Noise modelling

Real hardware is noisy. SimQ’s noise system (in simq-core::noise, exposed to Python as simq.noise) lets you attach realistic error channels to simulations.

Noise channels

Channel	Models
DepolarizingChannel	Uniform Pauli errors with probability p
AmplitudeDamping	Energy relaxation (T₁ decay)
PhaseDamping	Pure dephasing (T₂)
ReadoutError	Bit-flip errors at measurement time

Each channel has a Monte-Carlo counterpart (DepolarizingMC, AmplitudeDampingMC, …) used for trajectory-based sampling during shot simulation.

Hardware noise models

HardwareNoiseModel composes per-qubit and per-gate noise into a device profile:

• QubitProperties — T₁/T₂ times and per-qubit error rates

• GateTiming — gate durations, so damping is applied for the right time

• TwoQubitGateProperties and CrosstalkProperties — coupling-dependent errors

• GateNoise — attaches channels to specific gate types

Using noise from Python

import simq

# Build a device-like noise model
noise_model = simq.HardwareNoiseModel()
noise_model.add_gate_error("cx", simq.DepolarizingChannel(0.01))

# Simulate with noise
config = simq.SimulatorConfig(noise_model=noise_model, shots=1000)
simulator = simq.Simulator(config)
result = simulator.run(circuit)

A complete, commented walkthrough lives in simq-py/examples/noise_simulation.py — it builds noise channels, attaches them to a hardware model, and compares noisy versus ideal measurement distributions.

Tips

• Start with a single DepolarizingChannel on two-qubit gates — they dominate error budgets on real devices.

• Use SimulatorConfig(seed=...) for reproducible noisy runs when debugging.

• Readout error is often the largest effect on shallow circuits; model it separately from gate noise.