Q
QAtlas laboratory / local simulator

Noise Lab

Compare ideal output with noisy and readout-mitigated local simulations, then deliberately sweep a model until useful signal fails.

Shared circuit workspace

Circuit input

OpenQASM 2 static-circuit subset. Files stay in this browser; code is never executed.

Limits: 64 KB, 16 logical qubits, 512 operations. OpenQASM 3 and dynamic control are explicitly unsupported.
Noise model

Assumptions

Each enabled control changes the local simulation. Trajectory mode rejects damping channels rather than approximating them.

Load an example or paste OpenQASM 2, then validate it locally.

About Noise Lab

Test a circuit against stated noise assumptions

Noise Lab compares ideal distributions with genuine local noisy simulations and a bounded readout-mitigation method. It is designed for learning, development, and transparent what-if analysis—not for presenting synthetic results as device experiments.

Simulation modes

Exact density-matrix evolution supports up to five active qubits. Seeded trajectory simulation extends selected Pauli, depolarizing, gate, and readout models to six through eight qubits. Limits are displayed before a run.

Noise and mitigation

Depolarizing, flip, damping, gate, and readout controls alter the calculation. Readout mitigation uses an independent confusion model derived from supplied values and reports whether it improved, had negligible effect, or worsened similarity.

Break This Circuit

A bounded sweep locates where a selected simulated threshold is crossed. The result is conditional on this circuit, model, shot count, and threshold—not a universal hardware limit.

Method and reference

Readout mitigation uses bounded iterative Bayesian unfolding against a simulated independent confusion model. The readout-noise unfolding reference explains the method family; QAtlas does not use device calibration data or claim hardware mitigation performance.

Frequently asked questions

Are these hardware results?
No. Noise Lab does not contact real hardware and labels every output as simulated, estimated, assumed, or user supplied.
Why might mitigation worsen a result?
Mitigation can increase variance or be a poor fit for the selected assumptions. Noise Lab compares the computed result rather than declaring mitigation automatically successful.