Metrology, not Mythology: Making Flux-Tunable Qubits Work in Practice

Guest blog

How do you get rid of noise when building real experimental quantum systems? That’s something I’ve been working on for decades, and it remains a critical challenge in building quantum hardware that actually works.

More broadly, this challenge reflects a wider industry shift toward tightly integrated quantum-classical systems, with approaches like the Open Acceleration Stack bringing real-time control and low-latency feedback into the heart of quantum computing.

And that’s one of the questions I hope the next generation of physicists will be tackling alongside today’s researchers, working on building and operating real experimental systems. That’s why we at Qolab, together with Quantum Machines, have recently launched the John Martinis Prize, which provides access to superconducting qubit devices hosted at the Israeli Quantum Computing Center for experimentalists, software researchers, and educators.

The idea is to enable new experiments, new control approaches, and new ways of working with qubits, including hands-on access for both researchers and students. Even with a small system, there is a tremendous amount of physics to explore, especially as we move from a one-qubit device to a two-qubit device with tunable couplers and tunable frequencies, and provide access to state-of-the-art single- and two-qubit gates.

But access alone isn’t enough. To actually use these systems effectively, you have to understand how to build and operate them properly.

Controlling flux-tunable qubits means controlling noise

Take flux-tunable qubits, for instance. That’s when you apply DC signals to control the system. This is not how microwave circuits are typically built, as you’re going against standard electrical engineering practices, and that creates challenges. To make these systems work, you have tocontrol the noise environment of the entire setup.

Typically, you’re dealing with flux noise at about a micro-flux quantum per root hertz, and total integrated noise on the order of about 10 microvolts. Given that your electronics may be operating at around one volt per flux quantum, that means you need to control noise at the level of tens of microvolts. That’s a very small voltage. And if you don’t manage it properly, the system won’t behave as expected. You can tolerate somewhat higher levels in practice, maybe 30 to 50 microvolts, because some effects can be mitigated with techniques like spin echo. But you still have to be careful at that level.

When you build experimental systems, you’re dealing with real electronics, real power supplies, and real safety constraints. Your equipment is grounded for safety, connected through building infrastructure, and often spread across multiple racks. This creates loops in the grounding system. Magnetic fields from power lines can drive currents through these loops, and those currents generate voltages, exactly the kind of noise you’re trying to eliminate.

If you care about tens of microvolts, this becomes a serious problem. What people generally say is that you need a single-point ground to eliminate these loops, but you have to do this carefully and safely, because the grounding system is also what protects you from electrical hazards.

So the challenge is not just eliminating noise; it’s doing so while maintaining proper safety mechanisms.

You have to measure, not assume

One of the key ideas I tend to emphasize is what we call “metrology, not mythology.” It means that you can have a theory of how your system should work, but in practice, systems are more complicated, and things often don’t behave the way you expect. So you need to measure, and you need to be precise.

At Qolab, we use a setup where we tap into the control lines and directly measure voltage noise using a low-noise amplifier and a spectrum analyzer. To avoid introducing additional ground loops, the measurement system itself is battery-powered and floating. With this approach, you can quickly see whether your system is behaving correctly. You might see noise levels on the order of a nanovolt per root hertz, integrating to around a microvolt over the relevant bandwidth. But in other cases, for example, when the dilution refrigerator is running, you may see much higher noise, with clear spectral peaks. That tells you something is wrong, and you need to fix it.

This is why measurement is essential. There are many things that can go wrong, and the only way to know is to check. That also goes for the rest of your electronics, including control, as when you connect control electronics, you introduce additional effects. For example, control systems often use high-speed transistors, which can introduce more low-frequency (1/f) noise.Even when the overall noise level is acceptable – say, on the order of tens of microvolts – you may still see narrow spectral peaks from power supplies or other components. These peaks don’t always dominate the total noise, but they matter. And they are often difficult to eliminate.

Part of our work is to identify these issues and stress-test control systems. For example, when working with Quantum Machines hardware, we actively look for these kinds of noise features and push for improvements over time. If you’re aiming for a single-point ground, you also need a way to check that there aren’t unintended connections creating additional paths. We use a simple circuit that allows you to safely test this, ensuring that your system behaves as expected without compromising safety.

Similarly, you can measure voltage offsets, typically on the order of tens of millivolts, to detect unwanted 60 Hz noise injection. These are straightforward checks, but they make a big difference in practice. This kind of detailed measurement and debugging is a core part of making flux-tunable qubits work. And this is exactly why we launched the John Martinis Prize, together with Quantum Machines.

The goal is to enable new experiments and give people direct access to real quantum hardware, so they can explore how to control qubits and run experiments across full systems, including classical control and processing. Giving researchers and students that kind of access means they can work on real problems, from control and wiring to noise and measurement, and develop the skills needed to build and operate these systems.

Because in the end, progress in quantum computing comes down to what actually works, and that means dealing with noise, wiring, and measurement in real systems.