From Filtering to Control

Each stage addresses noise at a progressively deeper level of the quantum execution stack.

Stage 1 — Public

Trajectory Filtering

Operates on raw measurement outcomes after circuit execution. Identifies and suppresses noise-dominated trajectories to improve expectation-value estimates. Requires no hardware modification — works with standard Qiskit workflows and Level-2 (binary) measurement data.

Open source
Stage 2 — Non-Public

Calibration-Driven Observable Reconstruction

Uses device calibration data and structured circuit anchors to reconstruct observables with higher fidelity. By grounding reconstruction in calibration-informed references, this layer can extract more signal than binary outcome processing alone. Implementation specifics are patent-pending.

Patent-pending
Stage 3 — Non-Public

Real-Time Telemetry-Driven Control

The long-term architectural target. Uses runtime telemetry signals — analog IQ data and waveform-level information — to make control decisions during circuit execution rather than after. Requires hardware-level access and FPGA-adjacent deployment. Details remain confidential.

Patent-pending

Understanding the Measurement Stack

Quantum hardware produces information at multiple levels. Most software only sees the top.

Level 2

Binary Outcomes

The standard output: each qubit measurement yields 0 or 1. This is what most quantum software consumes. All analog information has been discarded by the discriminator.

Level 1

Analog IQ Telemetry

Before discrimination, each measurement produces a complex IQ point in the readout resonator's signal space. This analog data contains richer information about the qubit state — including confidence and proximity to decision boundaries.

Level 0

Waveform-Resolved Telemetry

The full time-resolved readout waveform before integration. Contains the most complete picture of the measurement process but requires direct hardware access and high-bandwidth data paths. Relevant for future runtime control.

The public qgate package operates on Level 2 data. Deeper measurement levels require hardware vendor partnerships for access and integration.

Weak Measurement Emulation

In quantum physics, a "weak measurement" extracts partial information about a system without fully collapsing its state. In practice, most quantum hardware performs strong, projective measurements — producing definitive 0/1 outcomes.

qgate's approach emulates aspects of weak measurement operationally: by leveraging indirect telemetry signals, calibration references, and structured filtering, it extracts more information from the measurement process than hard binary discrimination alone provides.

This is not a claim of performing true weak measurements. It is a signal-processing and calibration-informed strategy that recovers partial trajectory information that standard processing discards.

Clifford vs. Non-Clifford

Clifford circuits are a restricted class of quantum circuits that can be efficiently simulated classically. They serve as calibration anchors — because their ideal behavior is known exactly, deviations on real hardware reveal noise characteristics.

Non-Clifford circuits are the circuits that actually do useful quantum computation — but they are classically intractable, meaning we cannot predict their ideal outcomes.

qgate's architecture leverages this distinction: Clifford-based calibration data informs how the stack models and mitigates noise in the non-Clifford circuits that matter. The exact methods for bridging this gap are non-public.

Architecture Diagram

Quantum Hardware / FPGA Control Plane Qubits · Readout Resonators · Control Electronics · Waveform Generators Level 0 — Waveform Telemetry Level 1 — Analog IQ Telemetry Level 2 — Binary Outcomes Stage 1: Trajectory Filtering Public · Level 2 Input Stage 2: Reconstruction Patent-Pending · Level 1+2 Stage 3: Runtime Control Patent-Pending · Level 0+1+2 Calibration Data · Clifford Anchors · Device Characterization

Bridging Mitigation and Control

Most quantum error mitigation techniques operate entirely after execution — averaging over many shots to statistically cancel noise. This is useful but limited: it does not address the information lost at the measurement boundary, and it cannot adapt to runtime conditions.

qgate is designed to progressively close this gap. The public trajectory filtering layer works within the constraints of standard post-execution data. The non-public layers reach deeper — using calibration data to inform reconstruction, and ultimately using runtime telemetry to inform control decisions.

The architecture does not claim to solve the noise problem entirely. It is a structured approach to extracting more useful information from noisy quantum hardware at every level where information is available. The specific methods, algorithms, and transfer functions that enable this are non-public and patent-pending.