How do you do physical modelling on the G2?

[Antimon]

I thought it would be fun to try out a bit of physical modelling DIY on the G2 instead of just using the string module. After what I've read about the delay model resolution and bitrate, the G2 may not be the perfect tool for this, but I might get some interesting results anyway along the way.

I googled up a couple of articles and made this string thing (attached). There are two things that I hope that someone here might be able to help me with:

Noise

When you start up the patch and play a bit, after a while some kind of noise (or feedback tone or maybe DC?) builds up and distorts the sound, after a while it makes it go silent. Any idea about how to prevent this?

Pitch

If you look at the patch you'll see that I use control sequencers to individually tune each of the 32 notes that the patch supports (you can add more notes with more sequencers). It's a bit like tuning a piano. While it can be fun to have a few detuned notes for character, it would be nice to be able to implement a set of modules that mathematically calculated the perfect delay length for each note. I've tried mucking about with various shaper models, but I can't get it right.

Having said that, the first thing I did was amplify the pitch output from the keyboard module and pipe it into the delay lengths, which meant that you played the thing backwards on the keyboard and with pitch increasing too much at the top and too little at the bottom. Playing in that way can lead you to new ideas as well.

/Stefan

---

sbt_physmodel_string.pch2
Description:	Physical modelling experiment	Download
Filename:	sbt_physmodel_string.pch2
Filesize:	3.85 KB
Downloaded:	1163 Time(s)

---

[mosc]

You might find it helpful to check this PM Tutorial buy Chet. http://chet.getchwood.com/G2-Tutorial/Index.htm

Then check out all of his PM patches: http://electro-music.com/g2patches.php?owner=Chet

[Antimon]

Ah! Thanks. No strings in there yet, but I should be able to find lots of info.

About the noise issue, I realised that I should try a highpass filter and it seems to do the trick. That's something that should come without thinking, isn't it? Stuff you can't hear but the meters are still rising -> DC -> highpass filter. I need to write that on a note on my fridge or something.

/Stefan

[Robby]

IMO a DC offset indicates a flaw in a physical model.
Physical objects that resonate don't have DC offsets.

For example, a string that is plucked will bounce all the way to the
opposite side it was plucked away from - it doesn't just return to the
center line (0), it keeps bouncing to both sides of the center even
though it was plucked in only one direction.

Think of a pendulum - you pull it to one side only, but it will swing
all the way to the opposite side before coming back. It doesn't just
fall down to dead center then bounce back the way it came.

However, a simple damped delay string model will do exactly that.

Physical systems bipolarize unipolar input.
A DC offset indicates that the physical model doesn't have this property.

[Kassen]

bouncing tennis ball.

:-p

[Robby]

Does a bouncing tennis ball resonate?

[Blue Hell]

Yes it does.

[Robby]

OK, some resonant physical systems might have DC offsets.

A plucked string definitely woudn't,
neither would the classic mass-spring example.

If you're doing a "tennis ball off the floor" model on the G2,
you might want to consider placing the floor at -1 instead of 0
because you're already losing 8 bits of resolution when
you go through the delay memory and losing the sign bit also
is going to take you down to 15 bits.

[Blue Hell]

I'm not to sure about the details, but I remember a remark made by Chet Singer regarding his woodwind simulations. The way he modelled the air stream seemed to bring a DC offset with it as well, intinsically. Anyway, if that memory is correct it should be somewhere in Chet's tutorials (linked above by mosc).

You're right of course that a DC offset in the audio circuits of a patch could very well mean trouble