What is timbre and can we measure it?

[Cbdb2]

Yes, best instrument examples are cymbals, lots of frequencies no pitch but they have timbre.

 

[DonH56]

Musical definitions (from memory, hopefully correct):

• harmonic = multiples of the fundamental pitch (frequency, note)
• subharmonic = submultiples of the fundamental pitch
• enharmonic = different notation of the same pitch (e.g. D#, Eb are the same frequency but notated differently per the key signature, and may be slightly different in pitch depending upon the chord)
• inharmonic = discordant, not harmonically related (typical example is a drum or other percussion instrument, but other instruments can create inharmonic tones depending upon the instrument and the player)

There are a few other more esoteric harmonic terms (musical and otherwise) but those are the main ones IMO.

Formant refers to a region of resonance, or amplitude peak, with the center of the region generally referred to as the formant frequency. Usually refers to vocal sounds, though occasionally more disparate things like spread spectrum systems or even the grouping of strings for a piano note.

A reminder that timbre involves not only frequency but also time-varying amplitude information. To distinguish timbre, spectral analysis must reflect the time-varying nature of the waveforms and not just basic frequency information.

 

[Gringoaudio1]

Musical definitions (from memory, hopefully correct):

• harmonic = multiples of the fundamental pitch (frequency, note)
• subharmonic = submultiples of the fundamental pitch
• enharmonic = different notation of the same pitch (e.g. D#, Eb are the same frequency but notated differently per the key signature, and may be slightly different in pitch depending upon the chord)
• inharmonic = discordant, not harmonically related (typical example is a drum or other percussion instrument, but other instruments can create inharmonic tones depending upon the instrument and the player)

There are a few other more esoteric harmonic terms (musical and otherwise) but those are the main ones IMO.

Formant refers to a region of resonance, or amplitude peak, with the center of the region generally referred to as the formant frequency. Usually refers to vocal sounds, though occasionally more disparate things like spread spectrum systems or even the grouping of strings for a piano note.

A reminder that timbre involves not only frequency but also time-varying amplitude information. To distinguish timbre, spectral analysis must reflect the time-varying nature of the waveforms and not just basic frequency information.

Your last paragraph, as a non-musician, is what my rigid world view was not understanding. I was convinced that an instantaneous snapshot of the character of the sound was timbre. My apologies to all. The mix of various Fourier components changes through the entire duration of a note on an instrument. The piano strings when struck and as the sound decays give a changing mix of harmonics and whatever other frequencies make up the character of a piano sound. The entire event of a note is what identifies it as having the timbre of a piano created note. Gotcha.

 

[DonH56]

Your last paragraph, as a non-musician, is what my rigid world view was not understanding. I was convinced that an instantaneous snapshot of the character of the sound was timbre. My apologies to all. The mix of various Fourier components changes through the entire duration of a note on an instrument. The piano strings when struck and as the sound decays give a changing mix of harmonics and whatever other frequencies make up the character of a piano sound. The entire event of a note is what identifies it as having the timbre of a piano created note. Gotcha.

Yes! Each tone (harmonic or note) can have a different envelope, i.e. attack, decay, sustain, and release characteristics. That is part of timbre and one reason it is hard to analyze. Think of a sound from two different instruments, one that has the higher frequencies dying off quickly, and another the lower frequencies, for a handwaving thought problem. The percussive attack of a piano hammer vs. the softer attack of a flute playing the same pitch, for instance, and how the piano string causes the sound to decay less rapidly than when the flautist stops the note by stopping the air flow. Not only is the instantaneous spectral content different over time, but the overall envelope of individual frequencies within the sound varies.

 

[MaxwellsEq]

The piano strings when struck and as the sound decays give a changing mix of harmonics and whatever other frequencies make up the character of a piano sound.

It's a bit geeky, but I love listening to piano notes as they decay. The interplay, decay and sometimes the re-emergence of a previously faded harmonic is very satisfying...

 

[Galliardist]

Yes, best instrument examples are cymbals, lots of frequencies no pitch but they have timbre.

Cymbals do still have a fundamental, though in normal use it may be hard to discern.

 

[Galliardist]

Your last paragraph, as a non-musician, is what my rigid world view was not understanding. I was convinced that an instantaneous snapshot of the character of the sound was timbre. My apologies to all. The mix of various Fourier components changes through the entire duration of a note on an instrument. The piano strings when struck and as the sound decays give a changing mix of harmonics and whatever other frequencies make up the character of a piano sound. The entire event of a note is what identifies it as having the timbre of a piano created note. Gotcha.

Maybe, maybe not. There have been experiments done that show that we identify some instruments more from the initial transient than the decay of the note, by for example attaching the initial transient of a piano note to the decay of another instrument.

Once again, our brains take shortcuts. Sometimes we need to identify a sound instantaneously (our ancestors couldnt wait for the lion's entire roar), and it is also helpful to identify quickly the notes of one instrument in an ensemble.

 

[Titurel]

Cymbals do still have a fundamental, though in normal use it may be hard to discern.

Excite the cymbal's edge with a bow (bass bows are probably best) and you can readily hear the fundamental and the harmonics.

 

[Chr1]

Apologies for jumping in here, but would it be wrong to say that valve equipment used in audio has timbre?
If so, what would be the correct terminology for their effect on the signal?
Thanks.

 

[Titurel]

Maybe, maybe not. There have been experiments done that show that we identify some instruments more from the initial transient than the decay of the note, by for example attaching the initial transient of a piano note to the decay of another instrument.

Question: How were the listeners (test subjects) in these experiments "qualified*" or were they even qualified at all? Skilled musicians, those familiar with the instruments in the test, would be probably far more difficult, if not impossible, to duped by splicing the attack of one instrument onto the sustained tone of another. Hiding the splices, even if done in the digital domain, would be very challenging.

* Assessed for age, previous experience, relevant skills, aural health, etc.

 

[Galliardist]

Question: How were the listeners (test subjects) in these experiments "qualified*" or were they even qualified at all? Skilled musicians, those familiar with the instruments in the test, would be probably far more difficult, if not impossible, to duped by splicing the attack of one instrument onto the sustained tone of another. Hiding the splices, even if done in the digital domain, would be very challenging.

* Assessed for age, previous experience, relevant skills, aural health, etc.

I’ll need to find the details again, but I believe the abilities of the listeners weren’t specified in the summary I read for these tests.

 

[Guermantes]

Here is a short summary on research into "timbre spaces" -- attempting to ascribe acoustic descriptors to timbre perception. It's primarily focused on musical instrument acoustics and there are some papers cited in the references which may be useful.

Timbre Space — Timbre and Orchestration Resource

A sound’s timbre can be represented by a point inside a conceptual timbre space, a space where the axes are usually defined by the principal attributes (e.g., spectral centroid, attack time) that one would implicitly use to perceive or conceive any timbre in a given context. Timbre spaces are obtain

timbreandorchestration.org

Of particular interest might be this paper from the ASA journal:

https://pubs.aip.org/asa/jasa/article-abstract/118/1/471/539800/Acoustic-correlates-of-timbre-space-dimensions-A?redirectedFrom=fulltext

 

[antcollinet]

Very good point. I’d better find a more accurate term for those non harmonic frequencies. Sloppy of me. What do you call them? I’ve seen ‘inharmonic overtones’ used to describe them just now while searching.

Just "tones" for me.

 

[kemmler3D]

Maybe, maybe not. There have been experiments done that show that we identify some instruments more from the initial transient than the decay of the note, by for example attaching the initial transient of a piano note to the decay of another instrument.

Once again, our brains take shortcuts. Sometimes we need to identify a sound instantaneously (our ancestors couldnt wait for the lion's entire roar), and it is also helpful to identify quickly the notes of one instrument in an ensemble.

100% agree that often the attack is the most important part of timbre, but I'll also argue that it's pretty hard to identify most instruments based on a static snapshot.

With additive synthesizers, you can "freeze" the harmonics and listen to them without any variation over time. When you do that, most of the time the timbre is entirely lost. It's a lot like looking at a single row of pixels from a photo and trying to identify what it is.

 

[Titurel]

With additive synthesizers, you can "freeze" the harmonics and listen to them without any variation over time. When you do that, most of the time the timbre is entirely lost. It's a lot like looking at a single row of pixels from a photo and trying to identify what it is.

Question: Does your additive synthesizer produce (or account for) the non-harmonically correlated sounds typically associated with natural (I.e., physical) instruments? If not or if not systematically, it would non-sensical to assume poor timbral identification in any test based on simple additive synthesis to be anything other than the result of an incomplete (or inadequate) instrumental approximation.

 

[Titurel]

I’ll need to find the details again, but I believe the abilities of the listeners weren’t specified in the summary I read for these tests.

Pretty critical information, n'est pas?

 

[kemmler3D]

Question: Does your additive synthesizer produce (or account for) the non-harmonically correlated sounds typically associated with natural (I.e., physical) instruments? If not or if not systematically, it would non-sensical to assume poor timbral identification in any test based on simple additive synthesis to be anything other than the result of an incomplete (or inadequate) instrumental approximation.

Yes, the kind of thing I'm talking about is really just FFT based resynthesis, so it takes a recording of the instrument and lets you manipulate the partials, time-stretch, and so on.

I did a quick example, I won't tell you the instrument so as not to give it away, but even if you get it right away, I think you will agree it's not as recognizable as it would be if the harmonics varied over time, even though all the partials are present and accounted for.

 

[Axo1989]

I think there is a substantial difference.

When we 'hit' an instrument it produces a tone with a specific timbre (spectrum that has attack, sustain, decay and changes during that time).
A speaker, when 'hit' with a pulse (be it DC or Dirac pulse) generates a dry sounding 'tick' with very short attack, no sustain and quick decay.
When comparing speakers the 'tick' may well sound a bit different but does not produce a 'tone' with a clearly recognizable 'timbre'.

What a speaker will do, because of FR variations and resonances, is change (affect) the timbre of reproduced instruments but doesn't have one by itself.
A speaker does have a distinct tonal balance but not a timbre (it isn't an instrument).
A speaker can also change the timbre.
A (bass)guitar speaker in a cabinet can be said to be part of the instrument and determines the timbre (through FR irregularities and resonances) but those speakers are part of an instrument. One does not play a guitar amp with Dirac pulses to create a certain 'sound' either but uses them to change(modify) the sound of an instrument.

So have timbre does not equal affect timbre because timbre is from an instrument which a speaker isn't.

No, I meant that we can use the word "timbre" with respect to characteristics of a loudspeaker whether we consider they effect or affect it.

The final statement of your argument appears to beg the question—"timbre is from an instrument" is an assertion, and neither evidence nor conclusion.

As for the rest of the thread, it's another one that's becoming time-reversal symmetric so I've left it be. But @chervokas has made some interesting arguments, which I may explore further.

 

[IPunchCholla]

Yes, the kind of thing I'm talking about is really just FFT based resynthesis, so it takes a recording of the instrument and lets you manipulate the partials, time-stretch, and so on.

I did a quick example, I won't tell you the instrument so as not to give it away, but even if you get it right away, I think you will agree it's not as recognizable as it would be if the harmonics varied over time, even though all the partials are present and accounted for.

It’s clearly a 1987 Nissan Maxima horn.

 

[Titurel]

Yes, the kind of thing I'm talking about is really just FFT based resynthesis, so it takes a recording of the instrument and lets you manipulate the partials, time-stretch, and so on.

I did a quick example, I won't tell you the instrument so as not to give it away, but even if you get it right away, I think you will agree it's not as recognizable as it would be if the harmonics varied over time, even though all the partials are present and accounted for.

We were discussing the wonders of additive synthesis, not FFT based resynthesis. That kind of gets us off target. But for the sake of the game I'll play along. Your sample sounds like a poorly synthesized oboe - might I say it sounds very 1980s Casioesque? - but stripped of the tonal evolution that occurs with real instruments, it easily could be any number of things, even iPunchCholla's 1987 Nissan Maxima horn or the death cry of a gerbil. In short, it doesn't sound remotely like a conventional, real-world instrument, regardless of its ultimate origin. (BTW, I'm rooting for formant based analysis, not FFT.)

Now play fair. Get back to me with a timbral sample, created entirely with additive synthesis, which sounds identical to a fair length* sample from a physical instrument, when both samples are packaged in the same, generic* envelope. Then we can start talking seriously. (I will be honest with you. I'm not that interested in synthesis anymore. I've got a small stack of synthesizers (and a sampler) in my office closet gathering dust, where they've been for over a decade and where they will remain for the foreseeable future.)

* we can debate what fair length is and what a generic envelope might be, but I think you get the picture. We should neither hide the ball nor cover it with guacamole.