Device-Agnostic Microwave Noise Metrology for Nonlinear Cryogenic Quantum Devices

This paper presents a device-agnostic in situ noise metrology protocol that combines Planck spectroscopy with a variable temperature stage and S-parameter calibration to accurately characterize the gain and added noise of nonlinear cryogenic microwave devices, such as Josephson Traveling Wave Parametric Amplifiers, by referencing measurements to the device's input ports.

The Gist

Imagine you are trying to listen to a whisper in a very loud, freezing cold room. The whisper represents a single photon of microwave energy, and the room is a complex machine used to study quantum computers. To hear that whisper clearly, you need a super-sensitive amplifier. But here's the problem: the amplifier itself makes noise, and the cold room's equipment adds its own static. How do you know how much of the noise you hear is actually from the whisper, and how much is just the machine humming?

This paper presents a new, clever way to measure that "machine noise" without getting confused by the device you are testing.

The Problem: The "Serial" Trap

Think of the old way of doing this as a relay race.

1. You have a known noise source (a "loudspeaker" that plays a specific static sound).
2. You put the device you want to test (the "amplifier") right in front of it.
3. The sound goes: Loudspeaker → Amplifier → Microphone.

The problem is that if the amplifier is "weird" or "nonlinear" (meaning it reacts strangely to loud sounds, like a distorted guitar pedal), the sound coming out isn't just the original static plus the amplifier's noise. The amplifier might scramble the static in unpredictable ways. If you try to calculate the noise based on this scrambled sound, you get the wrong answer. It's like trying to measure how much a filter cleans water, but the filter itself changes the color of the water you're testing.

The Solution: The "Substitution" Switch

The authors propose a new method that acts like a smart switchboard.

Instead of forcing the sound to go through the device during the test, they use a set of cryogenic switches (tiny, super-cold traffic directors) to swap the device out.

1. Step 1: Calibrate the Chain. They connect the "loudspeaker" (a controllable noise source) directly to the microphone, bypassing the device entirely. They measure exactly how much the microphone and the cables add to the noise. This gives them a perfect baseline.
2. Step 2: Test the Device. They flip the switch, disconnect the loudspeaker, and connect the device. Now they measure the output.
3. Step 3: Compare. Because they know exactly how much noise the "chain" adds (from Step 1), they can subtract it from the total noise measured in Step 2. What's left is the true noise added by the device itself.

The "Variable Temperature Stage" (The Magic Heater)

To make this work, they needed a noise source that is perfectly predictable. They built a special device called a Variable Temperature Stage (VTS).

Imagine a small, super-cold block of metal with a tiny heater inside.

• When it's very cold, it emits almost no noise (like a silent room).
• When they turn up the heater, it gets slightly warmer and emits a predictable amount of thermal noise (like a room slowly filling with the hum of people talking).

By slowly heating this block and measuring the noise at every step, they can map out the "noise curve" with extreme precision. This is called Planck Spectroscopy. It's like tuning a radio by slowly turning the dial and noting exactly where the static starts, rather than guessing.

The Real-World Test: The "JTWPA"

To prove their method works, they tested it on a very tricky device called a Josephson Traveling Wave Parametric Amplifier (JTWPA).

• The Analogy: Think of this amplifier as a very sensitive microphone that uses magnets and superconductors to boost signals. However, when you push it hard (with a strong "pump" signal), it starts acting weird, creating extra noise channels that are hard to predict.
• The Result: Using their "switchboard" method, they were able to measure the amplifier's noise even while it was behaving chaotically. They found that as they pushed the device harder, the noise grew much faster than the signal.

Why This Matters

The authors aren't claiming this will fix quantum computers tomorrow or cure diseases. They are simply saying: "We have built a better ruler."

In the past, measuring the noise of these complex, non-linear quantum devices was like trying to weigh a feather while standing on a shaking boat. Their new method puts the boat on solid ground. It separates the measurement tools from the device being tested, ensuring that the "noise" you measure is actually coming from the device, not from your own confusion about how the machine works.

This allows scientists to trust their measurements of quantum devices, regardless of how complex or "weird" those devices behave.

Technical Summary

Technical Summary: Device-Agnostic Microwave Noise Metrology for Nonlinear Cryogenic Quantum Devices

Problem Statement
Cryogenic microwave quantum devices, such as near-quantum-limited amplifiers, mixers, and isolators, are essential for solid-state quantum technologies. However, characterizing their noise performance presents a significant metrological challenge. To assess suitability for single-photon signal processing, the "added noise" of a Device Under Test (DUT) must be measured in absolute units and referred to a precise reference plane at the DUT's ports.

Existing methods often place a calibrated noise source in series with the DUT. While effective for linear devices, this approach becomes problematic for nonlinear or parametrically driven DUTs. In such cases, the extraction of the DUT's contribution relies on assumptions about the internal noise-transmission model. If the DUT exhibits unmodeled noise channels (e.g., multimode scattering or pump-dependent losses), the serial calibration introduces systematic biases, distorting the estimated gain and added noise. Furthermore, separating the calibration of the readout chain from the internal dynamics of the DUT is difficult when the DUT is part of the calibration path.

Methodology
The authors propose a device-agnostic, in situ noise metrology protocol that decouples the readout-chain calibration from the DUT's behavior. The core of the methodology involves two main components:

1. Substitution Configuration: Instead of placing the noise source in series with the DUT, the protocol uses a cryogenic switch network to substitute the DUT with a controllable noise source. This ensures the readout chain is calibrated independently of the DUT's nonlinearities, satisfying the linearity assumption required for absolute noise calibration.
2. Variable Temperature Stage (VTS) and Planck Spectroscopy: The noise source is a custom-built VTS consisting of a matched attenuator mounted on a thermally isolated stage with a heater and thermometer. By sweeping the temperature of this stage, the system performs "Planck spectroscopy." Fitting the measured Power Spectral Density (PSD) against the theoretical thermal noise equation across multiple temperatures allows for the robust extraction of the readout chain's gain ($G_{sys}$) and noise temperature ($T_{sys}$).
3. S-Parameter Calibration: The same switch network supports a Short-Open-Load-Reciprocal (SOLR) scattering-parameter calibration. This ensures that both the absolute noise levels and the scattering parameters (such as gain) are referred to the same cryogenic reference planes ($R_{in}$ and $R_{out}$) at the DUT ports.

The workflow involves: (1) connecting the DUT, VTS, and calibration standards to the switches; (2) performing SOLR calibration; (3) measuring DUT scattering parameters; (4) measuring the DUT's output noise; (5) substituting the DUT with the VTS; (6) sweeping the VTS temperature to calibrate the readout chain; and (7) converting the DUT's measured PSD into absolute noise photon flux using the calibrated chain parameters.

Key Contributions

• Bias Mitigation: The paper demonstrates that the substitution configuration eliminates the model-dependent bias inherent in serial calibration for nonlinear devices. By keeping the DUT out of the calibration path, the method avoids assumptions about how noise propagates through unmodeled internal channels.
• Unified Reference Planes: The architecture aligns absolute noise measurements and scattering parameter measurements at the exact same cryogenic reference planes, a requirement often missed in heuristic approaches.
• Device-Agnostic Protocol: The protocol does not require a specific physical model of the DUT to perform the calibration. It provides a standardized framework for extracting observables (like input-referred added noise) regardless of the device's internal complexity.
• Experimental Validation: The authors implemented this protocol using a Josephson Traveling Wave Parametric Amplifier (JTWPA) based on SNAILs. The setup included a custom VTS and a Nonlinear Vector Network Analyzer (NVNA) for simultaneous S-parameter and PSD measurements.

Results
The protocol was applied to a JTWPA under pump conditions deliberately chosen to activate multimode nonlinear behavior, a scenario where standard models often fail.

• The system successfully extracted the gain and input-referred added noise as functions of pump power and signal frequency.
• Results showed that at higher pump powers, the added noise increased significantly faster than the gain, indicating the activation of additional pump-dependent noise channels (e.g., higher-order wave-mixing or pump-enhanced loss).
• Crucially, the extraction of these results did not rely on assuming a simplified two-mode amplifier model during the readout-chain calibration. The data revealed pump-dependent noise trends that would likely have been obscured or misinterpreted by serial calibration methods relying on simplified internal models.

Significance and Claims
The authors claim that this work provides a practical route toward the standardized, reproducible noise characterization of nonlinear cryogenic microwave devices. The significance lies in the separation of the readout-chain calibration from the DUT's internal dynamics, which allows for the reliable measurement of absolute noise metrics even when the device physics are complex or not fully understood.

The paper is modest regarding its status as a primary standard. It acknowledges that the accuracy of the current implementation depends on the thermal model of the VTS, impedance characterization, and the uncertainty budget of the Planck-spectroscopy fit. The authors state that a complete quantitative analysis of thermal gradients and traceability is a necessary next step. However, they assert that the proposed architecture supports the automated benchmarking of multiple devices within a single dilution-refrigerator platform, offering a robust alternative to model-dependent serial calibration methods.