An automatic detection method based on performance index of an optical quantum chip
By combining binary orthogonal time-domain encoded sequences and compressed sensing algorithms with real-time monitoring using a reference interferometer, the problem of the measurement of the transmission matrix of optical quantum chips being susceptible to environmental disturbances was solved, achieving high-precision and rapid detection and automated performance index extraction.
Patent Information
- Application Number
- CN202611370497.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-04
- Publication Date
- 2026-11-10
AI Technical Summary
The measurement of the transmission matrix of existing optical quantum chips is easily affected by environmental disturbances, resulting in poor repeatability and accuracy of the measurement results. Furthermore, the port-by-port scanning method is time-consuming and cannot meet the requirements for high-precision and rapid detection.
A binary orthogonal time-domain coding sequence is used to encode and excite multiple input ports. A reference interferometer is used to monitor environmental disturbances in real time and perform phase pre-compensation. The transmission matrix is reconstructed using a compressed sensing algorithm. The signal is collected by a single-pixel detector and Allan variance analysis is performed to extract performance indicators.
This approach achieves improved measurement accuracy and repeatability while reducing measurement time, and lowers hardware complexity and cost, enabling high-precision, high-speed, and fully automated detection of the performance indicators of optical quantum chips.
Smart Images

Figure CN123066535A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of optical quantum chip testing technology, specifically to an automated testing method for optical quantum chip performance indicators. Background Technology
[0002] Quantum optical chips hold significant promise for applications in quantum computing and quantum communication, and their performance metrics require evaluation through the calibration of their transfer matrices. The measurement accuracy of the transfer matrix directly impacts the acquisition of key parameters such as quantum gate fidelity and inter-channel crosstalk. Current methods for measuring the transfer matrix of quantum optical chips typically employ a port-by-port scanning approach, sequentially applying excitation signals to each input and acquiring responses at each output. This method requires multiple switching of excitation ports, and the measurement time increases linearly with the number of ports, making it difficult to meet the rapid detection requirements of quantum optical chips with a large number of ports. Furthermore, quantum optical chips are inevitably affected by global environmental disturbances such as temperature drift, mechanical vibration, and power supply ripple in their actual operating environment. These disturbances cause synchronous changes in the refractive index of waveguides on the chip, resulting in random phase shifts in the measurement results of the transfer matrix, thereby reducing the repeatability and accuracy of the measurement results. The impact of environmental disturbances is particularly pronounced in long-term or multi-batch measurements, and existing methods lack effective real-time compensation mechanisms, making it difficult to eliminate the interference of disturbances on the measurement results. Therefore, how to improve measurement accuracy and repeatability while shortening measurement time is a key technical problem that needs to be solved in the field of quantum optical chip performance testing. Summary of the Invention
[0003] To address the shortcomings of existing technologies, this invention provides an automated detection method for performance indicators based on optical quantum chips. This method solves the problems of poor repeatability caused by environmental disturbances in the measurement of the transmission matrix of optical quantum chips, and the time-consuming multi-port point-by-point scanning method, which is difficult to meet the requirements of high-precision and rapid detection.
[0004] To achieve the above objectives, the present invention provides the following technical solution: an automated detection method for performance indicators based on optical quantum chips, comprising the following steps: S1, Generate a set of binary orthogonal time-domain encoded sequences, and modulate the encoded sequences onto an optical carrier and apply them to the M input terminals of the chip; S2, monitors the chip's global environmental disturbances in real time and extracts the phase of the environmental disturbance at the current moment; S3, Apply phase pre-compensation to the excitation optical field of the M input terminals according to the phase of the environmental disturbance; S4, the signals from the N output ports of the chip are converged and sampled to obtain a one-dimensional measurement vector; S5. Based on the one-dimensional measurement vector, the environmental disturbance phase, and the pre-calibrated coupling matrix, the complex transmission matrix of the chip is reconstructed using a compressed sensing algorithm. S6. Extract the chip's performance indicators from the reconstructed complex transmission matrix. The performance indicators include static phase error, inter-channel crosstalk, dynamic phase noise statistical characteristics, and quantum gate fidelity. S7. Repeat steps S2 to S6 until the preset number of repetitions is reached. Perform Allan variance analysis on the dynamic phase noise and generate a chip performance test report.
[0005] The method of the present invention is based on the following principles: In step S1, binary orthogonal time-domain coded sequences are used to encode the excitations of the M input terminals. This sequence set has orthogonal properties, allowing the responses of different input ports to be distinguished at the output terminal through decorrelation, thus enabling simultaneous measurement of multiple input ports without port-by-port scanning. The length L of the coded sequence is a power of 2 and not less than M, ensuring that the columns of the measurement matrix are full rank, providing conditions for subsequent compressed sensing reconstruction.
[0006] In step S2, a reference interferometer integrated on the chip monitors global environmental disturbances in real time. This reference interferometer does not participate in the chip's quantum computing; it is only used to sense environmental disturbances. Since the environmental disturbances experienced by all waveguides on the chip are globally consistent, the phase changes sensed by the reference interferometer can represent the common phase disturbances of all channels on the chip. By acquiring the output optical power of the reference interferometer in real time, the phase of the environmental disturbance at the current moment is calculated, and the phase is de-ambigued to obtain a continuous phase sequence.
[0007] In step S3, phase pre-compensation is applied to the modulator array based on the environmental disturbance phase extracted in step S2. Specifically, the environmental disturbance phase is transmitted to a waveform generator, which adjusts the bias voltage of each modulator to superimpose an inverse compensation amount onto the phase of the excitation light field. This inverse compensation amount is equal in magnitude but opposite in sign to the environmental disturbance phase, thereby subtracting the influence of the environmental disturbance before the optical signal enters the chip. This closed-loop feedback control mechanism ensures that the equivalent phase of the excitation light field inside the chip is only affected by the chip's own state, eliminating the interference of environmental disturbances on the measurement results.
[0008] In step S4, N output ports of the chip are coupled via an optical fiber array, converged by a fiber bundle, and then connected to a single-pixel detector. The single-pixel detector superimposes multiple optical signals into a single composite optical signal, which is then acquired by an oscilloscope to obtain a one-dimensional measurement vector. This one-dimensional measurement vector contains the optical field information of all output ports, and the signals are distinguished in the time domain by differences in their encoded sequences. This acquisition method transforms a multi-channel measurement problem into a single-channel measurement problem, greatly reducing hardware complexity and cost.
[0009] In step S5, based on the one-dimensional measurement vector obtained in step S4, the environmental disturbance phase obtained in step S2, and the pre-calibrated coupling matrix, the complex transmission matrix of the chip is reconstructed using a compressed sensing algorithm. Specifically, this includes: constructing a measurement matrix whose elements are determined by the values of the encoded sequence; constructing a sparse transform basis using a discrete cosine transform basis to give the transmission matrix a sparse representation in the transform domain; calibrating the coupling matrix by applying a known disturbance sequence and solving using least squares to describe the linear effect of environmental disturbances on the output signal; establishing a complete measurement model, representing the one-dimensional measurement vector as the column vectorized form of the measurement matrix multiplied by the transmission matrix, plus the coupling matrix multiplied by the environmental disturbance vector, plus the measurement noise vector; solving the L1 norm optimization problem using an iterative soft thresholding algorithm to recover the sparse coefficient vector in the sparse domain; and converting the sparse coefficient vector into the column vectorized form of the transmission matrix through an inverse transform, and reshaping it into an N×M complex transmission matrix. This compressed sensing reconstruction method utilizes the sparsity of the transmission matrix in the transform domain to achieve high-precision reconstruction with far fewer sampling points than traditional methods, significantly shortening the measurement time.
[0010] In step S6, several performance indicators are automatically extracted from the reconstructed complex transmission matrix: static phase error, obtained by calculating the difference between the actual phase and the target phase of each channel, used to evaluate the phase accuracy of each channel of the chip; inter-channel crosstalk, obtained by calculating the square of the modulus of the off-diagonal elements, used to evaluate the isolation between channels; dynamic phase noise statistical characteristics, obtained by calculating the Allan variance of the phase sequence obtained from multiple repeated measurements, used to evaluate the phase stability of the chip; and quantum gate fidelity, obtained by dividing the square of the modulus of the trace of the reconstructed transmission matrix and the ideal transmission matrix by the square of the number of output ports, used to evaluate the degree of matching between the overall transmission characteristics of the chip and the design goals.
[0011] In step S7, steps S2 to S6 are repeated until the preset number of repetitions R is reached. Before each repeated measurement, step S2 is re-executed to detect environmental disturbances in real time, and step S3 is re-executed to perform phase pre-compensation, ensuring that each measurement is performed in the same equivalent environment and avoiding interference from environmental drift on long-term measurements. After reaching the preset number of repetitions, Allan variance analysis is performed on the phase sequence of each channel. The noise type is identified based on the slope of the Allan variance curve, and a detection report is generated that includes a static phase error list, crosstalk matrix, Allan variance curve, and quantum gate fidelity value. The chip is then judged to be qualified based on a preset threshold.
[0012] This invention achieves real-time sensing of environmental disturbances through an integrated reference interferometer, eliminates the impact of environmental disturbances on measurements through adaptive phase pre-compensation, enables multi-port parallel measurement through time-domain coding and single-pixel convergence, achieves rapid and high-precision reconstruction of the transmission matrix through compressed sensing algorithms, and achieves quantitative output of detection results through automatic extraction of performance indicators. The synergistic combination of these technologies enables the transmission matrix calibration and performance indicator detection of the optical quantum chip to possess the characteristics of high precision, high speed, high repeatability, and full automation.
[0013] This invention provides an automated method for detecting the performance indicators of optical quantum chips. It has the following beneficial effects: 1. This invention monitors global environmental disturbances in real time using a reference interferometer and applies reverse phase pre-compensation to the excitation light field to form closed-loop feedback control. This mechanism dynamically eliminates the influence of environmental disturbances on the measurement results, and significantly reduces the degree to which the measurement accuracy is affected by interference factors such as ambient temperature drift and mechanical vibration, thereby improving the repeatability and accuracy of the transmission matrix measurement results.
[0014] 2. This invention uses a binary orthogonal time-domain coding sequence to simultaneously excite multiple input ports, combined with a single-pixel detector convergence and compressed sensing reconstruction algorithm, to complete the reconstruction of the transmission matrix with fewer sampling points than the traditional port-by-port scanning method. This method shortens the time of a single measurement and reduces the complexity and cost of the hardware system.
[0015] 3. This invention automatically extracts static phase error, inter-channel crosstalk, dynamic phase noise statistical characteristics, and quantum gate fidelity from the reconstructed transmission matrix, and identifies noise type through repeated measurements and Allan variance analysis. This process requires no manual intervention, and the test results are output in the form of data files, making it suitable for batch automated testing of optical quantum chips. Attached Figure Description
[0016] Figure 1 This is a schematic diagram of the system architecture of the present invention; Figure 2 This is a flowchart illustrating the overall workflow of the present invention; Figure 3 This is a flowchart of the environmental disturbance monitoring and phase extraction process of the present invention; Figure 4 This is a flowchart of the adaptive phase pre-compensation control of the present invention; Figure 5 This is a flowchart of the multiplexed signal acquisition and compression sampling process of the present invention; Figure 6 This is a flowchart of the compressed sensing transmission matrix reconstruction process of the present invention; Figure 7 This is a flowchart of the automatic extraction process of performance indicators for this invention; Figure 8 This is a flowchart of the repeated measurement and dynamic characteristic analysis of the present invention. Detailed Implementation
[0017] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0018] See attached document Figure 1 , Figure 1 This is a schematic diagram of the overall architecture of an automated testing system for the performance indicators of a quantum optical chip according to an embodiment of the present invention. The present invention provides a rapid calibration system for the transmission matrix of a quantum optical chip based on adaptive phase compensation and time-domain compressed sensing. The system includes: an excitation module, a disturbance sensing module, an acquisition module, and a processing unit.
[0019] The excitation module includes a waveform generator and a modulator array. The waveform generator generates a set of binary orthogonal time-domain coded sequences of length L. The modulator array contains M modulators, each corresponding to an input of the chip, used to modulate the coded sequences onto an optical carrier. The disturbance sensing module includes a reference interferometer and a reference detector. The reference interferometer is integrated on the chip; its inputs and outputs do not participate in quantum computing, but are only used to sense global environmental disturbances. The reference detector is connected to the output of the reference interferometer for real-time monitoring of its output optical power. The acquisition module includes an optical fiber array, an optical fiber bundle, a single-pixel detector, and an oscilloscope. The optical fiber array is coupled to N output ports of the chip, converged by the optical fiber bundle, and connected to the single-pixel detector. The oscilloscope is used to acquire the converted electrical signals. The processing unit is connected to the waveform generator, the reference detector, and the oscilloscope, respectively, and is used to perform environmental disturbance feedback control, compressed sensing reconstruction algorithms, and performance index extraction.
[0020] This invention also provides an automated method for detecting the performance indicators of a quantum chip, comprising the following steps: S1, the processing unit initializes the system, sends instructions to the waveform generator to generate a set of binary orthogonal time-domain encoded sequences of length L, and sets the initial bias voltage of the reference interferometer.
[0021] S2, the disturbance sensing module monitors the output optical power of the reference interferometer in real time, and the processing unit calculates the environmental disturbance phase at the current moment according to formula (1). .
[0022] S3, the processing unit will process the environmental disturbance phase. The feedback signal is sent to the waveform generator, which adjusts the bias voltage of each modulator in the modulator array according to the disturbance signal, so that a reverse phase pre-compensation amount is superimposed on the optical carrier. This allows for phase pre-compensation of the excitation optical fields at the M input terminals.
[0023] S4, the acquisition module performs single-pixel convergence and compressed sampling on the signals of the N output ports of the chip, the oscilloscope acquires the one-dimensional measurement vector y, and transmits it to the processing unit.
[0024] S5, the processing unit is based on the measurement vector y and the environmental disturbance vector. And the pre-calibrated coupling matrix C, and the optimization problem shown in formula (7) is solved by using the iterative soft thresholding algorithm to reconstruct the complex transmission matrix Ĥ of the chip.
[0025] S6, the processing unit automatically extracts static phase error, inter-channel crosstalk, dynamic phase noise statistical characteristics and quantum gate fidelity from the reconstructed transmission matrix Ĥ according to formulas (9) to (12).
[0026] S7, the processing unit determines whether the preset number of repeated measurements R has been reached. If not, it returns to step S2 and repeats steps S2 to S6 until the preset number of repetitions is reached. If the preset number of repetitions has been reached, Allan variance analysis is performed on the dynamic phase noise to generate the final chip performance test report.
[0027] See attached document Figure 2 , Figure 2 This is a general flowchart according to an embodiment of the present invention. The automated detection method for optical quantum chip performance indicators provided by the present invention performs initialization and encoding sequence generation in step S1.
[0028] S101, the waveform generator generates a set of binary orthogonal time-domain encoded sequences of length L. This set of sequences uses Walsh-Hadamard sequences. The generation of Walsh-Hadamard sequences is based on the recursive construction of the Hadamard matrix. The Hadamard matrix H_{2k} is constructed recursively as follows: H_{2k}=[H_k,H_k;H_k,-H_k], where H_1=[1]. Each row of the matrix is an independent sequence with a length of L, where L is a power of 2. The cross-correlation function between sequences is zero, i.e., orthogonality is satisfied. In practical systems, due to device non-ideals, the cross-correlation peak may deviate from zero, but through design, the cross-correlation peak does not exceed 1% of the sequence length. This set of sequences is used to distinguish the excitation signals of different input ports, ensuring that the response of each channel can be recovered at the output through decorrelation.
[0029] S102, the modulator array contains M high-speed electro-optic intensity modulators, each corresponding to one input terminal of the chip. The electrical signal output from the waveform generator drives the modulator to modulate the coded sequence onto the optical carrier. The optical carrier is provided by a continuous-wave laser with power P_0, angular frequency ω_0, and initial phase φ_0. The modulation method is intensity modulation; when the coded sequence value is "1", the modulator output optical power is P_0; when the coded sequence value is "0", the output optical power approaches zero. The modulator extinction ratio is not less than 20dB to ensure the distinguishability of the coded sequence.
[0030] S103, the processing unit sets the initial bias voltage of the reference interferometer. The reference interferometer is a Mach-Zehnder interferometer structure, containing two directional couplers and two waveguide arms, one of which integrates a phase shifter for adjusting the bias point. The initial bias voltage causes the reference interferometer to operate at the quadrature bias point. The quadrature bias point corresponds to the midpoint of the linear region of the interferometer's output power versus phase relationship, i.e., the output power is ( + At point ) / 2, this bias point maximizes the rate of change of the output power phase with respect to environmental disturbances, dP / d\phi, thereby improving phase detection sensitivity.
[0031] S104, the length L of the encoded sequence and the number of input terminals M satisfy L ≥ qM. When L = M, the sequence set forms a full-rank orthogonal matrix, achieving interference-free multiplexing. When L > M, redundancy is provided by selecting a subset of the orthogonal matrix to enhance anti-interference capability. Specifically, L takes the value of a power of 2 and is not less than M. For example, if M = 8, then L takes the value of 8 or 16. The sequence set is generated recursively, and the generation method is a well-known technique in this field, which will not be elaborated here. See attached document Figure 3 , Figure 3 This is a general flowchart according to an embodiment of the present invention. The automated detection method for the performance indicators of optical quantum chips provided by the present invention performs real-time monitoring of environmental disturbances and phase extraction in step S2.
[0032] The S201 reference interferometer is integrated on the chip. The reference interferometer is a Mach-Zehnder interferometer structure, containing two directional couplers and two waveguide arms, one of which integrates a phase shifter. The phase shifter, either thermo-optical or electro-optical, is used to adjust the optical path difference between the two arms of the interferometer. The input port of the reference interferometer is connected to the chip's input optical path to receive continuous wave light signals from the laser. The output port of the reference interferometer is connected to a reference detector. Both output ports of the reference interferometer are monitoring ports and do not participate in the chip's quantum computing operations. The reference interferometer is used to sense global environmental disturbances such as temperature drift, mechanical vibration, and power supply ripple. When an environmental disturbance occurs, the refractive index of all waveguides on the chip changes synchronously, causing a change in the optical path difference of light propagating between the two arms of the reference interferometer, thereby altering its output optical power.
[0033] S202, the reference detector is an InGaAs PIN photodiode, whose response wavelength covers the chip's operating wavelength, with a bandwidth of at least 1 GHz. The reference detector converts the output optical signal of the reference interferometer into an electrical signal, which is acquired in real time by the processing unit at a sampling rate of at least 100 kHz. The processing unit records the output optical power value corresponding to each time t. . The unit is milliwatt.
[0034] S203, Pre-calibrate the interference fringe parameters of the reference interferometer. The calibration process is as follows: The processing unit applies a scanning voltage to the phase shifter of the reference interferometer, with the scanning voltage range from 0V to the half-wave voltage V_pi of the phase shifter, and the scanning step size not exceeding 0.01V. During the scanning process, the reference detector records the output optical power corresponding to each voltage value, obtaining a power-voltage curve. Since the output optical power of the Mach-Zehnder interferometer has a cosine relationship with the phase difference, the power-voltage curve exhibits a cosine waveform. The processing unit extracts the maximum optical power value from this curve. and minimum optical power value . and The unit is milliwatts. The standard for successful calibration is: the values obtained from two consecutive calibrations... and The relative changes were all less than 1%. After calibration, and These parameters are stored as fixed parameters in the processing unit for subsequent phase extraction calculations.
[0035] S204, the processing unit calculates the environmental disturbance phase at the current moment according to formula (1). : ; In the formula: Let φ_ref(t) be the output optical power of the reference interferometer acquired at time t, and P_max and P_min be the maximum and minimum optical power of the interference fringes obtained in step S203. The unit of φ_ref(t) is radians. This formula is derived based on the cosine relationship between the output power of the Mach-Zehnder interferometer and the phase difference. The derivation process is well known in the art and will not be elaborated here.
[0036] S205, Processing unit to Phase deambiguation is performed. Since the output range of the arccos function is [0, π], and the phase of actual environmental disturbances may exceed this range or change continuously, directly using the arccos function output value will cause phase jumps. The specific method for deambiguation is as follows: set the initial phase value to 0, and for each subsequent time t, calculate the phase difference Δ between adjacent time points. = - ,in This is the original phase value obtained directly according to formula (1). If Δ If Δ > π, then a positive ambiguity has occurred in the phase, and the current phase value is subtracted by 2π; if Δ If the value is less than or equal to π, then a negative phase ambiguity has occurred, and the current phase value is increased by 2π. Through this process, a continuous, non-jumping phase sequence is obtained. The processing unit will The output is sent to the subsequent step S3 in real time.
[0037] See attached document Figure 4 , Figure 4 This is a general flowchart according to an embodiment of the present invention. In step S3, the automated detection method for the performance indicators of optical quantum chips provided by the present invention performs adaptive phase pre-compensation control.
[0038] S301, the processing unit processes the continuous phase sequence calculated in step S2. Real-time data is transmitted to the waveform generator. The processing unit and the waveform generator are connected via a high-speed data interface, either PXIe or PCIe, with a data transmission rate of at least 1Gbps. The waveform generator integrates a digital signal processing module for receiving and processing data. The data processing forms a closed-loop control circuit: the reference interferometer senses environmental disturbances, the reference detector converts optical power into electrical signals, the processing unit calculates the phase and sends it to the waveform generator, and the waveform generator adjusts the output of the modulator array, ultimately compensating for environmental disturbances in the phase of the excitation light field. The control period of this closed-loop circuit, i.e., the time interval from phase sensing to the completion of compensation, does not exceed 10 microseconds. The control algorithm uses a proportional-integral controller, and the proportional and integral coefficients are determined through system identification experiments to ensure stable closed-loop response without overshoot.
[0039] S302, the waveform generator, based on the received... The value is used to adjust the bias voltage of each modulator in the modulator array. The modulator array contains M high-speed electro-optic intensity modulators, each corresponding to an independent drive channel. The waveform generator outputs two electrical signals for each drive channel: one is an coded sequence signal S. i(t) The other path is the bias voltage signal V. bias i(t). Bias voltage signal V bias i(t) and the encoded sequence signal S i(t) After being superimposed, they jointly drive the modulator. Bias voltage V biasi(t) The setting method is: V bias i(t) = V bias 0+ΔV bias (t), where V bias0 The initial bias voltage is ΔV. bias(t) For compensation voltage. Compensation voltage ΔV bias (t) and phase compensation amount − The relationship is determined through pre-calibration.
[0040] S303, the conversion relationship between bias voltage and phase offset is determined through pre-calibration. The calibration method is as follows: disconnect the coded sequence signal and apply only a DC bias voltage to the modulator. The processing unit controls the waveform generator to output a known DC voltage sequence, with the voltage value gradually increasing from 0V to the modulator half-wave voltage V. pi The step size is no greater than 0.01V. At each voltage value, the output optical power is measured using an interferometer structure, and the corresponding phase shift is recorded. After recording the data points of voltage and phase shift, the voltage-phase conversion coefficient K is obtained using a linear fitting method. phi The unit is radians per volt. K phi The calculation formula is: K phi=(Δφ) / (ΔV), where ΔV is the voltage change and Δφ is the corresponding phase change. The compensation voltage ΔV_bias(t) is determined according to the following relationship: ΔV_bias(t) = φ_ref(t) / K_phi, where φ_ref(t) is the phase of the environmental disturbance calculated in step S204. Since the compensation amount is inverted, the actual applied compensation voltage is −ΔV_bias(t). This calibration process is performed once during system initialization. If the temperature change of the modulator's operating environment exceeds a preset threshold, the calibration is re-executed.
[0041] S304, the matching relationship between the compensation bandwidth and the disturbance frequency satisfies the following condition: the compensation bandwidth is at least 5 times the disturbance frequency. The compensation bandwidth is determined by the bandwidth of the closed-loop control loop, including: the response bandwidth of the reference detector, the calculation processing rate of the processing unit, the communication rate of the data transmission interface, and the output update rate of the waveform generator. In this embodiment, the bandwidth of the reference detector is not less than 1 GHz, the calculation cycle of the processing unit is 1 microsecond, the data interface transmission rate is 1 Gbps, and the output update rate of the waveform generator is 100 MHz. The above parameters together determine that the compensation bandwidth is not less than 100 kHz. The main frequency components of environmental disturbances are concentrated in the range of 0 Hz to 1 kHz, including temperature drift (0.1 Hz to 10 Hz), mechanical vibration (10 Hz to 500 Hz), and power supply ripple (50 Hz to 100 Hz). The compensation bandwidth of 100 kHz is much higher than the upper limit of the disturbance frequency of 1 kHz, meeting the requirement of more than 5 times, ensuring sufficient suppression of the above-mentioned disturbance signals.
[0042] S305, after pre-compensation, the actual complex amplitude of the excitation optical field at the i-th input terminal is expressed by formula (2): ; In the formula: This represents the output optical power of a continuous-wave laser, measured in milliwatts. The value of the encoded sequence of the i-th input terminal at time t is 0 or 1; The environmental disturbance phase calculated in step S204 is expressed in radians, with a negative sign indicating reverse compensation. ω is the angular frequency of the optical carrier wave, expressed in radians per second. Let be the initial phase of the optical carrier, in radians. Equation (2) shows that, under environmental disturbance phase... Under the influence of this, the excitation light field pre-deducts the disturbance component, so that after the light signal enters the chip, its equivalent phase is only affected by the chip's own state and does not contain environmental disturbance components.
[0043] See attached document Figure 5 , Figure 5This is a general flowchart according to an embodiment of the present invention. The automated detection method for the performance indicators of optical quantum chips provided by the present invention performs multiplexed signal acquisition and compressed sampling in step S4.
[0044] In the S401, the chip's N output ports are coupled via a fiber optic array. The array consists of N single-mode fibers, each with a core diameter matched to the waveguide mode field diameter of its output port, ensuring a coupling efficiency of at least 90%. The input ends of the fiber optic array are aligned with the chip's output ports using end-face couplers or grating couplers, with alignment precision controlled within one micrometer. The output ends of the fiber optic array are converged via a fiber bundle. This bundle is made of multimode fiber with a core diameter of at least 50 micrometers and a numerical aperture of at least 0.22. The multimode fiber bundle combines N optical signals into a single output signal. During this combining process, the optical power of each signal is superimposed within the fiber bundle, forming a composite optical signal. This composite signal contains the optical field information of all output ports, and the individual signals are distinguished in the time domain by differences in their encoded sequences.
[0045] S402, the output end of the fiber optic bundle is connected to a single-pixel high-speed photodetector 33. The single-pixel high-speed photodetector 33 is an InGaAs PIN photodiode with a response wavelength covering 1300 nm to 1600 nm, matching the chip's operating wavelength. The detector's bandwidth is no less than 1 GHz, meeting the requirement for the code sequence symbol width. The code sequence symbol width is 1 nanosecond, corresponding to a bandwidth requirement of 1 GHz. The detector's responsivity is no less than 0.9 A / W, and the output current is no less than 0.9 mA when the chip's output optical power is 1 milliwatt. The detector's dark current does not exceed 1 nanoamp to avoid interference with weak signals. The detector converts the optical signal into an electrical signal, the amplitude of which is proportional to the optical power. The converted electrical signal is transmitted to an oscilloscope through an impedance matching circuit.
[0046] The S403 oscilloscope digitizes electrical signals. It is a high-speed digital storage oscilloscope with an analog bandwidth of at least 1 GHz, a sampling rate of at least 2.5 GSa / s, and a quantization bit depth of at least 8 bits. The oscilloscope has T sampling points, which is equal to the length L of the encoded sequence (T = L). The oscilloscope is externally triggered, with the trigger signal provided by a waveform generator, and the trigger time is synchronized with the start time of the encoded sequence. The data acquired by the oscilloscope is a series of discrete voltage values, each corresponding to a sampling time. This discrete voltage sequence constitutes a one-dimensional measurement vector y with dimensions T × 1. Each element y(t) of the measurement vector y represents the output voltage value at time t, which is proportional to the result of the superposition of the light field at the chip's output port.
[0047] S404, the oscilloscope transmits the acquired one-dimensional measurement vector y to the processing unit. The oscilloscope and processing unit are connected via a data interface, which can be either USB 3.0 or Ethernet. Data transmission uses a binary format to reduce data size. After receiving the measurement vector y, the processing unit stores it in memory for subsequent transfer matrix reconstruction calculations in step S5. The mathematical representation of the measurement vector y is: ; In the formula: A∈CT×(MN) is the measurement matrix, vec(H)∈CMN is the column vectorized form of the transmission matrix, and C∈CT×T is the coupling matrix. ∈CT represents the environmental disturbance vector, and n∈CT represents the measurement noise vector. A detailed explanation of this formula is given in step S5.
[0048] See attached document Figure 6 , Figure 6 This is a general flowchart according to an embodiment of the present invention. The automated detection method for optical quantum chip performance indicators provided by the present invention performs compressed sensing transmission matrix reconstruction based on structural priors in step S5.
[0049] S501, the processing unit constructs a measurement matrix A. The dimension of measurement matrix A is T×(MN), where T is the number of sampling points, M is the number of input ports, and N is the number of output ports. The specific construction method of measurement matrix A is as follows: the values of the encoded sequence corresponding to each time t are arranged in a fixed port order in a row. Let the value of the encoded sequence of the j-th input port at time t be... Then the output port index i corresponding to that moment ranges from 1 to N. The t-th row and (...)-th column of matrix A... The elements of column (4) are determined by formula (4): ; In the formula, The sampling time number. Let be the output port index and j be the input port index. Each row of matrix A corresponds to a sampling time, and each column corresponds to a specific input-output port pair. The number of rows T of matrix A is equal to the length L of the encoded sequence, and the number of columns MN is equal to the product of the number of input ports and the number of output ports.
[0050] S502, the processing unit constructs a sparse transform basis Ψ. The sparse transform basis Ψ adopts a discrete cosine transform basis with a dimension of MN×MN. Each element of the discrete cosine transform basis is determined by formula (6): ; Where k is the row index and l is the column index. Discrete cosine transform basis is a well-known technique in this field, and its properties and construction methods will not be elaborated here. A sparse transform basis Ψ is used to map the column vectorized form of the transfer matrix vec(H) to the sparse domain, i.e., there exists a sparse coefficient vector θ such that vec(H) = Ψθ. In the sparse domain, the absolute values of most elements in θ are close to zero, with only a few elements having large amplitude values. This sparsity stems from the structural characteristics of the chip transfer matrix in the phase and frequency domains; for example, the phase changes between adjacent channels are usually gradual, causing energy to concentrate in the low-frequency components in the transform domain.
[0051] S503, the processing unit calibrates the coupling matrix C. The coupling matrix C has dimensions T×T and describes the linear effect of environmental disturbances on the output signal at each sampling time. The calibration method is as follows: During the initialization process in step S1, a known disturbance sequence φ_ref_cal is applied to the reference interferometer. This disturbance sequence is a pre-generated pseudo-random sequence with a length equal to T and an amplitude range covering the linear response region of the reference interferometer. Simultaneously with applying the disturbance sequence, the responses of each output port of the chip are recorded to obtain the calibration measurement vector y_cal. The coupling matrix C is obtained by solving a least-squares problem: ; In the formula: The measurement vector is calibrated to have a dimension of T×1; Given a known perturbation sequence with dimension T×1; (·) + This represents a pseudo-inverse operation. The calibration is performed once during system initialization, and the calibration result C is stored in the processing unit. For chips produced in the same batch, since the chip layout and process parameters are consistent, the coupling matrix C can be reused, eliminating the need for recalibration for each measurement.
[0052] S504, the processing unit establishes a complete measurement model. The complete measurement model is represented by formula (3): ; In the formula: y is the one-dimensional measurement vector acquired in step S404, with a dimension of T×1; A is the measurement matrix constructed in step S501, with a dimension of T×(MN); vec(H) is the column vectorized form of the transmission matrix H, with a dimension of MN×1, where H is an N×M complex matrix representing the transmission characteristics of the chip; C is the coupling matrix calibrated in step S503, with a dimension of T×T; φ_ref is the environmental disturbance vector, composed of the continuous phase sequence φ_ref(t) obtained in step S205, with a dimension of T×1; n is the measurement noise vector, with a dimension of T×1, assumed to be zero-mean complex Gaussian white noise, and its covariance matrix is σ. 2 I_T, where σ 2 Let I_T be the noise variance, and let I_T be the T×T identity matrix.
[0053] S505, the processing unit constructs a compressed sensing reconstruction problem. The reconstruction problem is achieved by solving a constrained L1 norm optimization problem, the form of which is shown in formula (7): ; In the formula: θ is a sparse coefficient vector with dimension MN×1, satisfying vec(H) = Ψθ; Ψ is the discrete cosine transform basis constructed in step S502; ‖θ‖1 is the L1 norm of θ, used to promote sparsity; ‖·‖2 2 ε is the square of the L2 norm; ε is the upper bound of the noise energy, which takes the value ε = T·σ 2 , where σ 2 The noise variance is estimated by measuring the output noise level when there is no signal input.
[0054] S506, the processing unit uses an iterative soft thresholding algorithm to solve the optimization problem shown in formula (7). The iterative formula for the iterative soft thresholding algorithm is shown in formula (8): ; Where k is the iteration number. (k) is the estimated value of the sparse coefficient vector in the k-th iteration; (·) is the soft threshold function, defined as follows: ; In the formula: It can be a real number or a complex number. return The symbol, | |Return The absolute value, This is a regularization parameter that controls the threshold value. The step size μ is set to μ = 1 / ‖A‖_F 2 Where ||A||_F is the Frobenius norm of the measurement matrix A, and its calculation formula is ||A||F=sqrt(Σ|A{ij}| 2 The regularization parameter λ is determined through cross-validation. Specifically, the measurement vector y is divided into a training set and a validation set. An iterative soft thresholding algorithm is then performed on a series of candidate λ values, and the λ value that minimizes the reconstruction error of the validation set is selected.
[0055] S507, the processing unit sets the iteration stopping criteria. The iteration stopping criteria include two conditions, and iteration stops when either condition is met: Condition 1 is reaching the maximum number of iterations, which is set to 1000; Condition 2 is that the relative change between two adjacent iterations is less than a preset threshold, and the formula for calculating the relative change is |||. (k+1)- (k)‖2 / ‖ (k)‖2, the preset threshold is set to 10 -6 .
[0056] S508, after iterative convergence, the processing unit recovers the transfer matrix through inverse transformation. First, the column vectorized form of the transfer matrix is recovered from the sparse coefficient vector θ: ; In the formula: The discrete cosine transform basis constructed in step S502, This is the sparse coefficient vector after convergence in step S506. Then, vec(Ĥ) is reshaped into an N×M complex matrix Ĥ. The reshaping method is as follows: the first N elements of vec(Ĥ) form the first column of Ĥ, the next N elements form the second column, and so on, with the last column consisting of the last N elements of vec(Ĥ). The reshaped matrix Ĥ is the reconstructed chip complex transmission matrix, which contains all transmission coefficients between the input and output ports, including amplitude and phase information.
[0057] See attached document Figure 7 , Figure 7 This is a general flowchart according to an embodiment of the present invention. The automated detection method for performance indicators of optical quantum chips provided by the present invention performs automatic extraction of performance indicators in step S6.
[0058] S601, the processing unit calculates the static phase error. The static phase error represents the deviation between the actual phase and the target phase of each channel of the chip. For the reconstructed transmission matrix Ĥ, the element in the i-th row and i-th column is denoted as... , representing the transmission coefficient from the i-th input channel to the i-th output channel, where i ranges from 1 to min(M,N). Static phase error. Calculate according to formula (9): ; In the formula: arg(·) is the operation of taking the principal argument of a complex number, and the range of the principal argument is (-π, π]. The target phase value for the i-th channel is given in the chip's design file and is expressed in radians. The target phase value in the design file is determined based on the chip's quantum computing algorithm. For example, for a specific quantum gate operation, the phase difference between different channels should satisfy a specific relative relationship.
[0059] S602, the processing unit calculates inter-channel crosstalk. Inter-channel crosstalk represents the degree of non-ideal coupling between different channels. For the reconstructed transmission matrix Ĥ, the element in the i-th row and j-th column is denoted as... , where i≠j, represents the crosstalk transmission coefficient from the j-th input channel to the i-th output channel. (Channel crosstalk) Calculate according to formula (10): ; In the formula: |·| 2 Represents the square of the modulus of a complex number. This is a dimensionless power ratio. If expressed in dB, then... (dB)=10log10( The processing unit organizes the calculated crosstalk values into a crosstalk matrix with dimensions N×M and diagonal elements set to zero. The crosstalk matrix is used to evaluate the isolation between channels of the chip; the higher the isolation, the smaller the crosstalk value.
[0060] S603, the processing unit calculates the statistical characteristics of dynamic phase noise. Dynamic phase noise is caused by environmental disturbances and manifests as a phase change over time. After repeating steps S2 to S6, a total of R measurement results are obtained, with each measurement yielding a transmission matrix Ĥ. For the i-th channel, the phase value is extracted from the transmission matrix obtained in the r-th measurement. =arg( (r)), where r=1,...,R. The processing unit calculates the Allan variance of the phase sequence, which is calculated according to formula (11): ; Where τ is the average time in seconds, and the value of τ is determined by the sampling period, which is the sampling interval of the oscilloscope in step S403; R is the number of repeated measurements. Let be the phase value of the i-th channel in the r-th measurement, in radians. Allan variance. The unit is square radians. The processing unit calculates the Allan variance for different values of τ, and obtains... The curve shows the variation of noise with τ. By analyzing the slope of this curve, the noise type can be identified: a slope of -1 corresponds to white noise, a slope of 0 corresponds to random walk noise, and a slope of 0.5 corresponds to flicker noise. The above noise type identification method is a well-known technique in this field and will not be elaborated further here.
[0061] S604, the processing unit calculates the quantum gate fidelity. The quantum gate fidelity is used to measure the degree of matching between the actual transmission characteristics and the ideal transmission characteristics of the chip. For the reconstructed transmission matrix Ĥ, its dimension is N×M; the dimension of the ideal transmission matrix H_ideal is also N×M, which is determined by the chip's design topology and the target quantum gate operation. Its elements are known complex numbers, representing the transmission coefficients between each input port and each output port under ideal conditions. The quantum gate fidelity F is calculated according to formula (12): ; Where Tr(·) is the trace of the matrix, i.e., the sum of the diagonal elements of the matrix; Ĥ† is the conjugate transpose of the reconstructed transmission matrix Ĥ; For an ideal transmission matrix; N is the number of output ports; |·| 2 This represents the square of the modulus of the complex number. The fidelity F ranges from 0 to 1. The closer F is to 1, the closer the actual transmission characteristics of the chip are to the ideal transmission characteristics, and the better the chip performance.
[0062] In step S605, the processing unit summarizes the calculated static phase error, inter-channel crosstalk, Allan variance, and quantum gate fidelity, and generates a corresponding data file. The static phase error is stored in list form, containing the error value for each channel. The inter-channel crosstalk is stored in matrix form, containing the crosstalk values for all off-diagonal elements. The Allan variance is stored as curve data, containing the variance values for different τ values. The quantum gate fidelity is stored as a single numerical value. This data file serves as the basis for generating the detection report in the subsequent step S7.
[0063] See attached document Figure 8 , Figure 8 This is a general flowchart according to an embodiment of the present invention. The automated detection method for the performance indicators of optical quantum chips provided by the present invention performs repeated measurements and dynamic characteristic analysis in step S7.
[0064] S701, the processing unit determines whether the preset number of repeated measurements R has been reached. The value of R is set according to the required dynamic noise analysis accuracy, and the value ranges from 100 to 10000. If a higher Allan variance confidence level is required, a larger value is used, such as 10000 times; if only rapid detection is required, a smaller value is used, such as 100 times. The value of R is set by the user through the host computer software during system initialization and stored in the non-volatile memory of the processing unit.
[0065] S702, if the current number of repeated measurements has not reached the preset number R, the processing unit control flow returns to step S2, and steps S2 to S6 are re-executed. The time interval between each repeated measurement is determined by the sampling period, which is between 1 / 100 of a second and 1 second, with the specific value set by the user according to the rate of change of environmental disturbances. Before each repeated measurement, the processing unit re-executes the environmental disturbance sensing and phase extraction in step S2, and the adaptive phase pre-compensation control in step S3. This closed-loop feedback process ensures that the excitation light field undergoes phase pre-compensation during each measurement, eliminating the impact of environmental disturbances accumulating over time on the measurement results and avoiding interference from environmental drift on long-term measurements. After each repeated measurement is completed, the processing unit stores the transfer matrix K obtained from this measurement in memory for subsequent Allan variance analysis.
[0066] In S703, if the current number of repeated measurements has reached the preset number R, the processing unit stops the loop. The processing unit performs Allan variance analysis on the phase sequence of each channel. For the i-th channel, the phase sequence is extracted from R measurement results , ,..., , where =arg( (r)), which represents the phase value of the i-th channel in the r-th measurement. The Allan variance The calculation formula has been given in step S603 and will not be repeated here. The processing unit calculates the Allan variance under different values of the averaging time τ, and plots the variation curve with τ (on a double logarithmic coordinate). By measuring the slope of this curve, the noise type is identified: the noise type corresponding to a slope of -1 is white noise, the noise type corresponding to a slope of 0 is random walk noise, and the noise type corresponding to a slope of 0.5 is flicker noise. The above noise type identification method is a well-known technology in the art, and will not be elaborated here.
[0067] In S704, the processing unit generates a chip performance test report. The test report includes the following contents: a list of static phase errors of each channel, which lists the value of each channel in the form of a table, with the unit being degree or radian; a crosstalk matrix heat map, which presents the dB value in the form of a grayscale map or a color map, to visually show the coupling degree between channels; the Allan variance curve of each channel, which shows the relationship with τ in the form of a double logarithmic coordinate graph, and marks the noise type; the quantum gate fidelity value, which is represented by a single scalar F with a value ranging from 0 to 1.
[0068] In S705, the processing unit performs a qualification judgment on the chip performance according to preset thresholds. The preset thresholds include: the maximum allowable value of static phase error, for example ±5 degrees; the maximum allowable value of crosstalk, for example −20dB; the maximum allowable value of Allan variance at a specific τ value, for example 10 -4 square radians; the minimum allowable value of quantum gate fidelity, for example 0.99. If all indicators meet the threshold requirements, the chip is judged to be qualified; if any indicator exceeds the threshold, the chip is judged to be unqualified. The judgment result is written into the test report. The processing unit outputs the test report in PDF or CSV format, and stores it locally or uploads it to a server.
Claims
1. An automated detection method for performance indicators based on optical quantum chips, characterized in that, Includes the following steps: S1, Generate a set of binary orthogonal time-domain encoded sequences, and modulate the encoded sequences onto an optical carrier and apply them to the M input terminals of the chip; S2, monitors the chip's global environmental disturbances in real time and extracts the phase of the environmental disturbance at the current moment; S3, Apply phase pre-compensation to the excitation optical field of the M input terminals according to the phase of the environmental disturbance; S4, the signals from the N output ports of the chip are converged and sampled to obtain a one-dimensional measurement vector; S5. Based on the one-dimensional measurement vector, the environmental disturbance phase, and the pre-calibrated coupling matrix, the complex transmission matrix of the chip is reconstructed using a compressed sensing algorithm. S6. Extract the chip's performance indicators from the reconstructed complex transmission matrix. The performance indicators include static phase error, inter-channel crosstalk, dynamic phase noise statistical characteristics, and quantum gate fidelity. S7. Repeat steps S2 to S6 until the preset number of repetitions is reached. Perform Allan variance analysis on the dynamic phase noise and generate a chip performance test report.
2. The automated detection method for performance indicators based on optical quantum chips according to claim 1, characterized in that, In step S1, the binary orthogonal time-domain encoded sequence adopts the Walsh-Hadamard sequence, the sequence length L is a power of 2 and not less than M, and the cross-correlation function between the sequences is zero.
3. The automated detection method for performance indicators based on optical quantum chips according to claim 1, characterized in that, In step S2, the global environmental disturbance is monitored in real time by a reference interferometer integrated on the chip. The phase of the environmental disturbance is calculated based on the output optical power of the reference interferometer, and the phase defuzzification process is performed on the environmental disturbance phase to obtain a continuous phase sequence.
4. The automated detection method for performance indicators based on optical quantum chips according to claim 1, characterized in that, In step S3, the phase of the environmental disturbance is transmitted to the waveform generator, which adjusts the bias voltage of each modulator in the modulator array according to the phase of the environmental disturbance, so that the phase of the excitation light field is superimposed with the reverse compensation amount, thereby realizing closed-loop feedback compensation for the environmental disturbance.
5. The automated detection method for performance indicators based on optical quantum chips according to claim 1, characterized in that, In step S4, the N output ports of the chip are coupled through an optical fiber array, converged by an optical fiber bundle, and connected to a single-pixel detector. The one-dimensional measurement vector is acquired by an oscilloscope, and the length of the one-dimensional measurement vector is equal to the length of the encoded sequence.
6. The automated detection method for performance indicators based on optical quantum chips according to claim 1, characterized in that, Step S5 includes: Construct a measurement matrix and arrange the coded sequence values at each time step into rows according to port order; Construct a sparse transform basis, wherein the sparse transform basis is a discrete cosine transform basis; The coupling matrix is calibrated by applying a known perturbation sequence and solving it using least squares. A complete measurement model is established, which is represented as the one-dimensional measurement vector being equal to the measurement matrix multiplied by the column vectorized form of the transmission matrix, plus the coupling matrix multiplied by the environmental disturbance vector, plus the measurement noise vector; The L1 norm optimization problem is solved using an iterative soft thresholding algorithm to recover the sparse coefficient vector. The sparse coefficient vector is converted into a column vectorized form of the transfer matrix by inverse transformation and reshaped into an N×M complex transfer matrix.
7. The automated detection method for performance indicators based on optical quantum chips according to claim 1, characterized in that, Step S6 includes: calculating the difference between the actual phase and the target phase of each channel as the static phase error; calculating the square of the modulus of the off-diagonal elements as the inter-channel crosstalk; calculating the Allan variance of the phase sequence obtained from multiple repeated measurements to obtain the dynamic phase noise statistical characteristics; calculating the square of the modulus of the reconstructed complex transmission matrix and the trace of the ideal transmission matrix divided by the square of the number of output ports as the quantum gate fidelity; and summarizing the above calculation results into a data file.
8. The automated detection method for performance indicators based on optical quantum chips according to claim 1, characterized in that, Step S7 includes: setting a preset number of repeated measurements R, where R ranges from 100 to 10000; re-executing steps S2 and S3 before each repeated measurement; after reaching the preset number of measurements, performing Allan variance analysis on the phase sequence of each channel, identifying the noise type based on the slope of the Allan variance curve; generating a test report containing a static phase error list, crosstalk matrix, Allan variance curve, and quantum gate fidelity value; and determining the chip's pass / fail status based on a preset threshold.
9. The automated detection method for performance indicators based on optical quantum chips according to claim 1, characterized in that, In step S5, the step size of the iterative soft thresholding algorithm is taken as the reciprocal of the square of the Frobenius norm of the measurement matrix. The regularization parameter is determined by cross-validation. The iteration stopping condition is that the maximum number of iterations is reached or the relative change between two adjacent iterations is less than a preset threshold.
10. The automated detection method for performance indicators based on optical quantum chips according to claim 1, characterized in that, In step S6, the static phase error is calculated as the difference between the actual phase and the target phase of each channel, the inter-channel crosstalk is expressed as the square of the modulus of the off-diagonal elements and in dB, the Allan variance is calculated as the mean square of the phase difference over average time, and the quantum gate fidelity is calculated as the square of the modulus of the trace of the product of the conjugate transpose of the reconstructed transmission matrix and the ideal transmission matrix divided by the square of the number of output ports.
