Quantum Instrumentation Control Kit QICK
The requirement for complex and accurate control systems for an ever-increasing number of qubits has become critical as the field of quantum computing quickly changes. The Quantum Instrumentation Control Kit (QICK), an open-source platform at the core of many superconducting quantum computing labs, has greatly streamlined the challenging task of qubit control. QICK, created at Fermilab, is a well-known illustration of an open-source platform intended to lower the expense and complexity of handling complicated quantum systems.
Utilising RFSoC Technology to Simplify Qubit Control Strong AMD RFSoC (Radio Frequency System-on-Chip) components form the foundation of QICK. These integrated devices enable QICK to directly produce microwave pulses for qubit manipulation by combining ARM embedded CPUs, programmable logic, and high-speed data converters. By doing away with the necessity for conventional frequency upconversion through local oscillators and IQ mixers, this direct digital synthesis capability significantly reduces the complexity of the entire control stack and makes qubit control easier.
You can also read MnBi6Te10 Semiconductor: Thinnest Junction For Quantum Tech
Three main firmware components make up the QICK architecture: readout blocks, signal generators, and a specialised timing processor (tProcessor).
- Signal Generators: There are various kinds of these parts, which are connected to the RF Digital-to-Analogue Converters (DACs):
- Full-speed generators are perfect for producing high-fidelity gates that need accurate pulse shaping with sub-nanosecond resolution, and they run at the maximum DAC sampling rate (up to 9.85 GS/s on the ZCU216 board). They employ digital carriers modified by arbitrary waveform envelopes through Direct Digital Synthesis (DDS). For accurate waveform shaping in two-qubit gates, the Manarat system uses full-speed generators specifically designed for RF flux pulses.
- By creating pulse envelopes at a reduced sample rate (1/16 of the DAC rate) and then digitally interpolating the signal, interpolated generators minimise the amount of FPGA resources used. In the Manarat system, these are utilised for drive lines. In order to enable simultaneous multitone control for frequency-multiplexed qubits utilising a single output channel, multiplexed generators digitally aggregate multiple DDS channels prior to the DAC output. In a multi-qubit arrangement, one such generator is usually assigned per board.
- Readout Blocks: QICK uses two primary types of readout blocks to handle signals received by RF Analogue-to-Digital Converters (ADCs) on the acquisition side:
- Digital downconversion (DDC) is used by standard readout blocks to convert the signal to baseband, after which it is filtered and decimated. They allow findings from several acquisitions to be combined and averaged because they are optimised for single-tone measurements and lengthy integration durations.
- A polyphase filter bank (PFB) is used by multiplexed readout blocks to demultiplex the digital signal into several subbands, each of which includes a unique DDS/NCO for tone-specific demodulation. This allows up to eight channels per ADC stream to be used for the simultaneous readout of multiple qubits. Each board in the Manarat system has a single multiplexed readout block.
- Timing Processor (tProcessor): a small, specialised CPU built inside the FPGA fabric. Low-level assembly instructions, pulse and acquisition scheduling, register manipulation, loop implementation, and conditional branching are all carried out by the processor. It ensures deterministic and cycle-accurate control by carrying out its commands in a predetermined number of cycles. Through register writes, the processor dynamically modifies waveform parameters and activates the signal generators. There are two main modes of operation that it supports:
- Real-time execution: the experiment is run entirely on the FPGA, resulting in low latency and overhead and no need for external connectivity.
- Near-real-time execution, in which Python running on the ARM processor dynamically changes some parameters, but communication overhead is kept unusually low due to the close integration between the processing system and programmable logic.
You can also read Microsoft PQC ML-KEM, ML-DSA algorithms for windows & Linux
Impact and Software Interface A simple Python software interface for creating and carrying out quantum experiments is made available by QICK. The foundation of this interface is PYNQ, an open-source project that enables programmers to use Python to program Zynq System-on-Chip (SoC) devices. Waveform data and configuration registers are uploaded with control programs, which are written in Python and compiled into a special assembly language for the tProcessor.
A number of high-impact experiments in superconducting qubit research have been made possible with in large part to QICK. Significant improvements in coherence and control were made possible by it, for example, by supporting the demonstration of transmon qubits with energy-relaxation and dephasing times exceeding 1 millisecond and single qubit gate fidelities over 99.99%. Additionally, it made it easier to create novel parts, such as the first Floquet-mode traveling-wave parametric amplifier made using a superconducting qubit method. These accomplishments highlight QICK’s performance and adaptability as a fundamental platform for qubit control.
The Most Important Problem
Synchronisation Across Boards Notwithstanding its advantages, QICK has a significant drawback that gets worse as quantum computers get bigger: it doesn’t have native support for multi-board synchronisation. Despite having a significant number of input and output channels (16 14-bit, 2.5 GSPS ADCs and 16 14-bit, 9.85 GSPS DACs), boards such as the AMD ZCU216 are not enough for mid- and large-scale quantum processors. For instance, a single RFSoC board cannot handle the resources required to manage a 10-qubit flux-tunable transmon device in a ladder arrangement since it requires 10 drive lines, 10 flux lines, and 2 readout lines.
You can also read Belenos: Quandela’s Photonic Quantum Computing Innovation
Since each board’s processor operates independently, QICK’s existing design is predicated on local execution and synchronisation. It lacks integrated support for essential features like global clock distribution, processor synchronisation, and program distribution over several boards, despite having preliminarily implemented techniques like external triggering. The implementer is in charge of features like orchestrating multi-board experiments and configuring Multi-Tile Synchronisation (MTS). Furthermore, new flux control methods are necessary because the conventional analogue front-end (XM655) that comes with the ZCU216 evaluation board is made for communications and is not appropriate for the precise control needed by superconducting qubits.
Manarat: Extending QICK for Scalable Quantum Computing
The Manarat platform was created specifically to address this basic constraint in multi-board synchronisation. Manarat incorporates substantial hardware, firmware, and software improvements and is specifically developed as a scalable control architecture based on the QICK framework. Achieving timing alignment of less than 100 picoseconds across several RFSoC boards is its main innovation. A genuinely distributed control architecture for quantum processors is made possible by this degree of precision, which is essential for coherent multi-qubit control.
Manarat is an important step in resolving the scaling issues that superconducting quantum computing is now encountering by expanding QICK’s capabilities. By directly expanding on the strong foundation offered by QICK, it confirms a viable approach for scaling pulse-level control systems to enable mid-scale quantum processors, paving the way for the creation of larger and more potent quantum computers.
You can also read λambeq Gen II: QNLP Quantum Natural Language Processing
