The Eat at Joe's Kawai K5000 Message Board Digest What is Timbre? ------------------------------------------------------------------------------------------ ------------------------------------------------------------------------------------------ What is Timbre? Saturday, 02-Jan-99 19:46:48 199.199.157.34 writes: Okay, time to get some musical discussion going again-- What is timbre? I'm thinking it's like the figure-ground issue in visual perception. Somehow our brain decides to group a bunch of visual stimuli together as an object--seperate from the background--based on whether they are near, whether they move together, and whether they fit a known pattern. Timbre seems to be similar--our brain groups a bunch of overtones together as "one sound" based on whether they're harmonically related, whether they move together, and whether they fit a known pattern. Has anybody studied this? leiter ------------------------------------------------------------------------------------------ Re: What is Timbre? Sunday, 03-Jan-99 10:57:07 140.254.113.111 writes: Ahhh... A philisophical discussion... This should provoke a discussion. Though maybe not as "musical" as you'd like. >Timbre seems to be similar--our brain groups a bunch of overtones together as "one sound" based on whether they're harmonically related, whether they move together, and whether they fit a known pattern. Though I have not studied much about the relation between the actual physics of sound and the PERCEPTION of that sound, I think that one obvious supporting arguments for your statement about our brains checking to see if a sound fits a known pattern presents itself in attack transients (AT). Apparently, people cannot distinguish between a variety of instruments if the AT has been completely removed from it and only the sustained portion is played. Likewise, if the AT from one instrument is pasted on the sustained cycle of another (e.g. a violin AT is used with a saxophone sustained loop) in a skillful manner, people will identify the sound as purely the AT instrument (violin in our example). So, even though physically the violin and sax have very different "sounds" (the overtone series and partials are NOT equivalent), they may be perceived as the same sound if the AT is the same for both. So much for physics, eh? I find it frustrating since I'm a physicist (not specializing in acoustics though), but it's a reality I can't escape. It can't be helped, but it's a bit embarrasing when I can't produce a realistic sound from a bona fide Fourier spectrum ;-) Then again, the limitations of our silver buddy seem to put us all on a level playing field. I've been very impressed with some of the stuff from the SHARC database, but some of it just doesn't translate into realistic sounds on the K5k--and a heck of a lot more knowledge and work have gone into those timbres than I'll probably ever put into mine. Our brains can be fooled in some ways, but are pretty amazing in others. We can determine the fundamental of an overtone series even if it has been removed! As long as 2 or more higher harmonics of an overtone series are present, our brains will fill in the missing fundamental by calculating (unbeknownst to our concious selves) the frequency relationship of those harmonics. It knows they are multiples of one fundamental tone--that's pretty amazing to me. I guess what K5000-related timbre comes down to for me is the challenge of reproducing those "recognizable patterns" for our brains. I think the K5k's strengths are more (but not soley) suited for "new" sounds, and those are the sounds I will use most in actual recording. However, additive synthesis is still one of the best methods for attempting to recreate acoustic instruments because it is direct implementation of the method of most timbre research (spectral analysis). Bottom line: Despite it's limitations (no cosines, controllable phase, inharmonicities/partials, etc.), the K5000 is the best, if not only, commercially available (and affordable) synth for your own "timbre research." I love it. Bob ------------------------------------------------------------------------------------------ Re: Re: What is Timbre? Monday, 04-Jan-99 17:52:12 192.28.2.49 writes: That's a good point--our brain groups together "sound events" over time, as well as over the frequency spectrum. Make a moving timbre static, or preceed it by a different attack, and it can be perceived as a different timbre. leiter ------------------------------------------------------------------------------------------ Re: Re: What is Timbre? Sunday, 03-Jan-99 12:41:31 153.35.141.176 writes: Nice reply! I didn't know that about the attack playing a dominant role in the perception of timbre. In some ways it doesn't suprise me. I say this because of the way most music is composed... Instruments with loud distinctive attacks seem to do alot of the work in music, while steady state fade-in types of instruments are usually relegated to some kind of support role (So it's no coincidence that the K5k is popular as a "pad" instrument AND it has a weakness with attacks). Instruments that are capable of both styles seem to reserve their heavy attack style for carrying the melody or for creating rythmic definition (I don't feel that this is a supporting role for a lot of music, if you remove this it totally alters the character of a composition. Radio and dance clubs are filled with music for which this is the case). Here are some examples of traditional instruments that are usually used to define the character of a composition: Piano, Trumpet, Guitar, Flute.... All of these have some serious stuff going on when a note is first sounded. All of this makes me wonder if what really is going on with this situation is that the mind's memory (or attention span) is too short to remember all of the sound at once. When it hears the attack it registers it as an interesting sonic event and then hears the steady state portion of the sound and considers it as both less interesting and unrelated to the attack (which has already been "forgotten" at this point). In other words, your mind perceives two sounds, one that is loud and complex and one that is softer and simpler. Becuase it can't correlate the two portions as being related it picks one to consider as more relevent and soft pedals the importance of the other one. This is all just speculation on my part though. What do you think? jon jonwest@vcu.org ------------------------------------------------------------------------------------------ Re: Re: Re: What is Timbre? Monday, 04-Jan-99 12:08:56 164.107.171.54 writes: Interesting thought about the attention span, though I don't know enough about psychoacoustics to comment on its legitimacy. Just to play devil's advocate though, did you ever notice that after a particularly loud sound with a sharp attack (bells and horns come to mind) leave a sort of "afterimage" in your ear long after it has decayed away. That would suggest that the brain's attention span may be too long in some cases. However, that afterimage is only for the steady-state portion of the sound, which would lend creedence to the idea that the brain may be processing the AT and SS as two (or more?) totally seperate sounds. Once again, just a thought. I'm sort of out of my depth here. Neat ideas. Good discussion. BTW Leiter, do you write a column for Keyboard magazine? Bob ------------------------------------------------------------------------------------------ Re: Re: Re: Re: What is Timbre? Monday, 04-Jan-99 17:55:45 192.28.2.49 writes: >BTW Leiter, do you write a column for Keyboard magazine? Nope! Someone else writes like this? leiter ------------------------------------------------------------------------------------------ Re: Re: Re: What is Timbre? Monday, 04-Jan-99 17:54:20 192.28.2.49 writes: >All of this makes me wonder if what really is going on with this situation is >that the mind's memory (or attention span) is too short to remember all of the >sound at once. When it hears the attack it registers it as an interesting sonic >event and then hears the steady state portion of the sound and considers it as >both less interesting and unrelated to the attack (which has already been >"forgotten" at this point). In other words, your mind perceives two sounds, one >that is loud and complex and one that is softer and simpler. Becuase it can't >correlate the two portions as being related it picks one to consider as more >relevent and soft pedals the importance of the other one. In visual perception, there are different systems that work in parallel: some detect edges, some detect motion, some use color and some don't. Op Art painting plays with some of these. For example, if a painting contains two colors that have the same absolute reflectivity, your color perception says you're looking at a figure while your black-and-white perception says there's nothing there, it's all the same shade of grey. So maybe "attacks" and "sustains" are dealt with seperately in the brain. (That might correspond with the distinction between consonant and vowel sounds in speech, too.) How can you send them contradictory signals? Time-reversed sounds? leiter ------------------------------------------------------------------------------------------ Re: Re: Re: Re: What is Timbre? Tuesday, 05-Jan-99 11:48:42 205.227.43.15 writes: When you examine a curve of human hearing response, the brain is fine tuned to listen to the frequency componants of consonants rather than vowels, correct (upper mid-range frequencies)? Also, at least in the English language, most 'accents' are really changes in the vowel pronunciations without any (or only very slight) changes in the consonants. The vowel sounds can be changed within a certain range and the words will still be completely comprehensible. I think the human ear is oriented around distinguishing between different consonants, which share many characteristics with attack transients. Not to steer too off topic, but I think that the importance of speech in human evolution could present a physiological explanation for the relative importance of attack transients. Kenji ------------------------------------------------------------------------------------------ Re: Re: Re: Re: Re: What is Timbre? Friday, 08-Jan-99 16:16:51 192.28.2.49 writes: >When you examine a curve of human hearing response, the brain is fine tuned to >listen to the frequency componants of consonants rather than vowels, correct >(upper mid-range frequencies)? I think so--vowel formants are between 250 and 2k, which is a sensitive hearing range, but the most sensitive hearing range is about 2k-5k with the peak around 4k. leiter ------------------------------------------------------------------------------------------ Re: Re: What is Timbre? Friday, 08-Jan-99 12:16:34 192.28.2.49 writes: >Our brains can be fooled in some ways, but are pretty amazing in others. We can >determine the fundamental of an overtone series even if it has been removed! As >long as 2 or more higher harmonics of an overtone series are present, >our brains will fill in the missing fundamental by calculating (unbeknownst to >our concious selves) the frequency relationship of those harmonics. It knows they >are multiples of one fundamental tone--that's pretty amazing to me. From what I've read, our ears hear the missing fundamental as if it were there because the "beat" frequency between the two harmonics can excite the corresponding short hairs in our ears, even though there's no actual signal at that frequency. It's possible to make some very good bass patches with no fundamental at all; in fact, it can be better that way. I probably wrote this already, but I think this "filled-in" harmonics thing accounts for the "hollow" or "dead" sound of square and triangle waves. Their harmonics are all odd multiples of the fundamental, but the difference frequencies are all even multiples. So, none of the difference frequencies support the actual harmonics, and vice-versa. The second series of harmonics that does this goes: 1,4,6,9,11,14,16,... (It's used in my "Marimba" patch in the archive.) It's related to a 40% pulse. leiter ------------------------------------------------------------------------------------------ Re: What is Timbre? Monday, 11-Jan-99 08:59:19 194.172.230.108 writes: Hi all! I'm back from holidays and happy to see such an interesting discussion on the message board. Look at this collection of timbre definitions! http://sparky.parmly.luc.edu/sandell/sharc/timbredef.html Jens Groh ------------------------------------------------------------------------------------------ Re: Re: What is Timbre? Monday, 11-Jan-99 21:04:57 199.199.157.62 writes: That's a nice site. I agree with the last few authors that the ANSI definition is off base. It basically says that if you have two notes of the same pitch, loudness and duration, then any remaining difference is "timbre". I think this is backwards: timbre comes _first_ rather than last. Here's what I think timbre may be: The distinction of timbre in hearing is analogous to the distinction of figures in visual perception. In both cases, a limited number of stimuli are selected out of a continuous stream and grouped together as one object, seperate from the background. Timbre refers to the group of sound stimuli collected in a single sound object. This grouping of sound stimuli is mostly based on experience, and for the most part accurately reflects reality. That is, the overtones grouped together as a single sound object usually did come from a single source. The distinction of timbre allows our brain to reduce vast amounts of sound stimuli to simple causes: I hear all these particular tones becuase someone is playing a violin in a cathedral. The distinction of pitch is a further analysis of a sound object. It can be made only after the sound object is defined. It is not always possible to distinguish a single fundamental, or any pitch at all, in a sound object. Distinguishing pitch is the last refinement to distinguishing timbre. Thus, timbre is prior to pitch and not the reverse. leiter ------------------------------------------------------------------------------------------ Re: Re: Re: What is Timbre? Wednesday, 13-Jan-99 03:01:13 153.35.108.125 writes: Interesting viewpoint! But if it is true and timbre is a primary quality and pitch is secondary then why do we build up timbre via a series of pitches? Does this mean that the standard Fourier treatment of these things is deficient? I guess that the granular/wavelet crowd would say so..... jon jonwest@vcu.org ------------------------------------------------------------------------------------------ Re: Re: Re: Re: What is Timbre? Wednesday, 13-Jan-99 11:38:27 192.28.2.49 writes: >Interesting viewpoint! But if it is true and timbre is a primary quality and >pitch is secondary then why do we build up timbre via a series of pitches? Does >this mean that the standard Fourier treatment of these things is deficient? Fourier analysis is a useful approximation but it is deficient because it works for repeating waves, whereas natural sounds are always changing. There is usually a "noise" element left over. For example, there is one method of resynthesis that works by performing a Fourier analysis, resynthesizing the analysed sound from sines, subtracting the resynthesized sound from the original, then taking additional steps to characterize and resynthesize the "noise" portion that is left. (Someone cited a website for this a few months ago.) Also, that's why the K5k has attack transient PCM's and noise PCM's. Maybe some kind of Fourier treatment would give you a complete representation of sound if you took tiny slices of time, but that's beyond my knowledge of mathematics. To ramble on a little more, part of the reason "timbre" is slippery to define is because the timbre of any particular sound is partly arbitrary. I think the natural definition, the one our brains seek because it has survival value, is to lump together all of the sounds eminating from a single object as one timbre. But we can also subdivide a timbre, e.g., distinguish between the vibration of a piano string and the sympathetic vibration of the others. Or we can mentally join instruments, like an orchestra or that Zappa guitar+sax sound. leiter ------------------------------------------------------------------------------------------ Re: Re: Re: Re: Re: What is Timbre? Thursday, 14-Jan-99 00:08:28 153.35.108.178 writes: I guess that there have been many interesting experiments in timbre mixing (human voice + sympathetic piano string vibration ala "Lumpy Gravy" & "Civ Phz III"), have you heard "Thing Fish"? It has a wicked vocoder part on it! About the "noise", this sounds more like the error signal left over from using a finite number of terms in the series. It could be that using two series (resynth & noise) to represent the signal is, in practice, easier to do than to use enough terms to eliminate the error from the replication in one go. Mathematically this doesn't make sense but throw in some psychoacoustic effect and it may work in practice. I think that any time you attempt to represent a natural sound with a mathematical construct you will wind up with this same problem (ie a "deficiency") in regard to time evolution. The reason is pretty simple: the natural tone generator is responding to a causal stimulus that cannot be accurately (or sometimes even approximated) with the mathematical construct. So unless you can duplicate the all the factors influencing the natural tone generator with a numerical representation, the facsimile will not be accurate. Obviously its going to take alot more than a 4 or 5 step EG to do this. I think that physical modeling is an attempt to replicate those influences more accurately. I don't know much about it though. I heard that the Yamaha VL's weren't very programmable.... Just playing the devils advocate here but also if timbre is a primary quality then why can't it be measured like pitch and amplitude? You can build a device to measure these qualities, but, so far no one has done it to measure timbre. To me, the real question is, why doesn't "timbre" rhyme with "December"?:) Thanks for the good discussion! jon jonwest@vcu.org jon ------------------------------------------------------------------------------------------ Re: Re: Re: Re: Re: Re: What is Timbre? Thursday, 14-Jan-99 14:17:22 192.28.2.49 writes: >I think that physical modeling is an attempt to replicate those influences more >accurately. I don't know much about it though. I heard that the Yamaha VL's >weren't very programmable.... I think Yamaha gets out the stats for the DX7 every time they design a new synth, and they conclude that 99% of the public wants good presets and no programing. They may be right, but they lose me! IMHO, PM and sampling are best at tackling the "objective" end, i.e. what's going on in the instrument or what's going on in the air (respectively). So far, additive is the closest to tackling the "subjective" end, i.e. what's going on in your ear, since your ear performs something like a Fourier analysis. --Hey! Someone should do a PM of the ear! You would input what neurons should fire and when, and it would calculate backwards to the sound that caused that reaction. Direct brain music. >Just playing the devils advocate here but also if timbre is a primary quality >then why can't it be measured like pitch and amplitude? You can build a device >to measure these qualities, but, so far no one has done it to measure timbre. It can't be represented by a single number like pitch and amplitude. There are people doing work for robotic vision that I think is analogous, trying to develop rules that computers can apply to distinguish objects from their background. There can be more than one right answer, though. >To me, the real question is, why doesn't "timbre" rhyme with "December"?:) That's how I say it; maybe we could ask Bret Favre . . . >Thanks for the good discussion! Thanks for the response! I'm glad someone else is into it. leiter ------------------------------------------------------------------------------------------ Re: Re: Re: Re: Re: Re: Re: What is Timbre? Thursday, 14-Jan-99 15:29:48 140.254.113.177 writes: I appreciate this discussion too :-) Depending on your definition of noise, there is another way to strip noise from a frequency spectrum. If noise is a stochastic process that is not directly part of the phenomenon of sound production, you can record a sound source with two different microphones as two independent frequency spectra. Then you can cross-correlate the two spectra in the frequency domain such that only similar parts of the spectrum are emphasized/kept. You can then subtract this spectrum from the original to get the noise component which you can attempt to resynthesize. This can perhaps, in effect, give you the formants of an instrument. Especially if different frequency notes are cross-correlated. Hmmm... I'd like to try that. Then again, if you're defining noise as a constant offset in frequency-space (a horizontal line with positive intercept in the case of white noise, or a line segment in the case of band-limited white noise), then you don't really need to be this sophisticated. You can just try to subtract any offsets out of the frequency spectrum. At least that's what springs to mind. There may be some errors in my thought process, but this is not my area of speciality. Hopefully it's close enough. Bob ------------------------------------------------------------------------------------------ Re: Re: Re: Re: Re: Re: Re: Re: What is Timbre? Friday, 15-Jan-99 02:17:44 209.160.126.93 writes: Maybe we should ask Sony what timbre is. I was just reading an article on the net about the ATRAC format they use for their minidiscs. According to the article, %80 percent of the sound recorded is rejected as being unimportant for the human ear. This is done so that the recording takes 1/5 as much memory to store. I haven't listened to a Minidisc recorder, so I can't say how succesful this approach is, but if it works, then it would seem that Sony has a pretty good handle on what is important from a psychoacoustic perspective for humans to hear. Here's the article if you would like to read this yourself: http://209.224.145.9/misc/md.htm Leslie ------------------------------------------------------------------------------------------ Re: Re: Re: Re: Re: Re: Re: Re: Re: What is Timbre? Friday, 15-Jan-99 10:30:08 192.28.2.49 writes: I think Sony underestimated how much information is unimportant, because I've read a lot of comments in newsgroups about how ATRAC is audibly lossy. Apparently they've improved it, but the difference is still audible according to what I've read. leiter ------------------------------------------------------------------------------------------ Re: Re: Re: Re: Re: Re: Re: Re: Re: Re: What is Timbre? Friday, 15-Jan-99 10:40:50 153.35.140.89 writes: I heard the SAME thing about MD! To the point where it was strongly advised that the format not be used for any sort of professional recording. Still wouldn't mind seeing Sony's reports on the subject.... jon jonwest@i-c.net ------------------------------------------------------------------------------------------ ------------------------------------------------------------------------------------------