The myth of generic optimum room dimension ratios

[FrantzM]

microstrip

You provided your own. My take on it is an unusually well researched and with thorough argumentation original and subsequent posts. Replies are clear to the point and backed up by clear quotes taken within the context of the original question. To me a model of clear and educative posts. Very far from the usual "I think" , "I believe I heard", or "I have a hunch" posting. It suggests that treatments, positioning have a lot more to do with final sound of a system than the mere dimensions on which we, audiophiles, tend to eternalize.

Such quality postings are rare, yet sorely needed in advancing our music reproduction system of which the room is the most important element. Making of it a sticky is a personal or comity decision not mine to make, only to suggest.

 

[Bruce B]

So if all modes are equally energized and are equally audible in the corners (Toole 2006), then why don't you just stick the source (speakers) in the corners and fuggetaboudit? Does this only work when the listener is in the opposite corner?

 

[GaryProtein]

Optimal room proportions are real, not the entire story, but as others have stated, a cubic room, or one with dimensions that are multiples of each other are less than perfect.

There are a few ratios that work well.

The golden proportion 1:1.6:2.5 (1.6:2.5 is a 1:1.6 ratio) works very well. Standing wave and room modes with that ratio are evenly spaced every 1/3 octave throughout the audible range--all things being equal.

10H x 16W x 25L (or my dream room of 16H x 25W x 40L) would be fantastic and would need minimal room treatment for a pretty even room response.

 

[KlausR.]

So if all modes are equally energized and are equally audible in the corners (Toole 2006), then why don't you just stick the source (speakers) in the corners and fuggetaboudit? Does this only work when the listener is in the opposite corner?

There is an AES paper where active modal control methods have been tested, by a listener in the room center, with music: Fazenda et al, „Subjective preference of modal control methods in listening rooms“, JAES 2012, p.338

One important result was that not all program material is equally well suited for such tests, which means that with some tracks modes may be excited, with other tracks they won't.

Tested were 8 configurations: 1 sub with and without EQ, 2 subs in phase/opposite phase, 4 subs in opposite phase + DSP. The configuration sub in corner + EQ was ranked 4th.

To answer your question: a sub in a corner will drive all modes equally loud. Only a listener in a corner will hear all excited modes equally loud, anywhere else there will be differences in perceived level between the modes. Fazenda has shown that for that centre position a single sub in a corner + EQ is a valid solution. The sub in a corner without EQ was the worst configuration.

Klaus

 

[JackD201]

Hi Klaus,

Red flags typically show up from 70 to 80Hz even in a "worst case scenario", the cube. As I did post earlier, I said the construction is more important. This dovetails with what you've posted about calculators not taking into account that surfaces are actually not perfectly reflective. In a real world test then, it isn't surprising at all that a cube might not be scored particularly worse if the material played did not have much sustained content in the range mentioned. Let's say the three rooms tested were all bare and reflective. I conjecture that all three rooms would probably hinder speech intelligibility and music reproduction. If we take the three rooms, finish and furnish them, they would all likely improve as well. Nothing much to do with the ratios. Hence my agreement that ratios are overstated.

Now take content that does have long sustains and decays, bowed basses, pipe organs, long reverb tails and the relevance increases. One could work out that it takes less sabines for ITU RT60 with a room with a better ratio than one with dimension multiples given identical room volume. That's what I meant by "making the job easier". Again, I don't want to overstate this, just adding another data point to the discussion. In practice, you take the room you have and deal with it ratios be damned. This of course includes the natural consequence of limiting choices for loudspeakers, their positioning and the listening position. If going ground up however, one might as well plan ahead especially if one is planning for a large system. There's no harm.

The sad fact of the matter is that there is one dimension that we don't have much wiggle room. Room Height. While these aren't standardized, there are conventions. The variance is often small with heights over 3m a rarity. Given that peg, realistically many a room will fall within the infamous room ratio "ameoba" chart, especially since architects rarely design square rooms anyway because flow of occupant movement is awkward (and they don't look good on paper ). Again this supports that the importance of ratios are overstated.

 

[Ki Choi]

my dream room of 16H x 25W x 40L

Hi Gary:

How much variance can this ideal room dimensions take? +/-5"?

If the first reflection surfaces would be 'treated' with bookshelves for an incident, would you add the depth of the bookshelves to the inside room dimension?

I really do like the idea of having a space and not having to 'treat' with stuff and still sound optimized.

Thanks,
Ki

 

[microstrip]

microstrip

You provided your own. My take on it is an unusually well researched and with thorough argumentation original and subsequent posts. Replies are clear to the point and backed up by clear quotes taken within the context of the original question. To me a model of clear and educative posts. Very far from the usual "I think" , "I believe I heard", or "I have a hunch" posting. It suggests that treatments, positioning have a lot more to do with final sound of a system than the mere dimensions on which we, audiophiles, tend to eternalize.

Such quality postings are rare, yet sorely needed in advancing our music reproduction system of which the room is the most important element. Making of it a sticky is a personal or comity decision not mine to make, only to suggest.

Frantz,

IMHO, you spotted the critical point - suggesting something that is never proved, but just as it appears mixed with quality posts from knowledge people, makes it credible. I have read many of the original texts, not only quotations, and, mixing with what I know about treatments, coming mostly from the usual "I think" , "I believe I heard", or "I have a hunch" postings, I would doubtless choose a golden ratio listening room, if I had to decide know.

 

[Ethan Winer]

But, how do you define "nearby"? What's the bandwidth of modes, is this bandwidth a strict limit, meaning that every neighboring mode within this limit is excited and every mode outside is not?

Sure, the Q of the modes is a factor, but so is their absolute frequency. A mode spacing of 20 Hz is less relevant at 300 Hz than at 40 Hz! My ModeCalc program takes this into account.

--Ethan

 

[FrantzM]

microstrip

I should have been more careful with my choice of words. I differ on the "never proved" and the "make it credible". If you can infirm the post by all means do. You haven't yet. The references are peer-reviewable. They are not the anecdotal references that litters audiophiles boards.
The posts never said also that choosing these ratios was a mistake simply that they are not the end of it all. With rather strong references I repeat not based what on people think they heard or they think they know.
Building a room from scratch I would sure take the safe road and not build a cubic or round room. The post does however reinforce my hunch that one can work and treat a given room to make it more than adequate. YMMV

 

[KlausR.]

IMHO, you spotted the critical point - suggesting [It suggests that treatments, positioning have a lot more to do with final sound of a system than the mere dimensions on which we, audiophiles, tend to eternalize.] something that is never proved...

If the experiments of Fazenda and Wankling are no proof, then what is? Also read chapter 13.2.1 (Optimizing Room Shape and Dimensions) in Toole's book, which you appear to have on your bookshelf. The Fazenda paper is listed in the book, the Wankling paper is not. If you want to read those papers yourself, feel free to ask for a copy.

We had the opportunity to build our listening room from scratch, so I used Bonello's criterion for determining the room's dimensions. The builders built the room wider than planned, and when you put the new dimensions (8.4 x 4.6 x 2.5) into Bob Gold's calculator you see that the criterion is no longer fulfilled. Is it a disaster? No, it isn't, and Welti's AES paper demonstrates why.

Klaus

 

[KlausR.]

Hi Jack,

Red flags typically show up from 70 to 80Hz even in a "worst case scenario", the cube.

That would be a common first order floor-to-ceiling mode, but I suppose that, unlike mine, most speakers do not have their woofer close to the floor, so that they would not drive that mode at maximum amplitude. Also, when being seated, the ears are at about half room height, hence close to the pressure minimum of that mode.

In a real world test then, it isn't surprising at all that a cube might not be scored particularly worse if the material played did not have much sustained content in the range mentioned.

Wankling used a method where an audio sample is convolved with a real room impulse response. The part below 250 Hz is removed and replaced by a modeled room, whereby decay times relate to those of that real room. Used were “short clips of average 6 seconds, chosen specically to excite low frequency energy whilst not being too complex in nature. Sample 1 is a short refrain with most of the low frequency energy in the drum track, and a few sustained bass notes. Sample 2 is a short double bass riff with a number of notes while sample 3 is a modern remix of a jazz track, which has clean drum and bass notes.”

Now take content that does have long sustains and decays, bowed basses, pipe organs, long reverb tails and the relevance increases.

Obviously the right content will do a better job than the wrong one, but it still depends on positions of speaker and listener. When you place the source in a pressure minimum of a mode, you can play that sustained note all day long but you won’t get that mode going. And even with the right content, will it be able to drive all modes equally and at the same time? Because that is what the concept is about: ALL modes simultaneously, not a few here and there.

One could work out that it takes less sabines for ITU RT60 with a room with a better ratio than one with dimension multiples given identical room volume.

If one calculates the number of reflections per second (equations 10.19, 10.21 in Beranek 1954) for a room of 100 cbm, for the golden ratio you have 118, for a room of 8 x 5 x 2.5 you have 123. If I take a ratio recommended by Ethan (1 : 1.6 : 2.33), it's 116 reflections.

Hence, on paper, for the same RT60 you need more sabines in the good room than in the bad room.

In practice, you take the room you have and deal with it ratios be damned. This of course includes the natural consequence of limiting choices for loudspeakers, their positioning and the listening position.

As I've mentioned before, my speakers (30 Hz - 20 kHz ± 1.5 dB) have been positioned without any regard to room acoustics, the listening sofa is against the wall. When I play pipe organ pieces (Bach, Händel, Saint-Saëns) I'm not troubled by room modes, those few tracks where the (0,2,0) mode is disturbingly driven are with synthesizer and bass guitar (funnily enough, that particular track is "Pressure Points" from Camel's "Stationary Traveller").

Klaus

Beranek, "Acoustics; Chapter 10: Sound in enclosures", MacGraw-Hill 1954

 

[microstrip]

If the experiments of Fazenda and Wankling are no proof, then what is? Also read chapter 13.2.1 (Optimizing Room Shape and Dimensions) in Toole's book, which you appear to have on your bookshelf. The Fazenda paper is listed in the book, the Wankling paper is not. If you want to read those papers yourself, feel free to ask for a copy.

We had the opportunity to build our listening room from scratch, so I used Bonello's criterion for determining the room's dimensions. The builders built the room wider than planned, and when you put the new dimensions (8.4 x 4.6 x 2.5) into Bob Gold's calculator you see that the criterion is no longer fulfilled. Is it a disaster? No, it isn't, and Welti's AES paper demonstrates why.

Klaus

I have read the F. Toole book when I was researching for my room, and know of your findings since you posted them in Audiosylum some years ago - I appreciated the original version of your post and your contributions in many forums! I have not read the Fazenda and Wankling, I will send you a pm with my mail address, as I am very interested in them and thank you for your kind offer.

Toole'a approach to bass frequencies reproduction is well know - equalization and proper placement of several subwoofers can solve any problem and get optimum bass in a for reasonable number of people in a multichannel system, covering a significant part of the room area. I do not know your system, but you probably use this approach. This is completely different from a typical audiophile situation, that places his stereo full range speakers for optimum tonal properties, imaging, enveloping and enjoyment, and simultaneously wants good bass. I am not a acoustics expert, but it is my feeling that having a room with the golden ratio dimensions, that avoids superposition of room modes can create an environment where we will have less constrains to speaker placement, and logically lead to better sound reproduction.

Just to be clear, my position is mostly grounded in the refusal of using any kind of equalization in my full range stereo system - that I think is the approach of the majority of WBF audiophiles.

I am happy that your builder mistake was not a disaster, but how do you know if the room was built with the initial dimensions it would not sound better?

 

[JackD201]

If one calculates the number of reflections per second (equations 10.19, 10.21 in Beranek 1954) for a room of 100 cbm, for the golden ratio you have 118, for a room of 8 x 5 x 2.5 you have 123. If I take a ratio recommended by Ethan (1 : 1.6 : 2.33), it's 116 reflections.

Hence, on paper, for the same RT60 you need more sabines in the good room than in the bad room.

Quick time out Klaus, why are you counting the number of reflections per second for a sabine calculation? A bare room with multiples will have an inherently higher RT60 than one without, assuming identical surface construction. So that should require less sabines to achieve ITU RT60 or any target RT60.

Lets say 3m (H) x 6m (W) x 6m (L) vs. 3m (H) x 5m (W) x 7 (L) with the same construction. For simplicity let's assume no windows and doors. The difference in RT60 is .02 or so at 125Hz.

You might say sure but that's .2 out of an RT60 that is over 1 second. To put things in prospective domestic targets are less than half the bare RT60, critical environments less than a third.

I did say, construction is more important than dimensions these figures are simulated with code you would find here in the pacific rim of fire (concrete) but the difference will still be there if it's drywall or gypsum. I also agree that placement can override many effects e.g. find the best spots in whatever room you have. Doing so however does not eliminate ringing it just puts you in a zone with less reflections. The energy dissipation remains the same.

Perhaps we just may have a difference in priorities? I like even coverage throughout a space not one spot in a room. That makes dimensions just a little bit more relevant for me.

 

[GaryProtein]

Hi Gary:

How much variance can this ideal room dimensions take? +/-5"?

If the first reflection surfaces would be 'treated' with bookshelves for an incident, would you add the depth of the bookshelves to the inside room dimension?

I really do like the idea of having a space and not having to 'treat' with stuff and still sound optimized.

Thanks,
Ki

I am sure plus or minus a few percent would not be a problem. Computing measurements where bookshelves were used for first reflections, it would probably depend on how filled up the bookshelves were. If they were completely filled, it seems like it would just be like a narrower room, but if the shelves were sparsely filled, leaving spaces between clusters of books, they would be better absorbers of first reflections.

 

[KlausR.]

Toole's approach to bass frequencies reproduction is well known - equalization and proper placement of several subwoofers can solve any problem and get optimum bass in a for reasonable number of people in a multichannel system, covering a significant part of the room area. I do not know your system, but you probably use this approach.

My system is 2-channel, and since the rears can’t be placed where they should be I never investigated multichannel. As I see it, the angles for multichannel are based on old research (Damaske et al., “Interaural crosscorrelation for multichannel loudspeaker systems, Acustica 1972, p.232) and should be respected for optimum (psychoacoustic) results.

I am not a acoustics expert, but it is my feeling that having a room with the golden ratio dimensions, that avoids superposition of room modes can create an environment where we will have less constraints to speaker placement, and logically lead to better sound reproduction.

It is true that in rooms where one dimension is equal to another or a multiple of another the chances of getting boom are greater than in other rooms since there are spots where the two modes of the same frequency can be excited together. But even in that unfortunate case, how often would that happen? You need tracks that contain notes of that frequency and the duration of that note has to be long enough (or be played in fast repetition) to get the modes going.

However, the main argument has still not been invalidated: optimisation concepts aim at all modes, and you won't drive all modes at the same time at the same strength, except in a corner, and even then, you won't be able to do that with music.

Just to be clear, my position is mostly grounded in the refusal of using any kind of equalization in my full range stereo system - that I think is the approach of the majority of WBF audiophiles.

My speakers have an on-board parametric equalizer, which is used at the factory for the basic settings (flat frequency response, zero group delay, pair matching etc.) so it's constantly operational. Yet this speaker got very positive rewiews in the pro-mags (absolutely neutral, extremely precise imaging) and in the consumer mags (effortless, natural, extremely precise bass, exceptional speaker). Note that that EQ is FIR, which means that amplitude and phase are adjusted independently.

I am happy that your builder mistake was not a disaster, but how do you know if the room was built with the initial dimensions it would not sound better?

As I said, modes are a problem only very occasionally (found in 10 years only 3 tracks that drive the (0,2,0) mode), and I don't think that optimized dimensions as such would have made things better. Of course, with a different width dimension, the frequency of that (0,2,0) mode would have been different (79 instead of 74 Hz), and would probably not been driven by those tracks, but since it's a very rare event, and the solution is to simply move the head (or move the speakers, or use that on-board EQ for a narrowband intervention), it's no big deal.

Klaus

 

[microstrip]

KlausR,

We must thank you for bringing this debate to WBF. I will not go on debating as your implementation is very different from what I usually consider (you are using speakers with a parametric equalizer) and this completely changes the perspectives.

As you refer to listening impressions to support your views in your room it would be nice to know the exact dimensions, speaker and listener positions, and perhaps know the exact model of speakers you are using.

This debate was carried some time ago in several audio forums with much more knowledge participants than me. After going through them, my current views in these matters are more in line with the arguments of David Aiken in this debate. A Google search and a few keywords will find many more interesting debates.

http://www.audioasylum.com/cgi/t.mpl?f=rives&m=6590.

If you're building a house from scratch and designing a room, I'd consider one of the ideal dimension ratios that have been documented, not because I think it will automatically give better results but simply because the ratios I've seen deliver rooms with good proportions for placing speakers and the listening position. They're not long and thin which can cause placement problems and they aren't square which I find not a particularly visually pleasing shape for a room, and which also doesn't give you much in the way of options for placement. A well proportioned rectangular room gives you both long and short wall placement options. That's a plus in my view, and enough of a reason to consider using one of the ratios if you're designing and building the room. If you're buying an existing house you don't get the option and, provided the room isn't a long thin one, I think you can get reasonable results with any room, including a square one, if you're prepared to put a bit of effort into finding the best placements for the speakers and listening position. The problem with a room that's too long and thin is that there are no good placement options—the speakers end up too close together or too close to the walls if placed on the short wall, and you end up too close to the plane of the speakers if they're placed on the long wall. A long, thin room simply doesn't let you place the listening position and speakers in a reasonably proportioned triangle for good stereo results, regardless of how the room behaves acoustically.
David Aiken

 

[KlausR.]

Jack,

Quick time out Klaus, why are you counting the number of reflections per second for a sabine calculation? A bare room with multiples will have an inherently higher RT60 than one without, assuming identical surface construction. So that should require less sabines to achieve ITU RT60 or any target RT60.

The number of reflections per unit of time is an indicator of how fast sound decays: the more reflections, the faster the decay, hence the lower RT60.

Another way is to look at how fast sound decays. Sound pressure level decays at a rate of 1.085 (cS/V) [-2.3 log10 (1-?)] in dB/sec (equation 10.27 in Beranek 1954)

with
c = speed of sound in air
S = total area of absorbing surfaces
V = room volume
? = absorption coefficient

As can be seen, for our hypothetical case of ? and room volume remaining unchanged, RT60 entirely depends on S, which is different for different dimension ratios. In my example (100 cbm, golden ratio vs 8 x 5 x 2.5), S for the golden ratio room is smaller than for the room with multiples (140 sqm as compared to 145), hence there is less decay per second, which means that RT60 is higher than in the room with multiples.

Lets say 3m (H) x 6m (W) x 6m (L) vs. 3m (H) x 5m (W) x 7 (L) with the same construction. For simplicity let's assume no windows and doors. The difference in RT60 is .02 or so at 125Hz.

These two rooms have volumes of 108 and 105 cbm. If we change that to 2.78 x 6 x 6 and 2.86 x 5 x 7 (to get a volume of 100 cbm), then S is 138.72 for the square and 138.64 for the rectangular. RT60 would be slightly higher in the rectangular.

If I take a 100 cbm cube, then S is 130, RT60 is hence higher.

To summarize and rank from low to high RT:
1. Multiple: S = 145
2. Golden ratio: S = 140
3. Square: S = 138.7
4. Rectangular: S = 138.6
5. Cube: S = 130

Unless I've made mistakes during number-crunching, these examples show on paper that your assumption "a bare room with multiples will have an inherently higher RT60 than one without” is wrong.

You might say sure but that's .2 out of an RT60 that is over 1 second. To put things in prospective domestic targets are less than half the bare RT60, critical environments less than a third.

FYI, threshold of detection for differences in RT60 has been determined to be 0.05 sec:

Niaounakis et al., “Perception of reverberation time in small listening rooms”, J. of the Audio Engineering Society 2002, p.343.

A 0.02 second difference would probably go unnoticed.


Klaus

 

[KlausR.]

As you refer to listening impressions to support your views in your room it would be nice to know the exact dimensions, speaker and listener positions, and perhaps know the exact model of speakers you are using.

The exact dimension are 8,52 x 4,60. Height is tricky because we’ve got that stretched fabric as ceiling at 2.5m (Bosmans, “Sound absorption of stretched ceilings with an impervious synthetic membrane”, J. of the Acoustical Society of America 1999, p.233), see Toole’s book, p.211.

Speakers are on the long wall, right is 2.1m, left is 3.85 from the side wall, distance woofer – front wall is 0.60. Listening place is on the opposite wall, with sofa against the wall. Speakers are Klein+Hummel O500C.

This debate was carried some time ago in several audio forums with much more knowledge participants than me. After going through them, my current views in these matters are more in line with the arguments of David Aiken in this debate.

I basically agree with what David said, i.e. that a well proportioned rectangular room is more user-friendly in terms of options for placing speakers and listening chair than a square or a cube.

Klaus

 

[audioguy]

Hi Klaus,

Red flags typically show up from 70 to 80Hz even in a "worst case scenario", the cube. As I did post earlier, I said the construction is more important. This dovetails with what you've posted about calculators not taking into account that surfaces are actually not perfectly reflective. In a real world test then, it isn't surprising at all that a cube might not be scored particularly worse if the material played did not have much sustained content in the range mentioned. Let's say the three rooms tested were all bare and reflective. I conjecture that all three rooms would probably hinder speech intelligibility and music reproduction. If we take the three rooms, finish and furnish them, they would all likely improve as well. Nothing much to do with the ratios. Hence my agreement that ratios are overstated.

Now take content that does have long sustains and decays, bowed basses, pipe organs, long reverb tails and the relevance increases. One could work out that it takes less sabines for ITU RT60 with a room with a better ratio than one with dimension multiples given identical room volume. That's what I meant by "making the job easier". Again, I don't want to overstate this, just adding another data point to the discussion. In practice, you take the room you have and deal with it ratios be damned. This of course includes the natural consequence of limiting choices for loudspeakers, their positioning and the listening position. If going ground up however, one might as well plan ahead especially if one is planning for a large system. There's no harm.

The sad fact of the matter is that there is one dimension that we don't have much wiggle room. Room Height. While these aren't standardized, there are conventions. The variance is often small with heights over 3m a rarity. Given that peg, realistically many a room will fall within the infamous room ratio "ameoba" chart, especially since architects rarely design square rooms anyway because flow of occupant movement is awkward (and they don't look good on paper ). Again this supports that the importance of ratios are overstated.

White papers and research notwithstanding, my experience in three different rooms suggests room dimension play a huge role (NOT the only role) in the final outcome.

Room #1 was 16 x 16 x 8. As you might quickly determine, I had a huge issue in the bass and while EQ helped some, it did not help a lot.

Room #2 was 8 x 16 x24. And had 15db bump at about 65Hz (ish). Corner traps helped as did digital room correction but ONLY dead center of the main listening position. I tried moving speakers, subs and listening position and while it helped some, other portions of the audio spectrum got worse.

As suggested, there are other factors at play. In room number 2, I intentionally built ridged walls (studs on 1 foot centers, double dry-wall, etc).

In my current room, while the dimensions are similar to room #2, they are enough different to have more spacing along the frequency axis of the "bumps" so that multiple occurrences of the same "bump" are not right next to each other (unlike room 1 and 2) I further addressed the issue by purposely allowing the bass to leak through the walls (1/2 drywall on 16 inch centers) and my bass response is MUCH smoother.

But I still maintain the room dimension should certainly not be ignored.

 

[JackD201]

As is my position. "Irrelevant" is much too strong a word and I believe this word was taken a bit out of context. Everything matters.