"Statistics of natural reverberation enable perceptual separation of sound and space"

[Thomas_A]

I just read this one - and according to the results we decode sound and space as long as the reverberation is natural for the specific environment. Adding synthetic reverberation not consistent with those natural ones were detected as "synthetic".

https://www.pnas.org/content/113/48/E7856

Any comments or thoughts? Are we overdoing "corrections for rooms" sometimes?

 

[Blumlein 88]

Well typically we don't add reverberation. You might dampen a room too much, but even then it isn't adding or changing it to something wholly synthetic the way additive reverb would do.

Interesting paper confirming yet again that our ears/brain can largely hear thru the room at least for a few milliseconds.

 

[Thomas_A]

Well typically we don't add reverberation. You might dampen a room too much, but even then it isn't adding or changing it to something wholly synthetic the way additive reverb would do.

Interesting paper confirming yet again that our ears/brain can largely hear thru the room at least for a few milliseconds.

Yes. The questiion I raised however is if room resonances are more natural to us even if they mask details. Synthetic resonances added identified as ”synthetic” are those that do not follow the frequency pattern of a natural one.

 

[j_j]

This needs to be divided into several questions.

1) Do we use artificial reverberation in producing, mixing, etc?
2) Do we use it in a home listening environment.

The answer to 1 is yes, yes, and yes, and yes some more.The whole "natural vs. unnatural" thing is somewhat of a misnomer, when you use an actual room response (or a good synthetic one), but that's on the artistic side. I'm somewhat limited in what I can say about that, but there are many options, and many ways to go about that, including in our production tools.

To 2, we generally learn our home environment. Since we know it, we learn how to "hear through" it, unless it's really, really bad. So adding anything else adds a new task for the listener until the listener adjusts. I do have some issues with the comments in the paper, in particular the "generally linear", that's true only in a room with no air movement, but the effects are not extreme in a listening room. In a concert hall, they are quite noticeable, which is evident by just capturing half a dozen impulse responses with the same exact locations for source and mike(s).

Synthetic reverbs can be natural sounding, or not, but if you're adding it at the listening room end, and comparing it to the room the person is used to, that's kind of an odd idea.

In a few very special cases, some short-period natural-sounding reverb might be added by systems that are doing 2-loudspeaker virtualization, this in order to provide front/back cues, but that's not a normal audiophile environment, and I wouldn't generally advocate that it should be.

 

[Thomas_A]

As I understand it the reverb (articifial or not) is "natural" as long as it mimics the natural one. Learning seems to be adapted to many environments, not only our homes. So is there a limit to what we should compensate for, even below Shroeder? Adding reverb by means of Auro3D etc is another way of room re-make though.

 

[j_j]

Compensating for things that are not going to hold 6" away from where you sit are not going to do much good. It is better to control the overall response above about 1kHz or so (based on hearing, not schroeder frequency) rather than the fine details in most cases. In cases where that doesn't work, it's better to work on the acoustics.

Or so I conclude. Consider, at 1khz, moving 3" away can change the response of the room by a huge amount. Correcting the details in a 1Hz wide range at 1kHz is not going to be meaningful, nor is that what the ear is primarily sensitive to.

 

[Josq]

Interesting paper. Lots and lots of information, but the key take-away seems to be: don't have too much distortion of the acoustics of natural spaces.

The acoustics of a natural space is included into any recording of live music. When playing the recording, the original acoustics is getting mixed with the acoustics of the listening room. This could potentially lead to undesirable distortion of the original acoustics. I think the best cure is already well-known here: let the reflected sound be similar to the direct sound.

 

[Kal Rubinson]

The acoustics of a natural space is included into any recording of live music.

True but is it distorted simply due to the inability of a loudspeaker to reproduce the spatial pattern than impinged on the microphone diaphragm?

 

[j_j]

True but is it distorted simply due to the inability of a loudspeaker to reproduce the spatial pattern than impinged on the microphone diaphragm?

Now THAT is a horse of another color, yes.

 

[Josq]

True but is it distorted simply due to the inability of a loudspeaker to reproduce the spatial pattern than impinged on the microphone diaphragm?

The paper states:

Our IR measurements revealed four characteristics common to almost all of the IRs we surveyed: (i) a transition from high kurtosis, produced by sparse early reflections, to Gaussian statistical properties within ∼50 ms of the direct arrival; (ii) exponential decay; (iii) frequency-dependent decay rates, with the slowest decay between 200 Hz and 2,000 Hz and faster decay at higher and lower frequencies; and (iv) decay-vs.-frequency profiles that varied with the overall magnitude of the reverberation (decay rates in more reverberant spaces tended to vary more with frequency). (...)

We found that human listeners are sensitive to all four regularities and that they are necessary for the perception of reverberation and the accurate separation of a sound source from reverberation. We also found that realistic reverberation could be synthesized simply by imposing these four regularities on noise

If I am correct, all of these factors relate to frequence response over time, rather than the ability to detect spatial patterns. Maybe those spatial patterns are important too, but at first glance that aspect is not included in this study.

 

[j_j]

Maybe those spatial patterns are important too, but at first glance that aspect is not included in this study.

Hm. Let me check on something before I comment.

 

[Thomas_A]

The paper states:


If I am correct, all of these factors relate to frequence response over time, rather than the ability to detect spatial patterns. Maybe those spatial patterns are important too, but at first glance that aspect is not included in this study.

Indirectly, it would be included I think.

 

[Thomas_A]

Compensating for things that are not going to hold 6" away from where you sit are not going to do much good. It is better to control the overall response above about 1kHz or so (based on hearing, not schroeder frequency) rather than the fine details in most cases. In cases where that doesn't work, it's better to work on the acoustics.

Or so I conclude. Consider, at 1khz, moving 3" away can change the response of the room by a huge amount. Correcting the details in a 1Hz wide range at 1kHz is not going to be meaningful, nor is that what the ear is primarily sensitive to.

I was mainly thinking about lower frequencies where room dominates, where you can apply gentle EQ to bring down one or a few of those nasty peaks if you have them, or choose to linearise the response up to 300-400 Hz or so. Probably there is a limit somewhere where sound starts to be unnatural to us, including acoustic "overdamping" of our living rooms.

 

[GimeDsp]

There are MANY different types of reverbs, some useIR info and have room simulation adjustments in the GUI.

Then there is also the recording/sound. Does it already have some spatial info, if it does even an OG plate reverb will just add to what spatial info is there.

How we naturally hear sounds is one reason mixdown/mastering rooms keep the floor live and absorb ceiling reflections.

What's amazing is how many rear reflections impact spatial awareness.

You wouldn't think adding absorption behind the LP would mess up the spatial until you do it and it and it does. That is why for critical listening set ups its so hard to balance timings, spatial, and tone. The best set up is proper tone, accurate spatial that ends up being pinpoint accurate and just a smidge wider than reference HP. where as for listening I want a LOT more reflections for everything over 300hz.

But that's just me.