Noise models and fake backends

How a simulator imitates a device: depolarising, thermal relaxation and readout channels built from calibration data.

You will learn to

Name the three channels in a fake backend's noise model.
Explain what a density matrix adds over a statevector.
Choose between fake backends for a given circuit.

A fake backend is a snapshot of a real machine: its coupling map, basis gates and a day's calibration. Qiskit Aer turns the calibration into a noise model with three channels: depolarising error after each gate (scaled to the measured gate error), thermal relaxation during each gate's duration (from T1, T2) and a readout confusion matrix. Simulating with it predicts hardware histograms surprisingly well.

ρ \mapsto (1 - p) ρ + \frac{p}{3} (X ρX + Y ρ Y + Z ρZ)

The single-qubit depolarising channel. Noise maps pure states to mixed ones, so the simulator must track a density matrix ρ (2ⁿ × 2ⁿ) instead of a statevector (2ⁿ). That is why noisy simulation stops around 12–14 qubits.

Try it

Three chips, one circuit

Preset '3 noisy chips' runs the same Bell pair on three fake backends. Compare the total variation distances in Metrics; then open Hardware → calibration for each to see which number explains the ranking.

OPENQASM 3.0;
include "stdgates.inc";
qubit[2] q;
bit[2] c;
h q[0];
cx q[0], q[1];
c = measure q;

Open in the IDE Open in the Playground

Quiz 1 · 10 XP

Why does a noisy simulation need a density matrix?

Quiz 2 · 10 XP

Which calibration number most affects a deep circuit with few two-qubit gates?

Key takeaway

Fake backends turn a device's calibration into depolarising, relaxation and readout channels. They need density matrices, cost more to simulate, and are the honest reference before spending hardware minutes.

Docs: noisy targets

Up next: Error mitigation

You are reading without signing in. Sign in or continue as guest to take quizzes and keep XP.

02Noise and ErrorsAdvanced12 min· 45 XP