posted 05 April 2002 15:39            

mbmodulator.kym

I decided to try my hand at synthesizing a steel drum, since the only samples of it around here are rather thin. I did a spectrum analysis on the drum samples and found that the dominant harmonics are 1, 2, 4, 5.3, and 8. These harmonics come in at slightly different delays. The fundamental appears first, and then about 11 ms later you get harmonics 2 and 5.3. At 19 ms you get 4, and then at 30 ms you get harmonic 8.

I also wanted to settle, once and for all, how to get Kyma to obey a keyboard "trigger". Being percussive sounds, I need the attack to last the full 10 ms and then go into a decay of around 0.5 sec. This envelope needs to be triggered, as opposed to gated, by the keyboard. So the way I worked it out is to use the keyboard to trigger a FunctionGenerator running through a positive ramp over 0.5 sec. This is fed into a threshold detector to get a 1 or 0 output.

The attack is assured by feeding this thresholded binary value into a subtractor with 10 ms delayed version of it. That produces a 1 for 10 ms followed by a zero. Since the FunctionGenerator actually treats the trigger source (keyboard) as a trigger, it can't retrigger until it runs its course over the 0.5 sec ramp. That prevents key-up from cancelling the decaying envelopes on the oscillators.

[You could feed the !KeyDown trigger into this subtracted delay, and that works fine for the attack being stretched to 10 ms and then killed so the envelope goes into decay. But keyup events cancel that decay, which sounds all wrong. So the FunctionGenerator ramp into a threshold solved that keyup problem as well.]

So now, we have a polyphonic keyboard operated percussive sound synthesis. If the rest of you are as troubled as I have been about getting these Kyma envelopes to obey the keyboard then this is a solution. No doubt SSC has even better ones. But this is one area that I have been stubbing my toes on for some time.

BTW, the steel drum sounds reasonable, over the range C2-C4. Try adjusting the relative levels of the harmonics. I put the 5.3 harmonic at -9 dB, and all the others at 0 dB. This is all very approximate.

- DM

BTW: This additive synthesis with delays is something I can only imagine doing on Kyma. Those delays apparently help the sound to be a bit more realistic, and prevents the synthesis from sounding too much like an "organ". What other additive synths allow you to delay the partials? Maybe there is one, but I haven't seen such a thing...

[Well, I should correct this statement... I just managed to make this same Sound happen in Reaktor. Tassman also presents some interesting possibilities, but from a completely different angle.

In Tassman you have to specify the physical dimensions of plates and tubes, as well as the stiffness and strength of the mallet. Without a solid physical understanding of these items, it can take a long time to narrow in on the correct sound -- even if you do understand vibrating plates, it can still take a while... The dimensionality of the parameter space is just so large. It is a lot easier to synthesize this sound directly from an examination of the spectrum.]

posted 07 April 2002 00:27            

I have to go back to Morse and review the vibrating plate, but I seem to recall that a metal bar has its first overtones around 6.4 times the fundamental. This metal drum would be more of a plate and so perhaps that explains its lower anharmonicity? I have always been fascinated by these metal drums, ever since seeing and hearing them in the Carribean. A single surface deformed into multple platelettes, spanning an entire scale or more of tones, but with distinctly mistuned intervals. They are almost always played rapidly in alternating platelette pairs producing a triad interval (did I say that correctly?)

I was sitting out back a while ago musing about the delayed onset of the higher partials. These delay times are far too long to represent acoustic wave propagation and reflection phenomena. 1 ms = 1 foot. So 11 to 30 ms would be 5 to 15 feet there and back.

So I began to wonder about the mallet used and whether or not the way they beat and or brush against the metal top has something to do with these delays. While the mallet is in contact with the metal, it can certainly damp some vibration modes in the plate. But perhaps by beating with a swiping action this can produce the higher harmonics at short delays after the initial strike?