Overview

Know what your hardware actually does.

Quantum hardware vendors publish specifications, but real device performance depends on calibration state, environmental conditions, and interactions between qubits that specifications rarely capture fully. Quantum tomography is the practice of learning actual device behavior from experimental measurements rather than from documentation — running systematic probing circuits and analyzing the results to reconstruct what the hardware is actually doing to quantum states.

The φCoherent tomography package implements three complementary benchmarking protocols: Randomized Benchmarking (RB) for quick per-gate fidelity estimates; Cross-Entropy Benchmarking (XEB) for calibration against known theoretical predictions; and Gate-Set Tomography (GST) for the most complete characterization — a full process matrix for every gate in the device's native gate set. Together these three protocols cover the full range from rapid daily calibration to deep one-time characterization of a new device.

/ Capabilities

What it provides.

Three complementary benchmarking protocols — from rapid daily calibration checks to deep one-time device characterization — covering every stage of a hardware deployment lifecycle.

  • / 01

    Randomized Benchmarking

    Estimates average gate fidelity quickly by running random sequences and measuring how fidelity decays with sequence length; the standard protocol for ongoing hardware calibration.

  • / 02

    Cross-Entropy Benchmarking

    Measures how closely hardware results match the ideal distribution for random circuits, calibrated against the theoretical prediction; the protocol used by major hardware providers to report fidelity milestones.

  • / 03

    Gate-Set Tomography

    Reconstructs the full process matrix for each gate in the native set, providing the most complete available description of hardware behavior at the cost of substantially more measurement runs.

  • / 04

    Hardware fidelity profiling

    Aggregates results across protocols into a structured fidelity profile that informs error budget allocation, QEC code selection, and ZNE calibration for the specific device.

Measured benchmarks (GCC 13, -O3, Linux x86-64)

RB at Fibonacci lengths {2,3,5,8,13,21,34,55}: EPC matches theory 2p/3 to 4 decimal places (p=0.01 → EPC=0.00667, theory=0.00667) in 9.89 ms. Interleaved RB isolates T-gate error (0.02964) from Clifford baseline (0.00333). Linear XEB: F_XEB = 1.532 noiseless (n=6, depth=4), dropping monotonically to 0.0227 at p=0.10. GST: per-gate diamond distances in 0.01 ms, zero external dependencies — no pyGSTi required. Qiskit Experiments: RB available, no GST; Cirq: no GST; Stim: no characterization protocols.

Related packages
Licensing

Open under AGPLv3.

Published under the GNU AGPLv3 with whitepaper and reference implementation. Commercial licensing is available for closed-source deployments.

← Quantum Stack Commercial License