What is timbre and can we measure it?

[Randyman...]

Sure it's a performance, but a singer can change the timbre of his/her voice and a violin can have different timbres if the musician bows or picks the strings.

And the timbre of a drumset will be VERY different depending on who is playing it. Same instrument - same room - same mics. Can sound like a totally different kit w/o changing anything but the "source" that is exciting the instrument (the musician).

 

[solderdude]

And we all know recording a sound and listening to it in a room is not the same as being in the room with the original sound. We all know how recordings of our systems sound weird - even with 1 pass. Our ear/brain filters all that out when we are "in the environment". Mic's and reproduction won't allow us to filter that data out as the spatial cues we use for that are totally gone in the recording.

I argue that is not timbre. But I'm the minority How would you compartmentalize the timbre of a human voice, and do different systems fundamentally change that timbre? Or is the voice and performance itself what defines the timbre?

And put a "hack" drummer on a ideal drum kit and record. Put a seasoned player on the same kit and record. The timbre will be vastly different I guarantee - stick technique, power, dynamics, "oomph" with make it sound like a different kit with totally different impact and feel. Same for any guitar. I don't think that can be "measured".

All that can be recorded can be measured. Recorded sound can be analyzed in lots of different ways. Just FR and distortion won't capture that but is already pretty sufficient for electronics.

There is a huge difference between a live recording in a room and a live performance in a room.
It is because there is a limit to what a microphone and its placement can do. A recording is a (more often than not manipulated) glimpse of what has been.
One would need 'holographic' recordings with the room excluded to get the same person singing in a different room.
Maybe, in the future AI could assist when everything is recorded with multiple mics per instrument/voice so that might become a reality in the future but won't be cheap.

There are also people who can mimic other peoples voices to such a perfection that when you would not see who was talking you would have a hard time figuring out who was the real deal. I don't think one person can mimic all voices though and takes quite a lot of analyzing and practice to do so.

Every instrument (especially electric instruments) can be fully imitated and analyzed. Not so easy with anything acoustic.

 

[Randyman...]

Every instrument (especially electric instruments) can be fully imitated and analyzed. Not so easy with anything acoustic.

Makes it somewhat difficult then, doesn't it? I still say the timbre is not an isolated characteristic of the "instrument" (even a biological one made by God). And if the musician will directly impact the resulting timbre - what's the "benchmark" to compare against?

The musician driving the instrument with their most "energetic" take?
The musician driving the instrument with the most "controlled" take?
An "Average" take by the musician?
An overly-dynamic take?
A quiet take?
Let a robot play the "instrument" to be scientifically repeatable? (hard when it's a voice)

The resulting timbre will be different in all cases. But it's the same instrument. What's the reference?

If it's a sample or Synthesis - then yes - it's cookie-cutter at that point and CAN be exactly recreated (well - some analog synthesis and OSC's are "free running" and would be technically hard to exactly recreate some of that time-based stuff verbatim). But listening to the same cookie cutter sounds in different rooms still does not alter their root timbre from my argument.

 

[Axo1989]

I don't remember where I did that or what the context was, but if we do that again and look at the differences, what we see is relatively small differences in frequency response (so the relative level of different frequencies) and dispersion characteristics (how the sound spreads around the speakers).

Timbre I guess to a greater degree is the makeup and mix of frequencies and harmonics that make up the sound beyond the fundamental (I haven't followed this discussion). I don't think either my speakers or the Genelecs add a lot of harmonics to the sound, so while they don't measure exactly the same, they don't really change the timbre of the sound, the way I understand timbre. But I guess variations in frequency response could be perceived as changes in timbre.

So I think to an extent I agree with @sonitus mirus . But all speaker manufacturers don't necessarily agree on which route to take to get there. What I have argued in the past is that flat anechoic response doesn't necessarily equate to a natural in-room representation of the source, even though that would instinctively make sense. But that's perhaps a digression on the topic.

Our SBS.1 vs 8361 for reference:
View attachment 420485

I had a quick look for the posts I remembered when writing that post, but was unsuccessful. My recollection was that you posted similar graphs, and you or another poster commented that the two speakers sounded a bit different, but it was hard to argue one was flawed. Apologies if I have it wrong.

Re timbre, tonality (as opposed to pitch) is a component, among others. Thinking of the spectrogram conventionally in three dimensions where x = time, y = frequency, and z or the usual colour ramp = intensity/volume) any variation in frequency response will change the relative levels of the spectral components, altering timbre.

In the "musical" definition, no. It's a property of the instrument. There is I guess an assumption that the performer will get a sufficiently good performance from the instrument. And let's face it, most of us are not buying an expensive audio system to accurately reproduce the sound of a kid's first violin lesson.

Timbre is associated with an instrument (or anything that makes a sound) but isn't entirely fixed. A violin produces a range of sounds—there are 25 bowing techniques apparently. Producing the range of timbres the instrument is capable of. So, properties of the instrument, and the techniques used to play it.

It may seem a strange definition, but if you think about it, like many things it's not about us. The "musical" definition of timbre and the sound of an instrument is actually of most importance to the builders of the musical instruments.

An example of a(n instrument-maker's) perspective on timbre, but not a definition that excludes all others.

We are indeed interested in the performer-instrument combination, of course. But that is called a "performance", not "timbre". And the stuff that measures how that works - mentioned in the question - are indeed the standard measurements of distortion and frequency response that have been in use since - well, Edison, I suppose.

It's entirely common to understand that different playing styles and techniques alter the timbre of the sound produced by an instrument. Performance means several things, but doesn't circumscribe or exclude timbre in the way you suggest. Nor do specific measurements of sonic elements.

 

[solderdude]

The resulting timbre will be different in all cases. But it's the same instrument. What's the reference?

Each instrument/voice can still be analyzed in all the mentioned conditions. Its attack, sustain, decay, harmonics at each point in time can be captured.
In all cases we talk about timbre.
The reference is the measurement as long as that is done well.
With anything acoustic microphone placement becomes paramount. This is what the studio has to do. The playback chain should not change anything.

In case of a DAC or amp and even transducers we do not speak of timbre. Transducers can change the frequency response which affects the reproduced signal and thus its timbre but it does not have a timbre.
If it had all instruments and voices would start to sound the same as it would be defined by the timbre of gear. It doesn't. The timbre is altered (due to FR).

The violin remains the violin, the acoustic guitar remains the acoustic guitar, the voice remains the voice and even when the tonality is changed we can still hear the timbre of what has been recorded. It may sound a little 'off' due to modifications of the frequency amplitude.
So the examples below will still sound as the examples below and the 'timbre' will be preserved.

The musician driving the instrument with their most "energetic" take?
The musician driving the instrument with the most "controlled" take?
An "Average" take by the musician?
An overly-dynamic take?
A quiet take?
Let a robot play the "instrument" to be scientifically repeatable? (hard when it's a voice)

Now... if you told me their most "energetic" take will be altered by a playback system into the most "controlled" take then we could say gear alters the timbre.

So... reproductive gear does not have a timbre.... it may have a deviating tonality or add audible amounts of distortion but not timbre.

 

[Axo1989]

Each instrument can still be analyzed in all the mentioned conditions.
In all cases we talk about timbre.

In case of a DAC or amp and even transducers we do not speak of timbre. Transducers can change the frequency response which affects the reproduced signal and thus its timbre but it does not have a timbre.
If it had all instruments and voices would start to sound the same as it would be defined by the timbre of gear. It doesn't. The timbre is altered (due to FR).

The violin remains the violin, the acoustic guitar remains the acoustic guitar, the voice remains the voice and even when the tonality is changed we can still hear the timbre of what has been recorded. It may sound a little 'off' due to modifications of the frequency amplitude.
So the examples below will still sound as the examples below and the 'timbre' will be preserved.

Now... if you told me their most "energetic" take will be altered by a playback system into the most "controlled" take then we could say gear alters the timbre.

So... reproductive gear does not have a timbre.... it may have a deviating tonality or add audible amounts of distortion but not timbre.

When you say "the violin remains the violin" I think you are confusing timbre (the complex characteristics of the actual sound) with the categorical memory shortcut (that enables us to recognise sound sources, etc).

But the argument whether anything in the reproduction chain "has a timbre" or "affects the timbre" isn't very meaningful. Timbre is the complex character of the sound, which we can examine at any stage including production and reproduction depending on our purpose and area of interest.

 

[Randyman...]

So is the recording the reference? Or is the source before it was recorded the reference? And how do you reference that which is ephemeral and impossible to fully capture (being in the room with the timbre as it is created)? Timbre is exclusive of that IMO, and is intermingled with the musician and their performance on the instrument. Timbre is not ambiance or soundfield or even frequency response from my seats.

Each instrument can still be analyzed in all the mentioned conditions.
In all cases we talk about timbre.

In case of a DAC or amp and even transducers we do not speak of timbre. Transducers can change the frequency response which affects the reproduced signal and thus its timbre but it does not have a timbre.
If it had all instruments and voices would start to sound the same as it would be defined by the timbre of gear. It doesn't. The timbre is altered (due to FR).

The violin remains the violin, the acoustic guitar remains the acoustic guitar, the voice remains the voice and even when the tonality is changed we can still hear the timbre of what has been recorded. It may sound a little 'off' due to modifications of the frequency amplitude.
So the examples below will still sound as the examples below and the 'timbre' will be preserved.

Now... if you told me their most "energetic" take will be altered by a playback system into the most "controlled" take then we could say gear alters the timbre.

So... reproductive gear does not have a timbre.... it may have a deviating tonality or add audible amounts of distortion but not timbre.

And we are totally in agreement that reproductive gear does not have a timbre! That happens WAY upstream (with the musician as you know I will say )

 

[solderdude]

When you say "the violin remains the violin" I think you are confusing timbre (the complex characteristics of the actual sound) with the categorical memory shortcut (that enables us to recognise sound sources, etc).

Why would there be confusion about the timbre of anything and how that is processed by an individual ?
Those are totally different things.

 

[solderdude]

So is the recording the reference? Or is the source before it was recorded the reference? And how do you reference that which is ephemeral and impossible to fully capture (being in the room with the timbre as it is created)?

For playback there is ONLY one reference... the recording as there is no access to what original was and how that was recorded and above all manipulated.

 

[Randyman...]

For playback there is ONLY one reference... the recording as there is no access to what original was and how that was recorded and above all manipulated.

So we will be comparing the recording - as there is no access to what original was and how that was recorded and above all manipulated.

And then we will use that MANIPULATED recording to define the instrument's absolute timbre? Seems backwards - but OK. So at this point - then it really does come down to FR and Spectral analysis of your reproduction system since the recording will be the same waveform played back everytime. So what are we measuring if not playback accuracy? Now I'm lost...

This is not timbre IMO...

 

[Axo1989]

Why would there be confusion about the timbre of anything and how that is processed by an individual ?
Those are totally different things.

Perhaps because you think a "thing" has timbre, when in fact a sound does. You appear to use recognition of "violin" to argue against this. But maybe I've misunderstood you? Recognising "violin" after a change in tonality doesn't mean that spectral balance isn't part of timbre, it just means that there are other elements to timbre as well.

 

[Galliardist]

Sure it's a performance, but a singer can change the timbre of his/her voice and a violin can have different timbres if the musician bows or picks the strings.

And every note may be different, Changing the point where the bow contacts the strings by a couple of millimetres changes the sound. So musicians have an understanding of the different timbres their instruments should produce, Instrument makers (as I referred to before) have to create instruments that can adequately produce the expected sounds from different playing techniques. A classical guitar that produces weak sound when the player plays harmonics is not much use to a player performing music that requires that technique, after all.

At some point in this micromanagement of the word, though, the term timbre becomes useless for any purpose, especially ours, as we seek to reproduce infinitely different combinations of sounds on our systems.

I think the key point here is the second part of the question, can we measure it? I presume that what is intended is the question of whether we can measure the ability of our systems to reproduce different timbres: and standard measurements are how we do that.

If you want to measure instruments, that's been done for as long as we've had tools to measure sounds. Science and Music by Sir James Jeans was published, with measurements, as we know them, in 1937. Music was one of the first uses of mathematics, and the base mathematics for the Western music system was known in Mesopotamia as long ago as 1500 BC - a thousand years before Pythagoras, who we all know "invented" it. There's a whole literature on the subject.

As for voice, the singer is still restricted by the size of the diaphragm, lung capacity, the physical characteristics of their vocal chords and palate, and so on. If you know the voice of a singer well enough, you will recognise it no matter what vocal technique they are using.

 

[killdozzer]

I've seen this mentioned several times. Does anyone realize how lousy a speaker has to be before it resonates enough to be audible? Midrange leakage through a port is audible, but is it audible in a manner that it affects the timbre of instruments on the recording? Has anyone done a scientifically controlled test of any sort whatsoever to see whether a listener can identify speaker timbre? Does anyone have numbers or measurements?

This site supposedly advocates principles of science. Not only that, but we have criticized anecdotes that are not backed up by controlled tests. It seems that we should produce more data here, and make fewer assumptions.

Otherwise, we're just fumbling around in the dark.

Well, OK, but at one point in time speakers were deliberately made to resonate. Like these old Saba speakers:

And just to make it clear; YES, the same instrument sounds different when played by different musicians and YES speakers do differ. I just wouldn't use the term timbre for that to keep terms more precise. It's not that we couldn't stretch it to the point where it would equate with sound signature, but I don't think that is a goal worth achieving.

 

[solderdude]

Perhaps because you think a "thing" has timbre, when in fact a sound does.

An instrument has a timbre (distinct sound).
That timbre can change depending on how it is played even if the tuning of the fundamental remains the same.
Timbre is the spectrum at each moment in time which changes during attack, sustain and decay and all those aspects determine how it sounds.
How it 'sounds' also depends on acoustics, distance, angle even if the timbre of the played note is the same.

 

[killdozzer]

The musician driving the instrument with their most "energetic" take?

My problem with this:
The musician driving the instrument oboe with their most "energetic" take // vs. // the musician driving the instrument trumpet with their most "energetic" take.

This helps to underline how 'timbre' was taken from instruments to be applied to style, tempo or such 'non-instrument-related-impacts-on-sound'.

I'd say needlessly obfuscating.

 

[killdozzer]

Perhaps because you think a "thing" has timbre, when in fact a sound does.

This is a very strange notion to me. It's almost like saying language has ideas, not humans. Language carries/expresses human ideas.

 

[solderdude]

And then we will use that MANIPULATED recording to define the instrument's absolute timbre?

No, the timbre is what the instrument does, the recording including its manipulations, studio monitor sound and final production defines how the recording should sound.

That recording thus becomes the reference and we can't undo what has been done to the original spectrum, attack, sustain and decay nor the acoustics and microphone placement at the time of recording.

When this is done pretty well we still hear the oboe as oboe and not as a trumpet or piano and differentiate and even recognize its timbre.
The extremely small change in the spectrum that HD adds won't change how it sounds, modifications in the frequency response will.
We can't say what it sounded like at the moment the instrument was played so that reference is gone and at best in the memory of the performer.

 

[Galliardist]

So we will be comparing the recording - as there is no access to what original was and how that was recorded and above all manipulated.

And then we will use that MANIPULATED recording to define the instrument's absolute timbre? Seems backwards - but OK. So at this point - then it really does come down to FR and Spectral analysis of your reproduction system since the recording will be the same waveform played back everytime. So what are we measuring if not playback accuracy? Now I'm lost...

This is not timbre IMO...

So, what can we do that destroys timbre? As far as I can tell, to destroy the timbre of a particular musical instrument, say a Smallman classical guitar with a decent player hitting a couple of rest strokes, or a particular Steinway model piano, is hard to do during recording and playback. It will survive dynamic compression. It will survive most lossy compression. It will survive being recorded with an unsuitable microphone or the predations of a hamfisted engineer. It will survive turning the treble control where you have one, up to 11. It sounds like a Smallman guitar through any of the big speaker brands, including the ones regarded as garbage in these parts. It will survive being played back through non-transparent amps. It will sound like a Smallman guitar or a Steinway piano when played over iPhone speakers, cheap earbuds... on vinyl, played back on a 1950s radiogram... you'd probably have to go to low bitrate AAC with an unsuitable spectral replication setting to destroy the timbre of that instrument (I discovered this accidentally, courtesy of a new digital radio and ABC Classic FM, via DAB+ in that way). Sure, as you go down the levels of quality, you get poorer playback of that sound, but timbre it turns out is surprisingly robust.

Adele's voice is still Adele's voice at DR 4 - I'm not going through that list again, I'm sure you are getting the picture by now. Timbre is what makes the initial source of the sound recognisable.

We need to reproduce the infinite set of sounds that may be produced by anyone with anything. For the purposes of reproducing the sound on the recording, timbre is irrelevant. The source, amplifier and speakers have no concept of it.

So, we measure playback accuracy. When we do so, we measure the ability to reproduce the set of possible timbres. We need to be able to recognise timbre to relate to the art on the recording, but the system doesn't.

See? By measuring playback accuracy, you're not lost at all. It's the right answer!

 

[solderdude]

It would be nice to make a cheap acoustic guitar recording to sound like an expensive highly valued acoustic guitar with certain manipulation.
Perhaps AI could assist once it learns the difference and to apply that to the poor guitar sound recording making the recording sound like a M$.

In fact there are already AI generated songs that appear to be sung by original artists but aren't... that's a nice example of faked timbre of the the original voice.

 

[Galliardist]

It would be nice to make a cheap acoustic guitar recording to sound like an expensive highly valued acoustic guitar with certain manipulation.
Perhaps AI could assist once it learns the difference and to apply that to the poor guitar sound recording making the recording sound like a M$.

In fact there are already AI generated songs that appear to be sung by original artists but aren't... that's a nice example of faked timbre of the the original voice.

People thought it would be nice to fix pitch problems with people's singing, didn't they? Now all we get are complaints about that sort of manipulation. Ho hum...