Physical Modelling - Sitar

[JoeDeF]

heck i'm postin' (w/o thinking about it..)

i'm sure you've sussed this, to solidify it for you i'll parrot, bidirectional is only necessary if you've got anything happening at the terminii, otherwise you might as well treat it as a single delay loop.

throwing primitive, reactional thinking at it, the jawari (thanks for link!) could at least be modeled as a FIR, at a fixed sample rate as the jawari is of a fixed length??

I am not a dsp instrument builder (I'm more of a physical instrument builder), but the jawari is definitely something "happening at the terminii," (really at one terminus). Since it is a ramp, I believe what happens is this: larger amplitude pulses (transverse waves) encounter it earlier than smaller amplitude pulses. The larger amplitude pulses are not terminated by it (they don't reflect at that point); rather, they are altered by it as they approach the tippy-top portion of it (highly scientific lingo there ), which is the actual termination. By altered, I mean that they buzz against it. I have observed this kind of sitar-like sound in other, malfunctioning stringed instruments including pianos and guitars when the (malformed) termination is a ramp instead of a clear point. In fact, you can buy a "sitar bridge" for your electric guitar that works along the same principle.

Again, not being a dsp guy, that's the best I can describe it in the physical world, and I don't know how to model it in the virtual world. I think Dave might already have a handle on it.

Speaking of the link, did you go to the link within the link?

http://www.buckinghammusic.com/sitar/sittut/jaw.html

Cool stuff....

[MisterEko]

heck i'm postin' (w/o thinking about it..)

i'm sure you've sussed this, to solidify it for you i'll parrot, bidirectional is only necessary if you've got anything happening at the terminii, otherwise you might as well treat it as a single delay loop.

throwing primitive, reactional thinking at it, the jawari (thanks for link!) could at least be modeled as a FIR, at a fixed sample rate as the jawari is of a fixed length??

What I have at the moment is that each strings length is dynamically changing with respect to single delay line loop. How much energy that is in the delay line is dependent on the velocity of say the midi note. Also the velocity is mapped to how much the length of the string will initially change by. I have have also fed a small bit of energy from the main string back into single delay line loop. If anyone wants to look at my max patch so far I can email it to them.

Cheers,

Dave

[xoxos]

thanks for that - this is a project i can justify development for

i don't know how much time you've had to go through pasp - if you want to add 'dimensionality' to it, add orthagonal coupling - vertical vibration and horizontal vibration. my implementation was simple.. add so much of the other and subtract the same percentage, sounded alright to me i assign the horizontal wg a higher 'feedback' coefficient.

i use extra, non-regenerating delay lines to feed the impulse to each termini for plucking position (or, if your string is simplified to one delay line, the two delay lines stand for pick->bridge, and pick->nut->bridge, which can simplify to one delay line and one direct signal, the delay line being the longer length minus the shorter length..).

do you have tuning information for strings? there's a sitarist who sometimes posts over at dancetech.com who may be able to provide some insight

[MisterEko]

thanks for that - this is a project i can justify development for

i don't know how much time you've had to go through pasp - if you want to add 'dimensionality' to it, add orthagonal coupling - vertical vibration and horizontal vibration. my implementation was simple.. add so much of the other and subtract the same percentage, sounded alright to me i assign the horizontal wg a higher 'feedback' coefficient.

i use extra, non-regenerating delay lines to feed the impulse to each termini for plucking position (or, if your string is simplified to one delay line, the two delay lines stand for pick->bridge, and pick->nut->bridge, which can simplify to one delay line and one direct signal, the delay line being the longer length minus the shorter length..).

do you have tuning information for strings? there's a sitarist who sometimes posts over at dancetech.com who may be able to provide some insight

Nice one, I'll look into that so. I got the tuning information for the sympathetic strings strings off of wikipedia for the moment. Any chance you'd look at my patch and let me know what you think of it?

[xoxos]

i don't have any of the MAX versions :/ do you have audio?

atm i'm going thru my 'extended karplus strong' for improvements.. then hopefully i'll be able to add a jawara effect.

my previous e.k.s. (available as pling3) used -

linear interpolation w/ 2x oversampling (academically you can improve on this )
-an extra delay line to model plucking position
-damping (simple 1st order lpf eg. w * [x] + (1-w) * [y-1])
-allpass filtering with phase correction for pitch
-orthagonal coupling (linear crossfading)

all straightforward and gleaned mainly from julius smith with my math poor comprehension. pling3 also inherited a 'touch harmonics' function from v.2 that was useless bcs i didn't run the horizontal waveguide past the touch position

fwiw to you, on reevaluation i'm adding the following calculations: (and removing the touch harmonic crap bcs it requires 2x the cpu to work, having 2x as many 'string' sections)

-computing the phase response of pitch for the lowpass filter and adjusting the delay length accordingly (-arctan(cutoff / signal))
-using pow(2,octave) to adjust the gain across the scale so notes have similar decay length
-hopefully finding a way to get the same allpass response across the scale (no clue atm.. math!)

without being conceited, if you're grasping half these concepts, you're over par with most vst developers and most university thesis writers (i've been through lots of them..) from what i can tell, most thesis evaluators are looking more for your approach than your results. sure, there are designs that are far beyond me (eg. valimaki and smith) but most don't go anywhere near as far as i, as a gross amateur have.. so calibrate accordingly.

i've used other functions elsewhere that you may want to consider for detail, tho as said i'll guess your grade is based on analytical procedure more than result.. smith has laid it all out if you can understand him.. using a mass-spring (or other system) for bridge resonances.. using a mass-spring to modulate the length of the string to model rigidity, or lack thereof..

if you want to get twee and have loads of computational power for oversampling, you could divide the waveguide where the string contacts the jawara.. these few cm in steel are way under 1 sample @ 44.1k..

[Mumbai]

If oyu need any more info about real Sitars, I play that thing and can help - PM. Greetings, Ingo

[MisterEko]

i don't have any of the MAX versions :/ do you have audio?

atm i'm going thru my 'extended karplus strong' for improvements.. then hopefully i'll be able to add a jawara effect.

my previous e.k.s. (available as pling3) used -

linear interpolation w/ 2x oversampling (academically you can improve on this )
-an extra delay line to model plucking position
-damping (simple 1st order lpf eg. w * [x] + (1-w) * [y-1])
-allpass filtering with phase correction for pitch
-orthagonal coupling (linear crossfading)

all straightforward and gleaned mainly from julius smith with my math poor comprehension. pling3 also inherited a 'touch harmonics' function from v.2 that was useless bcs i didn't run the horizontal waveguide past the touch position

fwiw to you, on reevaluation i'm adding the following calculations: (and removing the touch harmonic crap bcs it requires 2x the cpu to work, having 2x as many 'string' sections)

-computing the phase response of pitch for the lowpass filter and adjusting the delay length accordingly (-arctan(cutoff / signal))
-using pow(2,octave) to adjust the gain across the scale so notes have similar decay length
-hopefully finding a way to get the same allpass response across the scale (no clue atm.. math!)

without being conceited, if you're grasping half these concepts, you're over par with most vst developers and most university thesis writers (i've been through lots of them..) from what i can tell, most thesis evaluators are looking more for your approach than your results. sure, there are designs that are far beyond me (eg. valimaki and smith) but most don't go anywhere near as far as i, as a gross amateur have.. so calibrate accordingly.

i've used other functions elsewhere that you may want to consider for detail, tho as said i'll guess your grade is based on analytical procedure more than result.. smith has laid it all out if you can understand him.. using a mass-spring (or other system) for bridge resonances.. using a mass-spring to modulate the length of the string to model rigidity, or lack thereof..

if you want to get twee and have loads of computational power for oversampling, you could divide the waveguide where the string contacts the jawara.. these few cm in steel are way under 1 sample @ 44.1k..

Sounds very interesting! I don't have audio yet but I'll send it on to you as soon as I am finished. Under serious pressure now to get it finished. The practical and written part are due in on the 16th of this month. Thanks for all your help anyway. Also, what is pling3?

Cheers,

Dave

[MisterEko]

If oyu need any more info about real Sitars, I play that thing and can help - PM. Greetings, Ingo

I'm pretty good on the mechanics and tuning side of things. Thanks for your help anyway!

[xoxos]

ingo - i'll probably check in with you at some point probably for temperament..

pling3 is an ongoing vst pl(ucked str)ing model.. it's used in such wonders as..
http://www.xoxos.net/music/unicornvalley.mp3
http://www.xoxos.net/music/theartofnoise.mp3
http://www.xoxos.net/music/payola.mp3 (properly videoed here )
http://www.xoxos.net/music/3notes.mp3

[taotekid]

ingo - i'll probably check in with you at some point probably for temperament..

pling3 is an ongoing vst pl(ucked str)ing model.. it's used in such wonders as..
http://www.xoxos.net/music/unicornvalley.mp3
http://www.xoxos.net/music/theartofnoise.mp3
http://www.xoxos.net/music/payola.mp3 (properly videoed here )
http://www.xoxos.net/music/3notes.mp3

Pling3 sounds wonderful

I'm looking forward to the day when you'll deliver it to the world

which of your instruments did you use for the flute in Unicorn Valley ?
it sounds amazingly lovely

your tracks are awesome

[xoxos]

pling3 is old front page, down the bottom (many of these aren't listed on the 'vst' page). i think the flute is a vst called 'reed5c' from a while back. it doesn't tune well and hasn't got much flexibility at all to it ../temp/reed5c.dll but don't tell - there's *got* to be somethnig better than that.. *the percs in u.v. are mass vst and stoq vst.. the horse is fauna vst*

u.v. is based on an old track i wrote for dancetech.com in the '90's
http://www.xoxos.net/temp/email.mp3 (warning: not for polite company)

[taotekid]

pling3 is old front page, down the bottom (many of these aren't listed on the 'vst' page). i think the flute is a vst called 'reed5c' from a while back. it doesn't tune well and hasn't got much flexibility at all to it ../temp/reed5c.dll but don't tell - there's *got* to be somethnig better than that.. *the percs in u.v. are mass vst and stoq vst.. the horse is fauna vst*

u.v. is based on an old track i wrote for dancetech.com in the '90's
http://www.xoxos.net/temp/email.mp3 (warning: not for polite company)

silly me ! of course, I've already got Pling3 & it IS wonderful,
(it shows up as Pling3b in my folder)
for some reason (maybe some amp model that you used in your tracks)
I thought that the current version that I owned was Pling2 ...

I've just tried it in amp sims with effects for the 1st time
& it sounds divine especially the Lawn Furniture patch

I'll try Reed5c now, thanks Rurik

(your Email song sounds priceless but it only lasts 27sec here,
maybe the upload was interrupted, where's the rest of the song ? )

[xoxos]

i've just roughed my jawara..

this has got a ways to go.. maybe nowhere.. it's unstable because the modulation affects the gain, i may not be able to suss it (prolly will this isn't oversampled..)

hows that??

http://www.xoxos.net/temp/sitar.wav

[MisterEko]

i've just roughed my jawara..

this has got a ways to go.. maybe nowhere.. it's unstable because the modulation affects the gain, i may not be able to suss it (prolly will this isn't oversampled..)

hows that??

http://www.xoxos.net/temp/sitar.wav

That sounds really good man. It sounds like mine except I've got a small bit of beating between the strings and I've got all the sympathetic strings as well. Definitely sounds like your on to a winner though.

Question for you though. It's with regard to orthogonal string coupling. I'm not exactly too sure how I go about this with a bi-directional waveguide. How I think it works is that say you have two bi-waveguides and then they both terminate at the same point and you basically sum there output into one or is it different? Maybe you have to feedback one into the other?

[xoxos]

i was using linear crossfading. that worked (or, more correctly, was moot) in the last situation due to differences in the structure of each axis. so i tried panning law. eventually i figured out what i wanted to use