Whispering Gallery Acoustics: A Psychoacoustic Observation

[Deleted member 89570]

First post here, so forgive me if I miss a convention. Wanted to share an observation I’ve been documenting in my own system.


My current system consists of two cherry B&W DM7 MK2 loudspeakers powered by a Rotel RB985, two MartinLogan Motion 40i towers powered by an Arcam AVR5 (which also handles processing), and a MartinLogan Dynamo subwoofer. And yes — I even had room to seamlessly integrate my boyfriend’s PS5.


When left uncorrected, this system produces a transient bloom that sustains in the 500–3 kHz range before decaying into a lingering tail that mimics vaulted reverberance. The psychoacoustic effect is striking: in the opposite corner of the room, backing vocals localize directly into the listener’s ear, as though carried across a whispering gallery.


It’s often assumed that room correction systems exist precisely to resolve such irregularities. And in one sense, they do — microphones record pressure anomalies at discrete points, and algorithms flatten them. Yet what the microphone reads as excess energy, the ear experiences as immersion. Engaging Dirac Live damped that modal bloom, narrowed the dispersion, and collapsed the soundstage into a tin-can image fixed to the screen. The frequency response graphs looked cleaner; the sound became smaller.


This is not to suggest the system is ‘accurate’ in the strict, target-curve sense. It is not flat. It is architectural. The MartinLogans and B&Ws project a spectral field that conforms not to Harman’s 2013 contour, but to the peculiar acoustics of the room itself, with the Rotel and Arcam providing structural support.


And still, dialogue remains clear and immersive. Late-night talk shows, horror films with unusual sound design — all retain their intelligibility and space.


So while discussions here often focus on the advantages of bass shelves or incremental target adjustments, I would simply suggest: not every psychoacoustic phenomenon can be meaningfully represented by a correction curve. Some effects are emergent — and in my case, they transformed a duplex into something closer to St. Paul’s Cathedral.

Illustrative comparison: uncorrected psychoacoustic bloom vs. Dirac’s target curve with +6 dB bass shelf. The spectrogram-inspired trace reflects the sustained midband energy observed in practice — the very energy Dirac flattened into a tin-can image.

 

[Chris A]

This is not to suggest the system is ‘accurate’ in the strict, target-curve sense. It is not flat. It is architectural. The MartinLogans and B&Ws project a spectral field that conforms not to Harman’s 2013 contour, but to the peculiar acoustics of the room itself, with the Rotel and Arcam providing structural support...And still, dialogue remains clear and immersive. Late-night talk shows, horror films with unusual sound design — all retain their intelligibility and space.

This reads like an LLM/GPT was used to copyedit or even generate the text. Is that the case? (One clue is that few people actually use semicolons in their casual written text nowadays, and even fewer "engineers" use them. And the choice of wording in the quote above looks like a copyeditor got a hold of your draft.)

Don't get me wrong--if this is the way that you speak and write--that's great.

When left uncorrected, this system produces a transient bloom that sustains in the 500–3 kHz range before decaying into a lingering tail that mimics vaulted reverberance. The psychoacoustic effect is striking: in the opposite corner of the room, backing vocals localize directly into the listener’s ear, as though carried across a whispering gallery.

Apparently, this is the meat of what you're saying, above. Unless you're using "Gunness focusing" to correct transient response (a patented process--for a patent still in effect), I doubt that Dirac is doing transient (time domain) correction without risking patent infringement lawsuit(s) with the patent holder.

I've found that Dirac corrects the transfer function response (amplitude and phase) using IIR and FIR filtering, which are also completely hidden from user view... and edit.

So what is likely occurring is that Dirac is correcting your amplitude response (mostly) and the phase response (mostly at higher frequencies where the added time delays don't become prohibitive for synchronization with video/TV). If this is the case, then what you're experiencing is likely modal characteristics of your listening room, and mostly frequencies below the "backing vocals". It is often the case that if some portion of the loudspeaker/room amplitude response is not flat, the subjective listening effect is that some portion of the frequency spectrum sounds out of balance, but not necessarily that portion of the amplitude spectrum that is generally above or below the average of the balance of the frequency spectrum. So if there is a lack of midbass amplitude response in-room, this could sound like the backing (female) vocals predominate.

Chris

 

[Deleted member 89570]

Structured writing isn’t artificial—it’s the residue of academic discipline. I went through the same gauntlet anyone in research does—college, then peer review—and that way of writing never really leaves you. I’ll happily let AI generate a quick graph, but the words are mine. If semicolons are the evidence of machine authorship, then we’ve set an oddly low bar for human expression.


On Dirac: my description was psychoacoustic, not algorithmic. I wasn’t suggesting they’ve secretly adopted Gunness focusing. Dirac does what it does—FIR/IIR correction of amplitude and phase. The “transient bloom” I described is an emergent property of the room and placement, not hidden DSP code.

 

[Keith_W]

Welcome to ASR. Your post hints at why you are experiencing problems with your correction - you are attempting to correct too high. As wavelengths get shorter, the correction becomes even more specific to the microphone position. If you are taking measurements for correction with nearby reflective surfaces, e.g. listening chair or rear wall, this is even more so.

The usual advice is to correct all minimum-phase deviations up to the Schroder frequency. If you want to correct above that, you have to take quasi-anechoic measurements of your speaker and apply "broad, low-Q, tone-control like" corrections for the upper frequencies. If you take a measurement from the listening position without controlling for reflections and attempt to correct that, the result is rarely satisfactory.

@Chris A made an important point - Dirac corrections are hidden from user view, and also edit. Simply put, I don't think that Dirac provides you with a workflow that allows you to do this. It might be better to set the "curtain" between 20Hz to 200-300Hz and leave it at that.

 

[Deleted member 89570]

Welcome to ASR. Your post hints at why you are experiencing problems with your correction - you are attempting to correct too high. As wavelengths get shorter, the correction becomes even more specific to the microphone position. If you are taking measurements for correction with nearby reflective surfaces, e.g. listening chair or rear wall, this is even more so.

The usual advice is to correct all minimum-phase deviations up to the Schroder frequency. If you want to correct above that, you have to take quasi-anechoic measurements of your speaker and apply "broad, low-Q, tone-control like" corrections for the upper frequencies. If you take a measurement from the listening position without controlling for reflections and attempt to correct that, the result is rarely satisfactory.

@Chris A made an important point - Dirac corrections are hidden from user view, and also edit. Simply put, I don't think that Dirac provides you with a workflow that allows you to do this. It might be better to set the "curtain" between 20Hz to 200-300Hz and leave it at that.

Thank you!

Dirac doesn’t perceive. It applies IIR and FIR filters; it doesn’t hear a room. My description was in situ — documenting how the system presents itself in space — not an attempt to rewrite DSP workflows. The limits of correction above Schroeder are well established, as is the effect of the curtain. But perception doesn’t halt where equations stop behaving neatly. The transient bloom and vaulted reverberance I described are emergent properties of the room itself, not hidden algorithms — and perhaps even a brush with qualia, those irreducibly subjective features of sound that remain outside the graph. At least until someone nails down a formula for sentience...

 

[Keith_W]

I suspect your "bloom" will be exposed if you look at the spectrogram. Do a before and after sweep with REW and post the MDAT. You will need to zip it first.

 

[Chris A]

BTW: you're still welcome for my reply to your initial post.

I had seen that no one was responding on the forum in over 12 hours (check the post times), so I decided to give you a break and take a chance that your post wasn't written synthetically via some prompt written by some prankster teenager (or worse) randomly poking around the web for "a good time" if you know what I mean. It's good to know that this likely isn't the case. My doubts were still quite high as I wrote my initial reply, however. So in balance, I think it fortuitous that my initial reservations were apparently not true in this instance.

Your chosen forum handle (not apparently being derivative of your real name) was also a bit of a red flag for me--so I proceeded with caution, hoping that those reservations were unfounded. I'm relieved to know that they were in that case, too, but I still wonder why one would choose such a handle for this forum. It seems a little off the beaten path, at least to me. I don't need an explanation--just acknowledgement that my reservations could have been well founded responding to a first-time poster here with no history.

As far as your response and your psychoacoustic "observation"--that was actually addressed in my last two sentences. I would encourage you to re-read it again in order to compare to your response, which appears to be a bit abusive and not well intentioned for anyone responding to the particular initial assertions you posted.

On the subject of your core assertions, I remain puzzled in why you would choose that subject to introduce yourself to the forum using only assertions (but no questions). You didn't do an introduction, and you didn't simply respond to someone else's questions in another pre-existing thread to be helpful. You instead chose to post a new thread--and without asking questions over whether your observations had possible correlation to past observations by others.

Your post hints at why you are experiencing problems with your correction - you are attempting to correct too high.

I think this is a bit wide of the mark for what the OP asserts.

I also disagree with your advice here--unless you're the casual "punch the button and stand back" type of in-room transfer function correction user. I think you likely know what I'm referring to: the type of person that wishes to learn nothing but get all of his/her problems automagically resolved instantly and without effort...only applying money,

But that's another matter that diverges from the apparent subject of this thread.

The transient bloom and vaulted reverberance I described are emergent properties of the room itself, not hidden algorithms — and perhaps even a brush with qualia, those irreducibly subjective features of sound that remain outside the graph. At least until someone nails down a formula for sentience...

I'm curious as to what set you on this path of assuming that others here are not aware of that which you speak already. I do not read that into any responses that I've seen on this forum. All listening rooms have limitations somewhere inside in terms of sound reproduction capabilities.

However, many here I've found seem to assume:

1) stereo reproduction only, and
2) a single listener sitting in the so-called "sweet spot".

Both of those assumptions are not valid for my setup and room acoustics.

Chris

 

[Deleted member 89570]

The suspicion that my post was written by an algorithm is almost flattering! If a machine could capture the difference between modal bloom and vaulted reverberance with semicolons, then we’d have to accept it had surpassed most human writers already. As for Dirac-- I’m not denying what it does, nor am I mistaking it for magic. It applies IIR and FIR filters with precision, and the graphs look cleaner. That’s fine. Neat. But what I described wasn’t algorithmic — it was psychoacoustic. A transient bloom sustained between 500–3 kHz, a tail of sound that moves through the room like a whispering gallery. It’s not hidden DSP code, it’s the room itself speaking back. If we insist that nothing is real unless it can be reduced to REW MDATs, then we’ve missed something essential about sound. Some effects are emergent. They don’t collapse neatly into a target curve, yet they remain undeniable in situ. I don’t expect everyone here to share that fascination, but to dismiss it outright is to conflate “unmeasured” with “imaginary.” The two are not the same.

 

[Deleted member 89570]

And as for the username: it isn’t a red flag, it’s a lens. Postmortem observer is exactly what I do — noticing what remains after collapse, whether in sound, in systems, or in sentience.

 

[Deleted member 89570]

@Keith_W Thanks for weighing in constructively.