not be audible, but eq surely is. So we can ignore time and eq the result...but that surely is using FR domain to correct things from the time domain, not gonna work really, OR we can set the timing right and then apply the EQ to what the eq should be applied to.

Part of Earl's approach uses the technique of adjusting the delay in each sub channel individually using the delay adjustment feature of the DCX2496. There's also an all-pass filter in the DCX2496 for individual sub phase adjustments too. Poster jtwtrace over at audiocircle is using both those adjustments in his setup on a per-sub basis.

However, it's not clear to me what happens in regard to the relative delays of mains and subs when an analog preamp is used in this scenario. Because analog preamps have no way to adjust the mains delay, I don't think it's possible in general to equalize the delay of subs and mains for that case. Rather, there may be some kind of phase matching in the crossover region modulo 360 degrees going on. It's hard to tell.

And usually, the FR looks far smoother with correct timing anyway, so we are left with less eq to apply and not filling in holes made by incorrect timing (which would not exist if the time was right).

I think what Earl was trying to get across is that there's a lot more than relative delay going on here. He gave the example that, at low frequencies, in the time it takes to accumulate a cycle or two, there's a strong possibility that multiple reflections from different boundaries are all combining to form the resultant pressure vs. time seen by the mic. In that sense, the room modes dominate and the impact of relative delay between subs and mains could get swamped out. I'm not an expert in this by a long shot, but I think he's been saying something along those lines.

 

[DS-21]

Not sure if I fully agree andy.

One thing that earle does not recognise it 'time' in the bass region. I think his comments go something like 'there is no evidence for audibility of time in the sub region' (only saying that to try and ensure I am not misquoting him)

I don't want to speak for Earl, but I suspect he agrees with Toole.

To paraphrase, in bass in a small room, if you fix the FR you've fixed timing problems too.

 

[terryj]

well, that was how I understood earles opinion on the matter to be.

when andy mentioned just above about the delays on the dcx, I assumed he'd come to that position later in time than when I was discussing it with him (four or five years ago now???)

(hope I didn't offend you andy, none meant)

To paraphrase, in bass in a small room, if you fix the FR you've fixed timing problems too.

This is the point I am 'disputing'. I mean we can take it to silly extremes for sake of illustration, one sub five seconds behind the mains, well maybe you can fill that gap with eq on a sweep, but that would hardly fix the time issue.

Sure, as I said, a silly extreme example. The other 'perfect' example is when the subs and mains play at the same time (whatever definition you want for 'same time')

The question then becomes WHERE in that range of silly example time to perfect time do we hit the 'central' point?

I dunno, but if the mindset is to never check time at any stage...we cannot even begin to answer that question.

 

[andy_c]

I'm starting to agree with a lot of what you're saying, Terry. A distance difference between two sources (say one sub and one mains speaker) of about 1.72 meters is enough to cause a 180 deg phase shift in the direct component of the sound from each source at 100 Hz. That's more sensitive than I thought.

So just because the sound pressure at the measuring microphone is the resultant of the direct sound and many reflections from each of the two sources, does this mean that the delay difference between the direct components of the sound of each source is rendered irrelevant? Like you, I'm thinking it's not irrelevant at all.

At one point, I was thinking of using an analog preamp in a multi-sub setup, but now I'm starting to question the wisdom of that. Just taking the case of 1 sub, say in a corner, it's likely this sub is further from the listener than the mains are. So a delay would need to be introduced into the mains to compensate the difference in delay of the direct sound components of each - not possible with an analog preamp.

One interesting experiment might be to use an HT pre/pro and tell it that the mains are closer to the listener than they really are, by an amount that corresponds to, say, 5 msec. Then the pre/pro will insert a delay into the mains that's 5 msec too long. Then, when looking at the system response with only one sub active, initially set the delay in the DCX2496 sub channel to 5 msec also. Tweak the DCX2496 delay up and down by small amounts and observe the system response. Is the 5 msec value close to optimum? If it is, I think that would say something.

Edit: I should have added that the experiment above might best be done with the pre/pro configured for a conventional crossover arrangement, with the DCX2496 LPF configured for a very wide bandwidth.

It looks like I'll be getting an HT pre/pro after all, rather than DIYing an analog preamp.

Edit 2: Scratch that 5 msec number. It's probably best for that experiment to use a distance number for the mains that's too short by an amount corresponding to a delay error of 6.25 msec. That results in a phase error of 180 deg at 80 Hz for the direct sound.

 

[DS-21]

This is the point I am 'disputing'. I mean we can take it to silly extremes for sake of illustration, one sub five seconds behind the mains, well maybe you can fill that gap with eq on a sweep, but that would hardly fix the time issue.

In any reasonable sweep, you'd see that difference as a giant hole.

Here's what Dr. Toole actually wrote:

"However, because we know that low-frequency room resonances generally behave in a minimum-phase manner, we know that if there are no prominent peaks protruding above the average spectrum level, there will not be prominent ringing in the time domain. It is this indirect, inferential knowledge that permits us to confidently use frequency responses as a primary source of information about room behavior at low frequencies."

Sound Reproduction, at 245.

Also, Toole's comments on waterfall graphs, which is what most people use when they think they're looking at "time" issues:
"Before moving on, some things should be said about waterfall diagrams:
-They are highly decorative.
-They contain a lot of information.
-That information is compromised in both time domain and frequency domain axes, and the compromise can be manipulated to favor one or the other, but not both. In other words, one can have high resolution in the frequency domain and sacrifice resolution in the time domain, or the reverse. All of this is most relevant at low frequencies."
ibid. at 240.

 

[terryj]

hi andy

first off (and this is obviously addressing ds as well) I am not here to say the IS or WILL BE a problem, I am simply saying 'if we can confirm that the subs are time aligned then why not?'

I mean, look at it the other way. If it is true that there are no adverse consequences audibly from non rime aligned subs, then it stands to reason that equally there are no adverse consequences when they are perfectly aligned.

So, if you have the capability, all you have lost is a few minutes of your time. A potential costless upgrade, who can complain about that? When I view it from that angle I can't see why the objection.

As mentioned, bring it back to a purely EQ issue. I am 'telling' you that the response to which you will apply eq WILL vary depending on the relative distances. And (in my case here at least, can't speak for every system and every room) the smoothest graph requiring the least amount of EQ had always been when the distance/time is the right one.

We all agree that the best way to eq is to use the least possible?

Andy, I will in the next few days be setting up another profile, and will have to go thru this rigmarole again, so I'll save a few graphs and post them. That may or may not help. Watcha reckon?

It MAY be solely due to different distances (as in your example with the passive pre), or it can get more complicated than that analysis. Let's say you are using a deq or dcx on the subs (no stinking unit messing with my pure mains type stuff). Well, there is a thru time on the unit as well! Just by the use of that unit (let alone any processing) we have introduced delay.

Well, that's fine if the subs lead the mains! We can always delay them to match. We can't do the other way, delay the mains can we.

Anyway, often all you do is just eq whatever FR you have and not worrry about it.

Again, I am not saying this is a problem that must be handled, it is more a theoretical thing coupled with only applying to those that can do something about it. If you can, why not? If you can't, then carry on.



Hi DS, thanks for the toole quote.

I think we are talking about two different phenomena actually! I can play each sub, and mains, and they will have no ringing yada yada, ie minimum phase. However, back to the five second extreme example.... (knowing what it was, WHY did you try and tell me that hole could not be filled with EQ? C'mon. Think with what I am trying to say rather than reject it as you did when you said that)....each driver etc STILL reproduces it's 'perfect' minimum phase signal, just at different times.

the result can be wildly varying FR which will then be eq'd.

I am NOT talking about waterfalls. Yes, I can perhaps see the confusion, as you said waterfalls deal in the time domain too. But that is the time response of the room when stimulated with a signal, NOT the same 'time' as when each sub plays compared to the mains.

Same word, different meaning..time.

The simplest way to show this is with results. I will be clear about what changes where, and all can make a personal judgement of it's effects.

We could very well see that in normal rooms with 'normal' delays that we DO cross that point I mentioned above, the point where going from five second delay towards perfect is reached.

As I said, I do it because I can, and cannot have negative consequences, only positive.