Calibratable reciprocal and non-reciprocal random dynamics measurement platform and method

Abstract

The present application relates to the technical field of random dynamics measurement, and particularly relates to a calibratable reciprocal and non-reciprocal random dynamics measurement platform and method, comprising: a state node unit comprising at least two state nodes; a coupling switching unit configured with a bidirectional voltage-controlled current source branch between each pair of state nodes; a calibration and dissipation adjustment unit comprising a common calibration bus, a standard sampling resistor, a calibration switching switch, a baseline pre-parallel resistor and a forward additional resistor; a noise and test excitation unit comprising an independent analog noise source and an optional known test signal source, for injecting random noise current or known disturbance current into the state nodes; a control and isolation unit comprising a main controller, a switching drive circuit and a state latch interface; the coupling states share the same physical platform and calibration chain, and can attribute observation differences to coupling directions or dissipation changes, eliminating confusion factors introduced by replacing devices or rewiring.

Description

Technical Field

[0001] This invention relates to the field of stochastic dynamics measurement technology, and in particular to a calibrable reciprocal and non-reciprocal stochastic dynamics measurement platform and a calibrable reciprocal and non-reciprocal stochastic dynamics measurement method. Background Technology

[0002] In statistical physics, nonequilibrium thermodynamics, and stochastic thermodynamics, nonreciprocal coupling is a crucial mechanism for generating probabilistic circulation, directional transport, and entropy generation. Experimental measurements of stochastic dynamical systems involving nonreciprocal coupling typically require switching between reciprocal and nonreciprocal states and comparing the system's statistical response in both states.

[0003] Existing measurement devices often require rewiring or device replacement when switching between reciprocal and non-reciprocal states. This causes transconductance, damping, noise, and acquisition conditions to change simultaneously, making it difficult to truly attribute observed differences to the non-reciprocal coupling itself. Furthermore, existing devices generally suffer from the following problems: transconductance is difficult to calibrate precisely on a per-path basis; digital control noise is easily mixed into random trajectories; relative positive and negative dissipation is difficult to implement with safe passive devices; and analysis results typically only provide mathematical matrices without a corresponding relationship to physical branches and data acquisition processes.

[0004] Therefore, there is a need for a stochastic dynamics measurement platform and corresponding measurement method that can perform reciprocal and non-reciprocal state switching, transconductance calibration on a path-by-path basis, relative positive and negative dissipation using passive positive resistors, and synchronous acquisition under digital silent conditions on the same physical platform. Summary of the Invention

[0005] To address the technical problem of switching between reciprocal and non-reciprocal states, this invention provides a calibrable reciprocal and non-reciprocal stochastic dynamics measurement platform and a calibrable reciprocal and non-reciprocal stochastic dynamics measurement method.

[0006] Firstly, a calibrable reciprocal and non-reciprocal stochastic dynamics measurement platform is provided, comprising: A state node unit includes at least two state nodes; each state node includes a state capacitor, a leakage branch, and a signal junction point; the state capacitor is used to store the state of the state node. A coupling switching unit is used to configure a bidirectional voltage-controlled current source branch between each pair of state nodes; each bidirectional voltage-controlled current source branch includes a transconductance amplification module, an on / off control module, and a polarity switching module; through the polarity switching module and the on / off control module, any one of four coupling states can be formed between the same pair of state nodes: off, reciprocal, forward non-reciprocal, and reverse non-reciprocal. The calibration and dissipation adjustment unit includes a common calibration bus, a standard sampling resistor, a calibration switching switch, a baseline pre-parallel resistor, and a positive additional resistor. The calibration switching switch is used to switch the bidirectional voltage-controlled current source branch between the state node and the common calibration bus. The standard sampling resistor is connected between the common calibration bus and ground. The baseline pre-parallel resistor and the positive additional resistor are used to form three passive dissipation states—relatively negative dissipation, zero dissipation, and positive dissipation—within a passive network containing only positive resistance. The noise and test excitation unit includes an independent simulated noise source and an optional known test signal source, used to inject random noise current or known disturbance current into the state node; The control and isolation unit includes a main controller, a switch drive circuit, and a state latch interface, which is used to issue and latch the topology, polarity, dissipation, and calibration configuration. After the configuration is completed, the main controller enters a sleep state or is powered off by the digital power supply. The synchronous acquisition unit includes a buffer amplifier, an anti-aliasing filter, and a multi-channel synchronous analog-to-digital converter; it is used to continuously and synchronously record the node voltage, calibration voltage, and configuration status code of all status nodes during the main controller's sleep or digital power-off period; the node voltage represents the system state and forms a random dynamic trajectory; The calculation and output unit is used to reconstruct the drift matrix, dissipation matrix, non-reciprocal matrix and noise matrix based on the trajectory data output by the synchronous acquisition unit, calculate the response quantity, entropy generation and quality index, and map the results to specific nodes, branches and hardware states.

[0007] Preferably, in the coupling switching unit, each bidirectional voltage-controlled current source branch is composed of a discrete transconductance range group, a latching switch array, and a polarity selection bit cascaded in three stages.

[0008] Preferably, the coupling switching unit, through the same set of bidirectional voltage-controlled current sources, can achieve the following four coupling modes by only changing the polarity position of the polarity switching module and the on / off position of the on / off control module: Off mode, where the transconductance of both branches is zero and the non-reciprocal matrix is ​​zero; Reciprocal self-test mode, where the transconductance of both branches has the same sign and is non-zero; Forward non-reciprocal mode, where the transconductance amplitudes of the two branches are equal but opposite in sign, forming a non-reciprocal cyclic matrix in the first direction; and Reverse non-reciprocal mode, where the transconductance amplitudes of the two branches are the same as in the forward non-reciprocal mode but simultaneously reversed in sign, forming a non-reciprocal cyclic matrix opposite to the first direction.

[0009] Preferably, in the calibration and dissipation adjustment unit, the baseline pre-parallel resistor and the positive additional resistor are configured around the same baseline conductance. The three passive dissipation states are: baseline state, where the baseline pre-parallel resistor is connected, the positive additional resistor is disconnected, and the total node conductance is equal to the baseline conductance; relative negative dissipation state, where both the baseline pre-parallel resistor and the positive additional resistor are disconnected, and the total node conductance is equal to the baseline conductance minus the conductance modulation amplitude; and positive dissipation state, where both the baseline pre-parallel resistor and the positive additional resistor are connected, and the total node conductance is equal to the baseline conductance plus the conductance modulation amplitude. The baseline pre-parallel resistor and the positive additional resistor are both implemented with positive resistors, without introducing active negative resistance.

[0010] Preferably, the calibration switching switch is a single-pole double-throw switch, with the common terminal connected to the output terminal of the bidirectional voltage-controlled current source branch, and the two throw terminals connected to the target status node and the common calibration bus, respectively; all the calibration switching switches adopt a first-off-then-close method for switching, and a non-overlapping dead zone is set.

[0011] Preferably, after completing the configuration and calibration, the control and isolation unit performs the following operations: confirms that all latching switches are in the locked state; stops USB data transmission and reception, serial port polling, PWM output, LED status refresh and all periodic timer interrupts; removes the relay coil drive current; puts the main controller into sleep mode or cuts off the digital power supply; and waits for a preset stabilization time to allow the analog node to reach a steady state again.

[0012] Preferably, the synchronous acquisition unit operates independently of the main controller, with a sampling rate not less than 20 times the highest characteristic frequency of the system, and a channel synchronization error of less than 0.1 sampling cycles; each data file simultaneously saves the original voltage trajectory, latch switch and relay status bit mask, calibration coefficient table, capacitance value, temperature, sampling rate, channel mapping, and configuration hash.

[0013] Preferably, the calculation and output unit performs the following: independently preprocessing trajectory data for each hardware state, deleting transient segments and analog-to-digital converter saturation segments, converting to node voltages and removing the mean; estimating the steady-state covariance matrix; reconstructing the continuous-time drift matrix using the time-delay covariance matrix and the finite-time propagator; decomposing the drift matrix into a symmetric dissipation matrix and an antisymmetric non-reciprocal matrix; independently estimating the noise covariance intensity matrix using both steady-state Lyapunov identities and discrete innovation residuals; and calculating the linear damping response matrix and steady-state entropy generation rate based on the estimation results.

[0014] Preferably, the calculation and output unit also performs nonlinear three-state analysis: when the linear propagation residual, the difference between the two noise estimates, or the actual dissipation intervention residual exceeds the threshold, three sets of matching trajectories—coupled off, forward non-reciprocal, and reverse non-reciprocal—are acquired under the same damping, noise, temperature, and sampling rate conditions; nonlinear drift, steady-state mean, and covariance are estimated; the Poisson equation is solved to obtain the nonlinear damping response matrix; and the direction-dependent response that is reversed with the direction of the cycle and the direction-independent response that is not reversed with the direction after the baseline is subtracted are separated.

[0015] Preferably, the calculation and output unit also performs quality gating: it judges stability, positive covariance definiteness, noise double estimation consistency, analog-to-digital converter linear region, aperiodic crosstalk, cyclic inversion matching, calibration drift, model residuals, and physical verification; if any check fails, the corresponding measurement unit is marked as invalid.

[0016] Secondly, a calibrable method for measuring reciprocal and non-reciprocal stochastic dynamics is also provided, employing the measurement platform described in any of the above technical solutions, comprising the following steps: Step S1: Arrange state nodes and bidirectional voltage-controlled current source branches, so that each state node is connected to a state capacitor, main leakage resistor, switchable dissipation branch, coupling current injection terminal and noise current injection terminal; arrange two voltage-controlled current source branches in opposite directions for each pair of selected state nodes. Step S2: Common calibration for each branch. Switch the voltage-controlled current source branch to be calibrated to the common calibration bus through the calibration switch. Apply at least two known DC voltages to the input terminal of the branch under the condition that the noise source is turned off. Measure the voltage across the standard sampling resistor and convert the injected current. Fit the signed transconductance, zero-point offset current and linear residual of the branch. Step S3: Based on the target measurement task, configure the turn-off, reciprocal self-test, forward non-reciprocal, or reverse non-reciprocal coupling modes by changing the polarity and on / off positions of the bidirectional voltage-controlled current source branch. Step S4: Configure passive dissipation state by switching the connection state of the baseline pre-parallel resistor and the positive additional resistor to form a zero dissipation, relative negative dissipation, or positive dissipation state: Baseline state: the baseline pre-parallel resistor is connected, the positive additional resistor is disconnected, and the total node conductance is equal to the baseline conductance; Relative negative dissipation state: both the baseline pre-parallel resistor and the positive additional resistor are disconnected, and the total node conductance is equal to the baseline conductance minus the conductance modulation amplitude; Positive dissipation state: both the baseline pre-parallel resistor and the positive additional resistor are connected, and the total node conductance is equal to the baseline conductance plus the conductance modulation amplitude; wherein, both the baseline pre-parallel resistor and the positive additional resistor are implemented by positive resistors, without introducing active negative resistance; Step S5: Enter digital silent state, latch all switch states, stop the digital activity of the main controller and put it into sleep or power off, and wait for the analog node to reach steady state; Step S6: During the period when the main controller is in sleep mode or the digital power is cut off, all node voltages, calibration voltages and configuration status codes are collected by an independent multi-channel synchronous analog-to-digital converter to form raw trajectory data; Step S7: Reconstruct the drift matrix, dissipation matrix, non-reciprocal matrix, and noise matrix from the same data block, and calculate the response, entropy generation, and quality indicators; Step S8: Change only the coupling direction or dissipation branch to be compared, and repeat steps S3 to S7. By differentiating the measurement results of the same platform, the same calibration, the same noise, and the same acquisition chain, the measurement of reciprocal and non-reciprocal stochastic dynamic characteristics is completed.

[0017] Preferably, in step S2, the stability margin of the candidate drift matrix is ​​calculated using the latest calibration value, and the connection between each branch and the state node is restored only when the maximum real eigenvalue is less than a preset stability threshold.

[0018] Preferably, step S7 further includes: taking the unit vector corresponding to the largest eigenvalue of the steady-state covariance matrix, solving the adjoint equation to obtain the adjoint matrix, and then calculating the linear damping response matrix; and using the drift matrix, noise covariance intensity matrix and steady-state covariance matrix to calculate the steady-state entropy generation rate.

[0019] Preferably, in step S7, when the linear propagation residual or the difference between the two noise estimates exceeds a threshold, a nonlinear three-state analysis is added: under the same conditions, three sets of matching trajectories—coupled off, forward non-reciprocal, and reverse non-reciprocal—are collected, the nonlinear drift and steady-state distribution are estimated, the Poisson equation is solved to obtain the nonlinear damping response matrix, and the direction-dependent response and direction-independent response are calculated.

[0020] Beneficial effects: (1) Each coupling state shares the same physical platform and calibration chain, which can attribute observation differences to changes in coupling direction or dissipation, eliminating confusion caused by replacing devices or rewiring.

[0021] (2) The signed transconductance, bias and linear residual of each voltage-controlled current source branch can be accurately measured one by one through the common calibration bus to ensure that the transconductance parameters are traceable.

[0022] (3) Only positive resistors are used to achieve three dissipation states: negative, zero, and positive. No active negative resistance circuit is introduced, which improves measurement safety.

[0023] (4) After configuration, the controller is put into sleep mode or powered off, and the data is collected by an independent synchronous analog-to-digital converter, which eliminates the interference of clock crosstalk, digital ground bounce and pulse width modulation ripple on random trajectories.

[0024] (5) The drift matrix, dissipation matrix, non-reciprocal matrix and noise matrix can be reconstructed from the same data block, and the noise covariance intensity can be estimated independently using two methods, providing a means of verifying the self-consistency of the model.

[0025] (6) Retain the complete data traceability chain of the original trajectory, calibration value, switch status and calculation results, so as to facilitate the reverse lookup of the results to the specific physical branch. Attached Figure Description

[0026] Figure 1 Figure 1 shows a first embodiment of a calibrable reciprocal and non-reciprocal stochastic dynamics measurement platform; Figure 2 Figure 1 shows a second embodiment of a calibrable reciprocal and non-reciprocal stochastic dynamics measurement platform; Figure 3 This is an example diagram of a calibrable reciprocal and non-reciprocal stochastic dynamics measurement method.

[0027] Explanation of reference numerals in the attached figures: 1. State node unit; 2. Coupling switching unit; 3. Calibration and dissipation adjustment unit; 4. Noise and test excitation unit; 5. Control and isolation unit; 6. Synchronous acquisition unit; 7. Calculation and output unit; 11. A first embodiment of a calibrable reciprocal and non-reciprocal stochastic dynamics measurement platform; 12. A second embodiment of a calibrable reciprocal and non-reciprocal stochastic dynamics measurement platform. Detailed Implementation

[0028] Firstly, such as Figure 1 As shown, a calibrable reciprocal and non-reciprocal stochastic dynamics measurement platform 11 is provided, comprising: State node unit 1 includes at least two state nodes; each state node includes a state capacitor, a leakage branch, and a signal junction point; the state capacitor is used to store the state of the state node. The coupling switching unit 2 is used to configure a bidirectional voltage-controlled current source branch between each pair of state nodes; each bidirectional voltage-controlled current source branch includes a transconductance amplification module, an on / off control module, and a polarity switching module; through the polarity switching module and the on / off control module, any one of four coupling states can be formed between the same pair of state nodes: off, reciprocal, forward non-reciprocal, and reverse non-reciprocal. The calibration and dissipation adjustment unit 3 includes a common calibration bus, a standard sampling resistor, a calibration switching switch, a baseline pre-parallel resistor, and a positive additional resistor. The calibration switching switch is used to switch the bidirectional voltage-controlled current source branch between the state node and the common calibration bus. The standard sampling resistor is connected between the common calibration bus and ground. The baseline pre-parallel resistor and the positive additional resistor are used to form three passive dissipation states: relative negative dissipation, zero dissipation, and positive dissipation, in a passive network containing only positive resistors. The noise and test excitation unit 4 includes an independent simulated noise source and an optional known test signal source, used to inject random noise current or known disturbance current into the state node; The control and isolation unit 5 includes a main controller, a switch drive circuit and a state latch interface, which is used to issue and latch the topology, polarity, dissipation and calibration configuration. After the configuration is completed, the main controller enters a sleep state or is cut off from the digital power supply. The synchronous acquisition unit 6 includes a buffer amplifier, an anti-aliasing filter, and a multi-channel synchronous analog-to-digital converter; it is used to continuously and synchronously record the node voltage, calibration voltage, and configuration status code of all status nodes during the period when the main controller is in sleep mode or the digital power is cut off; the node voltage represents the system state and forms a random dynamic trajectory; This embodiment uses two state nodes as an example to illustrate the hardware configuration of the platform. The platform has two state nodes, numbered Node 1 and Node 2.

[0029] The following five types of components are connected to the signal busbar of each state node: (i) State capacitors C1 and C2, with a value of 100 nanofarads, using C0G / NP0 temperature-stable ceramic capacitors or film capacitors. The capacitance value should be measured and recorded before installation. (ii) Main leakage resistor, 100 kΩ (corresponding to conductance gm_1=10 micro Siemens) for node 1, and 14.3 kΩ (corresponding to conductance gm_2=70 micro Siemens) for node 2, using precision metal film resistors, permanently connected to the circuit. (iii) Switchable dissipative branch group, including baseline pre-parallel resistor and forward additional resistor. (iv) Coupling current injection terminal, from the output of each voltage-controlled current source branch in the coupling switching unit. (v) Noise current injection terminal, connected to the output of the noise and test excitation unit and the input of the high-impedance buffer amplifier.

[0030] The voltage vi(t) across the state capacitor is the system state variable. Under random noise, the joint evolution of the node voltages constitutes a stochastic dynamic trajectory. The coupling between two nodes is achieved by the bidirectional voltage-controlled current source branch in the coupling switching unit.

[0031] The noise and test excitation unit employs two independent analog noise sources, injected into node 1 and node 2 respectively. The noise sources utilize Zener diode breakdown noise, amplified by a low-noise operational amplifier, and then connected to the nodes via injection resistors. The noise spectrum is determined after calibration, and boundary verification is performed only after the noise scale is fixed. A programmable precision voltage source can be selected as the known test signal source, outputting a known DC level during the calibration phase.

[0032] The control and isolation unit uses an ARM Cortex-M4 microcontroller as the main controller. The microcontroller sends configuration codes to the switch driver circuit via the Serial Peripheral Interface (SPI) bus. The switch driver circuit latches the configuration codes into the drive coil of the magnetically latched relay. After one pulse drive, the relay state is self-holding, and the microcontroller no longer needs to maintain the drive current. After configuration, the microcontroller enters a deep sleep mode, stopping all peripheral clocks.

[0033] The synchronous acquisition unit employs a synchronous analog-to-digital converter (ADC) module independent of the main controller, comprising two buffer amplifiers, two fourth-order Butterworth anti-aliasing low-pass filters, and a multi-channel synchronous ADC chip. The buffer amplifiers utilize JFET input-stage operational amplifiers with an input bias current of less than 1pA and a bandwidth no less than ten times the system's highest pole frequency. The ADC has a sampling rate of 25kHz, a resolution of 16 bits, and a synchronization error between the two channels of less than 0.1 sampling cycles. The ADC has an independent crystal oscillator clock source and does not share a clock with the main controller.

[0034] Preferred, such as Figure 2 As shown, it also includes: a calculation and output unit 7, which is used to reconstruct the drift matrix, dissipation matrix, non-reciprocal matrix and noise matrix based on the trajectory data output by the synchronous acquisition unit, calculate the response quantity, entropy generation and quality index, and map the results to specific nodes, branches and hardware status.

[0035] This embodiment illustrates the specific implementation of the calculation and output unit.

[0036] In this embodiment, the calculation and output unit employs a general-purpose offline computer processing scheme. The synchronous acquisition unit stores the raw trajectory data as binary data files via a secure digital (SD) card. Each data file contains: the raw voltage trajectory (16-bit code values ​​for each channel, arranged in order of sampling time), latching switch and relay status bit masks (32-bit unsigned integers, each bit corresponding to the on / off state of a switch or relay), a calibration coefficient table (signed transconductance cij, zero-point offset current bij, and linear residual eij for each transconductance branch, stored as floating-point numbers), measured capacitance values, ambient temperature, sampling rate, channel mapping table, and configuration hash (SHA-256 digest value).

[0037] On a general-purpose computer, the compute and output unit reads the data file and executes the following process: The first step is data traceability and verification: Calculate the SHA-256 hash of the original trajectory data and compare it with the configuration hash stored in the file to confirm data integrity. After successful verification, bind the calibration coefficient table with the corresponding branch number, node number, and acquisition time.

[0038] The second step involves processing the trajectory data for each hardware state (a combination of OFF, R, +NR, -NR and dissipation 0 / + / -) independently, without merging the original trajectories from different states. Specific processing steps include preprocessing, steady-state covariance estimation, drift matrix reconstruction, matrix decomposition, noise dual estimation, and response calculation.

[0039] The third step is to bind and save all the reconstructed drift matrices, dissipation matrices, non-reciprocal matrices, and noise matrices with their corresponding state node numbers, transconductance branch numbers, and hardware status codes. For example, the metadata record for a certain drift matrix L_hat is: node 1-node 2, forward non-reciprocal mode, baseline dissipation state, transconductance branch 1 is +5.013 microSiemens, and transconductance branch 2 is -4.987 microSiemens.

[0040] The fourth step is to output a complete data packet for each valid measurement unit. This data packet includes the original trajectory, calibration table, configuration hash, L_hat, A_hat, N_hat, Q_hat, Sigma_hat, and all quality gating flags. Those skilled in the art can trace any calculation result back to its corresponding original physical branch and hardware state.

[0041] Preferably, in the coupling switching unit, each bidirectional voltage-controlled current source branch is composed of three cascaded levels: a discrete transconductance range group, a latching switch array, and a polarity selection bit; the discrete transconductance range group is composed of fixed precision resistors and is used to set the transconductance amplitude; the latching switch array is used to maintain the range without holding current after selection; and the polarity selection bit is used to determine the sign and on / off state of the transconductance.

[0042] This embodiment details the three-stage cascade structure of each bidirectional voltage-controlled current source branch.

[0043] Stage 1: Discrete transconductance range group. Composed of fixed precision resistors, providing at least two non-zero ranges and one zero range. In this embodiment, the transconductance range group includes six precision resistor values, corresponding to transconductance amplitudes of 0, 0.5, 1, 2.5, 5, and 10 micro Siemens, respectively. Each range consists of a fixed resistor and an operational amplifier forming a transconductance amplification stage, with a resistance accuracy of 0.1% or higher and a temperature coefficient of less than 10 ppm / degree Celsius. The zero range is achieved by disconnecting the feedback loop of the operational amplifier, making the output current of this branch zero.

[0044] The second stage: a latching switch array. Between the first and third stages, this consists of magnetically latched relays or an analog switch array. The function of the latching switch array is to lock the signal path for the selected transconductance range, maintaining the range thereafter without requiring a driving current. This design eliminates charge injection noise caused by digital cycle refresh. In this embodiment, each branch uses three sets of double-pole magnetically latched relays to achieve six-to-one range switching. The relay drive coil is only subjected to a pulse current at the moment of switching, and the coil current is zero after switching is complete.

[0045] Level 3: Polarity Selection Bit. Implemented through a set of differential analog switches. The polarity selection bit can be +1, 0, or -1. When the polarity bit is +1, the positive output terminal is connected, and the output current is +|cij|vj; when the polarity bit is -1, the reverse output terminal is connected, and the output current is -|cij|vj; when the polarity bit is 0, both the positive and reverse output terminals are disconnected, which is equivalent to the on / off control module shutting down this branch.

[0046] Taking the branch from node 1 to node 2 as an example, the workflow of the three-level cascade is as follows: First, the latching switch array selects the 5µSiemens range, determining the transconductance amplitude |c12| = 5µSiemens. Then, the polarity selection bit is set to +1, making the output current direction to be positive current injected into node 1 after reading the voltage from node 2, i.e., c12 = +5µSiemens. If it is necessary to switch to the reverse non-reciprocal mode, simply switch the polarity selection bit from +1 to -1, while keeping the transconductance range unchanged, at which point c12 = -5µSiemens. Both the transconductance amplitude and polarity are latched by hardware and do not require digital control to maintain.

[0047] Preferably, the coupling switching unit, through the same set of bidirectional voltage-controlled current sources, can achieve the following four coupling modes by only changing the polarity position of the polarity switching module and the on / off position of the on / off control module: Off mode, where the transconductance of both branches is zero and the non-reciprocal matrix is ​​zero; Reciprocal self-test mode, where the transconductance of both branches has the same sign and is non-zero; Forward non-reciprocal mode, where the transconductance amplitudes of both branches are equal but opposite in sign, forming a non-reciprocal cyclic matrix in the first direction; and Reverse non-reciprocal mode, where the transconductance amplitudes of the two branches are the same as in the forward non-reciprocal mode but simultaneously reversed in sign, forming a non-reciprocal cyclic matrix opposite to the first direction. Based on the above three-level cascaded structure, the coupling switching unit implements four modes through the same set of bidirectional voltage-controlled current sources: (i) Off Mode: The polarity selection bits of both branches (node ​​1 to node 2 and node 2 to node 1) are set to 0, and the transconductances of both branches c12=c21=0. At this time, the non-reciprocal matrix N=0, and the system only has dissipation and random fluctuations. This mode serves as a pure dissipation baseline.

[0048] (ii) Reciprocal self-test mode (R): The polarity selection bits of both branches are set to +1, c12 = +5 microSiemens, c21 = +5 microSiemens. The transconductances of the two branches have the same sign and are non-zero, forming a symmetrical coupling. Since the symmetrical transconductance will change the dissipation matrix A at the same time, this mode is mainly used for self-testing rather than as a pure non-reciprocal comparison.

[0049] (iii) Positive Non-Reciprocal Mode (+NR): The polarity bit of the branch from node 1 to node 2 is set to +1 (c12 = +5 micro Siemens), and the polarity bit of the branch from node 2 to node 1 is set to -1 (c21 = -5 micro Siemens). The transconductance amplitudes of the two branches are equal but opposite in sign, and c12 = -c21, forming non-zero elements of the non-reciprocal cyclic matrix N.

[0050] (iv) Reverse Non-Reciprocal Mode (-NR): The two polarity bits of the forward non-reciprocal mode are reversed simultaneously, that is, the polarity bit of the branch from node 1 to node 2 is set to -1 (c12=-5 micro Siemens), and the polarity bit of the branch from node 2 to node 1 is set to +1 (c21=+5 micro Siemens). The transconductance amplitude remains unchanged, and N changes sign.

[0051] Switching between the four modes does not change any parameters of the state node, state capacitor, leakage resistance, noise source, or acquisition channel; it only changes the value of the polarity selection bit. Therefore, it can be ensured that the differences between modes only come from changes in the coupling direction.

[0052] In the preferred calibration and dissipation adjustment unit, the baseline pre-parallel resistor and the positive additional resistor are configured around the same baseline conductance. The three passive dissipation states are: baseline state, where the baseline pre-parallel resistor is connected, the positive additional resistor is disconnected, and the total node conductance equals the baseline conductance; relative negative dissipation state, where both the baseline pre-parallel resistor and the positive additional resistor are disconnected, and the total node conductance equals the baseline conductance minus the conductance modulation amplitude; and positive dissipation state, where both the baseline pre-parallel resistor and the positive additional resistor are connected, and the total node conductance equals the baseline conductance plus the conductance modulation amplitude. The baseline pre-parallel resistor and the positive additional resistor are both implemented using positive resistors, without introducing active negative resistance.

[0053] Preferably, the calibration switching switch is a single-pole double-throw switch, with its common terminal connected to the output terminal of the bidirectional voltage-controlled current source branch, and its two throw terminals connected to the target state node and the common calibration bus, respectively; all the calibration switching switches adopt a first-off-then-make switching method and are equipped with a non-overlapping dead zone. This embodiment details the specific implementation of the calibration switching switch.

[0054] Each voltage-controlled current source branch output is connected to a single-pole double-throw calibration switch (kij). This switch uses a magnetic latching relay; the common terminal is connected to the voltage-controlled current source output, throw terminal 1 is connected to the target status node, and throw terminal 2 is connected to the common calibration bus. The common calibration bus connects a standard sampling resistor R0 to ground; R0 is 1 kΩ, with an accuracy of 0.1% and a temperature coefficient less than 5 ppm / °C. The calibration measurement terminal is connected across R0 to a channel of the synchronous acquisition unit.

[0055] The calibration switching process is as follows: First, disconnect kij from throw terminal 1 (target node). At this point, the branch output terminal is left floating, confirming that the potential at the measurement terminal has been released to zero.

[0056] The second step involves waiting through a non-overlapping dead time. In this embodiment, the dead time is set to 2 microseconds. During this period, the common terminal of kij is not connected to any throw terminal, ensuring that there is no transient connection between the node and the bus simultaneously.

[0057] Third, connect kij to throw terminal 2 (common calibration bus). At this point, the branch output terminal is grounded via R0, and calibration measurements can begin.

[0058] The above-described disconnect-then-connect switching process ensures that at any given time, the branch output will not be simultaneously connected to the target status node and the common calibration bus. This prevents two consequences: first, it prevents test voltage on the calibration bus from interfering with normal measurements by entering the measurement node; second, it prevents node voltage from being injected into the ground loop through the calibration bus, affecting other branches that are being measured.

[0059] Preferably, after completing configuration and calibration, the control and isolation unit performs the following operations: confirming that all latching switches are locked; stopping USB data transmission and reception, serial port polling, PWM output, LED status refresh, and all periodic timer interrupts; removing the relay coil drive current; putting the main controller into sleep mode or cutting off the digital power supply; and waiting for a preset stabilization time to allow the analog node to reach a steady state again. This embodiment details the operation sequence for the control and isolation unit to enter sleep mode or cut off the digital power supply.

[0060] After configuration and calibration are complete, the main controller performs the following operations in sequence: (a) Traverse the status registers of all latching switches, read the readback value and compare it with the target configuration code. If they all match, confirm that latching is complete and that the transconductance position and polarity position are stable.

[0061] (b) Call the microcontroller's peripheral shutdown function to shut down the USB peripherals in sequence (stop USB data transmission and reception), UART peripherals (stop serial port polling), timer module (stop all periodic timer interrupts), PWM module (stop PWM output), and GPIO module (stop LED status refresh). The shutdown order is arranged in descending order of peripheral power consumption and interference intensity, shutting down high-power and high-interference peripherals first.

[0062] (c) Send a release command to the relay drive circuit. In this embodiment, since a magnetic latching relay is used, the release command only turns off the drive MOSFET, and there is no continuous current in the coil. If a non-magnetic latching relay is used, this step releases the coil current.

[0063] (d) The microcontroller's power management unit is invoked to switch the system clock to the internal low-speed oscillator, configure the wakeup pin level to trigger the wake-up condition, and then enter deep sleep mode. In this embodiment, after the microcontroller enters deep sleep mode, the core clock stops, all digital peripheral clocks stop, and the static power consumption drops to the microampere level.

[0064] (e) Wait for the simulated node to reach steady state again. In the two-node system of this embodiment, the time constant of node 2, tau2 = C2 / g2 = 100nF / 70uS, is approximately 1.43 milliseconds. The preset settling time is set to 50 times the maximum time constant, that is, after waiting for about 75 milliseconds, the transient state of the simulated node decays to a negligible level.

[0065] During the digital silence period, the independent crystal oscillator of the synchronous acquisition unit continues to run, and the analog-to-digital converter continuously records at a preset sampling rate. Spectrum analysis confirms that there are no significant peaks of the main controller clock (72 MHz in this embodiment) and its harmonics in the trajectory, verifying the effectiveness of the digital silence.

[0066] Preferably, the synchronous acquisition unit operates independently of the main controller, with a sampling rate not less than 20 times the highest characteristic frequency of the system, and an inter-channel synchronization error of less than 0.1 sampling cycles. Each data file simultaneously stores the original voltage trajectory, latch switch and relay status bit masks, calibration coefficient table, capacitance value, temperature, sampling rate, channel mapping, and configuration hash. In this embodiment, the highest characteristic frequency of the system is determined by node 2: fmax = g2 / (2 pi C2)=70uS / (2 pi 100nF is approximately equal to 111 Hz. The sampling rate of the synchronous acquisition unit is set to 25kHz, which is 225 times the highest characteristic frequency, far exceeding the minimum requirement of 20 times. The analog-to-digital converter has a resolution of 16 bits and a full-scale range of ±2.5 volts. The synchronization error between the two channels was measured to be less than 0.05 sampling periods, i.e., less than 2 microseconds.

[0067] The anti-aliasing filter uses two independent fourth-order Butterworth low-pass filters with a cutoff frequency set to half the sampling rate, i.e., 12.5kHz. The passband ripple is less than ±0.1 dB, and the stopband attenuation is greater than 80 dB.

[0068] Each data file is stored in the following structure: (i) File header: contains the acquisition start timestamp, sampling rate, number of channels, resolution per channel, channel mapping table (correspondence between channel number and node number), measured capacitance value, ambient temperature (read by the temperature sensor on the board), and configuration hash. (ii) Latch switch status bitmask: a 32-bit integer, each bit corresponding to the current state of a switch or relay. (iii) Calibration coefficient table: each branch records three floating-point numbers (signed transconductance, zero-point bias current, and linear residual). (iv) Raw voltage trajectory: the 16-bit code values ​​of each channel are interleaved in time, and each sampling moment contains data from all channels.

[0069] The data files are stored in binary format and accompanied by a plain text metadata summary file, allowing experimenters to quickly view configuration information without parsing the binary data.

[0070] Preferably, the calculation and output unit performs the following: independently preprocessing trajectory data for each hardware state, deleting transient segments and analog-to-digital converter saturation segments, converting to node voltages and removing the mean; estimating the steady-state covariance matrix; reconstructing the continuous-time drift matrix using the time-delay covariance matrix and the finite-time propagator; decomposing the drift matrix into a symmetric dissipation matrix and an antisymmetric non-reciprocal matrix; independently estimating the noise covariance intensity matrix using both steady-state Lyapunov identities and discrete innovation residuals; and calculating the linear damping response matrix and steady-state entropy generation rate based on the estimation results.

[0071] Preferably, the calculation and output unit also performs nonlinear three-state analysis: when the linear propagation residual, the difference between the two noise estimates, or the actual dissipation intervention residual exceeds the threshold, three sets of matching trajectories—coupled off, forward non-reciprocal, and reverse non-reciprocal—are acquired under the same damping, noise, temperature, and sampling rate conditions; nonlinear drift, steady-state mean, and covariance are estimated; the Poisson equation is solved to obtain the nonlinear damping response matrix; and the direction-dependent response that is reversed with the direction of the cycle and the direction-independent response that is not reversed with the direction after the baseline is subtracted are separated.

[0072] Preferably, the calculation and output unit also performs quality gating: judging stability, positive covariance definiteness, noise double estimation consistency, analog-to-digital converter linear region, aperiodic crosstalk, cyclic inversion matching, calibration drift, model residuals, and physical verification; if any check fails, the corresponding measurement unit is marked as invalid. This embodiment takes the quality gating execution process of one measurement unit as an example.

[0073] (i) Stability check: Calculate all eigenvalues ​​of the reconstructed drift matrix L_hat and take the largest real eigenvalue. In this embodiment, the eigenvalues ​​of L_hat are -100.8 and -698.2 (both real numbers), and max Re(eig(L_hat)) = -100.8, which is less than the preset stability threshold of -20 per second.

[0074] (ii) Check for positive covariance definiteness: Perform Cholesky decomposition on Sigma_hat; if the decomposition is successful, it is positive definite. Check the semi-positive definiteness of Q_hat to confirm that all eigenvalues ​​are greater than or equal to zero.

[0075] (iii) Noise double estimation consistency check: Calculate ||Q_L-Q_R|| / max(||Q_L||,||Q_R||). In this embodiment, the ratio is 0.03, which is less than the preset threshold theta_Q=0.1.

[0076] (iv) Linear region check of analog-to-digital converter: Traverse all sampling points to confirm that no code value touches the continuous interval above 98% of the full scale and the timestamp is continuous without jump.

[0077] (v) Periodic Crosstalk Check: Perform a Fast Fourier Transform on the trajectory data and search for the main controller clock frequency (72 MHz) and its integer multiples of harmonics in the spectrum. If the power spectral density at these frequencies does not exceed 3 times the background noise, it is determined to be periodic crosstalk.

[0078] (vi) Cycle Reversal Matching Check: Compare the N_hat values ​​of the forward non-reciprocal state and the reverse non-reciprocal state. The relative difference between |N_hat(+NR)| and |N_hat(-NR)| must be less than 5%, and N_hat(+NR) must be approximately equal to -N_hat(-NR). In this embodiment, the difference is 1.6%.

[0079] (vii) Calibration drift check: Compare the calibration values ​​before and after this round of measurement. The transconductance change |delta c| / |c| is less than theta_cal=0.5%, and the bias change is less than 0.1% of the full scale.

[0080] (viii) Model residual check: Calculate the linear propagation residual (the difference between the predicted and actual values ​​of F_dt), whose root mean square is less than 1% of the trajectory standard deviation.

[0081] (ix) Physical verification check: Under positive dissipative conditions, the direction of change of the largest eigenvalue of the actual covariance is decreasing. The damped response matrix predicts tr(MK)>0, indicating that the covariance in the main direction decreases after the dissipative branch is connected. The signs of both are consistent.

[0082] All nine checks were passed, and the measurement unit was marked as "valid" and included in subsequent cross-state comparisons.

[0083] Secondly, such as Figure 3 As shown, a calibrable method for measuring reciprocal and non-reciprocal stochastic dynamics is also provided, employing the measurement platform described in any of the above technical solutions, comprising the following steps: Step S1: Arrange state nodes and bidirectional voltage-controlled current source branches, so that each state node is connected to a state capacitor, main leakage resistor, switchable dissipation branch, coupling current injection terminal and noise current injection terminal; arrange two voltage-controlled current source branches in opposite directions for each pair of selected state nodes. Step S2: Common calibration for each branch. Switch the voltage-controlled current source branch to be calibrated to the common calibration bus through the calibration switch. Apply at least two known DC voltages to the input terminal of the branch under the condition that the noise source is turned off. Measure the voltage across the standard sampling resistor and convert the injected current. Fit the signed transconductance, zero-point offset current and linear residual of the branch. Step S3: Based on the target measurement task, configure the turn-off, reciprocal self-test, forward non-reciprocal, or reverse non-reciprocal coupling modes by changing the polarity and on / off positions of the bidirectional voltage-controlled current source branch. Step S4: Configure passive dissipation state by switching the connection state of the baseline pre-parallel resistor and the positive additional resistor to form a zero dissipation, relative negative dissipation, or positive dissipation state: Baseline state: the baseline pre-parallel resistor is connected, the positive additional resistor is disconnected, and the total node conductance is equal to the baseline conductance; Relative negative dissipation state: both the baseline pre-parallel resistor and the positive additional resistor are disconnected, and the total node conductance is equal to the baseline conductance minus the conductance modulation amplitude; Positive dissipation state: both the baseline pre-parallel resistor and the positive additional resistor are connected, and the total node conductance is equal to the baseline conductance plus the conductance modulation amplitude; wherein, both the baseline pre-parallel resistor and the positive additional resistor are implemented by positive resistors, without introducing active negative resistance; Step S5: Enter digital silent state, latch all switch states, stop the digital activity of the main controller and put it into sleep or power off, and wait for the analog node to reach steady state; Step S6: During the period when the main controller is in sleep mode or the digital power is cut off, all node voltages, calibration voltages and configuration status codes are collected by an independent multi-channel synchronous analog-to-digital converter to form raw trajectory data; Step S7: Reconstruct the drift matrix, dissipation matrix, non-reciprocal matrix, and noise matrix from the same data block, and calculate the response, entropy generation, and quality indicators; Step S8: Change only the coupling direction or dissipation branch to be compared, and repeat steps S3 to S7. By differentiating the measurement results of the same platform, the same calibration, the same noise, and the same acquisition chain, the measurement of reciprocal and non-reciprocal stochastic dynamic characteristics is completed.

[0084] Preferably, in step S2, the stability margin of the candidate drift matrix is ​​calculated using the latest calibration value, and the connection between each branch and the state node is restored only when the maximum real eigenvalue is less than a preset stability threshold.

[0085] Preferably, step S7 further includes: taking the unit vector corresponding to the largest eigenvalue of the steady-state covariance matrix, solving the adjoint equation to obtain the adjoint matrix, and then calculating the linear damping response matrix; and using the drift matrix, noise covariance intensity matrix and steady-state covariance matrix to calculate the steady-state entropy generation rate.

[0086] Preferably, in step S7, when the linear propagation residual or the difference between the two noise estimates exceeds a threshold, a nonlinear three-state analysis is added: under the same conditions, three sets of matching trajectories—coupled off, forward non-reciprocal, and reverse non-reciprocal—are collected, the nonlinear drift and steady-state distribution are estimated, the Poisson equation is solved to obtain the nonlinear damping response matrix, and the direction-dependent response and direction-independent response are calculated.

[0087] Beneficial effects: (1) Each coupling state shares the same physical platform and calibration chain, which can attribute observation differences to changes in coupling direction or dissipation, eliminating confusion caused by replacing devices or rewiring.

[0088] (2) The signed transconductance, bias and linear residual of each voltage-controlled current source branch can be accurately measured one by one through the common calibration bus to ensure that the transconductance parameters are traceable.

[0089] (3) Only positive resistors are used to achieve three dissipation states: negative, zero, and positive. No active negative resistance circuit is introduced, which improves measurement safety.

[0090] (4) After configuration, the controller is put into sleep mode or powered off, and the data is collected by an independent synchronous analog-to-digital converter, which eliminates the interference of clock crosstalk, digital ground bounce and pulse width modulation ripple on random trajectories.

[0091] (5) The drift matrix, dissipation matrix, non-reciprocal matrix and noise matrix can be reconstructed from the same data block, and the noise covariance intensity can be estimated independently using two methods, providing a means of verifying the self-consistency of the model.

[0092] (6) Retain the complete data traceability chain of the original trajectory, calibration value, switch status and calculation results, so as to facilitate the reverse lookup of the results to the specific physical branch.

[0093] Finally, it should be noted that any modification or equivalent substitution of some or all of the technical features based on the device structure and the technical solutions of the embodiments of the present invention, without departing from the corresponding technical solutions of the present invention, shall fall within the patent scope of the device structure and the embodiments of the present invention.

Claims

1. A calibrable reciprocal and non-reciprocal stochastic dynamics measurement platform, characterized in that, include: A state node unit includes at least two state nodes; each state node includes a state capacitor, a leakage branch, and a signal junction point. The state capacitor is used to store the state of the state node. A coupling switching unit is used to configure a bidirectional voltage-controlled current source branch between each pair of state nodes; each bidirectional voltage-controlled current source branch includes a transconductance amplification module, an on / off control module, and a polarity switching module; through the polarity switching module and the on / off control module, any one of four coupling states can be formed between the same pair of state nodes: off, reciprocal, forward non-reciprocal, and reverse non-reciprocal. The calibration and dissipation adjustment unit includes a common calibration bus, a standard sampling resistor, a calibration switching switch, a baseline pre-parallel resistor, and a positive additional resistor. The calibration switching switch is used to switch the bidirectional voltage-controlled current source branch between the state node and the common calibration bus. The standard sampling resistor is connected between the common calibration bus and ground. The baseline pre-parallel resistor and the positive additional resistor are configured around the same baseline conductance to form three passive dissipation states: relative negative dissipation, zero dissipation, and positive dissipation, in a passive network containing only positive resistance. In the baseline state, the baseline pre-parallel resistor is connected, the positive additional resistor is disconnected, and the total node conductance is equal to the baseline conductance. In the relative negative dissipation state, both the baseline pre-parallel resistor and the positive additional resistor are disconnected, and the total node conductance is equal to the baseline conductance minus the conductance modulation amplitude; In the positive dissipation state, both the baseline pre-parallel resistor and the positive additional resistor are connected, and the total node conductance is equal to the baseline conductance plus the conductance modulation amplitude; wherein, both the baseline pre-parallel resistor and the positive additional resistor are implemented by positive resistors, without introducing active negative resistance; The noise and test excitation unit includes an independent simulated noise source and a known test signal source, used to inject random noise current or known disturbance current into the state node; The control and isolation unit includes a main controller, a switch drive circuit, and a state latch interface, which is used to issue and latch the topology, polarity, dissipation, and calibration configuration. After the configuration is completed, the main controller enters a sleep state or is powered off by the digital power supply. The synchronous acquisition unit includes a buffer amplifier, an anti-aliasing filter, and a multi-channel synchronous analog-to-digital converter; it is used to continuously and synchronously record the node voltage, calibration voltage, and configuration status code of all status nodes during the period when the main controller is in sleep mode or the digital power is cut off; the node voltage represents the system state and forms a random dynamic trajectory.

2. The calibrable reciprocal and non-reciprocal stochastic dynamics measurement platform according to claim 1, characterized in that, Also includes: The calculation and output unit is used to reconstruct the drift matrix, dissipation matrix, non-reciprocal matrix and noise matrix based on the trajectory data output by the synchronous acquisition unit, calculate the response quantity, entropy generation and quality index, and map the results to specific nodes, branches and hardware states.

3. The calibrable reciprocal and non-reciprocal stochastic dynamics measurement platform according to claim 2, characterized in that, In the coupling switching unit, each bidirectional voltage-controlled current source branch is composed of a discrete transconductance range group, a latching switch array, and a polarity selection bit cascaded in three stages.

4. The calibrable reciprocal and non-reciprocal stochastic dynamics measurement platform according to claim 1, characterized in that, The calibration switching switch is a single-pole double-throw switch. The common terminal is connected to the output terminal of the bidirectional voltage-controlled current source branch, and the two throw terminals are respectively connected to the target status node and the common calibration bus. All the calibration switching switches are switched in a first-off-then-close manner and a non-overlapping dead zone is set.

5. The calibrable reciprocal and non-reciprocal stochastic dynamics measurement platform according to claim 1, characterized in that, The synchronous acquisition unit operates independently of the main controller, with a sampling rate no less than 20 times the highest characteristic frequency of the system, and a channel synchronization error of less than 0.1 sampling cycles. Each data file simultaneously saves the original voltage trajectory, latch switch and relay status bit mask, calibration coefficient table, capacitance value, temperature, sampling rate, channel mapping, and configuration hash.

6. The calibrable reciprocal and non-reciprocal stochastic dynamics measurement platform according to claim 2, characterized in that, The calculation and output unit performs the following steps: independently preprocesses trajectory data for each hardware state, removes transient segments and analog-to-digital converter saturation segments, converts the data to node voltages and removes the mean; estimates the steady-state covariance matrix; reconstructs the continuous-time drift matrix using the time-delay covariance matrix and the finite-time propagator; decomposes the drift matrix into a symmetric dissipation matrix and an antisymmetric non-reciprocal matrix; and independently estimates the noise covariance intensity matrix using both steady-state Lyapunov identities and discrete innovation residuals. And calculate the linear damping response matrix and steady-state entropy generation rate based on the estimation results.

7. The calibrable reciprocal and non-reciprocal stochastic dynamics measurement platform according to claim 6, characterized in that, The calculation and output unit also performs nonlinear three-state analysis: when the linear propagation residual, the difference between the two noise estimates, or the actual dissipation intervention residual exceeds the threshold, it acquires three sets of matching trajectories under the same damping, noise, temperature, and sampling rate conditions: coupling off, forward non-reciprocal, and reverse non-reciprocal; it estimates the nonlinear drift, steady-state mean, and covariance; and it solves the Poisson equation to obtain the nonlinear damping response matrix. It separates the direction-dependent response that is reversed with the direction of the cycle and the direction-independent response that is not reversed with the direction after the baseline is subtracted.

8. The calibrable reciprocal and non-reciprocal stochastic dynamics measurement platform according to claim 7, characterized in that, The calculation and output unit also performs quality gating: it judges stability, positive covariance definiteness, noise double estimation consistency, analog-to-digital converter linear region, aperiodic crosstalk, cyclic inversion matching, calibration drift, model residuals, and physical verification; if any check fails, the corresponding measurement unit is marked as invalid.

9. A calibrable method for measuring reciprocal and non-reciprocal stochastic dynamics, characterized in that, The measurement platform according to any one of claims 1 to 8 includes the following steps: Step S1: Arrange state nodes and bidirectional voltage-controlled current source branches, so that each state node is connected to a state capacitor, main leakage resistor, switchable dissipation branch, coupling current injection terminal and noise current injection terminal; arrange two voltage-controlled current source branches in opposite directions for each pair of selected state nodes. Step S2: Common calibration for each branch. Switch the voltage-controlled current source branch to be calibrated to the common calibration bus through the calibration switch. Apply at least two known DC voltages to the input terminal of the branch under the condition that the noise source is turned off. Measure the voltage across the standard sampling resistor and convert the injected current. Fit the signed transconductance, zero-point offset current and linear residual of the branch. Step S3: Based on the target measurement task, configure the turn-off, reciprocal self-test, forward non-reciprocal, or reverse non-reciprocal coupling modes by changing the polarity and on / off positions of the bidirectional voltage-controlled current source branch. Step S4: Configure passive dissipation state by switching the connection state of the baseline pre-parallel resistor and the positive additional resistor to form a zero dissipation, relative negative dissipation, or positive dissipation state: Baseline state: the baseline pre-parallel resistor is connected, the positive additional resistor is disconnected, and the total node conductance is equal to the baseline conductance; Relative negative dissipation state: both the baseline pre-parallel resistor and the positive additional resistor are disconnected, and the total node conductance is equal to the baseline conductance minus the conductance modulation amplitude; Positive dissipation state: both the baseline pre-parallel resistor and the positive additional resistor are connected, and the total node conductance is equal to the baseline conductance plus the conductance modulation amplitude; wherein, both the baseline pre-parallel resistor and the positive additional resistor are implemented by positive resistors, without introducing active negative resistance; Step S5: Enter digital silent state, latch all switch states, stop the digital activity of the main controller and put it into sleep or power off, and wait for the analog node to reach steady state; Step S6: During the period when the main controller is in sleep mode or the digital power is cut off, all node voltages, calibration voltages and configuration status codes are collected by an independent multi-channel synchronous analog-to-digital converter to form raw trajectory data; Step S7: Reconstruct the drift matrix, dissipation matrix, non-reciprocal matrix, and noise matrix from the same data block, and calculate the response, entropy generation, and quality indicators; Step S8: Change only the coupling direction or dissipation branch to be compared, and repeat steps S3 to S7. By differentiating the measurement results of the same platform, the same calibration, the same noise, and the same acquisition chain, the measurement of reciprocal and non-reciprocal stochastic dynamic characteristics is completed.

10. The calibrable reciprocal and non-reciprocal stochastic dynamics measurement method according to claim 9, characterized in that, Step S7 also includes: taking the unit vector corresponding to the largest eigenvalue of the steady-state covariance matrix, solving the adjoint equation to obtain the adjoint matrix, and then calculating the linear damping response matrix; and using the drift matrix, noise covariance intensity matrix and steady-state covariance matrix to calculate the steady-state entropy generation rate.

Images