▲ 139 points
back
108 comments
> "Timbre’s role in music has always been underrated, or even ignored, probably because it is an intangible that’s difficult to describe, hard to categorize, and so far, immune to measurement. "
In the mid-to-late 1800s, Herman Helmholtz was building spectrographs of musical sounds by listening to them through a bunch of resonators of varying sizes that acted as band-pass filters, and recording how loud each harmonic sounded.
His writings are still relevant today.
https://www.amazon.com/Sensations-Tone-Dover-Books-Music/dp/...
I've heard that the Sethares book (which is much more recent) is really good too, but I don't have a copy of that one.
https://www.amazon.com/Tuning-Timbre-Spectrum-William-Sethar...
The quoted section of the article is pretty out of touch with modern music. Many (most?) genres of EDM place the importance of timbre and texture above harmony and melody
To quote from Computer Music by Dodge and Jerse (an excellent reference),
"Helmholtz concluded that the spectral description of a sound had the most straitforward correlation with its timbre.... For instance the qualitative description of 'brilliant' or 'bright' characterizes spectra that have a great deal of energy at high frequencies [e.g. brass instruments] ... A spectrum with little energy in the even-numbered harmonics characterizes the particular timbre that is produced by the clarinet in its low register...." (2nd Ed. Section 2.6)
Understanding that each instrument produces a spectrum of harmonics peculiar to it ... and that those harmonics have relatively different pitches and amplitudes and evolutions ... that vary with how the instrument is played ... is key to quantifying its timbre.
(At first, not an easy thing to grasp let alone quantify. But Partch made a lot of progress!)
I visited conservatoire des arts et metiers museum in Paris a couple years ago. It is full of fascinating scientific devices from the past. Among then was the timbre analyser. I was blown away. A purely analogic spectrum analyser.
https://commons.wikimedia.org/wiki/File:Timbre_analyser-CNAM...
I've gotten a lot of use out of thinking of genre in relation to timbre. Not as the only relevant thing, but part of a constellation.
In most genres you have a palette of timbres to work with for each instrument based on the genre conventions. You can push at the edges or add things in, but have to balance audience expectations carefully as you do it. So for example the way an electric bass sounds in contemporary metal is just not normally going to "fit" into a chicago blues band, even though both are heavily dependent on the sound of electric bass.
Different genres have different relationships to this constraint, for example western classical has a huge palette available in an orchestra but is relatively averse to using instrumentation outside of that collection. Except in percussion, where there is a lot more flexibility! Jazz has a fairly small & rigid set of acceptable timbres for its core instruments, but is fairly tolerant of experimenting with new instrumentation.
So then the two genre-timbre relationships I find most interesting are electronic and pop. Pop is, more than any other genre I think, curious about how timbre effects emotional response in music. An album by the same artist could have a huge range of different sounds for recognizable instruments, using the tension between them for different effects.
And then large swaths electronic music are built largely around active change of timbre over time through a piece. Something you see used conservatively and carefully in most genres becomes almost the central practice.
IDK it's hard to articulate and I'm not trying to set a reductionist framework about how music sounds or anything. Just a line of musical thought I've been exposed to and found valuable.
Well there is Dimmu Borgir, with an orchestra and a choir: https://www.youtube.com/watch?v=El4zsUZjsDc - and it is just nuts. Yes that's 90 minutes, but watch 5 and then decide.
And that's why I love folk-ish metal so much.
Why shouldn't we have a lead/harmony violin on top of a metal rhythm section? https://www.youtube.com/watch?v=F17sxgZdXZw
Or a hurdy gurdy? https://www.youtube.com/watch?v=LT7HqIP55tI
Or a pipa, which was btw very hilarious in poland: https://www.youtube.com/watch?v=MwpKZtPBIEQ
I'm honestly thinking of dropping some money to get myself an A-tuned drone flute[0]. Drone flutes sound really cool, and the drone would be tuned to just about the highest two notes on the guitar... and if you work with it a bit, drums/bass/rhythm guitars helping with rhythms, chords and harmony could really push that somewhere really interesting.
I went to a interesting hackathon the other day focused on building tools for exploring timbre in sound https://comma.eecs.qmul.ac.uk/timbre-tools-hackathon/.
Was brilliant, a lot of groups focusing on the use of ML to characterise the "unexplainable" in sound synthesis.
We ended up submitting a tool for interacting directly with Ableton using LLM agents after becoming disenchanted with text2audio models, wrote about it here - https://montyanderson.net/writing/synthesis
Very cool. I will have to join "mod wiggler" lol.
I have been out of it so long muff wiggler is now mod wiggler. Come on, that is absurd. Muff Wiggler was the best name ever.
The lead developer of Melodyne, Peter Neubäcker, has one of the most precise and concrete understandings of timbre. He used that knowledge to create software which changes pitch without changing speed or timbre.
There's a neat documentary floating around which includes some scenes where he explains how he increased his understanding of waveforms in the frequency domain by physcially modeling them in the time domain.
The way a sound evolves in time contains a lot of timbrical information.
Different harmonics have different ADSR curves (Attack, Decay, Sustain, Release).
Above all, one cannot overstress the importance of the attack transient. There are famous experiments in psychoacoustics that show that, when deprived of their attack transient, the sounds of two different instruments may become hard to tell apart.
Personal anecdote: as a classic guitarist, it took me three years of experimentation to find the right way to cut my fingernails in order to have a better sound. The Electronic Engineer in me says that those were three years spent to look for how to improve 0.1 seconds of noise at the start of each of my notes.
>Above all, one cannot overstress the importance of the attack transient.
Roland used this fact to produce surprisingly realistic sounds at low cost. When memory was too expensive to sample the whole note, you could get excellent results from sampling just the transient and using traditional subtractive synthesis for the rest:
> when deprived of their attack transient, the sounds of two different instruments may become hard to tell apart
A neat trick is playing guitar through a volume pedal. You mute the sound, pluck a string and then swell in the volume pedal. The rest of the ASDR envelope can be created with other effects like delays.
Timbre is what makes the violin sound different from the bassoon, what makes the strings in general gives them their sound, and brass theirs. Timbre is what makes a singers voice unique.
Classical composers had a set of fixed timbres to play with, each instrument having their own. With synthesizers everything is possible.
There are two levels to "everything is possible." First, recorded music is a string of numbers, and any way of generating those numbers will sound like it's intended to. But synthesizing a desired sound is still an evolving process of developing useful interfaces and waveform generators.
Generating that sound in a live performance that is partially or fully improvised is a harder problem.
The evolution of solutions to either problem depends on which instrument you're interested in, and is based partially on Means, Motive, and Opportunity. There's a reason to synthesize a piano or Hammond organ, which is that they're hard to transport and maintain. Less motive for synthesizing a violin.
"everything is possible" except for a synthesized instrument that actually has the expressive range of a real violin or saxophone. Even a synthesized piano is noticeably lifeless compared to the real thing (synthesized piano bass notes are especially thin).
I've always thought of Jimi Hendrix as an example of timbre used for revolutionary purpose. A lot of what he acheived he acheived with timbre.
He definitely opened the door for timbral exploration on the guitar using electronics.
> A vocal timbre can also reveal economic and cultural background. Some singers...were encouraged to either accentuate their timbre...or to abandon what occurred naturally for them
Vocal timbre is a big identifier of different rap subgenres and artists, seems like. Certainly seems like much of the vocal delivery on a track explores fluctuation of the timbre, even when the pitch is constant or has little dynamics.
It's interesting to think how the timbral effects of a regional accent would interact with the beat construction/development. Is the trap hi-hat a result of the Southern drawl?
The reason I hate heavy autotune in recent hip-hop is because it destroys all of the interesting timbre that could be there, and reduces it all the same annoying voice
As a synth enthusiast I generally think of timbre in relation to "real" instruments, brass, strings, piano, etc. In the article they mention there are as many as 10 dimensions but don't actually define what those are. Timbre is something I've thought about a lot but nailing down a more general classification seems pretty challenging.
That’s funny, I’m more
likely to describe “real”
instrument timbres in synth terms, many of which I bet relate closely to the 10 dimensions. Oscillator waveform, envelopes, filter shape/cutoff, resonance…
Synth players have studied and emulated timbres (and created new ones) since the pipe organ (an additive synthesizer). Many books have been written about timbre in electronic terms -- a combination of frequencies and amplitude envelopes that either sound to our ears like classical instruments, or like something else entirely.
I would go so far as to say understanding and selection of timbre is 50% of the work of a synth player.
Tangentially, Sennheiser HD600 headphones are known as "timbre kings" for their ability to reproduce timbre accurately.
Pitch is the frequency of the oscillation, timbre is the internal structure of what's oscillating. Is that right at all?
"Everything is overtones", or so I've heard somewhere. The pitch, or note you hear from an instrument is the fundamental frequency, and the combination of overtones determines the timbre. These overtones are other sine waves at different amplitudes and frequencies, and the combination of these are instrument dependent. Additive synthesizers do this to approximate "real" instruments.
Timbre is the quality that's left after you factor out the pitch and intensity. How it "sounds".
Yes, that sounds mostly right to my understanding. Exploring the space of that internal structure is the point of the article. Both pitch and timbre get remarkably complicated the more they're looked in to.
Timbre can vary a lot day by day on the same instrument.
Violin and the bow are both made of wood and a rainy damp day can easily sound different than dry summer's day.
You can get good carbon fiber bows and they are much less weather sensitive.
You can also get carbon fiber violins but I haven't heard one that sounds as good as one made from wood.
We can get closer to the mathematical definition of timbre by talking about the strength of the harmonics https://www.youtube.com/watch?v=VRAXK4QKJ1Q
A fine article on the role of timbre which is indeed discussed little compared to the focus on harmony and form for classical music. Charles Rosen gave one possible reason for this neglect in his book "Piano Notes", where he traces it to a philosophical prejudice of composers from Hadyn onwards against variety in tone colour and in favour of more abstract qualities.
> The utility of the piano for composing was its neutral and uniform tone color: in theory (although not in reality) the tone quality of the bass is the same as the treble. In any case, the change in tone color over the whole range of the piano is, or should be, gradual and continuous (there are breaks, of course, when the notes go from one string in the bass to two and then to three in the treble). The monochrome piano might be used therefore just for its arrangements of pitches, and the quality of the sound could-absurdly in many cases-be considered secondary.
> What made it possible for composers to refuse to acknowledge the difference on the piano between treble and bass and leave whatever problems arose to be solved by the performer was the fact that the change in tone color over the span of the keyboard is not like the leap from a bassoon to a flute but continuous and very gradual when the instrument is properly voiced. These imperceptible gradations are the result of a deliberate policy of a unified sonority on the part of musicians and instrument makers. All attempts over the history of piano construction to incorporate anything analogous to the picturesque changes of registration in the organ and the harpsichord had little success, were not exploited by composers, and were finally abandoned. Radical contrasts of tone color were traded for the possibility of making a gradual crescendo or diminuendo. This was a decision that took place at the same time as the preeminence accorded to the string quartet over all other forms of chamber music; that, too, emphasized the importance of a unified tone color. Chamber music with wind instruments, while the occasion for several masterpieces, became the exception, an exotic form. That is why the use of colorful sonorities in the orchestra has so often been considered somewhat vulgar, as if calling attention to the sound were paradoxically to detract from the music. The prestige given to pure string sonority is part of the asceticism of nineteenth-century high culture. Contrasts of tone color were given a significantly lower place in the hierarchy of musical elements. This is one reason that only the piano repertory rivals the string quartet as the most respectable medium for private and semiprivate music-making from Haydn to Brahms.
Highly recommend the book “Tuning Timbre Spectrum Scale” for engineers interested in music theory
Arnold Schoenberg observed the deep connection between music, perception, and cognition; anticipated composers applying timbral control as is now commonly done with synthesizers; and apparently did not anticipate non-subtlety of the effect in EDM:
"I think the tone becomes perceptible by virtue of tone color [timbre], of which one dimension is pitch. ... Pitch is nothing else but tone color measured in one direction. Now, if it is possible to create patterns out of tone colors that are differentiated according to pitch, patterns we call 'melodies', progressions, whose coherence evokes an effect analogous to thought processes, then it must also be possible to make such progressions out of the tone colors of the other dimension, out of that which we call simply 'tone color', progressions whose relations with one another work with a kind of logic entirely equivalent to that logic which satisfies us in the melody of pitches. That has the appearance of a futuristic fantasy and is probably just that. But it is one which, I firmly believe, will be realized.
...
Tone-color melodies! How acute the senses that would be able to perceive them! How high the development of spirit that could find pleasure in such subtle things!"
-- Arnold Schoenberg, Theory of Harmony, 1911, toward the end of chapter 22.
Personally, I anticipate that if humanlike AGI is achieved, then it will understand the illusion of music. Not merely because music understanding is an attribute of humanlike, but because there must be something intrinsic in the process of human cognition that causes the illusion of music occur. Music understanding could be viewed as an essential waypoint on the path to AGI.
I had a simple thought a couple years ago that still fascinates me. If you do something as simple as "clap clap clap" (with equal measures of time), you can naturally relate it to any other set of sounds with that simple pattern (ABABA). moreover, there is a 1 to 1 relation of this abstraction in space, where one can represent the sounds as marks on a screen (as I have done here for example, as "clap clap clap, X X X, - - -, or whatever).
Despite the different sensory systems (sound, or visual) and the difference in actual data (I don't know how many photons or audio waves entering your brain), we can abstract them down and understand/relate them with a simple pattern.
Pop music really does do this as he described. There are lots of songs that play with expectations of timbre in the same way that normal music plays with expectations of pitch.
When I get mentally crystallized, I hear music as proof-of-work in the vein of Bitcoin hashes.
I very much appreciate your comment, and at the same time am interested to challenge this idea of "the illusion of music". What about music is an illusion? Maybe this was just an unfortunate phrasing, and yet it might touch on the heart of the matter. Music is a language of relationships, indeed it is nothing but relationships, in time, pitch, timbre, volume, voice, number and so on, non-symbolic and changing relationships of sound carrying aesthetic and emotional feeling and meaning. It is both objective and subjective -- objective in the sense that "4 + 4 = 8" is an undeniably more beautiful statement than "4 + 4 = 6", and subjective in the sense that you and I may have different favorite numbers and also we may react differently to various medications or prefer different foods. Music is external and internal, social and individual. What it is not ever is an illusion as then it would not exist. (It may express illusion, or become more like a creator of illusions, for example when the orchestra tries to conjure up the illusion of a thunder storm.) Now, one might be tempted to counter that the kind of "meaning" music carries is itself an illusion, an epiphenomenon maybe, but then why should one care about anything at all, including AGI, since the kind of meaning that music carries is the same as the meaning of love, of hate, of good and bad, pains and pleasures, and hopes and regrets, tragedies and successes?
Maybe a reason for calling music an illusion is to try to point out that it's all in our heads. Sure, the physical instruments are over there, but they are just some atoms and molecules rattling about creating vibrations that, like everything ultimately, share the meaninglessness of the physical universe of just more material stuff (ignoring the quantum and the unknown). In which case though life seems also to look more like an illusion. Ah but the meaningless vibrations enter our ears and are processed and become music, so like color there is none but in our heads, thus illusion? But these notions also fall apart. The redness of apples isn't all in our heads; it requires a certain relationship to light frequencies that we see as redness. It's not an illusion it's just what red looks like through a living human body. If not through our senses and brain, and then mind, where should we experience it?
Music is only ever meaningless if it is unheard, as apples can only not be red when they are unseen (or are Granny Smiths).
So music is only meaningless if there is nobody -- no body -- to listen. And maybe that's where all of this leads. Your AGI needs to be embodied and alive.
This is tangential to much of what this article discusses, but I've had this idea kicking around in my head for a while, that, to some degree, the artificial imitation of acoustic instruments using synthesizers is to music what skeuomorphism is to design, and I (generally) dislike it for all the same reasons. This isn't to diss synthesizers on the whole - I unequivocally love them, and they really opened up my musical world by a wide margin. Nor am I saying that there is no place for synth patches that attempt to "sound like" "real" instruments (for lack of a better word; I consider every "model" of synthesizer to be its own very real and distinct instrument).
Skeuomorphic design, to me, largely feels lazy and unattractive; rather than designing for the new, maybe unfamiliar, medium you're working in, you attempt some facsimile which inevitably cannot live up to the original, either out of lack of ambition, or lack of faith in your audience to understand or appreciate without the anchor of a common metaphor. It lacks idiomaticity - there are particular details, quirks, associated with different mediums that lend themselves to different sorts of designs, and a skeuomorphic design language ignores these. (The same is true of musical instruments. A piece written for the lute, say, played on piano, would feel very different than a piece written for piano, played on piano; what is easy, or possible, on one instrument, is not necessarily as sonorous on another.)
To me, the use of synthesizers to emulate acoustic instruments, where the express intent of the composer/producer/whatever is to evoke the sound of that acoustic original (read: where the creator would prefer the sound of the real instrument), is, at its most generous, telling of a lack of creativity and at worst laziness. (If what you want is expressly the sound of a synthetic imitation for the purposes of your art, none of this applies.) It is a lack of creativity in the case where the creator simply does not have access to "real" instruments they'd rather use, and it's laziness where you have access and ability but opt for something you yourself deem inferior.
Of course, this all presupposes that there is something lost when say, a synthesizer plays a violin patch in an earnest effort by the creator to emulate a violin - where a real violin is actually what's desired. I think this isn't controversial to say, but of course it's a spectrum; I have much less of a problem in cases where the delta is smaller; e.g., simple legato harmonies from a string section can be emulated more convincingly (by orders of magnitude) than some virtuosic cadenza by a solo instrumentalist.
At least this article will reinforce the concept, and the correct spelling, if not any appreciation for the quality!
Everything we can't describe in music" is a thought-provoking statement that highlights the limitations and strengths of music as a medium.
"Terroir" is a terrible way to metaphorize timbre. The only similarity is that they're both hard to describe. Terroir is there before the wine is made and can't really be controlled by the winemaker (except insofar as they choose where to plant or what grapes to blend). Timbre is absolutely under the control of the musician.
> Terroir is there before the wine is made and can't really be controlled by the winemaker (except insofar as they choose where to plant or what grapes to blend).
I've met winemakers who think of planting in terms of multigenerational experiments with highly intentional controls for different aspects of soil and micro-climate that may vary on a scale of a single meter, who work to find and train their successors because they will not see the results of some of these experiments in their lifetimes; winemakers who absolutely understand that blending is not just a linear combination of the characteristics of blended wines, but a truly complicated system where certain flavors may heighten others; winemakers who rely on centuries of experience in how funguses and other microorganisms can be used in safe and flavor-enhancing ways.
Winemaking is engineering, just as musical instrument design is (both physical and digital), and just as building a startup is. The presence of human subjectivity as an evaluating oracle in a complex system does not imply that its creators work without intention or without rigorous systems-level thinking.
As a string player, terroir strikes me as a beautifully appropriate way to describe timbre. My instrument was constructed in the 19th century, probably in Germany. The luthier, their choice of wood, the shape, the workmanship, the strings, the bow, the weather, the player, the room, the mood, the other players all affect the sound. A large portion of the job of an instrumentalist is choosing which tools to use. There are always compromises, and many variables that effect the timbre simply are before a note is played, and have to be worked with.
> Another way to think about timbre is by comparing instruments. In an old stand-up bit, Steve Martin strummed the banjo and mused, “You just can’t sing a depressing song when you’re playing the banjo . . . You can’t just go, ‘Oh, death, and grief, and sorrow, and murder.’
Anyone who's played Outer Wilds should disagree with this.[0]
To say we can't describe timbre is a bit misleading, because there are concrete mathematical ways to analyze sounds; they exist in a three-dimensional space of frequency, amplitude and time. But that's helpful in the same way that describing programming languages as collections of 1s and 0s is.
What's lacking for describing timbre, I suppose, are the steps between "this sound is a sum of a particular arrangement of sine waves" and "this sound is a piano". There are common terms such as ADSR or "brightness" and "warmth" but those don't tell the full story.
The question is, how valuable is that intermediate step when you could just say "this is a piano, this is a banjo"?
Tone is in the fingers, as they say for guitarists. Different woods, magnets, strings all impact tone for sure, but at the end of the day Jimmy Page sounds like Jimmy Page and Hendrix sounds like Hendrix cause tone is in the fingers.
That’s what stops music from being math.
This story was not submitted by tintinnabula.
I'm scandalized. /s
tldr: Timbre, the unique quality of a sound that distinguishes different voices or instruments from one another, is complex to define and measure and is often overlooked in music theory, which tends to focus more on pitch, rhythm, and harmony.