ACME is a Julia package for the simulation of electrical circuits, focusing on audio effect circuits. It allows to programmatically describe a circuit in terms of elements and connections between them and then automatically derive a model for the circuit. The model can then be run on varying input data.
ACME is based on the method described in M. Holters, U. Zölzer, "A Generalized Method for the Derivation of Non-Linear State-Space Models from Circuit Schematics".
To install ACME, start Julia and run:
This will download ACME and all of its dependencies.
We will demonstrate ACME by modeling a simple diode clipper. The first step is to load ACME:
Now we create the circuit description:
circ = @circuit begin j_in = voltagesource() r1 = resistor(1e3) c1 = capacitor(47e-9) d1 = diode(is=1e-15) d2 = diode(is=1.8e-15) j_out = voltageprobe() j_in[+] ⟷ r1 j_in[-] ⟷ gnd r1 ⟷ c1 ⟷ d1[+] ⟷ d2[-] ⟷ j_out[+] gnd ⟷ c1 ⟷ d1[-] ⟷ d2[+] ⟷ j_out[-] end
The first six lines inside the
end block instantiate circuit elements.
voltagesource() sets up a voltage source as an input, i.e. the
voltage it sources will be specified when running the model. Alternatively, one
can instantiate a constant voltage source for say 9V with
capacitor calls take the resistance in ohm and the
capacitance in farad, respectively, as arguments. For the
diode, one may
specify the saturation current
is as done here and/or the emission
η. Finally, desired outputs are denoted by adding probes to the
circuit; in this case a
voltageprobe() will provide voltage as output.
The remaining four lines specify connections, either among element pins as in
j_in[+] ⟷ r1, which connects the
+ pin of the input voltage to pin
the resistor, or among pins and named nets as in
j_in[-] ⟷ gnd, which
- pin of the input voltage source to a net named
gnd. Note that
naming nets is only for the sake of readability; there is nothing special about
them and the names are arbitrary. As can be seen in the last two lines, multiple
pins can be connected at once.
It is also possible to specify connections following the element definition (separated by commas), in which case the element name may be omitted. However, one can only connect to elements defined before. Thus, above circuit could also be entered as:
circ = @circuit begin j_in = voltagesource(), [-] ⟷ gnd r1 = resistor(1e3),  ⟷ j_in[+] c1 = capacitor(47e-9),  ⟷ r1,  ⟷ gnd d1 = diode(is=1e-15), [+] ⟷ r1, [-] ⟷ gnd d2 = diode(is=1.8e-15), [+] ⟷ gnd, [-] ⟷ r1 j_out = voltageprobe(), [+] ⟷ r1, [-] ⟷ gnd end
Now that the circuit has been set up, we need to turn it into a model. This could hardly be any easier:
model = DiscreteModel(circ, 1/44100)
The second argument specifies the sampling interval, the reciprocal of the sampling rate, here assumed to be the typical 44100 Hz.
Now we can process some input data. It has to be provided as a matrix with one row per input (just one in the example) and one column per sample. So for a sinusoid at 1 kHz lasting one second, we do
y = run!(model, sin(2π*1000/44100*(0:44099).'))
y now likewise is a matrix with one row for the one probe we have
added to the circuit and one column per sample.
In the likely event that you would like to process real audio data, take a look at the WAV package for reading writing WAV files.
Note that the solver used to solve the non-linear equation when running the model saves solutions to use as starting points in the future. Model execution will therefore become faster after an initial learning phase. Nevertheless, ACME is at present more geared towards computing all the model matrices than to actually running the model. More complex circuits may run intolerably slow or fail to run altogether.
There is some documentation
available for how
to use ACME. Additionally, you can take a look at the examples that can be found
examples directory below
If you would like to extend and improve ACME, that's great! But unfortunately, there is no developer documentation as of now, so you will to delve into the source code to figure out how things work, or try to ask on gitter.
7 days ago