Broadband oscillation characterization platform for multi-chip power modules and near-field signature evaluation method

Abstract

The application discloses a broadband oscillation characterization platform of a multi-chip power module and a near-field characteristic evaluation method, and belongs to the technical field of power electronics. Continuous operation conditions of a power converter are decoupled into a series of discrete switching points, transient near-field radiation data of a to-be-tested power module at the discrete switching points are acquired through double-pulse testing, and a near-magnetic field radiation signal is reconstructed; time-frequency analysis is performed on the reconstructed near-magnetic field radiation signal, and a two-dimensional time-frequency matrix is generated; radiation energy entropy values are calculated based on the two-dimensional time-frequency matrix, and peak amplitudes in the two-dimensional time-frequency matrix are fused, so that a comprehensive EMI index for evaluating broadband oscillation electromagnetic interference risk is generated, and EMI risk early warning is triggered when the comprehensive EMI index exceeds a threshold value. The application quantifies spectrum complexity through radiation energy entropy, performs multi-dimensional EMI risk evaluation by fusing peak amplitudes in a time-frequency matrix, and can accurately and efficiently characterize broadband oscillation inside a multi-chip parallel power module.

Description

Technical Field

[0001] This invention relates to the field of power electronics technology, and in particular to a wideband oscillation characterization platform and near-field characteristic evaluation method for multi-chip power modules. Background Technology

[0002] Currently, power modules are widely used in renewable energy fields such as photovoltaics, wind power, and electric vehicles, and are the core components of power conversion units. As industrial applications place increasingly higher power demands on power conversion units, they are continuously evolving towards multi-chip parallel connection, high switching speed, and high power density.

[0003] However, with the increase in the capacity and speed of power devices, wideband oscillations are prone to occur within multi-chip parallel power modules. The resulting near-field radiation easily couples with surrounding sub-components, reducing the common-mode rejection capability of the isolation driver, leading to severe high-frequency interference and device failure. Furthermore, in high-density power converters, near-field radiation couples with EMI filters near the power module, degrading the converter's filter performance and electromagnetic compatibility. Therefore, accurately characterizing the wideband oscillations within the power module is crucial.

[0004] Among existing broadband oscillation characterization methods, near-field measurement, as a non-contact method, can accurately characterize broadband oscillations within power modules. However, existing near-field measurement techniques are mainly applied to power converter scenarios and rarely used to characterize the operating status of power modules. Furthermore, the acquisition of near-field data is based on the continuous operating conditions of actual power modules, and continuous operation testing under full operating conditions takes tens of hours, resulting in long initial verification cycles, high costs, and the need for sophisticated heat dissipation and control design. Moreover, relying on spectrum analyzers to directly acquire the spectrum provides a single analytical perspective. Since the spectrum data is a comprehensive result of near-field radiation during continuous power module switching on / off, it is difficult to use to distinguish / identify the transient phases and devices that cause high-frequency oscillations. In addition, relying on peak amplitude to assess electromagnetic interference (EMI) ignores the spectral complexity of broadband oscillations, which may lead to an increased misjudgment rate of EMI risks.

[0005] In summary, for multi-chip parallel power modules, how to accurately characterize their internal wideband oscillations, how to conveniently, accurately, and efficiently obtain test results during the initial verification stage of the power module, and how to comprehensively assess the EMI risk of wideband oscillations in the power module have become research hotspots in this field. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of the prior art and provide a wideband oscillation characterization platform and near-field characteristic evaluation method for multi-chip power modules, so as to accurately characterize the wideband oscillation inside the multi-chip parallel power module, and to obtain test results conveniently and efficiently in the initial verification stage of the power module without the need for heat dissipation and control design.

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] In a first aspect, the present invention proposes a broadband oscillation characterization platform and a near-field characteristic evaluation method for a multi-chip power module, comprising the following steps:

[0009] (1) Decouple the continuous operation of the power converter into a series of discrete switching points. For each discrete switching point, obtain the transient near-field radiation data of the power module under test at the discrete switching point through a double pulse test.

[0010] (2) Based on the transient near-field radiation data of discrete switching points, the near-field radiation signal of the power module under test of the power converter under continuous operation is reconstructed by phase alignment and time-domain energy superposition processing.

[0011] (3) Perform time-frequency analysis on the reconstructed near-magnetic field radiation signal to generate a two-dimensional time-frequency matrix in order to identify the transient stage of high-frequency oscillation;

[0012] (4) Calculate the radiated energy entropy value representing the spectral complexity based on the two-dimensional time-frequency matrix, and combine the radiated energy entropy value with the peak amplitude in the two-dimensional time-frequency matrix to generate a comprehensive EMI index for assessing the risk of broadband oscillating electromagnetic interference. When the comprehensive EMI index exceeds the threshold, trigger an EMI risk warning.

[0013] Furthermore, the operating conditions of each discrete switching point are uniquely determined by three parameters: DC bus voltage, load current, and operating junction temperature. The operating junction temperature is calculated based on the DC bus voltage and load current using a device loss thermal impedance model.

[0014] Furthermore, the process of acquiring transient near-field radiation data of the power module under test at the discrete switching points includes:

[0015] (1.1) Initialize the test system, set the input bus voltage, load current, junction temperature, and set the probe displacement range and step size;

[0016] (1.2) The control system parses the displacement command of the operating system, drives the three-axis stepper motor to move the near-field probe along an S-shaped path, ensuring that the path fully covers the power module under test, and uploads the probe position coordinates to the control system in real time;

[0017] (1.3) After the near-field probe moves to a certain target position, it traverses a series of discrete switching point conditions and performs a double pulse test under each condition, and synchronously collects the transient near-field radiation data of the test position and each condition.

[0018] Furthermore, wavelet transform is used to perform time-frequency analysis on the reconstructed near-magnetic field radiation signal, and the wavelet transform uses the Morlet wavelet basis.

[0019] Furthermore, the calculation of radiative energy entropy includes:

[0020] For a fixed time point The radiated energy in a certain frequency band is normalized into a probability distribution. ;

[0021] The time-frequency matrix is ​​divided into sub-bands according to frequency bands, and the proportion of total radiated energy in each sub-band is calculated. ;

[0022] Entropy formula:

[0023]

[0024] Where num represents the total number of subbands; each test point is at a fixed time point. The next value corresponds to an energy entropy value.

[0025] Furthermore, the comprehensive EMI index is generated in the following way:

[0026] The peak amplitude is linearly normalized to the global maximum value;

[0027] The energy entropy is calculated as log2 of the theoretical maximum value. num Normalization, in which num Indicates the total number of sub-bands;

[0028] The normalized peak amplitude and the normalized radiant energy entropy are weighted and summed to obtain the comprehensive EMI index.

[0029] Furthermore, the test system for performing dual-pulse testing integrates the power module under test, an oil heater, a displacement platform, a near-field probe, a voltage probe, a current probe, and an oscilloscope. The oil heater provides a temperature-controlled environment for the power module under test, and the near-field probe is fixed on the displacement platform. The voltage probe is connected in parallel across the power terminals of the power module to capture transient switching voltages. The current probe surrounds a current path in the power circuit to measure load current changes. The near-field probe, voltage probe, and current probe are connected to different channels of the oscilloscope.

[0030] During the dual-pulse test, the oil temperature heater adjusts the power module under test to the target junction temperature, the high-voltage power supply outputs the set bus voltage to the DC bus, the real-time central controller generates a low-voltage test signal, which is amplified by the drive unit and triggers the switch of the power module under test; the real-time central controller sends a trigger signal to the oscilloscope, and simultaneously acquires waveform data from the voltage probe, current probe, and near-field probe. The waveform data is uploaded to the control system in real time and stored according to the position coordinates.

[0031] Furthermore, the displacement platform is controlled by a three-axis stepper motor, with an x / y axis resolution of 0.5mm and a displacement range of 200mm; and a z-axis resolution of 0.5mm and a displacement range of 500mm, adaptable to power modules of different sizes.

[0032] Secondly, this invention proposes a wideband oscillation characterization platform for a multi-chip power module, used to implement the above-mentioned method, including:

[0033] The testing system is used to perform double-pulse tests;

[0034] The operating system is used to receive user input of test conditions, near-field probe displacement range and step size, generate drive pulse trigger commands, and receive test data;

[0035] The control system performs operations through a real-time central controller: adjusting the output voltage of the high-voltage power supply to the DC busbar of the power module under test; generating a low-voltage test signal and transmitting it to the drive unit, which then amplifies it to drive the switch of the power module under test; and communicating and exchanging data with the operating system.

[0036] Preferably, the near-field probe supports modular replacement, and the probe interface supports customized extended frequency bands.

[0037] Compared with existing technologies, the present invention has the following advantages:

[0038] This invention decouples the continuous operating conditions of a power converter into discrete switching points using discrete operating condition reconstruction technology. Near-field radiation data at each discrete point is acquired through dual-pulse testing. Phase alignment and time-domain energy superposition are combined to achieve equivalent reconstruction of the continuous operating conditions, eliminating the need for heat dissipation and control design required by traditional methods and improving testing efficiency. Furthermore, wavelet transform is introduced to extract time-frequency features from the near-magnetic field radiation signal, generating a two-dimensional time-frequency matrix to accurately locate the high-frequency oscillation transient stage and corresponding devices, overcoming the limitations of single-peak analysis by spectrum analyzers. Further, spectral complexity is quantified through radiation energy entropy, and multi-dimensional EMI risk assessment is performed by integrating the peak amplitude of the time-frequency matrix. A normalized weighted comprehensive EMI index is generated, significantly reducing the misjudgment rate caused by traditional single-peak assessment and comprehensively optimizing the EMI risk prediction accuracy during the power module design stage. Attached Figure Description

[0039] Figure 1 This is an overall block diagram of a wideband oscillation characterization platform for a multi-chip parallel power module provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of near-field magnetic field scanning provided according to an embodiment of the present invention;

[0040] Figure 3 This is a flowchart illustrating the process of near-field probe displacement traversing all working conditions and all test positions during near-field magnetic field scanning according to an embodiment of the present invention.

[0041] Figure 4 This is an overall calculation flowchart of the near-field characteristic evaluation method for broadband oscillation of a multi-chip power module provided in an embodiment of the present invention;

[0042] Figure 5 This is a flowchart of a method for evaluating the near-field characteristics of wideband oscillations in a multi-chip power module according to an embodiment of the present invention;

[0043] Figure 6 It is a comparison of the cloud map distribution of peak values, entropy values, and comprehensive evaluation indices;

[0044] Figure 7 These are test data under different bus voltages;

[0045] Figure 8 These are test data at different junction temperatures and voltages. Detailed Implementation

[0046] The present invention will be further described and illustrated below with reference to specific embodiments. The embodiments described are merely examples of the content of this disclosure and do not limit the scope of the invention. The technical features of each embodiment in the present invention can be combined accordingly, provided that there is no mutual conflict.

[0047] This invention provides a wideband oscillation characterization platform for multi-chip power modules, such as... Figure 1 As shown, the system includes a test system, an operating system, and a control system. These three components interact through data flow and control commands to form an automated closed-loop test architecture.

[0048] The testing system includes a power module under test (DUT), an oil heater, a displacement platform, a near-field probe, a voltage / current probe, and an oscilloscope. The oil heater provides a temperature-controlled environment for the DUT. The near-field probe is fixed on the displacement platform, and by controlling the movement of the displacement platform in the X, Y, and Z axes, the near-field probe is precisely positioned above the power module in three-dimensional space. The voltage probe is connected in parallel to the DC bus or the two ends of the switching transistor of the power module to capture the transient voltage of the switch. The current probe is connected in series in the power circuit to measure the load current change and the current sharing status of the chip. The near-field probe and the voltage / current probe are connected to different channels of the oscilloscope.

[0049] The operating system serves as the human-computer interface, receiving user-input test conditions (including bus voltage, load current, operating junction temperature, etc.) through LabVIEW software, defining the displacement range and unit displacement distance (step size) of the near-field probe, generating drive pulse trigger commands, and simultaneously receiving and initially managing test data uploaded from the control system.

[0050] The control system is based on the NI CompactRIO real-time central controller, which communicates bidirectionally with the operating system to exchange test conditions and test data. Under the trigger command of the drive pulse, its execution logic includes: first, regulating the output DC voltage of the high-voltage power supply to the DC bus of the module. Here, the output voltage of the high-voltage power supply is directly connected to the positive / negative DC input terminals of the power module under test through a copper bus or high-voltage cable to provide it with the DC bus voltage required for operation; at the same time, a low-voltage test signal is generated, which is amplified by the power device drive unit and then directly drives the gate of the power device to realize the switching control of the power module under test; finally, a synchronous trigger signal is sent to the oscilloscope to collect data.

[0051] The displacement platform in the testing system is controlled by a stepper motor and a control system. A real-time central controller parses displacement commands from the operating system, driving the three-axis stepper motor displacement platform to move the near-field probe along an S-shaped path. Simultaneously, the current position information is uploaded to the control system to ensure the near-field probe accurately moves to the target position for testing. In this embodiment, the x / y axis resolution of the three-axis stepper control is 0.5 mm, and the displacement range is 200 mm, which meets the size requirements of existing power modules; the z-axis resolution is 0.5 mm, and the displacement range is 500 mm, which can adapt to the testing needs of power modules placed at different heights.

[0052] The user-input test conditions constitute the test operating conditions and can be adjusted online. The test bus voltage adjustment range is 0~10kV, and the test current adjustment range is 0~3.6kA. The test current refers to the transient load current flowing through the main power circuit of the power module under test during a double-pulse test. The device junction temperature adjustment range is 25~150℃, regulated by an oil heater. Near-field probes can be freely replaced according to test bandwidth requirements, such as ETS 901 (790MHz), ETS 902 (1.5GHz), and ETS 903 (2.3GHz). The platform supports customized probe interfaces to adapt to higher frequency band testing needs. 。 The near-field probe directly outputs a time-domain voltage waveform, which, based on Faraday's law of electromagnetic induction, is proportional to the intensity of the alternating magnetic field at the probe's location. Since the bandwidth of the near-field probe is much higher than the frequency range of the broadband oscillations within the power module, it can be used to characterize the broadband oscillation phenomenon within the power module.

[0053] In one specific embodiment of the invention, the near-field probe is connected to an oscilloscope via BNC test leads, and the measured time-domain waveform is displayed. The displacement range of the near-field probe is defined according to the size of the power module under test, and the displacement step size is defined according to the size and resolution requirements of the near-field probe.

[0054] like Figure 3 As shown, the displacement control of the near-field probe is tightly coupled with the dual-pulse test. During the near-field magnetic field scanning process, the displacement of the near-field probe is S-shaped with equal intervals. When the probe moves along the path to... Figure 2 After identifying a target location, a double-pulse test is performed at that location, and the corresponding device voltage, current, and near-field probe time-domain waveform are recorded. The device voltage and current are obtained through voltage / current probe testing, until all test locations are completed. After the oscilloscope captures the test waveforms, the near-field magnetic field test data corresponding to different operating conditions and locations are uploaded to the control system for storage to facilitate post-processing.

[0055] In this invention, the near-field probe moves at equal intervals in an S-shape along the XY direction on the power module under test to ensure that the probe accurately moves to each target test position. The control logic for the probe to traverse the test positions under all operating conditions is as follows:

[0056] After obtaining the platform hardware parameters, probe displacement parameters and test conditions (such as target voltage and current) set by the user, the real-time central controller generates a voltage regulation command, the high-voltage power supply outputs a precise DC voltage, and the busbar of the power module under test is loaded with the target voltage value to provide it with the DC bus voltage required for operation.

[0057] Based on the test conditions input by the user, the real-time central controller drives the near-field probe to initialize its position, i.e., moves it to the preset scan start position. At the same time, it calculates the on-state time of the power device and generates a low-voltage test signal. The low-voltage test signal is a series of digital pulses representing the turn-on / turn-off timing. It is transmitted directly to the power device drive unit through a shielded cable for signal amplification and conversion into a high-voltage drive waveform. The drive pulses directly drive the gate of the power device to realize the switching control of the power device.

[0058] The real-time central controller sends a synchronous trigger signal to collect test data, including voltage, current and near-field radiation data, and records the current position of the near-field probe. After the data of a single test position is collected, the system controls the near-field probe to move one step along an S-shaped path according to the preset displacement parameters to collect test data of a new spatial position. At the new test position, the trigger drive pulse, test data collection and storage, and near-field probe movement process are repeated until the probe has traversed all test positions.

[0059] Once all test positions in the entire scanning area have been tested, the system will further determine whether all preset test conditions have been executed. If there are still unexecuted conditions, new test conditions will be input and the above process will be repeated until all test conditions are completed.

[0060] This invention also provides a near-field characteristic evaluation method for broadband oscillations of multi-chip power modules, applied to the characterization platform described above. It solves the simulation problem of actual continuous operating conditions through equivalence mapping of discrete test data. The overall calculation process is as follows: Figure 4 As shown, the test consists of two parts: offline dual-pulse testing under multiple operating conditions based on a single scan position and online dual-pulse testing under single operating conditions based on the full scan surface. Ultimately, it achieves accurate characterization and risk assessment of the wideband oscillation of the power module.

[0061] (I) Offline testing of dual-pulse under multiple operating conditions based on a single scan position

[0062] This section focuses on deeply traversing various electrical conditions at a single fixed point to analyze the variation of near-field radiation with operating parameters. First, the continuous operation of the power converter is decoupled into multiple discrete switching points, each corresponding to a unique combination of DC bus voltage, load current, and junction temperature. During the discretization process, an operating model of the power converter is established. Based on the control strategy and load conditions, the specific operating point of the switching transistors (such as IGBTs) in the power module during each switching cycle can be theoretically calculated, yielding the bus voltage and transient current. Furthermore, a device loss thermal impedance model is introduced, which can calculate the chip power loss corresponding to the switching point and deduce the junction temperature based on given bus voltage, transient current, switching frequency, duty cycle, and other information. Therefore, each discrete switching point is uniquely determined by three key parameters: DC bus voltage, load current, and junction temperature. For each extracted discrete switching transient point, the electrical and thermal states of the target discrete switching point are reproduced on a test platform by adjusting the high-voltage power supply, load inductor, and oil heater within a short pulse cycle, triggering a switching action. During the specific switching transient triggered by the double-pulse test, the radiation signal from the near-field probe is synchronously acquired to obtain the time-domain radiation waveform. Based on the PWM timing of the actual converter, the accurate position of each discrete data point on the continuous time axis is determined. The occurrence time of each switching transient is calculated according to the actual control strategy. The time-domain radiation waveforms of each discrete transient are superimposed according to their correct time sequence to reconstruct the equivalent near-field radiation characteristics of the power module under continuous operating conditions. It should be noted that during the double-pulse test, the power commutation circuit and drive circuit of the power module under test are consistent with the actual converter, and the operating conditions of the power devices during the switching process are consistent with the specific switching transient process in the actual converter. After completing the traversal of all test positions across the entire scan surface, the test conditions are updated and the near-field probe is redriven. The above steps are repeated until the traversal of all test positions under all operating conditions is completed.

[0063] Since the near-magnetic field radiation of the power module mainly occurs during the switching process, it is closely related to the operating conditions corresponding to the switching moment, including DC bus voltage, load current, operating junction temperature, drive circuit, and power commutation circuit. Therefore, under the premise of ensuring the consistency of the above parameters, the specific moment of each switching transient is calculated according to the PWM modulation strategy of the actual converter. This allows for phase alignment and energy superposition of the discrete test data. By combining multiple discrete operating condition tests and Fast Fourier Transform (FFT) transformations, the broadband oscillation near-magnetic field characteristics of the power module during actual operation of the power converter can be reconstructed. Furthermore, the spectrum waveform generated after reconstruction is compared and analyzed with the spectrum waveform tested by the actual converter to verify the quasi-online feasibility.

[0064] (II) Online testing of dual pulses under single working conditions based on full-scan surface

[0065] After reconstruction, the data processing was primarily performed using MATLAB software. The data processing module integrates time-frequency analysis and radiation sub-band energy entropy calculation functions, generating radiation risk assessment indicators through a preset algorithm.

[0066] Specifically, since the test results are reconstructed continuous near-magnetic field voltage time-domain waveforms, wavelet transform is used to obtain the time-frequency spectrum of the near-magnetic field radiation signal as raw data to identify the moments of strong near-magnetic field radiation and their corresponding instantaneous frequencies. To evaluate the spatiotemporal distribution of radiation intensity, the raw near-magnetic field radiation signal is mapped into a two-dimensional time-frequency matrix. For the time-frequency matrix of each scanned test point, local peak amplitudes are searched point-by-point along the frequency axis within the transient time-domain window of the switch, realizing the extraction of time-domain and frequency-domain variation trends. This can be combined with the physical structure distribution on the power module surface for spatiotemporal frequency feature localization and analysis. An FFT transform is performed on the searched local peak amplitudes to obtain the frequency points closest to the characteristic frequency. The peak amplitude distribution throughout the entire scan area is statistically analyzed, and combined with the generated spatiotemporal amplitude cloud map, the evolution of radiation intensity with location and time can be intuitively reflected.

[0067] Furthermore, considering the broadband characteristics of the oscillations within the power module, the concept of radiative energy entropy is introduced to quantify the spectral complexity of the near-magnetic field radiation signal. Finally, combining near-field measurement results at different locations, an interpolation function is used to plot the electromagnetic frequency-space cloud map of the scanning surface, obtaining the near-field time-frequency-space distribution characteristics within the power module. This data, combined with the physical structure distribution on the power module surface, is used for oscillation radiation source analysis.

[0068] Taking a single double-pulse test result as an example, for near-magnetic field radiation signals... Its CWT mathematical expression is as follows:

[0069]

[0070] in, It is a scaling factor used to control the compression or stretching of wavelets and can reflect frequency components. It is a translation factor used to control the position of the wavelet in the time domain. These are wavelet basis functions. The wavelet basis used here is the Morlet wavelet basis, which is suitable for analyzing high-frequency electromagnetic radiation caused by the switching transients of power devices due to its good time-frequency localization characteristics. This refers to the complex conjugate of wavelet basis functions.

[0071] In this embodiment, CWT calculation can be performed using the Wavelet Toolbox in MATLAB. Its horizontal and vertical axes correspond to time and frequency, and the matrix elements represent the energy density distribution at a specific time and instantaneous frequency.

[0072] Furthermore, radiative energy entropy is introduced to quantify the complexity of energy distribution in the near-magnetic field frequency domain, and the energy is normalized based on the CWT time-frequency matrix to calculate the probability distribution for a fixed time point. The radiated energy in a certain frequency band is normalized to a probability distribution:

[0073]

[0074] The frequency band under study is then divided into multiple sub-bands, and the proportion of total radiated energy in each sub-band is calculated, which is called the sub-band. This is calculated by summing the probability distributions within each sub-band. Finally, the energy entropy value is calculated:

[0075]

[0076] Where num refers to the total number of sub-bands. Each test point, at a fixed time point... The next value corresponds to an energy entropy value.

[0077] Energy entropy can quantify the complexity of frequency domain radiated energy distribution to some extent. For example, an increase in energy entropy indicates that signal energy is spreading from a few main frequency bands to a wider frequency band, which may be caused by different parasitic parameter resonances or instability in the switching process. By mapping the radiated energy entropy value of each spatial point to an entropy distribution map, spatial hotspots of radiative complexity can be identified.

[0078] Existing research typically relies on peak amplitude measurements to assess EMI risk, which only reflects the maximum intensity at a specific frequency point and cannot reflect the global spectral information of broadband signals. In practical applications, EMI limitations in relevant standards are not limited to a single frequency point but span the entire frequency band. Therefore, relying solely on peak amplitude cannot provide a comprehensive understanding of the EMI risks posed by power modules. To comprehensively evaluate the intensity and complexity risks of near-field radiation, this invention uses a normalized weighted combination of peak amplitude and energy entropy exponent. Specifically, the peak amplitude at each scan point is linearly normalized based on the global maximum value, reflecting the proportion of local radiation intensity relative to the worst operating conditions; for energy entropy, it is calculated based on its theoretical maximum value log2. num Normalization is performed to quantify the complexity of the radiated energy distribution in the frequency domain. Finally, a comprehensive EMI index (ECI) is generated by weighting the above two normalized data, comprehensively reflecting the near-field radiation characteristics caused by the broadband oscillation of the power module. In this embodiment, an EMI risk warning should be triggered when the comprehensive EMI index exceeds a threshold. Based on this invention, an empirical database of radiation peak / entropy value - EMI exceedance probability can be further established for risk assessment.

[0079] In summary, based on the constructed wideband oscillation characterization platform, the overall process for evaluating the near-field characteristics of multi-chip parallel power modules is as follows: Figure 5 As shown, the specific steps include the following:

[0080] Step 1: Requirements Definition

[0081] Based on the operating conditions and control strategy of the power converter, obtain the bus voltage, load current and operating temperature information corresponding to each discrete switch of the power device to determine the test conditions of the power device; determine the scanning range of the displacement platform based on the size of the power module, and select different types of near-field probes according to the test bandwidth requirements; determine the scanning displacement step distance based on the probe size and resolution requirements.

[0082] Step 2: Data Collection

[0083] For the aforementioned test conditions (voltage / current / temperature), an oil heater is used to adjust the junction temperature to the required temperature. The voltage and current information to be measured are input into the LabVIEW control interface, which automatically calculates the required pulse width. Simultaneously, the scanning range and step distance of the near-field probe are input into the LabVIEW control interface. The displacement test platform interacts with the host computer to drive the displacement platform to perform a double-pulse test on a single test position. Based on the current near-field probe position information, the probe position is controlled, and the double-pulse test of each test position on the full scanning surface of the power converter under test is completed by traversing an S-shaped route.

[0084] During testing, the near-field probe is connected to the oscilloscope via BNC test leads to measure its time-domain waveform. Tests are performed at each location, and the corresponding device voltage, current, and near-field probe time-domain waveform are recorded until all locations are tested. After the oscilloscope captures the test waveforms, the near-field magnetic field test data corresponding to different operating conditions and locations are uploaded to the control system for storage and post-processing.

[0085] Step 3: Data Processing

[0086] After combining and reconstructing the test data, it is transmitted to MATLAB. A data processing program built using MATLAB can construct the near-field characteristics of power devices under continuous online operation of the full-scan surface of the power converter based on offline discrete test data, and plot the near-magnetic field distribution cloud map. Furthermore, it can present time-domain waveforms and time-frequency spectra under different operating conditions and spatial locations, perform sensitivity analysis, and plot the frequency-space distribution map of radiation peak values ​​and the distribution map of radiation energy entropy values ​​at different times. By weighting and comprehensively reflecting the near-field radiation characteristics caused by the broadband oscillation of the power module, it can further guide the optimization of chip and package layout during the design phase.

[0087] This embodiment uses a certain model of 650V / 400A multi-chip parallel power module: FS400R07A1E3 as the test object, and presents some test results. Figure 6 This comparison of peak amplitude, entropy, and comprehensive evaluation index (ECI) cloud map distributions reveals that for turn-on transients, the peak amplitude radiation cloud map shows a large area of ​​high radiation, covering most of the DBC chip. This broad coverage makes it difficult to pinpoint the exact radiation source and extract detailed features. However, the ECI cloud map narrows the high-radiation area down to the lower diode D1 and the higher 1GBT2, providing a more precise location of the radiation source. Similarly, for zero-current turn-on transients, the ECI distribution cloud map highlights the DC positive and DC negative power terminals, allowing for a more comprehensive and accurate identification of potential radiation sources. This is because it considers both peak amplitude and broadband spectral characteristics, providing a more comprehensive perspective for radiation analysis. In contrast, turn-off transients exhibit relatively low near-field radiation overall, indicating that they pose less of a threat to normal operation compared to turn-on transients.

[0088] Figure 7 These are test data under different bus voltages. Figure 8 The test data at different junction temperatures demonstrates the variation of ECI under different bus voltages and junction temperatures. This comparative analysis reveals a significant behavioral difference between ECI and conventional peak amplitude mapping: although the peak amplitude should monotonically increase with increasing bus voltage and decreasing temperature due to the influence of carrier mobility on the di / dt rate, the ECI distribution exhibits nonlinear anomalies under certain conditions. For example, at 300 V / 25°C, despite a lower peak amplitude, localized ECI hotspots still appear above some chips. This is due to the dispersion of radiative energy in the spectrum caused by parasitic LC resonance activation, which may be caused by voltage-dependent inter-electrode capacitance shift. Since the bus voltage affects the spectral envelope rather than the specific amplitude of a frequency, and also affects the parasitic capacitance between poles, thus affecting the resonant frequency of the commutation loop, energy is redistributed between each sub-band. Meanwhile, high temperatures may prolong the duration of IGBT tail current, but suppress high-frequency ringing due to reduced carrier mobility. This frequency-domain energy dispersion explains why certain operating conditions may exhibit decoupling intensity-complexity behavior. This multidimensional analysis enables targeted EMI mitigation, such as optimizing buffer networks for voltage-sensitive resonators.

[0089] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention. Those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention.

Claims

1. A broadband oscillation characterization platform and near-field characteristic evaluation method for a multi-chip power module, characterized in that, Includes the following steps: (1) Decouple the continuous operation of the power converter into a series of discrete switching points. For each discrete switching point, obtain the transient near-field radiation data of the power module under test at the discrete switching point through a double pulse test. (2) Based on the transient near-field radiation data of discrete switching points, the near-field radiation signal of the power module under test of the power converter under continuous operation is reconstructed by phase alignment and time-domain energy superposition processing. (3) Perform time-frequency analysis on the reconstructed near-magnetic field radiation signal to generate a two-dimensional time-frequency matrix in order to identify the transient stage of high-frequency oscillation; (4) Calculate the radiated energy entropy value representing the spectral complexity based on the two-dimensional time-frequency matrix, and integrate the radiated energy entropy value with the peak amplitude in the two-dimensional time-frequency matrix to generate a comprehensive EMI index for assessing the risk of broadband oscillating electromagnetic interference. When the comprehensive EMI index exceeds the threshold, trigger an EMI risk warning. The calculation of radiative energy entropy includes: For a fixed time point The radiated energy in a certain frequency band is normalized into a probability distribution. ; The time-frequency matrix is ​​divided into sub-bands according to frequency bands, and the proportion of total radiated energy in each sub-band is calculated. ; Entropy formula: ； Where num represents the total number of subbands; each test point is at a fixed time point. The next value corresponds to an energy entropy value; The comprehensive EMI index is generated in the following way: The peak amplitude is linearly normalized to the global maximum value; The energy entropy is normalized to the theoretical maximum value log2num, where num represents the total number of subbands; The normalized peak amplitude and the normalized radiant energy entropy are weighted and summed to obtain the comprehensive EMI index.

2. The broadband oscillation characterization platform and near-field characteristic evaluation method for multi-chip power modules according to claim 1, characterized in that, The operating conditions of each discrete switching point are uniquely determined by three parameters: DC bus voltage, load current, and operating junction temperature. The operating junction temperature is calculated based on the DC bus voltage and load current using a device loss thermal impedance model.

3. The broadband oscillation characterization platform and near-field characteristic evaluation method for multi-chip power modules according to claim 1, characterized in that, The process of acquiring transient near-field radiation data of the power module under test at the discrete switching points includes: (1.1) Initialize the test system, set the input bus voltage, load current, junction temperature, and set the probe displacement range and step size; (1.2) The control system parses the displacement command of the operating system, drives the three-axis stepper motor to move the near field probe along an S-shaped path, ensures that the path fully covers the power module under test, and uploads the probe position coordinates to the control system in real time; (1.3) After the near-field probe moves to a certain target position, it traverses a series of discrete switching point conditions and performs a double pulse test under each condition, and synchronously collects the transient near-field radiation data of the target position and each condition.

4. The broadband oscillation characterization platform and near-field characteristic evaluation method for multi-chip power modules according to claim 1, characterized in that, The reconstructed near-magnetic field radiation signal was subjected to time-frequency analysis using wavelet transform based on the Morlet wavelet basis.

5. The broadband oscillation characterization platform and near-field characteristic evaluation method for multi-chip power modules according to claim 1, characterized in that, The test system for performing dual-pulse testing integrates a power module under test (DUT), an oil heater, a displacement platform, a near-field probe, a voltage probe, a current probe, and an oscilloscope. The oil heater provides a temperature-controlled environment for the DUT, and the near-field probe is fixed on the displacement platform. The voltage probe is connected in parallel across the power terminals of the power module to capture transient switching voltages. The current probe surrounds a current path in the power circuit to measure load current changes. The near-field probe, voltage probe, and current probe are connected to different channels of the oscilloscope. During the dual-pulse test, the oil temperature heater adjusts the power module under test to the target junction temperature, the high-voltage power supply outputs the set bus voltage to the DC bus, the real-time central controller generates a low-voltage test signal, which is amplified by the drive unit and triggers the switch of the power module under test; the real-time central controller sends a trigger signal to the oscilloscope, and simultaneously acquires waveform data from the voltage probe, current probe, and near-field probe. The waveform data is uploaded to the control system in real time and stored according to the position coordinates.

6. The broadband oscillation characterization platform and near-field characteristic evaluation method for multi-chip power modules according to claim 5, characterized in that, The displacement platform is controlled by a three-axis stepper motor, with an x / y axis resolution of 0.5mm and a displacement range of 200mm; and a z axis resolution of 0.5mm and a displacement range of 500mm, adaptable to power modules of different sizes.

7. A wideband oscillation characterization platform for a multi-chip power module, used to implement the method according to any one of claims 1-6, characterized in that, include: The testing system is used to perform double-pulse tests; The operating system is used to receive user input of test conditions, near-field probe displacement range and step size, generate drive pulse trigger commands, and receive test data; The control system executes operations through a real-time central controller: adjusting the output voltage of the high-voltage power supply to the DC busbar of the power module under test; A low-voltage test signal is generated and transmitted to the drive unit, where it is amplified and used to drive the switch of the power module under test; it also communicates and exchanges data with the operating system.

8. The wideband oscillation characterization platform for a multi-chip power module according to claim 7, characterized in that, The near-field probe supports modular replacement, and the probe interface supports customized extended frequency bands.

Images