Photonic Computing API Reference
Complete documentation for all classes and methods
class RingResonator
Silicon microring resonator model for implementing photonic weights.
from photonic import RingResonator
ring = RingResonator(
radius=10e-6, # Ring radius in meters
coupling_gap=200e-9, # Waveguide-ring gap
waveguide_width=500e-9,
loss_db_cm=2.0, # Propagation loss
n_eff=2.4 # Effective refractive index
)Constructor Parameters
| Parameter | Type | Description |
|---|---|---|
radius | float | Ring radius in meters (typically 5-20 μm) |
coupling_gap | float | Gap between bus waveguide and ring in meters |
waveguide_width | float | Waveguide width in meters |
loss_db_cm | float | Propagation loss in dB/cm |
n_eff | float | Effective refractive index of waveguide mode |
Methods
transmission(wavelength) → float
Returns power transmission coefficient at given wavelength (meters).
phase(wavelength) → float
Returns phase shift in radians at given wavelength.
get_fsr() → float
Returns Free Spectral Range in meters.
get_quality_factor() → float
Returns loaded Q factor.
tune_resonance(delta_lambda) → float
Thermally tune resonance by delta_lambda. Returns required heater power in mW.
class PhotonicMVM
Photonic Matrix-Vector Multiplication unit using WDM crossbar architecture.
from photonic import PhotonicMVM
mvm = PhotonicMVM(
size=(32, 32), # Matrix dimensions
precision=4, # Bits per weight
wavelengths=8, # Number of WDM channels
clock_rate=10e9 # Operating frequency in Hz
)Constructor Parameters
| Parameter | Type | Description |
|---|---|---|
size | tuple | Matrix dimensions (rows, cols) |
precision | int | Weight precision in bits (2, 4, or 8) |
wavelengths | int | Number of WDM channels |
clock_rate | float | Operating frequency in Hz |
Methods
load_weights(matrix) → None
Load weight matrix (numpy array). Automatically quantizes to specified precision.
forward(input_vector) → ndarray
Perform matrix-vector multiplication. Returns output vector.
get_throughput() → float
Returns compute throughput in TOPS.
get_energy_per_mac() → float
Returns energy per MAC operation in fJ.
set_noise_model(model) → None
Attach a NoiseModel instance for realistic simulation.
class PCMWeight
Phase-change material model for non-volatile photonic weight storage.
from photonic import PCMWeight
pcm = PCMWeight(
material='gst', # 'gst', 'gsst', 'aist'
thickness=30e-9, # Film thickness in meters
wavelength=1550e-9, # Operating wavelength
levels=16 # Number of analog levels
)Methods
set_state(level) → float
Program to specified level (0 to levels-1). Returns energy consumed in nJ.
get_transmission() → float
Returns current optical transmission (0 to 1).
get_refractive_index() → complex
Returns current complex refractive index n + ik.
crystallize(pulse_width, power) → None
Apply SET pulse with given parameters.
amorphize(pulse_width, power) → None
Apply RESET pulse with given parameters.
class OpticalDAC
High-speed optical digital-to-analog converter for input encoding.
from photonic import OpticalDAC
dac = OpticalDAC(
bits=8, # Resolution
architecture='segmented', # 'binary', 'thermometer', 'segmented'
sample_rate=40e9 # Sample rate in Hz
)Methods
convert(digital_code) → float
Convert digital code to analog optical power level.
get_inl() → ndarray
Returns Integral Non-Linearity for all codes in LSB.
get_dnl() → ndarray
Returns Differential Non-Linearity for all codes in LSB.
get_enob() → float
Returns Effective Number of Bits.
class OpticalADC
Photonic analog-to-digital converter for output readout.
from photonic import OpticalADC
adc = OpticalADC(
bits=6, # Resolution
architecture='flash', # 'flash', 'sar', 'pipeline'
sample_rate=20e9 # Sample rate in Hz
)Methods
sample(optical_power) → int
Sample analog optical signal and return digital code.
get_sndr() → float
Returns Signal-to-Noise-and-Distortion Ratio in dB.
get_sfdr() → float
Returns Spurious-Free Dynamic Range in dB.
class NoiseModel
Comprehensive noise analysis for photonic computing systems.
from photonic import NoiseModel
noise = NoiseModel(
optical_power=1e-3, # Input power in Watts
bandwidth=20e9, # Detection bandwidth in Hz
temperature=300, # Temperature in Kelvin
responsivity=0.9, # Photodetector responsivity A/W
rin_dbhz=-150 # Laser RIN in dB/Hz
)Methods
shot_noise() → float
Returns shot noise current in A/√Hz.
thermal_noise(load_resistance) → float
Returns thermal noise current in A/√Hz.
rin_noise() → float
Returns RIN-induced noise current in A/√Hz.
total_noise() → float
Returns total noise current (RSS) in A/√Hz.
get_snr() → float
Returns Signal-to-Noise Ratio in dB.
get_enob() → float
Returns Effective Number of Bits.
Complete Example
import numpy as np
from photonic import PhotonicMVM, PCMWeight, NoiseModel
# Create 32x32 photonic MVM unit
mvm = PhotonicMVM(size=(32, 32), precision=4, wavelengths=8, clock_rate=10e9)
# Generate random weight matrix
W = np.random.randn(32, 32) * 0.1
mvm.load_weights(W)
# Add realistic noise model
noise = NoiseModel(optical_power=1e-3, bandwidth=20e9, temperature=300)
mvm.set_noise_model(noise)
# Forward pass
x = np.random.randn(32)
y = mvm.forward(x)
# Check performance
print(f"Throughput: {mvm.get_throughput():.1f} TOPS")
print(f"Energy/MAC: {mvm.get_energy_per_mac():.1f} fJ")
print(f"SNR: {noise.get_snr():.1f} dB")
print(f"ENOB: {noise.get_enob():.1f} bits")
# Compare with ideal result
y_ideal = W @ x
mse = np.mean((y - y_ideal)**2)
print(f"MSE: {mse:.2e}")