The QICK Box: A Modular RF Front-End System for Quantum Control and Readout

AI summary

Overview

Research area: Quantum computing hardware — classical control and readout electronics for quantum information science (superconducting qubits, atomic systems, trapped ions, spin defects, quantum sensors, and quantum networks).

Technical level: Intermediate. The paper is a hardware white paper: it is readable by anyone with a general physics/engineering background, but the dense RF specifications (noise temperature, Nyquist zones, SFDR, attenuator resolution) will be most useful to specialists.

Scope (one sentence): This white paper describes the QICK Box, a modular RF front-end companion board for the AMD ZCU216 RFSoC-based QICK control system, together with its measured RF performance and a demonstration of single-shot readout of a six-transmon device at Fermilab's QUIET testbed.

What This Paper Is About

The QICK (Quantum Instrumentation Control Kit), introduced in 2022, is an open-source, FPGA-based control system built on a commercial AMD ZCU216 RFSoC board that can synthesize microwave pulses directly, without analog upconversion. That board alone, however, still requires a great deal of external room-temperature electronics — amplifiers, attenuators, filters, bias supplies — which becomes unwieldy as channel counts grow.

This paper introduces a companion "RF Main" board that plugs into the ZCU216 and supplies the missing analog front end: modular daughtercards for RF and fast DC-coupled inputs and outputs, precision DC biasing, and digital I/O. Together, the QICK board and companion board form the QICK Box. The paper's goal is to document the hardware design and to quantify its real-world performance through a qubit readout experiment.

Key Contributions

1. A modular motherboard–daughtercard front end. The companion board acts as an extension of the ZCU216, interfacing through two RFMC connectors from which it also draws power, so the whole QICK Box needs only a single external 12 V DC supply (100 W minimum). Output daughtercards carry four channels each; input daughtercards carry two, a deliberate choice to match typical quantum system needs and reduce cost, power, and footprint.

2. Six daughtercard variants covering both RF and fast DC signals. The variants are: RF outputs (100 MHz–9 GHz), DC-coupled outputs (DC–1.5 GHz), RF inputs (50 MHz–9 GHz), DC-coupled inputs (DC–1.5 GHz), balun-only outputs, and balun-only inputs. The card flavor (RF or DC) is automatically detected.

3. System-level integration of biasing, timing, and digital control. The board adds eight 20-bit DC bias channels (±10 V, 1 ppm resolution, 1 ppm integral nonlinearity, 7.8 nV/√Hz noise with a 1/f knee at 4 Hz), sixteen general-purpose digital I/O ports for markers and triggers, and support for the XCOM network, which can connect up to 16 QICK boxes in a mesh, phase-synchronize PLLs to better than 50 ps, and deliver deterministic latency of 95 ns per 32-bit word.

4. An end-to-end validation on real qubits. The authors characterize output power, linearity, spectral purity, input gain, and noise temperature, then use the QICK Box to perform single-shot dispersive readout of a six-transmon device at the QUIET testbed, reaching an SNR of 4.4 ± 0.1 and a readout fidelity of (91.7 ± 0.4) % with only a HEMT amplifier and no quantum amplifier.

Main Findings

• Output chain performance: A randomly selected RF output channel was characterized. With the attenuators set to Att₁ = 20 dB and Att₂ = 30 dB, output power linearity was measured over a 0–45 dB attenuation sweep, with residuals within 0.22 dB RMS across the 1–9 GHz range. Step attenuators provide 0.25 dB resolution per step.

• Nyquist zone switching: The DAC output power is higher in the first Nyquist zone from 0.1–5 GHz, while selecting the second Nyquist zone improves gain for 5–9 GHz. The Nyquist zone is a software parameter in QICK, and the paper recommends switching from NZ=1 to NZ=2 at or above 5 GHz.

• Spectral purity: Excluding higher-order harmonics, the RF output shows a non-harmonic Spurious-Free Dynamic Range of approximately 75 dBc (RBW = 0.02 MHz), measured on a 1.9 GHz single tone with the 2nd through 5th harmonics visible. For a −20 dBm tone at 8 GHz, the noise floor 1 GHz from the carrier is below −150 dBc/Hz.

• Power limits: Maximum output power before compression is approximately 10 dBm, decreasing at higher frequency; at 9 GHz the maximum output power is greater than −10 dBm. A −2 dB power droop is observed at higher frequencies for each of the two cascaded amplifiers, and RF trace insertion loss further reduces gain.

• Input gain: With the step attenuator at its maximum of 30 dB, the system maintains high sensitivity across 1–8 GHz; with attenuation removed, maximum gain is approximately 90 dB within 1–5 GHz. At 9 GHz, gain of both the ADC and RF amplifiers drops significantly, and maximum system gain is limited to 37 dB even at zero attenuation.

• Input noise: Measured with a 50 Ω termination at room temperature, the system holds a noise temperature of approximately 100 K (1.4 dB noise figure) between 3 and 6 GHz. Above 6 GHz the noise temperature rises because reduced first-stage gain lets later stages dominate the cascaded noise figure.

• Programmable bandpass filtering: With a bandwidth of approximately 0.4 GHz the filter adds 10 dB of insertion loss, which is accounted for in the amplification chain, but it lowers the noise floor by 20 dB at 1 GHz from the carrier. Filtering also provides image rejection of 20 dB at 1 GHz from the carrier and acts as a tracking anti-aliasing filter for the 2.4576 GSPS ADC.

• Integrated noise reduction: From 132 × 10³ samples partitioned into 50 segments of 2.64 × 10³ points each (about 1 μs of integration at 2.4576 GSPS), enabling the BPF reduced RMS noise by 7.6 dB, from 0.596 mV to 0.249 mV.

• Single-shot readout at QUIET: Using 3000 single-shot measurements per state, the state separation was Δμ = 28 ± 0.2 a.u., with clean-side widths σ_g = 6.4 ± 0.1 a.u. and σ_e = 6.5 ± 0.1 a.u. (average 6.5 ± 0.1 a.u.), giving an electronics SNR of 4.4 ± 0.1 and a readout fidelity of (91.7 ± 0.4) %.

• Fidelity is limited by T₁, not the electronics: The device's median relaxation time is T₁ = 58.3 ± 4.1 μs and the readout pulse duration is 5.6 μs, giving a T₁-limited assignment error of (4.7 ± 0.3) %, which accounts for approximately 57 % of the measured total infidelity of 8.3 %. The remaining ≈3.6 % is attributed to residual thermal population at a qubit effective temperature of ≈44 mK, imperfect state-preparation pulses, and Gaussian overlap between the distributions.

• No degradation from the new hardware: The authors observed no change in T₁ or effective qubit temperature when switching from the QICK board setup to the QICK Box setup.

Methodology in Plain English

The team built a two-board system. The first board is a commercial AMD ZCU216 RFSoC evaluation board, whose converters (sixteen 14-bit 10 GSPS DACs and sixteen 12-bit 2.5 GSPS ADCs) can synthesize and digitize microwave signals directly, skipping analog mixing stages. The new second board is a custom "RF Main" motherboard that plugs into it and holds swappable daughtercards, each of which contains the amplifiers, step attenuators, filters, and baluns needed to condition signals on the way to and from the quantum device.

To characterize the hardware, the authors used a randomly selected channel as a baseline and measured output power versus frequency, linearity versus programmed attenuation, harmonic content, input gain, and input-referred noise temperature, using a 50 Ω termination for noise measurements and an RF output channel plus fixed attenuators (keeping input power between −78 and −82 dBm) to generate test signals. The bandpass filters were set to a 1 GHz bandwidth centered on the carrier during these tests.

For the qubit demonstration, they used a six-qubit tantalum-on-sapphire transmon device with Manhattan-style Josephson junctions, fabricated at the Pritzker Nanofabrication Facility and packaged at Fermilab, inside QUIET — a 10 mK dilution refrigerator located 100 m underground to reduce cosmic-ray interference. The readout chain contained only a single HEMT amplifier at the cryogenic stage, with no TWPA or JPA. Readout fidelity was extracted from the maximum cumulative histogram contrast after rotating the I-Q data to align the axis of maximum state separation, and the SNR was extracted by fitting only the "clean" outer side of each distribution — the side uncontaminated by T₁ decay — so as not to overstate the electronic noise.

Why This Matters

Control electronics are a practical bottleneck in scaling quantum processors: every qubit needs carefully conditioned microwave lines, and racks of discrete instruments do not scale. This paper shows that a single integrated, open-source box can replace much of that discrete electronics while still delivering low-noise, high-linearity performance, and it provides an honest accounting of what limits readout fidelity (in this case the qubit's own T₁, not the room-temperature electronics).

Real-world applications:

• Superconducting qubit control and dispersive readout in academic and national-lab quantum processors.
• Trapped-ion and neutral-atom experiments, where the same pulse-generation and timing infrastructure applies.
• Spin-qubit and quantum-dot experiments, using the DC-coupled output channels for baseband voltage pulses and fast flux pulses (the paper cites fluxonium as a motivating case).
• Quantum sensing and detector readout, including direct detector readout using QICK's Time Tagger IP, plus use as an auxiliary oscilloscope or spectrum analyzer.

Industry relevance: The design builds on a commercially available AMD RFSoC part rather than custom silicon, which makes the approach reproducible and lowers the barrier for commercial labs and startups; the open-source software and firmware (with a separate forthcoming paper on firmware and software) position the QICK Box as a viable alternative to proprietary arbitrary waveform generators and digitizers.

Future Directions

• Firmware and software advances. The paper explicitly states that these will be described in a separate paper, so the full capability of the platform is not yet documented here.
• Improving T₁ to raise fidelity. The authors note that improvements in T₁ are expected to increase readout fidelity without any change to the QICK Box electronics, making the qubit, not the instrument, the limiting factor.
• Infrared shielding and filtering studies. An ongoing study at QUIET is systematically adding infrared shielding and filtering across successive cooldowns to model how each stage affects coherence times and effective qubit temperature; the authors cite prior work on this effort.
• Scaling to larger channel counts and multi-board systems. The XCOM network's support for up to 16 boxes in a mesh with sub-50 ps phase synchronization and 95 ns deterministic latency sets up multi-board, large-fanout operation that this paper characterizes but does not demonstrate at scale.

Target Audience

Quantum computing experimentalists and lab engineers who need to build or scale a control and readout stack; RF and mixed-signal hardware engineers interested in what an RFSoC-based front end can achieve; procurement and instrument-selection teams at universities, national laboratories, and quantum hardware companies; and graduate students or newcomers who want a concrete, fully specified example of how a modern qubit control chain is architected and benchmarked.

Authors’ abstract

The QICK (Quantum Instrumentation Control Kit), first introduced in 2022, is a standalone open source control system for quantum information science. Since then, QICK has been migrated to new hardware platforms, expanded its functionality for applications with different types of qubits (superconducting, atomic, ion-trapped, and spin defects), as well as quantum sensors, and quantum networks. QICK is a highly optimized control kit, with a library of functions that can be used for more than one application. In particular, the software and firmware are customized to minimize latency and maximize control performance and throughput. In this white paper, we describe hardware details and performance of the QICK Box. The QICK Box features up to 16 outputs, up to 8 inputs, and 8 20-bit DC biases. Modular output and input daughtercards give the choice of signal chains optimized for fast DC-coupled signals or RF signals up to 10 GHz. Power levels, step attenuation, and bandpass filtering are controlled by software. We demonstrate the system's performance through single-shot readout measurements of a six-transmon device at the Quantum Underground Instrumentation Experimental Testbed (QUIET) at Fermilab, achieving (91.7 +/- 0.4)% readout fidelity limited by T1 decay during the readout. The single shot readout signal-to-noise ratio (SNR) is 4.4 +/- 0.1 without a quantum amplifier, confirming that the QICK Box is suitable for high-fidelity quantum control and measurement applications. Advances in firmware and software will be described in a separate paper.