Broadly speaking, the harmonic tremolo circuit works in the following way:
First, the audio is split into separate treble and bass signals with the
resistors and capacitors wired to the grids of the 7025 tube in the
center-left of the schematic. Then, subaudio control signals of opposite
phase are superimposed on these two audio signals via a network of 1M
resistors to the left of the frequency-splitting circuit. To reiterate,
these control signals are not within the human hearing range. The
superimposed signals reach the grids of the center-left 7025 tube, where
the subaudio and audio signals do different things. The subaudio signal
controls the gain of the valves, while the audio signal is acted upon by
that gain. The treble and bass signals are then combined and sent to the
rest of the guitar amp and eventually make their way to the speaker. The
control signals are rejected from the rest of the audio path by
series-coupling capacitors.
The aforementioned opposite-phase control signals come from the circuits
shown in the vertical center, all the way to the left in the schematic.
These circuits form a low-frequency phase-shift oscillator that drives a
simple phase inverter. This neat design requires only one tube for both
the oscillation part and the output drive part.
The guitar signal first arrives at the input stage, then makes its way to
the frequency splitter stage, followed by the VCA stages, and finally
comes to the current summing stage before making it to the output.
It may seem a bit obtuse, but the best way to explain this design is from
the middle first, where the Voltage-Controlled Amplifiers are found. The
VCAs are the most critical part of this circuit: They are the devices
used to modulate the volume of each frequency range. Most of this design
focuses on how to get the VCAs to play nice in a guitar-pedal setting.
Two VCAs are used: one for the higher audio frequencies, and one for the
lower frequencies. Only one VCA circuit is shown here, for simplicity.
The VCA I chose for this design is of the respected THAT218X series of
VCAs. Like many VCAs, this device operates on a current signal, but has
multiple voltage control ports that determine the specifics of that
operation. The specifics of VCA operation are beyond the scope of this
article, so curious readers should page through the device's datasheet for
a more thorough discussion. Here, I will go over the aspects of the
VCA's operation as they pertain to this design.
In practice, the current signal operation means that conversion from
one mode of signaling to another is required around the VCA. This sounds
more complicated than it is -- the simplicity will become apparent during
the discussion of following circuits.
The most important part to consider in this application is the
relationship between the voltage at the EC- pin and the amount of signal
current passed from pin 1 to pin 10. With zero voltage on EC- (relative
to ground), the VCA passes all of the input current to the output. If a
positive voltage is applied to EC-, the current passed from input to
output will be reduced exponentially with that voltage. In this way, the
VCA controls the current gain from 0 to -90dB.
In this application, a slowly varying DC signal is applied to each VCA's
EC- pin. This signal enters the VCA subcircuit via the LFO port
indicated in the schematic. How much of this DC signal is allowed to
affect the VCA gain is controlled by the HIGH FREQ. DEPTH CONTROL
potentiometer.
The "SYM" pin is also an important part of the circuit. By applying a
small voltage to this pin, the VCA can be adjusted to introduce the least
amount of noise possible to the circuit.
The VCAs control the volume of each frequency band, and the aptly named
frequency-splitter circuit separates the input signal into a
high-frequency and a low-frequency band -- one for each VCA.
Additionally, the frequency-splitter circuit converts each signal from a
voltage into a current and provides those signals to the appropriate VCA
input.
Truth be told, the frequency-splitter circuit is quite simple. The RC
network made up of R7, C4, C10, and C12 split the audio signal at about
320Hz. Higher frequencies take the path on the top half of Figure 4, and
lower frequencies take the path on the bottom half. This circuit is nearly
identical to that seen in the original fender schematic, but with some
minor changes in order to use component values that are easier to come by.
The voltage-to-current conversion step for each band is merely a
resistor (R2 and R22). The inputs to the VCAs look like ground, so the
values of these resistors determine the amount of signal current that
enters the VCA for each band. Lower resistor values mean more current
enters the VCA. R22 is smaller than R2 to correct for a perceived
reduction in volume of the lower frequencies while this effect is engaged.
Driving the frequency-splitter is a simple input stage. This circuit does
three things: It provides the guitar signal with a high-impedance load,
increases the amplitude of the guitar signal by about 12dB, and provides
the frequency-splitter circuit with a low impedance signal. The gain is
required in order to maintain a good signal-to-noise ratio in the rest of
the circuit. Even after trimming the SYM pin on the VCAs, the circuit's
noise may be unacceptable if the input stage doesn't provide enough gain.
12dB was chosen for the gain as a good compromise for a variety of
guitars and playing styles while still keeping the stage's output signal
within the limits of the voltage rails (+9V and -9V).
Moving to the end of the signal path, we have the current-summing circuit. As mentioned before, the VCAs output a current signal instead of a more standard voltage signal. This circuit adds the current signals from each VCA together and converts them to a voltage. The conversion ratio is controlled by R17 and the setting of the GAIN TRIM potentiometer. Generally, the conversion can be expressed mathematically as V_OUT = I_IN * R_FB where R_FB is the total feedback resistance seen by the OpAmp. The "GAIN TRIM" potentiometer is intended to be set so that, on the whole, the effect does not increase or decrease the perceived signal volume. C7 simply acts to remove some very high frequency noise. The signal out port is the output node of this effect, U1B is merely acting as a buffer to drive whatever circuits come after this circuit.
Instead of generating the phase-inverted low-frequency control signals for
the VCAs with a phase-shift oscillator and phase splitter as in the Fender
Super Amp, I went with a more modern approach of using a microcontroller
to generate a single oscillator signal via PWM, and then used some simple
analog circuits to process that PWM signal into something more useful. The
advantage to this approach is that the oscillator amplitude can be
independent of oscillator frequency. This means that the tremolo effect
can be very slow or very fast while retaining the same intensity -- a
feature absent from the original phase-shift oscillator! Additionally,
using a microcontroller means that multiple oscillator waveshapes can be
used without increasing the component count significantly. This gives the
user more control over how the effect sounds than was available in the
original Fender amplifiers.
The PWM processing circuits are shown in Figure 7. I've omitted the
microcontroller schematic here since it's trivial. The complete circuit is
described in the full schematic, of course.
The left-hand side of the diagram shows the PWM filter. This filter is a
passive fourth-order filter and does a great job of converting the
PWM signal from the microcontroller into a smooth analog waveform. Note
the scaling of RC values with each successive stage. This is done to
prevent the proceeding stage from heavily loading the proceeding stage.
The right-hand side of the diagram shows the oscillator buffer circuit and
the phase-reversal circuit which generates an opposite-polarity control
signal. This phase-reversal circuit is a textbook inverting OpAmp circuit
with a simple offset adjustment control tacked on. The offset adjustment
control is useful to ensure that both control signals reach the same upper
and lower bounds.
The power circuits in this design are simple, but carefully chosen. Most
important is the MAX1044 switching power supply circuit that provides the
-9V supply. Alternatives to the MAX1044 are popular and cheap, but only
the MAX1044 has a "boost" function that pushes the switching frequency
above the human hearing range. This turns an otherwise noisy switching
power supply into something perfectly acceptable for audio applications.
A positive and negative 9V supply was chosen for this circuit instead of
the more common single positive 9V supply found in guitar effects for a
few reasons. First, the VCAs require a bipolar power supply to function
correctly. Second, a bipolar power supply allows for greater signal levels
and therefore greater signal-to-noise ratio. Finally, using 9V means that
no special power adapter is required for this circuit.
The 5V linear regulator is simply there to power the microcontroller.