The myth of generic optimum room dimension ratios

[KlausR.]

Hi all,

at the time when I started this thread I was not aware of the possibility to attach pdf files. So here's the complete piece about room dimension ratios with drawings etc.

Klaus

View attachment Room dimensions.pdf

 

[j_j]

Indeed, nulls are usually the larger problem, except maybe in square and cube shaped rooms. Peaks are typically 6 dB or less, but nulls are often 30 dB deep or even deeper. Further, the worst null is usually due to reflections from the wall behind you. The frequency of that null is related more to your distance from that wall than the room dimensions.

--Ethan

One has to remember why nulls exist. A null in a room does not mean that there is no energy present in that room at that frequency, it means, rather, that the room stores an excessive amount of energy at that frequency, and the PRESSURE (for an Omni mike) or .5*Pressure+.5*velocity_in_one_direction (for a cardioid mike), is zero at that point. However, a full measurement of the soundfield at the same place will show a substantial volume velocity at the very same place, showing the actual storage of energy in the room, and showing that there is a lot of energy at that point, just not in the form of pressure, or whatever.

Note, by the way, the cardioid and the Omni will not show the null in the same place, either. What's more, at any frequency above a few hundred Hz, your head will hear a null in yet a third place.

Nulls are a problem, indeed, because the room stores too much energy at that frequency, resulting in a pressure (typically) null.

Since your eardrums respond to pressure, at very low frequencies (under 100Hz or so) an Omni more or less matches what your ears will hear. Above that, since your head interferes with the volume velocity field (which by the way is the source of head related transfer functions) your hearing apparatus will respond to a complex function of the 4 variables present, in fact, the 4 variables around your entire head.

N.B. To understand that it is necessary to realize that there are in fact 4 variables at any point in space for any measurement of an acoustic wave, the pressure and 3 (x,y,z) volume velocities. A microphone with one output only captures 1/4 of the information at even a single point in space. That's what the "Soundfield Microphone" is all about (Calrec, etc).

The way I would put the room shape question is that it is possible to have very BAD room dimensions, but that there is no magic GOOD set of dimensions, ratios, or whatever.

Hence room treatment. It's not just a good idea, it's physics.

 

[KlausR.]

Since your eardrums respond to pressure, at very low frequencies (under 100Hz or so) an Omni more or less matches what your ears will hear.

No, it won't, simply because your hearing, while processing the sound data, uses equal loudness curves and the Omni does not. That's why a measured 6 dB peak and 30 dB dip are not necessarily a bad thing, they may both be on one and the same curve of equal loudness and be hence be perceived as equally loud.

Klaus

 

[JackD201]

Where's the popcorn. This is gonna be good.

 

[amirm]

No, it won't, simply because your hearing, while processing the sound data, uses equal loudness curves and the Omni does not. That's why a measured 6 dB peak and 30 dB dip are not necessarily a bad thing, they may both be on one and the same curve of equal loudness and be hence be perceived as equally loud.

Klaus

JJ is not saying that the SPL meaning is the same as every frequency. But rather the fact that perceptually, measuring SPL and judging it at any one signal frequency in sub region, can pretty closely resemble what we hear. Boost a 50 Hz response by that 30 or 6 dB and it will sound different to your ears proportional to those numbers (more or less).

Also, equal loudness does not play a role here since presumably whoever created the content was subject to it just the same.

 

[Ethan Winer]

One has to remember why nulls exist. A null in a room does not mean that there is no energy present in that room at that frequency, it means, rather, that the room stores an excessive amount of energy at that frequency, and...
<snip>

Excellent points about the physics, JJ. That's why you get the big bucks!

--Ethan

 

[microstrip]

(...) The way I would put the room shape question is that it is possible to have very BAD room dimensions, but that there is no magic GOOD set of dimensions, ratios, or whatever. (...)

Now the question should be : can we have a software utility that can identify ALL the very bad rooms looking at their dimensions? As far as I know we only have a few very crude receipts for very bad rooms, such as as a cube, two equal dimensions or exact multiples.

 

[amirm]

If you do full modeling of the room like we did for our theater, then you can determine how "bad" the room is. You get neat looking animations like this:

I is expensive to do though. Currently it costs a few thousand dollars to create the model and run it with different speaker configurations to show the optimal one. My dream is that one day you can do it with an online tool where you draw your room and it runs the simulation for you.

 

[j_j]

No, it won't, simply because your hearing, while processing the sound data, uses equal loudness curves and the Omni does not. That's why a measured 6 dB peak and 30 dB dip are not necessarily a bad thing, they may both be on one and the same curve of equal loudness and be hence be perceived as equally loud.

Klaus

Sorry, Klaus, it's the same excitation your eardrums get. And that, ladies and gentlemen, is what you care about.

I am rather aware of equal loudness curves, to say the least, in fact the infamous Lucent vs. Microsoft MP3 lawsuit was quite about them, and I was, after all, the inventor of the patents being fought about.

My point is very simple, the eardrum responds to pressure, and at low frequencies, your head does not interact significantly with the volume velocity, ergo the pressure is what you hear.

I said nothing about loudness for a reason, that's an entirely different set of questions that don't need to be introduced here, since you're now comparing apples to oranges. In context, we are talking about the stimulus, NOT the perception. You're now trying to introduce something for what looks like the sole purpose of having an argument, so no, you're wrong, because your comment is irrelevant IN CONTEXT, and if you want to get into an argument, go somewhere else, m'kay? Why you've chosen to go quote-mining I don't know, but that's a classical propagandist maneuver, and I'm afraid with this old curmudgeon, you're just going to get called on your derail. So cheerio, pip pip, and have a nice day. Somewhere else.

 

[j_j]

Now the question should be : can we have a software utility that can identify ALL the very bad rooms looking at their dimensions? As far as I know we only have a few very crude receipts for very bad rooms, such as as a cube, two equal dimensions or exact multiples.

Unfortunately, there is roughly speaking an infinite set of such, as well as an infinite set of "kind of bad", and "a bit bad" and so on. So it's not something one is going to build a list of.

 

[j_j]

Excellent points about the physics, JJ. That's why you get the big bucks!

--Ethan

Bucks? I'm retired. I use bucks. I don't get any big bucks, unless you're sending me a winning lottery ticket!

 

[KlausR.]

Jack,

Where's the popcorn. This is gonna be good.

leave the popcorn in the cupboard, I'm unwilling to engage in discussions with a grumpy old man who behaves in such a disrespectful and contemptuous manner.

Amir,

JJ is not saying that the SPL meaning is the same as every frequency. But rather the fact that perceptually, measuring SPL and judging it at any one signal frequency in sub region, can pretty closely resemble what we hear. Boost a 50 Hz response by that 30 or 6 dB and it will sound different to your ears proportional to those numbers (more or less).

jj mentioned the pressure responsive Omni in the context of the sound wave interference pattern with peaks and nulls so I understood this as meaning that the Omni reading (i.e. room impulse response) will tell you what you hear.

Of course at any one frequency 80 dB SPL will be perceived as being much louder than 44 dB SPL. But jj quoted Ethan referring to 6 dB peaks and 30 dB nulls, which in the given context will occur at different frequencies: if you measure 80 dB at 30 Hz and 44 dB at 110 Hz, the measured difference is 36 dB yet you will perceive both tones as equally loud. In this context it's not about someone boosting the same frequencies by some degree, it's about the fact at the listening position you measure different pressure levels at different frequencies and then equal loudness comes into play to see whether or not these frequencies are perceived as louder.

If you do full modeling of the room like we did for our theater, then you can determine how "bad" the room is.

Does the model take boundary impedances of the real room into account, does it take furniture, carpet, rugs, drapes, does it take structural weaknesses such as doors and windows into account? Does it take the real radiation pattern of the speakers into account? Does it take the fact that you are not playing test tones but music or movie material into account?

I am asking because I did some experiments with sine tones and what I heard was not what I had expected to hear which I took as an indication that real rooms do not behave like rooms on paper.

Your graph shows the situation for one specific mode frequency, but the question is: when you play 5 minutes of a symphony, how often is that mode excited to a degree which is perceived as disturbing? And when you run the animation for all bass frequencies simultaneously, what does it look like when you convert the SPL to Sone? And further, maybe looking at magnitude alone is not enough, if you apply magnitude optimization you still excite those modes, what about their temporal behaviour?


microstrip,

Originally Posted by j_j
(...) The way I would put the room shape question is that it is possible to have very BAD room dimensions, but that there is no magic GOOD set of dimensions, ratios, or whatever. (...)

Now the question should be : can we have a software utility that can identify ALL the very bad rooms looking at their dimensions? As far as I know we only have a few very crude receipts for very bad rooms, such as a cube, two equal dimensions or exact multiples.

As has been pointed out, looking at only the dimensions will not tell you anything about what is going to happen in the room! If you don’t take positions of source and listener into account the exercise is pretty much useless.

Klaus

 

[microstrip]

I is expensive to do though. Currently it costs a few thousand dollars to create the model and run it with different speaker configurations to show the optimal one. My dream is that one day you can do it with an online tool where you draw your room and it runs the simulation for you.

Unfortunately, there is roughly speaking an infinite set of such, as well as an infinite set of "kind of bad", and "a bit bad" and so on. So it's not something one is going to build a list of.

It is something that puzzles me. An industry that is eager to debate the .001% of distortion and does not have a solution with a reasonable price when someone just asks what are the dimensions that I should have in a simple shaped rectangular room, but recognizes that there are many bad, kind of bad and bit of bad ones. What are the criteria for such classifications? Why it is not it possible to systematize them?

 

[microstrip]

(...)
microstrip,

As has been pointed out, looking at only the dimensions will not tell you anything about what is going to happen in the room! If you don’t take positions of source and listener into account the exercise is pretty much useless.

Klaus

Klaus,

I know about it since long, but we can consider that we are free to put the speakers and listener in any situation that defines a symmetrical triangle with 10 feet side in the room - a typical audiophile placement.

 

[JackD201]

Klaus,

Ok, in the cupboard it remains.

I do think that although seldom, just like artifacts on a sweeping shot in a movie, it can be enough to disturb the suspension of disbelief. This is not something I like after a long day, week or month. Granted I am more finicky than most just because audio is my primary hobby but isn't that the case with all hobbyists? I mean we pay attention to things that other people not in the hobby wouldn't care much about. I used to be particular with string patterns and tension when I played racquet sports, still pay attention to club fitting with golf, all things casual players hardly ever consider but things people who compete even at the lowest club levels do.

 

[KlausR.]

I know about it since long, but we can consider that we are free to put the speakers and listener in any situation that defines a symmetrical triangle with 10 feet side in the room - a typical audiophile placement.

In a room with optimized dimensions, if you move source and/or listener the result will change, what was good may become bad and vice versa. Since locations of source and listener are parameters that affect the result you have to consider the whole package, not the room on its own.

As is said in my piece, when you know the exact location of loudspeakers and listener in advance, you can select optimum dimension ratios for those positions. And then it still makes no sense since you never will excite all of the modes at the same time so that you won't benefit from the optimized distribution of the mode frequencies.

Klaus

 

[KlausR.]

Jack,

I do think that although seldom, just like artifacts on a sweeping shot in a movie, it can be enough to disturb the suspension of disbelief. This is not something I like after a long day, week or month. Granted I am more finicky than most just because audio is my primary hobby but isn't that the case with all hobbyists? I mean we pay attention to things that other people not in the hobby wouldn't care much about.

Of course, if there are perceptibly disturbing issues in a room, they should be addressed. The dimensions of my room were optimized on paper, but the builders messed it up. My speakers are placed where my wife told me to, same for the listening sofa against the wall, and it sounds great. Sure, that acoustic ceiling helps, but it does not take away everything.

What I mean is that paper and theory is one thing, reality is another. People who look at the paper and at measurements often forget that a room built from real materials probably does not behave like the one on the paper drawing, and they often don't consider the psychacoustics.

Klaus

 

[amirm]

Does the model take boundary impedances of the real room into account, does it take furniture, carpet, rugs, drapes, does it take structural weaknesses such as doors and windows into account?

Absolutely. Indeed that is the reason to do the modelling to get around idealized assumptions made in room mode calculators and such. Of course it gets progressively more expensive and sometimes very hard if you don't have the physical characteristics of some items. Keith tells me that the simulations routinely mach the real room measurement to 1 db difference.

Does it take the real radiation pattern of the speakers into account? Does it take the fact that you are not playing test tones but music or movie material into account?

The modeling that I was specifically referring to was for low frequencies which I thought was the topic of discussion. In that regard, directivity of the speaker is quite low so not as much of an issue.

Your graph shows the situation for one specific mode frequency, but the question is: when you play 5 minutes of a symphony, how often is that mode excited to a degree which is perceived as disturbing?

That is a frame out of full frequency sweep video. In that sense the test signal stimulates all room modes/resonances. It is true however that if you have very narrow dips for example that span 1-3 Hz, you may not hit them as a matter of probability. For that reason, it is best to go after the wider resonances that are much more audible because they are excited by far more real life signals.

And when you run the animation for all bass frequencies simultaneously, what does it look like when you convert the SPL to Sone? And further, maybe looking at magnitude alone is not enough, if you apply magnitude optimization you still excite those modes, what about their temporal behaviour?

The #1 goal in that simulation is to get even response across multiple seats. In that regard, it doesn't matter if we are measuring SPLs or Sones. If there is a difference between seats, there is a difference and we want to reduce that. Secondarily we want to also get smooth response for one seat. I am not following why we would not want to optimize the room further if our goal is highest levels of performance.

 

[amirm]

It is something that puzzles me. An industry that is eager to debate the .001% of distortion and does not have a solution with a reasonable price when someone just asks what are the dimensions that I should have in a simple shaped rectangular room, but recognizes that there are many bad, kind of bad and bit of bad ones. What are the criteria for such classifications? Why it is not it possible to systematize them?

There are two schools here:

1. This class says there are magic ratios like golden dimensions and tell you to chase them. So if you believe such a rule must exist, then people are providing the specifics.

2. This class says the analysis that group #1 performed is too simplistic and cannot be relied upon to matter in real life (e.g. you don't sit in a corner of a room and put the speaker in the other corner). The method that is proposed is to let the room be whatever it is, then come up with systems the give us the results we want through placement, use of subs and number/placement of those, plus electronic correction (and I don't just mean EQ). The fact that we give up on formulas and criteria for good/bad rooms is not an admission of defeat but rather focusing on solutions that really work, as opposed to simplistic formulas that don't.

The ultimate solution does exist which is to deal with all the laws of physics and fluids and fully model every aspect of the room and the speakers that excite it. Such complete modelling is too expensive and labor intensive to be practical for every room of even most high-end rooms. So we take short cuts and other tools per above.

 

[j_j]

leave the popcorn in the cupboard, I'm unwilling to engage in discussions with a grumpy old man who behaves in such a disrespectful and contemptuous manner.

Excuse me? You misquoted me and insisted I said something I didn't, and that makes me "disrespectful"? Really?

I disagreed with what you said, and pointed out both the extraction from context and the rhetorical misconduct. If you don't want to hear about that, don't do it.

Sorry, no, you just picked a fight, and did so inaccurately.

jj mentioned the pressure responsive Omni in the context of the sound wave interference pattern with peaks and nulls so I understood this as meaning that the Omni reading (i.e. room impulse response) will tell you what you hear.

Which it will, in context, meaning vs. a room that does not have said modes, which is the present context. Of course the modes need to be excited, etc. The point I made is that nodes in pressure are only 1/4 of the story in regard to the actual energy storage in the room. Remember that an Omni captures pressure, which is what your eardrums also capture, and pretty much all you capture at very low frequencies. A cardioid captures 1/2 pressure and 1/2 volume velocity in one given direction, which is why it's directional, and why a node found with a pressure microphone is not likely to be a node found with a cardioid. (all depending, of course, on which way the cardioid is pointed, too)

As frequencies get higher, HRTF's convert part of the velocity field into pressure in the ear canal, and of course in a directionally sensitive fashion. How much is literally what determines your HRTF's.

But the point remains, at low frequencies, under 100Hz or so, a pressure microphone captures the same stimulus as your ear does. We are staying strictly in the acoustic domain here, not entering that of perception. There is no need to introduce anything beyond physical acoustics at this point.

As has been pointed out, looking at only the dimensions will not tell you anything about what is going to happen in the room! If you don’t take positions of source and listener into account the exercise is pretty much useless.

Klaus

Well, actually, no. Storage in a room that does not show itself as pressure at one spot in a room may still have effects that you don't want, from rattling windows to uncomfortable secondary listening positions. Moving source and destination will move around the modal responses, but does little to address the overall energy storage. (Of course position and frequency may change.) Of course, the "size" of the room varies with frequency because of the wall impedance, but that's a secondary issue, except around wall resonance frequencies.

Using a full capture of the soundfield at one point, all 4 variables, it is possible to estimate the energy storage in a room quite accurately, even if you happen to be in a pressure node or a velocity node. (velocity nodes are pressure peaks, and vice versa, almost exactly.) The point is simple, too much storage, especially of the specular sort, is a problem, and diffusing bass is an interesting proposition to say the least. (remember, wavelength is ~1130/f in feet)