The Basic Reed Woodwind


Tutorial home




Reed woodwinds include saxophones, clarinets, oboes, and bassoons.  Other reed instruments include harmonicas, accordions, and some organ pipes.  These pages will concentrate on monophonic instruments, like saxophones.


Our discussion will assume some familiarity with the blown pipe model.  In fact, we’ll borrow some pieces of it.


Reed woodwinds can have two kinds of reeds:  a single reed, as used in a clarinet or saxophone, or a double reed, as used in an oboe or bassoon.  Since I can’t find a double-reed model, we’ll model a single reed, and then use EQ to attempt to mimic oboes and bassoons.


Reed woodwinds can have two kinds of bores:  a cylindrical bore, as used in a clarinet, or a conical bore, as used in saxophones, oboes, and bassoons.  Our bore will be able to simulate both kinds of bores.



The basic model


The basic model is below.  It doesn’t sound like much yet, but don’t be discouraged.  There are many improvements to be made.





How does it work?


The operation is similar to our blown pipe.  Once again, the combination of the jet driver and a tuned delay is the basis of the model.  In fact, the jet driver is identical to the one in the blown pipe model.


It’s the delay line that’s different.  In fact, there are two delays, one each in the Left Side Pipe and Right Side Pipe.


Let’s trace the signal flow.  We’ll begin with the Left Side Pipe.  This section contains a delay line, a lowpass filter, and a gain control.  The output then goes to the Right Side Pipe, and is inverted as it enters.  The Right Side Pipe is identical.  It contains another delay line, a lowpass filter, and a gain control.  The output then returns to the Left Side Pipe, and is again inverted as it enters.  Notice that the lowpass filters and gain controls have matching characteristics, but that the delay lines can have different lengths.


It turns out that this loop is, in fact, the standard physical model of a string.  One delay represents the length of string from the pick to the bridge, and the other delay represents the length of string from the pick to the nut.  The lowpass filters and inverting gain controls represent the reflections and energy losses at the bridge and nut.


But instead of “plucking” the string by injecting a noise pulse into it, we’ve placed a jet driver at the pick point.  We’ve summed the outputs of the delays, divided them in half, and then added them to the input air pressure.  This feeds the jet driver.  The output of the jet driver goes back into both delay lines.  This kind of model is called a “blown string”.  It’s an attempt to model a conical bore in an inexpensive way.




Cylindrical and conical bores


It’s the cylindrical bore of a clarinet that gives it that “woody” sound.  This is because the straight bore supports only odd harmonics (more or less).  The conical bores of saxophones, oboes, and bassoons are the reason that their sounds include even harmonics.


There are two common ways to model conical bores.  The “physical” way is to create something called a Kelly-Lochbaum filter, containing dozens of tiny delay lines, adders, and multipliers.  Using this method, any shape of bore can be created.


We don’t have a Kelly-Lochbaum module in the G2, so we’ll use the “non-physical” way: the blown string.  It can’t do everything a Kelly-Lochbaum filter can do, but it’s enough to create even harmonics, and we can do the rest with EQ.


We’ll control the “shape” of our bore by controlling the ratio of the delay line lengths, via the Pan knob on the “Splitter” Pan module.  If the knob is in the center, the two lengths are equal, and our pipe sounds woody, like a clarinet.  But if one length is longer than the other, the sound begins to include even harmonics.  Varying this control is a lot like varying the pulse width of a pulse-wave oscillator.  Regardless of the position of the Pan knob, the total length of the loop remains the same, and this is why the pitch doesn’t change when the Pan knob is moved.


The delays are tuned using an old trick from the original Nord Modular: the Level Scaler Module.  This module can be linked to the keyboard and adjusted to output a -6dB/octave slope.  Since -6dB represents a 50% reduction, and since we want the delay line to become 50% shorter with each increasing octave, this module is perfect for the task.




The control parameters


This model has 5 control parameters.  They are:


  • Pitch.  Pitch should be directed to the Note input of the Level Scaler.  In this patch, the pitch comes directly from the keyboard.


  • Pipe Shape.  The shape of the pipe is controlled by the position of the Pan module labeled Splitter.  Pipe shape can be modulated via the modulation input on the Splitter.


  • Air Pressure.  This drives the model, and represents the air coming from the player’s lungs.  Air pressure should be directed to the Chain input of the Pipe Mixer module.  This is typically provided by an envelope generator, or a breath controller.


  • Bell Frequency.  This is the cutoff frequency of the lowpass filters in the loop.  It represents frequency-dependent loss.  On the Yamaha VL1, this is called “Absorption”.  Lowering the cutoff frequency makes the sound duller.  It also lowers the pitch a little.  Because of this pitch-lowering effect, changing the Bell Frequency usually requires retuning the model.


  • Bell Gain.  This is the amount of loss in the loop, irrespective of frequency.  On the Yamaha VL1, this is called “Damping”.  Lowering the gain makes the sound quieter.  It does not affect the pitch.