How it Works, Theoretically Speaking
To best use the audio servo board, it is important to
explore the underlying theory of operation. Basically, the
board works by detecting the presence or absence of sound.
When sound is detected, the board instructs the servo motor
to move from its “home” position towards a “target” position.
When sound is absent, the servo is instructed to move back
to the home position. When the servo motor is coupled
to a jaw or mouth, the motions are surprisingly accurate.
If you have a look at the schematic (Figure 4), we
can go over the operation in detail so we can really
understand the parts and how they work. The audio signal
enters through the 1/8” stereo jack J3. Jack J2 is wired in
parallel so that the signal can exit the board to an external
speaker system. J4 and J5 control which channel is passed
to the next stage (i.e., left, right, or both).
The audio signal from J4/J5 passes through C2 and R3
which provide some isolation and the resulting signal is
fed to the U2.1 op-amp which is configured as an audio
amplifier. VR1 is used to adjust the input gain of the
op-amp section to match the incoming audio signal level.
The output of the first op-amp is then fed to the
second op-amp which is configured as a comparator. VR2
is used to set the threshold that triggers the comparator
output. Jumper J8 is used to select which damping capacitor
(C7, C8, or C9) is used to shape the decay of the trigger
signal that feeds the U5 quad switch. Larger capacitance
results in slower, smoother movement; smaller capacitance
results in faster motion. When the audio signal is above the
threshold of the comparator, the output of the comparator
is passed to all four inputs of the U5 quad switch. U5 is
then used to drive four separate outputs.
The first output (U5.1) drives a pair of remotely-mounted
LED eyes that light up when sound is present. The second
output (U5.2) drives an LED that is used as a guide to help
you visually calibrate the board to the incoming audio level
(i.e., it lights when the board is detecting audio). The third
output (U5.3) is used to drive the ULN2803a driver and
the fourth output (U5.4) is used to control the 555 timer.
The 555 timer is configured as an astable multivibrator
and when the trigger level changes, U5.4 changes the timing
resistor which shifts the PWM width from ~0.3 ms to ~ 3. 7 ms
(VR3 adjusts the 3. 7 ms down to 2.0 ms or so allowing you
to position the servo’s home position). The PWM output is
inverted and driven to the servo jack by R11 and transistor Q1.
The ULN2803a is used as a high-current driver allowing
you to control devices such as incandescent lamps, small
motors, relays, or solenoids. Though the ULN2803a has
built-in back-EMF diodes, if you plan to drive some larger
solenoids it might be a good idea to place an additional
back-EMF protection diode in the circuit. The ULN2803a
is a neat addition to the original prototype board and
comes in handy if you decide you would prefer to have
the audio servo board drive a solenoid valve connected to
a pneumatic cylinder. This way, you could use the board
to control the movement of very large puppet jaws or
mouths weighing tens or even hundreds of pounds!
To assist in getting the best performance out of the
circuit, the board has a series of adjustments that allow you
to adapt the device to the sound track you choose. For
example, you can choose which channel (left, right, or both)
is used to drive the servo motor. This is important since
the servo will respond to any sound it detects. Typically,
you would want to use a separate channel for dialog and
sound effects. Or, you can daisy-chain the devices together
and have each talking skull respond to a different dialog
track. This allows you to have a pair of skulls carry on a
conversation! You can also use the J4/J5 jumpers to select
which board should respond. The J8 jumper block is used
to fine tune how responsive you want the jaw to be.
The Head Bone’s Connected
to the Servo Bone
Okay, so enough theory. Let’s get to the application!
As a member of the TXfx group in Austin, TX, I thought
it would be fun to have a “Make and Take” with talking
skulls. I sent out an email message announcing the idea
to all my scary friends on the TXfx mailing list.
For those of you not familiar with the concept, a “Make
and Take” is part work-day and part party-day. Folks with
common interests and overlapping skill sets gather at a
predetermined time and place to take on the same challenge.
We meet up (usually on a weekend) and by cooperating
and teaching each other, we each make a working device
■ FIGURE 5. TXfx members working
at the skull assembly area.
■ FIGURE 6. TXfx members and
folks from The Robot Group at the
■ FIGURE 7. Eric Lundquist and Vern
Graner (cleverly disguised as a giant
pumpkin) at the system test area.