Can Loudspeakers Accurately Reproduce The Sound Of Real Instruments...and Do You Care?

[MattHooper]

Had yet another experience in terms of instrumental timbre today. (Or, at least, I think that still seems to be the characteristic I'm hearing).

Brought a bunch of music over to my pal's house and listened on his super-resolving system. It's crazy how clear the sound is through his speakers - you get the every bit of bow-on-string texture, a "right there in front of you" sense of vivid clarity, and the tiniest details at the furthest "distance" sonically effortlessly clear.

And yet, not a darned thing I heard sounded natural to me. Aside from the first bracing "wow that's clear" sensation, once that shock was over I just wasn't compelled to listen to it very long at all. Everything seemed blanched of believable timbre. And it had a consequence for me that complexity in a mix actually became harder work for my brain. In other words, when instruments started lining up in one speaker (these were some older recordings where you often had hard panned instruments), it became hard to disentangle what they were because of the timbral homogenization. "Something is ringing in there, something is being percussive or snapping, something is holding a tone..." but it felt like they were "all made of the same material" in a way.

When I came home and spun the same music on my system (Thiel speakers in my case) it wasn't as obviously "resolving" and clear. But things sounded timbrally more distinct - I was hearing instruments "made of different materials" so in the very spots that sounded confused to me, the pile up of instruments in one speaker, sounded easy to process: "Oh, that's a metal bell being struck, behind it a castanet! And that's a string, a cello, holding a note, and that is also a synthesizer note adding it's part in there too...."

So...less detailed and clear, yet timbrally differentiated. And unlike listening through my friend's system, this aspect of my system gave me the "ahh..this sounds right" sensation and it was hard to stop listening.

Again, it's hard to know exactly what difference in these systems (and of course possibly in my own brain) is responsible for what I perceive.

 

[MRC01]

I've had this experience many times. It's often the way music is recorded and mastered. And it is also often the way speakers are voiced. They add a bit of boost to the presence region (upper mids / lower treble). This adds a sense of enhanced detail, which of course is artificial but in the short term, first impression, it sounds more detailed. It doesn't take much, the ear is so sensitive to this frequency range even a 1 dB lift is noticeable.
So what is responsible? The simplest explanation is non-linear frequency response.
A better test for real (as opposed to artificial) detail is listening deep down into the layers of music: can you resolve a super quiet sound underneath simultaneous louder sounds? Like hearing a solo piano player breathing or humming, not only during rests, but also while he's playing, or the rustle of his clothes as he moves. In a symphony, can you hear the individual instruments, or is it just a wall of sound? Or hear an occasional page turn, the conductor's baton hitting the music stand, the HVAC system in the studio turning on or off, etc.

 

[cjfrbw]

Sounds like 'audiophile skeleton sound' as I like to call it. It's like looking an x-ray of a person with only shadows of the flesh and blood. Lots of interesting detail, but not exactly how you see the subject in ordinary light of day.

Texturized rendition will be a bit more affably muddled by tone and color. Getting both tone and exquisite detail is an uncommon synergy of equipment and recording.

 

[MattHooper]

MRC01,

Yes the first thing I would suspect with the description I gave would be "hyped top end" in terms of detail. But it wasn't really like that. I'm familiar with the effects of EQ (I use them every day) and it wasn't just a presence-region or high frequency lift. My friend's speakers also sound fairly rich. There was just a *sense* of clarity in the sense of low distortion: the whole spectrum seemed well controlled and clean. Though without measurements I can't be absolutely sure there wasn't some frequency sculpting.

cjfrbw,

Yes! "Audiophile skeleton sound" is an image I've long held too. I've heard many speakers like that, and my pal's speakers were to some degree exhibiting that "too see through" quality. I like a sense of density. And, yes, that combo of tone/detail is darned hard to find. Whenever I come across a speaker that does both I tend to grab it.

 

[Hifi]

For me . . . This is the goal of Two Channel audio.

 

[Hifi]

Again, various sins can be pinned to Pearson (his confidence in his auditory recall could be galling). But I see the above as part of the straw-manning I was mentioning. It paints any readers of TAS, or adherents to the principle, as gullible unthinking sheep.






When you say a realistic sounding reproduction would only bear resemblance to that instrument in real space "by coincidence" that strikes me...and I don't mean it to be insulting...as a bit facile. Surely if something sounds real, there is a reason for it. A reason that could be enlightening. If of "real instruments being played in real space" far more effortlessly than many other speakers I'm acquainted with. Their mids/tweeters operating in free space virtually eliminate the sense of instruments coming from a box or from a "speaker." Just like a real voice or acoustic guitar doesn't sound like it's coming from a box or speaker. There is a spacious dimensionality to the sound that few regular speakers can match. I have some vocal recordings (I think I mentioned) recorded in a natural space and if I close my eyes it's almost effortless to picture being near the front of a stage hearing live singers - in a way that any number of box speakers I can mention seem to struggle with. There are going to be real-world, physical/engineering reasons why they produce this different presentation. For someone who wants that sensation of hearing voices and instruments produced without box colorations reminding them it's artificial, and with a sense of those sounds happening in more "realistic unconstrained space" - putting it down to coincidence isn't going to help. There's a reason they may want to seek out some omni-speakers.
(Not saying the MBL omnis are perfect by any means...only using them as an example for my point).

I've seen more replies at this point stating people don't expect a perfect replication of the live event (orchestra or otherwise). My original question was meant to focus specifically on timbrel reproduction - can we expect speaker drivers to accurately reproduce the huge variety of real-world timbres in instruments and voices, and if so to what degree? I was trying to leave out dynamics, scale, soundstaging and all that.

But I guess we sort of got through that and moved on to TAS-like discussions. Fair enough.

I think there is more to “realization” in Two Channel audio than timbrel reproduction on its own and that’s why your thread has expanded. And furthermore... to dismiss all box speakers and their timbrel reproduction is going to bring more interest/opinion/op participation.

But let’s say ... the entire thread conversation is “somehow” limited to only timbrel reproduction of a wide range of instruments including human voice on Omni design speakers .... don’t we just have a discussion of my Omni speakers are better than your Omni speakers? Where does the conversation go now?

 

[MattHooper]

But let’s say ... the entire thread conversation is “somehow” limited to only timbrel reproduction of a wide range of instruments including human voice on Omni design speakers .... don’t we just have a discussion of my Omni speakers are better than your Omni speakers? Where does the conversation go now?

I'm not sure about the motivation to limit the conversation to omni speakers. I certainly would not want my original question limited to one type of speaker. In fact the point was to ask if ANY speaker could, in principle, reproduce as wide an array of timbres as exist in real life. And also to survey other people's ideas and experience in listening for accurate-sounding timbral reproduction.

It has nothing to do with a "my speaker is better than your speaker" conversation.

 

[Hifi]

First off ... I am certain your speakers are exceptional in many ways. I have never had an opportunity to listen to MBL. With that said they may do everything perfect.

Question for you regarding your listening set up . . . What instrument would you say is the most difficult for your speakers to reproduce with realism?

 

[MRC01]

... My original question was meant to focus specifically on timbrel reproduction - can we expect speaker drivers to accurately reproduce the huge variety of real-world timbres in instruments and voices, and if so to what degree? I was trying to leave out dynamics, scale, soundstaging and all that. ...

Dynamics is a critical part of timbre and can't be left out. Every natural sound (including musical instruments) has the same harmonic frequencies; the only difference between them is the relative levels/dynamics of the individual frequencies. This is one reason why dynamic compression makes acoustic instruments sound unnatural.

To produce accurate timbre, a speaker must have flat frequency response, low distortion, and linear, uncompressed dynamics. I would imagine flat/linear phase response is also important, because otherwise it can cause comb filtering which affects frequency response.

 

[andreasmaaan]

I would imagine flat/linear phase response is also important, because otherwise it can cause comb filtering which affects frequency response.

Could you explain what you mean by this point please?

 

[MRC01]

I'm hypothesizing that if a speaker has a poor phase response it could cause a comb filter effect. Consider frequencies F1 < F2; if due to poor phase response F2 is delayed relative to F1, their superposition can have uneven frequency response.

 

[MattHooper]

Dynamics is a critical part of timbre and can't be left out. Every natural sound (including musical instruments) has the same harmonic frequencies; the only difference between them is the relative levels/dynamics of the individual frequencies. This is one reason why dynamic compression makes acoustic instruments sound unnatural.

To produce accurate timbre, a speaker must have flat frequency response, low distortion, and linear, uncompressed dynamics. I would imagine flat/linear phase response is also important, because otherwise it can cause comb filtering which affects frequency response.

That sounds reasonable.

However, I personally haven't noticed a distinct relationship between dynamics and believable instrumental timbre. In fact, I've had speakers that had to my ears particularly accurate sounding timbre, but which also lacked in dynamics. (And I've heard very dynamic speakers have artificial-sounding timbres).

In fact last night I had set up a pair of old Spendor S 3/5 monitors - very small speakers. Hooked up to my tube amps, playing some vinyl, I was taken aback by how amazingly believable the timbral reproduction was - acoustic guitars, woodwinds, horns, sax, and human voice was just amazing. A test I often do is when there is some clapping on a track, or finger snaps, I close my eyes and clap lightly as well, or snap my fingers, to compare that distinct "human flesh" timbre to the reproduced version. Through the spendors human clapping sounded almost bang on - like flesh clapping flesh, vs sizzling bacon or electronic sounding. At one point a vocal was so warm and natural sounding I went in to another room where my wife was speaking on the phone, closed my eyes and listened and the timbre of her voice, the gestalt of "what a human voice sounds like" and the "color/tone" I see with my eyes closed was amazingly close between the sound from the speakers.

But...the Spendors are little guys and if anything they lack dynamics. I've heard speakers that are far more dynamic, and which even measure pretty neutral, but which don't produce to my ears the type of accurate-to-the-real-thing sensation as the little Spendor speakers.

Boy it would be interesting to do trully well-controlled blind tests between various speakers and the real thing.

 

[RayDunzl]

But...the Spendors are little guys and if anything they lack dynamics.

I would think dynamics (or at least dynamic accuracy) is represented by the correct reproduction of relative levels, not ultimate loudness.

 

[andreasmaaan]

I'm hypothesizing that if a speaker has a poor phase response it could cause a comb filter effect. Consider frequencies F1 < F2; if due to poor phase response F2 is delayed relative to F1, their superposition can have uneven frequency response.

I still don't follow

A comb filter occurs when a signal containing the same frequency components is delayed and then fed back to itself. This of course affects the amplitude response. The amplitude response should not be affected though when the phases of multiple tones are varied relative to each other, as is the case with an all-pass filter, or with any non-linear phase but nevertheless time-coherent loudspeaker.

Or perhaps I'm still misunderstanding what you're saying?

 

[MRC01]

... A comb filter occurs when a signal containing the same frequency components is delayed and then fed back to itself. This of course affects the amplitude response. The amplitude response should not be affected though when the phases of multiple tones are varied relative to each other ...

For comb filter effects to occur, the frequencies don't have to be exactly equal. Based on how comb filter effects arise, I'm speculating that you will start to see comb filter effects as F1 and F2 approach each other, before they are exactly equal. Also if one is an integer multiple of the other.

 

[andreasmaaan]

For comb filter effects to occur, the frequencies don't have to be exactly equal. Based on how comb filter effects arise, I'm speculating that you will start to see comb filter effects as F1 and F2 approach each other, before they are exactly equal. Also if one is an integer multiple of the other.

I still don't understand how this would be the case, but anyway, I'm interested so I thought I'd run some tests using sine waves and a DAW.

Here's the spectrum of a 1kHz and a 2kHz tone, neither signal delayed:

Now with the 1kHz tone delayed by 1ms:

And now with the 1kHz tone delayed by 10ms (significantly more than would be the case with any normal minimum phase crossover filter):

 

[RayDunzl]

Comb filtering:

The source - usually a sweep, for illustrative purposes.

The reflection or second source - has a constant delay, like unequal distance to two speakers, or wall reflections.

Various frequencies will then recombine at some listening point at various different phase angles to cause peaks (coherent) and nulls (out of phase relative to each other).

Wiki: "In signal processing, a comb filter is a filter implemented by adding a delayed version of a signal to itself, causing constructive and destructive interference."

Here are two frequency sweeps, the second delayed by 5 milliseconds, and combined:

This effect also occurs without reflections if two speakers are sounding, and the listener moves away from the central sweet spot. Move to the left or right to create the above condition.

I would think the room would be a far greater influence on this phenomenon than phase anomalies within the speaker, provided both speakers (stereo) exhibit the same fault.

 

[andreasmaaan]

Same again but with 1000Hz and 1050Hz (first graph: no delay, second graph: 1000Hz tone delayed by 10ms):

 

[MRC01]

Based on how the comb filter effect happens, it should also affect 2 frequencies that are an integer multiple of each other. For the intuition behind this, look at the image on the Wikipedia page I posted above, and imagine if one of those waves only had every other blip (half the frequency of the other). It's possible my intuition is wrong, so to see if it holds up numerically, I will test this in a spreadsheet and post the results.

 

[andreasmaaan]

Based on how the comb filter effect happens, it should also affect 2 frequencies that are an integer multiple of each other.

That's why I chose 1kHz and 2kHz in the first example.

For the intuition behind this, look at the image on the Wikipedia page I posted above, and imagine if one of those waves only had every other blip (half the frequency of the other). It's possible my intuition is wrong, so to see if it holds up numerically, I will test this in a spreadsheet and post the results.

Thanks Definitely interested here.

EDIT: FWIW BTW, I zoomed in much further using the spectrum analyser, and there was no discernible difference between the two tones when the lowr tone was delayed by any magnitude that I tried. Accuracy is to within 0.1dB.