Using the Vocoder?

Press the MASTER key
Press the MAST2 soft button to reach the MAST2 page:
The MAST2 page enables you to perform a hard reset of your instrument (this deletes
everything in RAM, so be careful!), and to turn the vocoder feature on and off.
The Vocoder
Vocoding is a special feature that allows you to use an input signal to control another audio
(slave) signal. Typically you would use a synthesizer for the input signal, although in fact you
can use any sound source. You must have the sampling option to be able to use the vocoder.
Using the Vocoder
Go to Setup Mode and select one of the setups in the memory bank where you just loaded the
vocoder -le. If you are using an external sound source for your slave, choose the setup Vocoder-
ExtSlave. If you are using the K2600 as the input source for the slave, then you can choose either
Vocoder-22 Band or Vocoder-20 Band. The 22-band vocoder will allow you to play up to 4
voices of polyphony on the slave program; the 20-band vocoder will allow you to play up to 8
voices of polyphony on the slave program.
Play a note or chord on your keyboard and speak into the microphone. You should be able to
hear what you are speaking, but the sound will be a string sound (assuming you are using the
K2600 as the slave source), pitched to the note or chord you are playing.
If you have a keyboard model, try moving Sliders A, B, and C, and listen for changes in the
sound. If you have a rack-mount model, you can send MIDI Controller numbers 6, 12, & 13 from
your controller. Since the setups contain entry values for these sliders, you may have to move
the slider across its full range before it begins to take effect.
Effects Issues and Output Issues
The studio assigned to the vocoder setups is con-gured in the following manner: If you are
using the K2600 for the slave signal, the slave program (in zone 3) has its output assigned to
KDFX-B, which is being routed to the FXBus2, with no effect. On the OUTPUT page in the Setup
Editor, Output B is set to FXBus2, thereby sending the signal from the slave program to the B
outputs and from B Right into the right side of the sample input.
The slave program has its output panned hard right within the program, so if you decide to try
using a different slave program, you will probably want to edit the program itself to pan its
output hard right, so you get 100% of the signal. You don’t need to worry about setting the output pair within the program, because the Out parameter on the CH/PRG page of the Setup
Editor is set to KDFX-B in zone 3, thereby overriding any settings from within the program.
The vocoder programs themselves are assigned to KDFX-A, which is being routed to FXBus1.
On the OUTPUT page in the Setup Editor, Output A is set to Mix. So the -nal output of the
vocoder programs is run through the effect and then comes out the A Outs and the Mix Outs.
DonÕt use the Mix audio outputs, however, or youÕll hear the slave program along with the
vocoder.
If you choose to change the effects, you may -nd it easier to edit the vocoder studio, and try
changing the effects assigned to FXBus1, FXBus2, and AuxFX. But if you want to change to a
different studio, you will need to make sure the following parameters are set correctly: on the
FXBUS page, for FXBus2, set the Level parameters for both Aux and Mix to Off, and on the
OUTPUT page, set Output B to FXBus2.
How Vocoding Works
A vocoder is a device that analyzes the time-varying audio spectrum of one signal (the master)
and imposes that spectrum as a -lter on a second signal (the slave.) The method we use is an
emulation of the traditional analog technique involving banks of bandpass -lters and envelope
followers.
The master signal is what you send from the microphone, and the slave signal is what you send
from an external synthesizer or other sound source, or a program from the K2600.
The master signal is sent to a number of bandpass -lters in parallel. The center frequencies are
spaced to cover the most useful frequencies. The lowest frequency -lter is a low pass rather than
a bandpass, which groups all low-frequency components together. Likewise, the highest -lter is
a high pass. The outputs of all these bandpass -lters go into individual envelope followers,
which detect the level of signal present in each band. The output of the envelope follower is then
used as a control for the slave signal.
The slave signal is also sent to the same number of bandpass -lters. These generally have the
same center frequencies as the master bandpasses. The output signals from the slave bandpasses
are multiplied, one by one, by the outputs of the envelope followers (from the master signal).
The resulting products are all added together for the -nal output.
Since each band requires two layers (one for master and one for slave), the largest number of
bands you can have for vocoding is 24. (24*2=48, which is your maximum polyphony.) The
programs in the Setup called Vocoder-ExtSlave use 24 bands. If you want to use the K2600 to
generate your slave signal, then you have to use either the 22- or 20-band vocoder setups, which
have fewer bands, and therefore leave 4 or 8 voices of polyphony available for the slave signal
program.
Since 48 (or 44 or 40) layers are used, and a drum program has a maximum of 32 layers, we use
two 24 (or 22 or 20) layer programs, on different MIDI channels, that are combined in a setup.
Each of the setups has 3 zones. In the 22- and 20-band vocoder setups, the -rst two zones are
used for the vocoding programs and the third zone plays the internal program that is used for
the slave signal. In Vocoder-ExtSlave, the third zone is set to transmit via MIDI only, on
Channel 1. (This allows you to play your external sound source, if it is a rack, but won’t play a
K2600 internal program.)
Layers are grouped in pairs, with the master signal going to the -rst layer, and the slave to the
second. All odd numbered layers are master and all even numbered layers are slave. If you look
at the algorithms in the vocoding programs, you will see that the -rst two DSP blocks (after
PITCH) of each layer are a bandpass -lter (or low pass or hi pass -lters for the -rst and last
bands). The -rst layer then has a DSP called MASTER, while the second layer has a DSP called SLAVE. These stages are then followed by an AMP stage. These DSP blocks perform the
function of an envelope follower and gain multiplication.
The signal ßows from the odd numbered (master) layer to its associated even numbered (Slave)
layer (for example, from layer 1 to 2), which is something that does not happen in other
algorithms. The low pass frequencies controlled by the third time slot for each layer set the
response speed of the envelope follower. They are normally set to the same frequency. The
master layer controls the frequency of one pole of low pass -ltering, and the slave layer controls
two more poles.
The AMP page on the master layer does nothing. There is no output from this layer, so any
settings on the OUTPUT page don’t matter. The slave layer’s AMP page does do an actual
amplitude control. The output pages for slave layers are active, and can be used to choose the
output group and set the step panning.
All of the master layers use the LiveIn Left keymap and all of the slave layers use the LiveIn
Right keymap. That is why you must plug the microphone into the left side of the sample input
and the slave source into the right side.
As is always the case with Live mode, a note message is required in order for an incoming signal
to be processed through VAST. Therefore, the two layers in the setup assigned to the vocoding
programs have Pswitch2 set to generate a C4 with a velocity of 127, as soon as the setup is
selected. That note remains on until you select a different setup. The setups are edited so that
none of the notes on an 88 note keyboard are assigned to either of the two vocoding programs
Real-time Control of the Vocoding Programs
The most important control parameter is the envelope follower speed, set by the third time slot
low pass parameters. These are set to C 6 on all the layers for the initial level. Slider A (MIDI 6)
lowers the cutoff up to 8 octaves (9600 cents). Therefore, the higher you raise the slider, the
slower the envelope follower speed. C 6, as a -lter cutoff, has a time constant on the order of one
millisecond. This is generally too fast. For best results, this should be lowered about 4 octaves to
C 2 (half the range of the Data Slider), to a time constant of 16 milliseconds. Too slow and the
vocoder will not respond to quick transients, like consonants, and too fast will result in a jittery
sort of sound, as the envelopes follow every little ßuctuation. At the fastest possible setting, the
envelopes follow the master audio signal itself, and an extremely harsh intermodulation is
heard between master and slave. The vocoder setups have an entry value of 64 for this slider, so
when the setup is selected it is the equivalent of having the slider halfway up.
Slider B (MIDI 12) is used to control the width of the band pass -lters (for all bands except the
lowest and highest). The vocoder setups have an entry value of 10 for this slider, the equivalent
of having the slider at the -rst dot above the bottom.
Slider C (MIDI 13) transposes the center frequencies of all the slave bandpasses upward
together. It gives you the same result as pitch shifting the master signal up. Vocal formants will
be munchkinized as you bring the slider up. The vocoder setups have an entry value of 0 for this
slider, the equivalent of having the slider at the bottom.
Additional Notes and Programming Suggestions
The classic application of a vocoder is to make instrumental sounds talk/sing. The slave signal
has to have a lot of high frequency content, or the consonants will not be heard clearly. However,
there is no rule set in stone that you must speak words into the microphone. Using the vocoder
just as a timbral control can be just as interesting. You can get very expressive results by using
your voice to control a lead line, doing the articulation and -lter control by talking, singing, or
just making various vocal sounds. You can get some of the same types of results you would by
using a breath controller. ItÕs a little like having a 24-band graphic equalizer, but instead of
controlling it with your hands, you use your voice. Furthermore, you don’t even have to use a microphone as the master. You can send a signal from
anything else that has varied timbral content and get interesting results. For example, the master
signal could be a drum loop or some other recorded sound that changes timbres regularly.
The analog sample inputs on the K2500 are line level, not mic level. This means you have to
boost the gain on the sample page to get a good signal. But this also increases the general noise
level of the input signal. If you have a mic preamp, or plug the mic into a mixing board before
sending the signal to the K2600, you can lower the Gain parameter and start with a much
cleaner signal. This is highly recommended.
In addition, you will -nd you get better results if you run the preamped mic signal into a
compressor before sending it to the K2600. This can also be done for the slave signal. Using
compressors will give you a much more even dynamic result, making it easier to play and
control your sound. This is because the dynamic range of the master and slave signals is added
together. For example, letÕs say both the master and slave signals have a dynamic range of 20 dB.
The resulting signal will have a dynamic range of 40 dB, giving you a very wide range between
the softest and loudest signals you can produce.
One way to improve intelligibility is to mix in a little of the master signal into the -nal audio
output. This can be done in a couple of ways. If you run the mic into a mixer, you can split the
signal, sending it both to the K2600 as well as to your -nal mix.
A second way is to include it in the vocoder program. You can do this by editing one of the
programs in the 22- or 20-band vocoder setups. You would want to add a layer to the program
(it doesn’t matter which one of the two programs you edit). Set the Keymap for the layer to
LiveIn L and choose Algorithm 1 with the DSP function set to NONE. You could then control the
amount of the signal by editing the Adjust parameter on the F4 AMP page (or even assign a
control source to vary the amount).
You could then try various algorithms and DSP functions to further modify the signal. Running
the signal through a high pass DSP to emphasize vocal articulations is one obvious example.
Just make sure that you don’t use the SHAPE 2 or AMP MOD OSC DSP functions. In that case,
the master signal won’t be output.
If you are using the K2600 for the slave signal, try editing the slave vocoder program. A simple
thing to try is to choose a different keymap. The AMPENV in this program has been set to User,
with a lengthy decay, so you can even choose decaying sounds such as guitar, and get
interesting results. And of course, you can choose other programs as the slave.
And of course, you should try making some of your own programs to use as a source. Just edit
the setup and change the program in zone 3 to your new program. For example:
¥ Use an LFO to modulate the center frequencies of the slave bandpasses, or the master
bandpasses.
¥ Try panning alternate bands of the slave layers to L and R to create a Òfake stereoÓ program.
¥ Try different center frequencies from the ones used in the preset programs.
¥ Currently the center frequencies of the slave layers match the master layers. Try scrambling
the slave frequencies relative to the master frequencies.
¥ If you are using the K2600 for the slave signal and need more polyphony, you can delete
some of the layers in the vocoding programs. Make sure to delete matching sets of master
and slave layers. You will probably want to readjust the frequencies and widths of the
remaining layers accordingly.
More applications
Instead of using a microphone or other external source for your master, you could use the K2600
to generate both the master and slave signals. There are two ways you could set this up. You can
either edit the setup to add another program on a 4th zone, or you could edit the slave source
program to add more layers. Then split the keyboard so that one side plays the master zone/
layers and the other side plays the slave zone/layers. On the OUTPUT page, make sure all the
master layers are assigned to B and panned hard left and the slave layers assigned to B and
panned hard right. You will then have to alter the wiring setup described at the beginning of this
document so that the B Left jack is going to the left side of the stereo sample input.
If you edit width of the master layers so that they are extremely narrow, and set the frequencies
to a speci-c scale pattern, then if you sing into the microphone, you will only hear sound as you
sing the speci-c pitches in that scale.
If you edit the width of the slave layers so that they are extremely narrow, then you will get a
very pure tonal sound, hearing only very speci-c pitches depending on the harmonic content of
the master.
Another possibility for using very narrow width master layers: Edit the slave layers so that
instead of using a series of bandpass -lters, each slave layer uses different DSP functions in the
F1 and F2 slots (remember that the F3 slot still needs to be set to LPCLIP in order for the
vocoding function to workÑyou can change algorithms as long as the algorithm allows LPCLIP
to be selected for the F3 slot). Now, if you sing various pitches, the slave signal will be played
through the various corresponding VAST algorithms.
It is actually possible to use samples in RAM (or ROM) instead of the Live Mode In for either the
master or slave signals (or even both of them). Just change the Keymap parameter on the
KEYMAP Page. (Remember that you need to edit the Keymap parameter on all master and/or
slave layers.) In this case, the keymap would be playing a single held sample, so you will want
to use a looped sample. Loops with changing harmonic content will work best. The note used in
the setups is C 4, so you would want the sample root at C 4 to hear it back without transposition.
You will need to edit the layers, save the programs, and reselect the setup before you will hear
the change. If both the master and slave layers call up samples in the unit, then as soon as you
select the setup, you will hear sound without even touching the keyboard! You might want to
assign a slider to the F4 AMP page on the slave layers to control the amount of output. If the
master and slave layers are loops of slightly different lengths, then you will hear a continually
changing sound that could appear to go in inde-nitely without changing.
Continuing with the previous suggestion, you could set the slave layers to different keymaps,
each layer assigned to a different sample loop. Edit the DSP functions on the slave layers so that
F1 and F2 are set to NONE, or some other DSP function. Set the master layers to very narrow
widths. Now, as your master signal changes frequencies you will hear different sample loops
fading in and out.

Knowledge Base