This circuit operates by using the LM13700 as a variable current source charging
a capacitor. It is configured as a non-inverting op-amp charging the capacitor
to a reference voltage V_ref at the + pin with a slew rate limited by the current
control. The capacitor's voltage will change linearly based on I_abc until it settles
at V_ref. Then a bit of digital logic to create a 3-state state machine to select the
correct V_ref and I_abc.
States:
- Attack:
V_gate=5V,V_ref=V_in,I_abccontrolled byR_attack - Decay:
V_gate=5V,V_ref=V_sustain,I_abccontrolled byR_decay - Release:
V_gate=0,V_ref=0,I_abccontrolled byR_release
The Attack, Decay, and Release knobs control I_abc in each state,
while the Sustain knob divides down V_in to set V_ref in the
Decay state.
A comparator, capacitor, and a few transistors create cheap analog
a flip-flop that turns on when the output hits V_in (flipping from
the Attack state to the Decay state), and turns off when V_gate
turns off.
Voltage control of the parameters could be possible by exposing
access to I_abc, but it doesn't make much sense to me to expose
those parameters with any control other than a knob - it's not the
kind of thing you need to modulate automatically.
Choice of the capacitor and the knob resistance control the time
range. Large values for both mean a longer charge time, but a large
capacitance effects the fastest charge time too, since I_abc is
capped at 2mA. Pot size effects dynamic range.
In practice, charge times less than 15ms are undetectable, and charge times above 2s are largely unnecessary.
A non-linear resistance would allow finer control at larger charge times. However linear seems to have plenty of expressiveness.
C=4.7u and a knob range of 1M gives a control range of 12ms to ~1.8s for a 0-5V change, i.e. a slew rate between 356 V/s and 2.4 V/s.
Could offer two caps, fast and slow, with a toggle switch. A 22u capacitor gives 58ms to 8.6s. A 1u capacitor gives ~5ms to ~55ms.
A current mirror isolates the 3 different reference voltages, without it they interfere with each other.
Rates are dependant on the input level. This means if the input gate level is variable (for example it depends on MIDI note velocity), then the times will be slower for harder hits. We'd probably want constant rates regardless of velocity for example. However this seems difficult to achieve with the current design, as during release the input voltage is off, so without memory you don't have a reference to scale the output by.
TODO: power usage