Circuit from YAML

Contents

Circuit from YAML#

Sometimes it’s useful to be able to define circuits from YAML definitions. To not re-invent the wheel, SAX uses GDSFactory’s YAML netlist spec to define its circuits. This makes it very easy to convert a GDSFactory layout to a SAX circuit model!

import jax.numpy as jnp
import matplotlib.pyplot as plt
import sax
import yaml

MZI#

Let’s first see how we can define a SAX circuit from YAML:

netlist = """
instances:
    lft:
      component: coupler
      settings:
          coupling: 0.5
    rgt:
        component: coupler
        settings:
            coupling: 0.5
    top:
        component: straight
        settings:
            length: 25.0
    btm:
        component: straight
        settings:
            length: 15.0

connections:
    lft,out0: btm,in0
    btm,out0: rgt,in0
    lft,out1: top,in0
    top,out0: rgt,in1

ports:
    in0:  lft,in0
    in1:  lft,in1
    out0: rgt,out0
    out1: rgt,out1

"""
yaml.safe_load(netlist)
{'instances': {'lft': {'component': 'coupler', 'settings': {'coupling': 0.5}},
  'rgt': {'component': 'coupler', 'settings': {'coupling': 0.5}},
  'top': {'component': 'straight', 'settings': {'length': 25.0}},
  'btm': {'component': 'straight', 'settings': {'length': 15.0}}},
 'connections': {'lft,out0': 'btm,in0',
  'btm,out0': 'rgt,in0',
  'lft,out1': 'top,in0',
  'top,out0': 'rgt,in1'},
 'ports': {'in0': 'lft,in0',
  'in1': 'lft,in1',
  'out0': 'rgt,out0',
  'out1': 'rgt,out1'}}
mzi, _ = sax.circuit(
    yaml.safe_load(netlist),
    models={"coupler": sax.models.coupler, "straight": sax.models.straight},
)
wl = jnp.linspace(1.5, 1.6, 1000)
transmission = jnp.abs(mzi(wl=wl)["in0", "out0"]) ** 2

plt.plot(wl * 1e3, transmission)
plt.xlabel("λ [nm]")
plt.ylabel("T")
plt.show()
../_images/7b5f06475aa208c22d365b48b6629787d460777cb2b6c0529b2ca87fb97e76a0.png

That was easy! However, during the above YAML conversion, only models available in sax.models were used. What if we want to map the YAML component names to custom models? Let’s say we want to use a dispersionless waveguide for the above model for example:

def waveguide_without_dispersion(wl=1.55, length=25.0, neff=2.34):
    phase = 2 * jnp.pi * neff * length / wl
    sdict = sax.reciprocal({("in0", "out0"): jnp.exp(1j * phase)})
    return sdict

We can regenerate the above circuit again, but this time we specify a models mapping:

mzi, _ = sax.circuit(
    yaml.safe_load(netlist),
    models={"straight": waveguide_without_dispersion, "coupler": sax.models.coupler},
)

The models= keyword in circuit_from_yaml can be a dictionary or an imported python module (like for example sax.models). Or a list containing multiple of such dictionary mappings and imported modules.

wl = jnp.linspace(1.5, 1.6, 1000)
transmission = jnp.abs(mzi(wl=wl)["in0", "out0"]) ** 2

plt.plot(wl, transmission)
plt.xlabel("Wavelength [nm]")
plt.ylabel("T")
plt.show()
../_images/64592a9627b1d2fea08bb52b88f594d52cf8c9ab200bf13c66eafa6fc1cffe11.png