Deconstructing a natural sound involves analyzing the spectrum of the sound in order to recreate it using the technique of ADDITIVE (Fourier) SYNTHESIS. The sound you will create in this tutorial is the sound of a flute-like instrument called a RECORDER. Click below to hear a note played on a recorder:

RecorderSound.aiff

After analyzing the spectrum of this sound and using that information to reconstruct it using additive synthesis techniques, you will create this sound:

SynthesizedRecorder.aiff

You will use two applications in order to deconstruct and then synthesize the recorder sound: Amadeus II and JSYD.

If you need help with JSYD, see the SYD Tutorials.

This tutorial will be divided into the following sections:

• Opening the Recorder AIFF sound file in Amadeus II
• Analyzing the spectrum of the Recorder sound file
• Converting the Decibel (dB) levels of the various frequencies to Amplitude values
• Creating the basic SYD patch
• Adding breath noise (to simulate the sound of the breath blowing through the Recorder mouthpiece)

OPENING THE RECORDER AIFF SOUND FILE IN AMADEUS II

Begin by downloading the recorder sound file:

• Hold down the OPTION key on your Macintosh computer keyboard while you click on RecorderSound.aiff
• Save the sound file to your Local Workspace.

Launch the application, Amadeus II.

From Amadeus II, open the sound file you just saved, RecorderSound.aiff

You should see the following:

Figure 1: Recorder Sound File as viewed from Amadeus II

ANALYZING THE SPECTRUM OF THE RECORDER SOUND FILE

Select a region of the sound file for which you will examine the frequency spectrum. Try to select a region which appears to be the MOST STABLE (in terms of amplitude fluctuation).

Figure 2: Recorder Sound File with region selected as viewed from Amadeus II

From the 'Analyze' pull-down menu in Amadeus II, select 'Spectrum':

Figure 3: Select 'Spectrum' in the Analyze pull-down menu in Amadeus II

Choose these options and then lick OK. It is best to select AVERAGE, which will return the MEAN frequencies and amplitudes.

You should see this graphic which is a representation of the individual frequencies (harmonics) along with their respective amplitudes. Here is the spectrum of the note, A4 played on a recorder:

Figure 4: Sound spectrum of the note A4 played on a recorder as viewed from Amadeus II

Place the corsair (corsairs) exactly over the leftmost frequency in the spectrum:

Figure 5: Frequency and Decibel (dB) levels of the note A4 played on a recorder as viewed from Amadeus II

AmadeusII shows the amplitude values in both DECIBELS and 100% amplitude values. Since SYD only utilizes 100% amplitude values, you will have to approximate those values as closely as possible.

In a convenient location, write down the values of the frequency and the amplitude values. Repeat this for each of the other two frequencies in the recorder spectrum. You should end up with these approximate values:

Frequency	Amplitude
905.34	1.0 (100%)
1798	.42 (42%)
2702	.11 (11%)

Table 1: Frequency and Amplitude values

for the note A4 played on a recorder

as viewed from Amadeus II

CREATING THE BASIC SYD PATCH

To synthesize the recorder sound using the frequency and amplitude values from the above table, launch JSYD. See the JSYD Tutorials if you need to review this application.

Create the following patch with a output duration of 4.0 seconds:

Figure 6: Basic Recorder Patch using JSYD

You could put an envelope on each of the three oscillators. But, because one of the amplitudes from the decibel (dB) to amplitude conversion table [Table 2. above] is > 1.0, it is best to patch the outputs from the oscillators through the mixer. This attenuates the amplitudes proportionally such that the overall amplitude coming out of the mixer is <= 1.0. Consequently, it is more convenient to put the envelope on the amplifier.

The recorder is a 'woodwind' instrument (wind instrument) which requires wind pressure from the breath to produce the sound. The basic woodwind envelope [which would also include such instruments as flute, clarinet, oboe, bassoon, or any reed instrument which requires the human breath to produce the sound] looks like this:

Figure 7: Simple 'woodwind' Envelope

Here are the specific parameter settings for the woodwind envelope for the above patch:

Figure 8: Parameter settings for 'woodwind' Envelope using JSYD

When you synthesize the patch, it should sound like this:

BasicRecorderPatch.AIFF

ADDING BREATH NOISE TO THE BASIC JSYD PATCH

Adding breath noise to the basic recorder patch involves using Butterworth filter set to BANDPASS mode. You will add three bands of noise roughly matching the three frequencies derived from the spectrum analysis above.

Here is the next stage of the Recorder Patch:

Figure 9: Basic Recorder Patch with Noise added using Butterworth Filters using JSYD

Look for the three Butterworth filters which have their output to a mixer and then to an amplifier, just like the Basic Recorder patch described above. In addition, the same envelope controls the amplifier for these additions.

In order to more easily control the noise level, an EXPRESSION operator has been incorporated (you used this operator in your COMPLEX BELL to easily control the frequency. Here you will use it to control the level of the noise). Place the EXPRESSION operator as shown above and make connections to each Butterworth filter. For the type of connection, choose the variable, CTL (for "control"). When you choose this type of connection, the line turns from blue to pink. Double-click the Expression operator and enter 150 as it's value:

Figure 10: Parameter settings for the Expression operator

The parameter field in the Expression operator will contain the basic BANDWIDTH for each filter and will pass that value to the Butterworth filter using the "ctl" variable. Here is the format for using the Expression value in the BUTTERWORTH filter:

Figure 11: Parameter settings for the Butterworth Filter using the CTL variable from the Expression operator

Each BUTTERWORTH filter will have similar settings where the value in the 'Freq' field corresponds to the value of the frequency from the spectral analysis above. The CTL variable is used n the BAND field. Here is a table which shows these values:

When you synthesize the patch, it should sound like this:

BasicRecorderWithNoise.AIFF

Compare with the Basic Recorder Patch:

BasicRecorderPatch.AIFF

To increase or decrease the amount of noise you would like in your Recorder sound, change the value in the Expression operator. A higher value (eg., 300) will give you a lot more noise; a lower value (eg., 100) will give you less noise.

Make your Recorder patch more "complex" by adding additional frequencies. Revisit the SPECTRAL ANALYSIS you did with AmadeusII to discover the additional frequencies.

Other qualities you might add:

1. Do a 3-D spectral analysis and notice how the amplitudes of individual frequencies differ from each other over time.

2. The frequency of the ATTACK and the DECAY vary because of the change in breath pressure at those points. Less pressure creates a lower frequency than a higher pressure. Consequently, the frequency of the ATTACK becomes gradually higher and the frequency of the DECAY becomes lower.

Try analyzing and then synthesizing the following musical instruments:

clarinet

flute

cello

[To download the sounds, hold the OPTION key down and then click on the link.]

Main Page

Back to the SYD Tutorials Index