A Broad Discussion of Speakers with Major Audio Luminaries

[testp]

It is quite amazing that we don't hear those as they are substantial.

its all subjective, but substantially better or worse/wierd..?

 

[Blumlein 88]

its all subjective, but substantially better or worse/wierd..?

Substantially worse and weird. It is something like if reverb were turned way up too far on the recordings. Way too far. If you are imagining hey a little more reverb or reflective space might help some recordings it is too much for that.

 

[Artsfols]

I looked at this somewhat with those 8th generation files. Distortion increased a small amount around the level you would expect from the gear used. Noise increased just about as you would predict. The noise floor would become audible before other aspects of it would if you continued doing more loopbacks.

I don't know if you are aware of pkane's Deltawave software. Lets you do null tests using music. So you can compare loopbacks, find out what the difference is and listen to that difference. I don't remember where the post is, but I used that and some recordings of speakers. The results are not great. However, I used various positioning of the microphones and clearly that made for much of the difference or more specifically how much room reflection corrupted the sound. I also at one time recorded the same song over my living room system with various miking techniques and let people hear it. It primarily demonstrated the ways that microphones don't hear like human ears. One of the large issues is our ears have a way of filtering out nearly all early reflections. So mostly direct sound is what we hear.

When you record speakers and then play the recording you hear the reflections clearly because in the recording they are coming from the speaker and our ears cannot filter it out since it is in the source. It is quite amazing that we don't hear those as they are substantial.

That is fascinating stuff. My problem is that I am curious but have so many subjects that I study out of interest, I am not sure I should do another deep dive. I think I would rather enjoy the fruits based on the bottom line and let someone else do the hard work. But I will look into the leads you provided here.

 

[Artsfols]

Substantially worse and weird. It is something like if reverb were turned way up too far on the recordings. Way too far. If you are imagining hey a little more reverb or reflective space might help some recordings it is too much for that.

If you ever record a voice in a busy restaurant or office, you realize how much we filter out background chatter and noise. I suspect this is the same effect as you have encountered. Thankfully, as I have 7x24 pulsatile tinnitus but still enjoy music a lot.

 

[Mnyb]

I looked at this somewhat with those 8th generation files. Distortion increased a small amount around the level you would expect from the gear used. Noise increased just about as you would predict. The noise floor would become audible before other aspects of it would if you continued doing more loopbacks.

I don't know if you are aware of pkane's Deltawave software. Lets you do null tests using music. So you can compare loopbacks, find out what the difference is and listen to that difference. I don't remember where the post is, but I used that and some recordings of speakers. The results are not great. However, I used various positioning of the microphones and clearly that made for much of the difference or more specifically how much room reflection corrupted the sound. I also at one time recorded the same song over my living room system with various miking techniques and let people hear it. It primarily demonstrated the ways that microphones don't hear like human ears. One of the large issues is our ears have a way of filtering out nearly all early reflections. So mostly direct sound is what we hear.

When you record speakers and then play the recording you hear the reflections clearly because in the recording they are coming from the speaker and our ears cannot filter it out since it is in the source. It is quite amazing that we don't hear those as they are substantial.

+1 …..and still people are are comparing DAC’s or amps with in room recordings of speakers sound on YouTube.

Or any speaker sound recorded on YouTube ?

And actual recordings, it can’t be stated enough that microphones are not ears you can not really document an acoustic event as a human mind hears the same event . Great production with the tools of the trade makes enjoyable ”fakes” that apear realistic over speakers, that’s also awesome that it actually works reasonably well.

But science knows some things about the human ear brain system I think hence a question.

Is there any plugin or sfx trying to mimic the human hearing and apply this filtering to a microphone signal ?
For headphones only there are these dummy head recordings that smartly moves the process to your own personal filter in your brain.

Simplest attempt to make microphones pick up the ”essential” stuff and disregard the ”unwanted” is probably different coverage patterns like omni or cardioids btw is not Blumlein an early stereo recording pioneer who invented a microphone placement technique among many other things ?

Just thinking if anyone has taken further steps like recording omni and have the polar pattern applied after the fact , but maybe the true directional information is lost ..

 

[Blumlein 88]

+1 …..and still people are are comparing DAC’s or amps with in room recordings of speakers sound on YouTube.

Or any speaker sound recorded on YouTube ?

And actual recordings, it can’t be stated enough that microphones are not ears you can not really document an acoustic event as a human mind hears the same event . Great production with the tools of the trade makes enjoyable ”fakes” that apear realistic over speakers, that’s also awesome that it actually works reasonably well.

But science knows some things about the human ear brain system I think hence a question.

Is there any plugin or sfx trying to mimic the human hearing and apply this filtering to a microphone signal ?
For headphones only there are these dummy head recordings that smartly moves the process to your own personal filter in your brain.

Simplest attempt to make microphones pick up the ”essential” stuff and disregard the ”unwanted” is probably different coverage patterns like omni or cardioids btw is not Blumlein an early stereo recording pioneer who invented a microphone placement technique among many other things ?

Just thinking if anyone has taken further steps like recording omni and have the polar pattern applied after the fact , but maybe the true directional information is lost ..

Certainly there are microphones that can be any pattern after the fact including the direction in which it is looking. Blumlein crossed figure 8's allow this to a limited extent. Something like the tetrahedral soundfield type mics allow more choice after the fact of recording.

I don't know how the filtering is done by our brain and ears to mostly ignore early reflections. Maybe @j_j knows the answer to this.

 

[j_j]

I don't know how the filtering is done by our brain and ears to mostly ignore early reflections. Maybe @j_j knows the answer to this.

A big part is cochlear dynamics. When a sound onset happens, it causes depolarization of some of the outer hair cells. This, to some extent, reduces the loudness (perceptual term, remember, not SPL) of the signal that arrives afterwards. So, to some extent, "leading edge wins". In addition higher parts of the CNS seem to identify somehow places where the leading edge lines up across frequency, providing some more reduction of "what comes immediately after". This inhibition happens at signal onset (it takes a millisecond for the inner hair cell to recharge) and then the outer hair cell has depolarizated, and you have a few more milliseconds of "less gain" before the system, as it were, resets.

This is part of the system. I refuse to even speculate on how the CNS handles this.

 

[testp]

i would add what does work for reviews, Erins audio corner youtube: speaker in-room behaviour comparison against original-signal test using convulation filter on test for speaker version or such, i do like to listen speaker vs original clip test there

 

[j_j]

i would add what does work for reviews, Erins audio corner youtube: speaker in-room behaviour comparison against original-signal test using convulation filter on test for speaker version or such, i do like to listen speaker vs original clip test there

How do you deal with direct vs. indirect sound in that case?

 

[testp]

How do you deal with direct vs. indirect sound in that case?

good question, probably not possible

 

[Lion♡]

I don't know how the filtering is done by our brain and ears to mostly ignore early reflections. Maybe @j_j knows the answer to this.

AES PNW Meeting Report - Auditory Mechanisms for Spatial Hearing

You or We can see more of what j_j mentioned in the presentation.

When you record speakers and then play the recording you hear the reflections clearly because in the recording they are coming from the speaker and our ears cannot filter it out since it is in the source. It is quite amazing that we don't hear those as they are substantial.

I'm not sure if it's exactly in the same context as what you wrote, but I remember seeing something in one of David Griesinger's papers discussing the difference between hearing something live and listening to a recording.
(I can’t seem to find which paper it was, though.)
It mentioned that the reflections in the recording sounded excessive and reverberant, but when heard in the actual space, it didn’t sound that way at all.

 

[j_j]

It mentioned that the reflections in the recording sounded excessive and reverberant, but when heard in the actual space, it didn’t sound that way at all.

This is true only if the complexity of the reverberation coming from at least 5, better 7 or more, sources, is not sufficient. If it is sufficient (and of the proper decorrelation and complexity) then you get the same sensation as the actual space.

Even with 5 microphones (See the PSR work from late 1990's from AT&T Research) this works pretty well.

 

[Floyd Toole]

A big part is cochlear dynamics. When a sound onset happens, it causes depolarization of some of the outer hair cells. This, to some extent, reduces the loudness (perceptual term, remember, not SPL) of the signal that arrives afterwards. So, to some extent, "leading edge wins". In addition higher parts of the CNS seem to identify somehow places where the leading edge lines up across frequency, providing some more reduction of "what comes immediately after". This inhibition happens at signal onset (it takes a millisecond for the inner hair cell to recharge) and then the outer hair cell has depolarizated, and you have a few more milliseconds of "less gain" before the system, as it were, resets.

This is part of the system. I refuse to even speculate on how the CNS handles this.

"leading edge wins". JJ is right. Put into acoustical terms, the first arriving, direct, sound has perceptual advantages over later arrivals. Evidence of this exists in the very first of my JAES publications, back in 1985/86 in which double-blind multiple-loudspeaker comparison results clearly showed that the loudspeakers receiving the highest sound quality ratings had the flattest, smoothest anechoic on-axis frequency responses. Those that were also well behaved off axis received even higher ratings. In the hundreds of tests done in subsequent decades this has just been reinforced. In Sean Olive's subjective/objective correlations it was statistically confirmed, and then, in a clever separate test, trained listeners were asked to draw the spectrum of the "sound" they heard from various loudspeakers. They drew the direct sound spectra. All of this is in AES papers and summarized in my books. More in the upcoming 4th edition.

This is why steady-state "room curves" are not definitive descriptors of what we hear above the transition/Schroeder frequency.

 

[Floyd Toole]

It mentioned that the reflections in the recording sounded excessive and reverberant, but when heard in the actual space, it didn’t sound that way at all.

The principle factor is binaural hearing, which allows us to perceptually discriminate between the first arriving sound from one direction, and later arriving sounds from different directions. In stereo recordings, microphones don't capture this information in the form required by the brain - binaural, dummy head, recordings are needed. These days good binaural recordings reproduced through "neutral" headphones equipped with head-position tracking can be amazingly realistic. Stereo is not capable of delivering the necessary long-delayed sounds of large spaces from the right directions - multichannel sound is required.

The long-delayed reflections in recordings tend to perceptually dominate the local reflections in the listening room, an effect that is more convincing the more channels there are to convey the large-space information. This is so powerful an effect that when multiple channels are simultaneously active listeners are less aware of resonant colorations in loudspeakers. This is why monophonic evaluations are more revealing of loudspeaker problems. The fact that there are periods of hard-panned sounds in stereo and multichannel recordings give us opportunities to hear the "raw" loudspeakers, so mono evaluations are still the place to start. Loudspeakers that win mono tests also win stereo and multichannel tests, but the reverse is not necessarily true. Fortunately, nowadays it is possible to recognize timbrally neutral loudspeakers in a comprehensive set of anechoic measurements, such as the ANSI/CTA 2034 spinorama presentations that are increasingly available, thanks to Amir, Erin, and others.

 

[Lion♡]

This is true only if the complexity of the reverberation coming from at least 5, better 7 or more, sources, is not sufficient. If it is sufficient (and of the proper decorrelation and complexity) then you get the same sensation as the actual space.

Even with 5 microphones (See the PSR work from late 1990's from AT&T Research) this works pretty well.

Yes. The content I saw was probably related to David's explanation about reflections lacking complexity, as you mentioned. Searching through the PDFs of materials from you and many other excellent researchers and trying to recall what I've seen feels like looking for a needle in a haystack. I am always grateful for your contributions.

The principle factor is binaural hearing, which allows us to perceptually discriminate between the first arriving sound from one direction, and later arriving sounds from different directions. In stereo recordings, microphones don't capture this information in the form required by the brain - binaural, dummy head, recordings are needed. These days good binaural recordings reproduced through "neutral" headphones equipped with head-position tracking can be amazingly realistic. Stereo is not capable of delivering the necessary long-delayed sounds of large spaces from the right directions - multichannel sound is required.

I can't believe I received a reply from Toole! I don't think I'll be able to sleep tonight. Thank you for your thoughtful response. I must apologize as the brief comment I made (the one I intended to quote but can't remember) might have been somewhat ambiguous and led to some misunderstanding. I also fully agree with everything you've said (especially regarding the lack of spatiality in stereo). I enjoy binaural virtualization(personalized Custom IEM base, considering neutral speaker+space response and pinna coloration as well.), drawing from the materials of j_j, David, and yourself, Toole.

 

[Floyd Toole]

Yes. The content I saw was probably related to David's explanation about reflections lacking complexity, as you mentioned. Searching through the PDFs of materials from you and many other excellent researchers and trying to recall what I've seen feels like looking for a needle in a haystack. I am always grateful for your contributions.


I can't believe I received a reply from Toole! I don't think I'll be able to sleep tonight. Thank you for your thoughtful response. I must apologize as the brief comment I made (the one I intended to quote but can't remember) might have been somewhat ambiguous and led to some misunderstanding. I also fully agree with everything you've said (especially regarding the lack of spatiality in stereo). I enjoy binaural virtualization(personalized Custom IEM base, considering neutral speaker+space response and pinna coloration as well.), drawing from the materials of j_j, David, and yourself, Toole.

Relax, enjoy your sleep . Related to this discussion are some facts that tend to be oversimplified. A truly diffuse sound is an academic concept. Even reverberation chambers struggle to approximate one, and when samples of absorbing material are introduced for measurement, whatever diffusion that existed is degraded. Concert halls have "relatively" diffuse sound fields, some more than others, and domestic listening rooms with typical reverb times of 0.5 s and less are incapable of supporting diffuse sound fields. In typical listening rooms the sound from typical forward firing loudspeakers arriving at a listener is absolutely dominated by a few early reflections, the direct sound being physically dominant only at the highest frequencies. We can closely predict room curves from anechoic data as proof. and it is not sound power (energy for a diffuse sound field) that dominates.

Perceptually, binaural hearing can "hear through" rooms to a substantial extent, attaching significant importance to the first arriving sound. It is a remarkable experience to hear synthesized spaces. In an anechoic room a direct sound followed by only a single reflection from a different direction, say 60 deg, is perceived as "spacious". In measurements a microphone tells us it is a horrible comb filter. More reflections of the right amplitude, timing and direction progressively build a recognizable room. A diffuse sound field is not necessary for the perception of envelopment - the sense of being in a large space - which audiophiles desperately seek.

As shown in my books, serious experiments show that the conventional 5-channel surround arrangement can closely approximate the illusions generated by a much larger number of channels - BUT only for a single listener. If the experience is to be shared more channels are necessary. Then one can walk around the room and the illusion does not disappear. Fortunately, most perfectionist audiophiles listen alone. The rest of the world, and me much of the time, is content to enjoy the musical content.

 

[j_j]

This is why steady-state "room curves" are not definitive descriptors of what we hear above the transition/Schroeder frequency.

I would argue that "not definitive" is all that's necessary, no qualification is necessary.

 

[Floyd Toole]

I would argue that "not definitive" is all that's necessary, no qualification is necessary.

Sorry JJ but In small rooms the resonances dominate bass quality, and it turns out that prominent resonances behave in a minimum-phase manner, meaning that the steady-state frequency response in that frequency range is a reliable indicator of the potential audibility of those resonances and of the frequency and Q of filters to attenuate them - but only for a single listener. Multiple subs can actively manipulate room modes, which is a more challenging, but reduces seat-to-seat variations. Todd Welti elaborates on this in the upcoming 4th edition, but the basics are in earlier editions.

 

[j_j]

Sorry JJ but In small rooms the resonances dominate bass quality, and it turns out that prominent resonances behave in a minimum-phase manner, meaning that the steady-state frequency response in that frequency range is a reliable indicator of the potential audibility of those resonances and of the frequency and Q of filters to attenuate them - but only for a single listener. Multiple subs can actively manipulate room modes, which is a more challenging, but reduces seat-to-seat variations. Todd Welti elaborates on this in the upcoming 4th edition, but the basics are in earlier editions.

Well, in fact, while you can't 'image' below 90Hz, you can get intense changes in envelopment. This absolutely requires more than one sub.
That's way, way below any transition frequency, and you can't claim "bass quality" until you make that listening room sound as "wide" as the church you started in.

Sorry, but we have to disagree, and there are discussions thereabouts this board detailing others' experiences with it.

Reducing seat to seat variations is only the start.

 

[Lion♡]

Relax, enjoy your sleep

Thank you for your kind response.

Perceptually, binaural hearing can "hear through" rooms to a substantial extent, attaching significant importance to the first arriving sound. It is a remarkable experience to hear synthesized spaces. In an anechoic room a direct sound followed by only a single reflection from a different direction, say 60 deg, is perceived as "spacious". In measurements a microphone tells us it is a horrible comb filter. More reflections of the right amplitude, timing and direction progressively build a recognizable room. A diffuse sound field is not necessary for the perception of envelopment - the sense of being in a large space - which audiophiles desperately seek.

Thanks again. Based on that, I have experimented with adjusting the number and angles of reflections, modifying the shape of the measured room impulse's ITDG or ETC, and intentionally applying slight decorrelation, among other things. It was quite an interesting test.

In particular, by using the direct sound from the front stereo speakers and the reflected sounds from the room, and referring to David's ideal reverb shape, I was able to observe the difference that occurs when modifying the ITDG/ETC, extending from the foreground to the background. While this is just a small test done casually as a hobbyist, it allowed me to understand and relate to the original purpose of multichannel systems and upmixers, reflections resulting from stereo and radiation characteristics, and the perceptual sensations based on time differences.

and then, in a clever separate test, trained listeners were asked to draw the spectrum of the "sound" they heard from various loudspeakers. They drew the direct sound spectra. All of this is in AES papers and summarized in my books. More in the upcoming 4th edition.

I am very excited about the results of such tests.

Reducing seat to seat variations is only the start.

I agree. And your words carry a similar nuance to a statement made by Thomas Lund, while also providing a powerful idea: "Disregard for inter-aural time domain coherency at low frequency. In case LF inter-aural time and magnitude differences have been recorded across channels, and made it safely through a reproduction chain, it is such a pity to kill Auditory Envelopment (AE) at the last stage, by using mono sub(s) with bookshelf/nearfield monitors. That’s game over before even started."