Ok, tougher one. Piano modeling!

[FarleyCZ]

Hi all.

Few monts ago I started here topic about modeling plucked guitar using Zebra's comb filter. http://www.kvraudio.com/forum/viewtopic ... highlight=
I like the result. http://dl.dropbox.com/u/19028571/guitar_model_prew2.mp3 (thanks Howard ) It's not extremely acurate, but it's cool startpoint for any "in project sounddesign".
Now here comes another challenge.
Is any of you guys capable of thinking way how to model a hammer? I mean behaviour of string is achivable by that comb filter, but that impulses I'm sending in are still creating plucked sound. I'd need a hammer for Piano obviously.
I know there was a piano preset on KVR some time ago, but it used classical synthesis. I'm more interested in modeling it by Karplus-Strong-ish method by comb filter.

It's damn easy in Zebra to do dozens of bell sounds. Even with no FM. But if you run them by comb, it still doesn't sounds like a piano. May be higher notes are a bit similar, but overall result is not good at all. Clue must be in something else.

So here is first (really bad) atempt. http://dl.dropbox.com/u/19028571/piano_atempt_1.mp3
It's square OSC for bottom and bellish "just a few harmonics" osc for top. All routed to comb. And clearly bottom and mid notes are really bad. I did from this preset few cool hawaii malet drums. But no piano. There is absolutely no definition of the hammer or piano behaviour what so ever.

Any ideas?

[bmrzycki]

Is any of you guys capable of thinking way how to model a hammer? I mean behaviour of string is achivable by that comb filter, but that impulses I'm sending in are still creating plucked sound. I'd need a hammer for Piano obviously.

Howard has a preset in the factory bank:
Factory 2.5 -> 4 Pianoids -> HS Harpsichord

In lane 1 the comb module is modelling the string sound while in lane 2 the FMO has a very short sound (ENV2 closes FMO's volume) emulating the initial pluck sound. While not exactly a hammer I'd imagine the same method can be used to get a good piano hammer sound and let the comb handle the modelling of the string.

Hope that helps.

[FarleyCZ]

While not exactly a hammer I'd imagine the same method can be used to get a good piano hammer sound and let the comb handle the modelling of the string.

Hope that helps.

Thx, gonna check it out.
Edit: Nope, still not much progress. It creates similar noise/wave combination I used in that guitar model.
There must be some other aspects we're missing. :-/
I tried to study some materials about piano synthesis on internet, but I got lost in their mathematical descriptions. :-/

[FarleyCZ]

Ok. I loaded up piano multisample and played C0 on it. Looked on the spectrogram expecting information about harmonical content. There was something about 64 Hz as lowest sine, which is C1. But, next very dominanet sine in begining of tone was something about 97 Hz which is G! Wel I didn't expect that.
Let's see if that information is any helpful.
Edit: Nope. Even E and A frequencies discovered at the bump. Interesting that no metter of velocity, freqs are there. It's acoustic pianos from IK and I know for sure it has velocity layers, so if that frequencies was just random noise, it would change in other sample, but it's still here. Hmmmm. I'm trying to "rewrite it" into spectrablend oscilators. Will see what will come out after throwing it to comb.

Edit2: Nope, no progress. :-/ But still interesting. I'd never expect any onther harmonics then the ones from fundamental in the sound of such a clear instrumens like a Piano.

[benj3737]

Piano modeling could be a cool patch contest!

[Jafo]

Ok. I loaded up piano multisample and played C0 on it. Looked on the spectrogram expecting information about harmonical content. There was something about 64 Hz as lowest sine, which is C1. But, next very dominanet sine in begining of tone was something about 97 Hz which is G! Wel I didn't expect that.
Let's see if that information is any helpful.
Edit: Nope. Even E and A frequencies discovered at the bump. Interesting that no metter of velocity, freqs are there. It's acoustic pianos from IK and I know for sure it has velocity layers, so if that frequencies was just random noise, it would change in other sample, but it's still here. Hmmmm. I'm trying to "rewrite it" into spectrablend oscilators. Will see what will come out after throwing it to comb.

Edit2: Nope, no progress. :-/ But still interesting. I'd never expect any onther harmonics then the ones from fundamental in the sound of such a clear instrumens like a Piano.

I think that maybe you've just discovered that the hammer hitting the strings causes a sound on its own. If every sample has a thump at 97Hz, and every sample has a hammer hitting a string, and all piano hammers are the same (in a given piano, at least), maybe 97Hz is the frequency created by the hitting of the hammer? After all, acoustic instruments are full of such nonharmonic, constant components -- it's part of why certain instruments sound better in certain keys.

Anyway, try filtering it out and seeing if it still sounds like a piano; if it doesn't, it's an essential component of the sound and probably the hammer.

I'll bet that frequency tends to fade quickly, in a percussive manner.

[hakey]

The piano is perhaps considered the most difficult acoustic instrument to accurately model.

SOS's Synth Secrets goes into some detail as to why:

http://www.soundonsound.com/sos/Oct02/a ... rets10.asp

In short:

A piano has a very large fundamental range.

The high and low strings are of different construction.

The hammer position is different for high and low strings.

For some notes the hammer strikes three strings (+ constructive/destructive phase stuff going on).

The action of the hammer is more complex than a simple pluck - the hammer rests on the string for long enough to dampen some of the harmonic modes.

Other strings than those which have been struck will resonate sympathetically when a key is played.

The body resonance of a piano and the interaction of strings with it is complex.


That said, I've heard some more-or-less convincing piano patches, so it's by no means impossible. Though there is a tendency for such patches to sound good only over a limited range.

Piano modeling could be a cool patch contest!

Anybody fancy that for the next ZPC?

[hakey]

Ok. I loaded up piano multisample and played C0 on it. Looked on the spectrogram expecting information about harmonical content. There was something about 64 Hz as lowest sine, which is C1. But, next very dominanet sine in begining of tone was something about 97 Hz which is G! Wel I didn't expect that.

AFAIK if the fundamental is C0 then G1 is close to (2cents off) the third harmonic (C1 being the 2nd harmonic) - and you would expect a noticeable 3rd harmonic from a hammered or plucked string.

[Gino Cortesi]

Anybody fancy that for the next ZPC?

I like it! It's a true challenge, that's for sure!

Best regards,

Gino

[3ee]

Anybody fancy that for the next ZPC?

+1

[FarleyCZ]

Anyway, try filtering it out and seeing if it still sounds like a piano; if it doesn't, it's an essential component of the sound and probably the hammer.

I'll bet that frequency tends to fade quickly, in a percussive manner.

I will try the filtering. It's gonna need a really sharp filter though.

Btw bring the contest on! I'm slowly out of time and ideas, but I can't wait what do you guys come up with!

[mkastrup]

Good luck, getting 1 key right is fairly simple but getting the entire keyboard range is a nightmare. Harmonics changes almost every 6 notes.

This is my best shot so far doing a Zebra piano version:



/Michael

[pdxindy]

That is a good piano for a synth... and yes, it is easy to get a small range sounding good... some patches I have a fine lower octave but as soon as the notes start going they get wrong... or a higher octave but the lows are crap... even so it is fun to play with