A transformer and its anti-saturation method

Abstract

This invention relates to the field of switching power supply transformer technology, and discloses a transformer and its anti-saturation method. The transformer includes a cavity frame, within which multiple magnetic core units and winding assemblies are embedded. The winding assembly includes: a primary winding comprising multiple series-connected primary helical coils; and a secondary winding comprising multiple series-connected secondary helical coils. The primary and secondary helical coils are staggered and stacked, with the central axes of all helical coils collinear, and the distance from the central axis of each helical coil to each magnetic core unit is equal. This invention's transformer and its anti-saturation method employ a structure of staggered, multi-layered helical coils with collinear central axes and equidistant magnetic core units to ensure uniform magnetomotive force received by each magnetic core unit. By acquiring multi-dimensional transformer parameters and extracting observation indicators, intervention is performed before the magnetic core units enter the nonlinear region, thereby preventing saturation.

Description

Technical Field

[0001] This invention relates to the field of switching power supply transformer technology, and in particular to a transformer and its anti-saturation method. Background Technology

[0002] In the field of switching power supply transformer technology, especially in high-frequency, high-density multiphase voltage regulation modules that power high-performance computing chips (such as CPUs and GPUs), the magnetic saturation problem of power transformers has always been a core bottleneck restricting the improvement of system performance, reliability and power density.

[0003] Traditional antisaturation methods are mainly divided into two categories:

[0004] One approach is based on the configuration of the magnetic core itself, such as introducing an air gap in the magnetic circuit or using a distributed magnetic core, to reduce the effective permeability and widen the linear operating range. However, this type of method has inherent contradictions: air gaps increase leakage inductance and iron loss, leading to decreased efficiency and increased electromagnetic interference; while traditional distributed magnetic cores (such as discrete magnetic rings) are difficult to achieve extremely high space utilization and perfect parametric symmetry.

[0005] Secondly, there are circuit-based control strategies, such as peak current mode control, which prevent saturation by detecting and limiting the peak primary current in each switching cycle. However, this method relies on current signals for detection, and a significant rise in current indicates that the core has entered or is beginning to saturate, resulting in an inherent control delay. Under extreme conditions of high switching frequencies (MHz level) and ultra-large dynamic load current changes (>500A / μs), this delay may cause the controller to only react after magnetic saturation has actually occurred and damaged the switching transistor, making prevention impossible. Summary of the Invention

[0006] Therefore, the purpose of this invention is to overcome the technical problems in multiphase parallel applications, such as asymmetrical magnetic circuit layout, which easily leads to local magnetic saturation and thermal runaway, and the reliance on current peak detection for anti-saturation control, resulting in large response delays and difficulty in preventing saturation. This invention provides a transformer and its anti-saturation method, which adopts a structure of multi-layer spiral coils with interlaced and collinear central axes and equidistant magnetic core units to ensure that the magnetomotive force received by each magnetic core unit is uniform and consistent. By acquiring multi-dimensional transformer parameters and extracting observation indicators, intervention can be performed before the magnetic core unit enters the nonlinear saturation region to prevent saturation.

[0007] To address the aforementioned technical problems, in one aspect, the present invention provides a transformer, including a cavity frame, wherein multiple magnetic core units and a winding assembly are embedded within the cavity frame, and the winding assembly includes:

[0008] Primary winding, including multiple series-connected primary helical coils;

[0009] Secondary windings, including multiple series-connected secondary helical coils;

[0010] The primary spiral coil and the secondary spiral coil are stacked alternately, with the central axes of each spiral coil being collinear, and the distance from the central axis of the spiral coil to each magnetic core unit is equal.

[0011] Preferably, the primary helical coil and the secondary helical coil are respectively distributed on corresponding substrates, and their rotation directions are opposite and mirror symmetrical.

[0012] Preferably, the magnetic core unit has an air gap groove, the length of which is equal to the height of the magnetic core unit.

[0013] Preferably, the cavity frame is a thick-walled circular ring;

[0014] Along the axial direction of the thick-walled ring, multiple annular grooves are formed on the inner ring surface for mounting the primary helical coil and the secondary helical coil; the centers of each annular groove are equally spaced on the central axis of the inner ring surface;

[0015] Multiple mounting through holes are equally spaced along the circumference of the thick-walled ring for mounting the magnetic core unit;

[0016] A heat dissipation hole array is provided inside the thick-walled circular ring. One side of the heat dissipation hole array is connected to the mounting hole, and the other side of the heat dissipation hole array is connected to the heat dissipation component. The heat dissipation component includes multiple heat dissipation fins, which are evenly spaced on the outer ring surface of the thick-walled circular ring.

[0017] On the other hand, the present invention provides a method for preventing saturation of the above-mentioned transformer, comprising the following steps:

[0018] Obtain transformer parameters; the transformer parameters include the primary current signal of the primary winding, the magnetic state signal of the core of each phase, and the operating temperature of the core of each phase;

[0019] Based on the transformer parameters, observation indicators are extracted; the observation indicators include the amplitude of odd harmonics and the magnetic state imbalance.

[0020] Based on the observed indicators, it is determined whether the core unit is about to enter the nonlinear saturation region. In response to the core unit entering the nonlinear saturation region, targeted control is performed to keep the operating position of the core unit within the linear region of the BH curve and avoid magnetic saturation.

[0021] Preferably, the magnetic state imbalance includes the following formula:

[0022] ;

[0023] In the formula, The magnetic state imbalance of the core unit; The magnetic state signal of any magnetic core unit within one switching cycle; for The maximum value; for The minimum value; for The average value.

[0024] Preferably, determining whether the magnetic core unit is about to enter the nonlinear saturation region based on observation indicators includes:

[0025] The amplitude of the odd harmonic is compared with the first preset threshold. When the amplitude of the odd harmonic is greater than the first preset threshold, it is determined that the magnetic core unit is about to enter the nonlinear saturation region.

[0026] The method for determining the first preset threshold includes: using a temperature threshold relationship table, dynamically updating the first preset threshold according to the core operating temperature; the temperature threshold relationship table includes the third harmonic amplitude of the power phase when the BH curve of the core unit reaches the nonlinear inflection point at each calibration temperature.

[0027] Compare the magnetic state imbalance with the second preset threshold. When the magnetic state imbalance is greater than the second preset threshold, it is determined that the magnetic core unit is about to enter the nonlinear saturation region.

[0028] The second preset threshold is determined based on the mean and standard deviation of the sample set of the average growth rate of the second observation index over multiple consecutive switching cycles.

[0029] Preferably, the targeted regulation includes:

[0030] Step SS31: When the amplitude of the odd harmonic is greater than the first preset threshold, determine the first target control phase based on the primary current signal of the primary winding, and adjust the peak current reference voltage of the first target control phase, as well as the duty cycle of other power phases.

[0031] Step SS32: When the magnetic state imbalance is greater than the second preset threshold, determine the second target control phase according to the magnetic core magnetic state signal, and adjust the phase of the second target control phase so that the second target control phase deviates from the adjacent power phase with a larger magnetic flux density in time.

[0032] Preferably, step SS31 includes:

[0033] Based on the first observation index, the power phase with the largest average current is selected as the first target control phase;

[0034] Query the overload index reduction coefficient mapping relationship table, and determine the current limit reduction coefficient based on the overload value of the first observed index;

[0035] Update the peak current reference voltage of the first target control phase based on the current limit reduction coefficient;

[0036] Based on the original average current of the first target control phase and the current limit reduction coefficient, determine the total compensation current of the transformer and update the duty cycle of other power phases.

[0037] Preferably, step SS32 includes:

[0038] Based on the magnetic state signal of the magnetic core, the magnetic core unit with the highest magnetic field strength is selected as the second target unit;

[0039] Query the phase magnetic circuit mapping table to identify one or more power phases with the highest coupling weight to the second target unit, and mark them as the second target control phase;

[0040] Based on the second observation index, calculate the phase adjustment amount of the second target control phase;

[0041] Based on the phase adjustment amount, the phase angle of the second target control phase is updated, causing the switching event of the second target control phase to move in the opposite direction to the adjacent power phase with a larger magnetic flux density.

[0042] Compared with the prior art, the above-described technical solution of the present invention has the following advantages:

[0043] In high-frequency, high-density, multiphase applications, traditional transformers suffer from uneven magnetic reluctance and flux distribution across phases due to asymmetrical magnetic circuit layouts, which can easily lead to localized magnetic saturation and thermal runaway. The transformer of this invention employs a structure of multi-layered, staggered spiral coils with collinear central axes and equidistant magnetic core units. This ensures that each core unit receives a uniform magnetomotive force, achieving symmetrical magnetic circuit parameters across phases and providing a structural foundation for current sharing and saturation resistance.

[0044] Traditional anti-saturation control relies on current peak detection, which has a large response delay and is difficult to prevent saturation. The anti-saturation method of this invention acquires multi-dimensional transformer parameters (such as primary current signal, core magnetic state signal, and core operating temperature) and extracts early warning indicators (i.e., observation indicators, such as the amplitude of odd harmonics and magnetic state imbalance) to predict saturation trends and intervene before the core unit enters the nonlinear saturation region, thereby preventing saturation. Attached Figure Description

[0045] To make the content of this invention easier to understand, the invention will be further described in detail below with reference to specific embodiments and accompanying drawings, wherein:

[0046] Figure 1 This is a partial structural diagram of the transformer in an embodiment of the present invention.

[0047] Figure 2This is a schematic diagram of a winding assembly in an embodiment of the present invention.

[0048] Figure 3 This is a schematic diagram of the structure of the primary helical coil in an embodiment of the present invention.

[0049] Figure 4 This is a schematic diagram of a cavity skeleton in an embodiment of the present invention.

[0050] Figure 5 This is a flowchart illustrating a method for preventing transformer saturation in an embodiment of the present invention.

[0051] Explanation of reference numerals in the accompanying drawings: 1. Cavity frame; 11. Outer annular surface; 12. Inner annular surface; 13. Mounting through hole; 14. Heat dissipation fins; 15. Annular groove; 2. Magnetic core unit; 3. Winding assembly; 30. Substrate; 31. Second through hole; 32. First through hole; 33. Primary helical coil; 34. Secondary helical coil. Detailed Implementation

[0052] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, so that those skilled in the art can better understand and implement the present invention. However, the embodiments described are not intended to limit the present invention.

[0053] Existing predictive saturation methods, such as detecting excitation current distortion or using pulse injection, either still rely on indirect analysis of electrical signals, resulting in limited early warning; or they interfere with normal power transmission, making them difficult to apply to scenarios with stringent power quality requirements, such as high-frequency, high-density multiphase parallel applications. Therefore, this invention proposes a transformer and its anti-saturation method.

[0054] Example 1: This example discloses a transformer.

[0055] To address the issues of transient saturation and uneven current caused by parasitic parameter asymmetry, thermal imbalance, and control delay in high-frequency, ultra-high-density multiphase parallel applications at the MHz level, the transformer in this embodiment includes: a cavity frame 1, multiple magnetic core units 2, and a winding assembly 3.

[0056] refer to Figure 1 In this embodiment, the magnetic core unit 2 is disposed within the cavity frame 1. In this embodiment, multiple magnetic core units 2 are equally spaced on the circumference of the same circle (forming a ring magnetic core), and all magnetic core units 2 are from the same production batch and have undergone magnetic performance screening and pairing to ensure that the deviation of the initial permeability and saturation magnetic flux density of the magnetic core units 2 is less than ±2%.

[0057] The initial permeability refers to the permeability of core unit 2 in the low magnetic field strength, unsaturated linear operating region, reflecting the ease with which core unit 2 is magnetized in the initial stage. A higher initial permeability means a stronger magnetic field (magnetic flux density) can be generated under the same excitation (ampere-turns), and the inductance is usually also larger. This embodiment ensures consistent magnetic impedance across all magnetic circuit branches by selecting core units 2 with consistent initial permeability, facilitating current sharing.

[0058] Furthermore, saturation magnetic flux density refers to the magnetic flux density value when the magnetization of core unit 2 reaches its limit. When the applied magnetic field strength continues to increase, the magnetic flux density almost stops increasing, core unit 2 enters a saturation state, and the permeability drops sharply. Saturation magnetic flux density is the absolute upper limit of the operation of core unit 2. This embodiment, by selecting core units 2 with consistent saturation magnetic flux density, ensures that all core units 2 have the same anti-saturation capability ceiling, avoiding premature saturation of individual core units 2 that could become a system bottleneck.

[0059] In application, to improve the efficiency and stability of the transformer at MHz-level switching frequencies, the magnetic core unit 2 in this embodiment is made of high-frequency, low-loss, high-saturation magnetic flux density high-performance ferrite (e.g., TDK's PC95). Furthermore, the magnetic core unit 2 in this embodiment has a columnar structure. In some preferred embodiments, the magnetic core unit 2 includes an inner arc surface and an outer arc surface, with the outer arc surface close to the outer ring surface 11 and the inner arc surface away from the outer ring surface 11 and close to the inner ring surface 12. In addition, the radii of curvature of the inner and outer arc surfaces are the same as those of the inner ring surface 12 and the outer ring surface 11 to ensure a tight fit between the magnetic core unit 2 and the cavity frame 1, reducing magnetic circuit discontinuities and leakage flux. Furthermore, the longitudinal section of the magnetic core unit 2 is rectangular, and the cross-sectional area is determined based on the single-phase power and switching frequency. Preferably, after calculating the minimum cross-sectional area based on the ratio of single-phase power to switching frequency, the cross-sectional dimensions of a single magnetic core unit 2 are determined based on the total length of the magnetic circuit and the number of magnetic core units 2 to achieve a balanced optimization of magnetic flux density and temperature rise.

[0060] In practical applications, the BH curve, with magnetic field strength on the horizontal axis and magnetic flux density on the vertical axis, is used to describe the magnetization characteristics of magnetic materials. Specifically, it represents the degree of magnetization of a material (e.g., core unit 2) under different magnetic field strengths, including the linear and nonlinear saturation regions. When the operating position of core unit 2 is in the linear region, the BH curve is a straight line. As core unit 2 begins to approach saturation, the BH curve enters the nonlinear saturation region.

[0061] To introduce high reluctance, reduce effective permeability, and soften the slope of the BH curve to widen the linear operating region and delay saturation, the core unit 2 in this embodiment is provided with an air gap groove. In some preferred embodiments, the air gap groove is located at the center of the inner arc surface of the core unit 2. Furthermore, the length of the air gap groove is equal to the height of the core unit 2, and the width of the air gap groove can be from 0.5 mm to 1 mm.

[0062] The magnetic reluctance of the air gap groove is much greater than that of the magnetic core unit 2, which reduces the effective permeability of the entire magnetic circuit. Reducing the effective permeability makes the slope of the BH curve gentler, meaning that under the same magnetic field strength, a lower magnetic flux density is generated, moving the operating point away from the saturation region.

[0063] In actual implementation, the air gap groove is ground out. In order to avoid the diffusion flux at the edge of the air gap from inducing eddy currents on the surface of the exposed magnetic core unit 2 under high-frequency alternating magnetic field, causing additional eddy current loss and local heating, it is necessary to coat the magnetic core unit 2 with ceramic paste after the air gap is ground out. After the ceramic paste is cured, it forms an insulating layer, which can block the eddy current path and reduce high-frequency loss.

[0064] refer to Figure 1 In this embodiment, the winding assembly 3 is disposed within the cavity frame 1. (See reference...) Figure 2 Along the axial direction (e.g., the Z-axis direction) of the thick-walled annulus, the winding assembly 3 of this embodiment includes multiple layers (e.g., 12 layers) of overlapping primary spiral coils 33 (e.g., L1, L3, L5, L7, L9, L11 layers) and secondary spiral coils 34 (e.g., L2, L4, L6, L8, L10, L12 layers). The central axes of all spiral coils are collinear, and the distance from the central axis of the spiral coil to each magnetic core unit 2 is equal. Furthermore, the magnetomotive force generated by the current in each turn of the primary spiral coil 33 and the secondary spiral coil 34 is uniformly and proportionally coupled to each magnetic core unit 2 to ensure a uniform distribution of magnetomotive force in each magnetic circuit branch.

[0065] In application, the primary helical coil 33 and the secondary helical coil 34 rotate in opposite directions (see reference). Figure 3 For example, the primary spiral coil 33 rotates clockwise, and the secondary spiral coil 34 rotates counterclockwise, meaning they are mirror images of each other. Furthermore, the spiral coil patterns of the primary spiral coil 33 and the secondary spiral coil 34 are distributed (e.g., printed) on corresponding substrates 30. The substrate 30 can be a circular ceramic substrate, and the centers of each layer of substrate 30 are evenly spaced along the central axis of the cavity frame 1. Even further, each spiral coil pattern is an Archimedean spiral, with its inner end starting near the center of the substrate 30 and its outer end spiraling outwards from the substrate 30.

[0066] In practical applications, to improve both space utilization and coupling efficiency, a first through-hole 32 is provided at the center of the substrate 30 in this embodiment, and a second through-hole 31 is provided on the outer disk of the substrate 30. Furthermore, the outer ends of all primary helical coils 33 are connected in series through the first through-hole 32 and the second through-hole 31 to form a primary winding; and the inner ends of all secondary helical coils 34 are connected in series to form a secondary winding. Specifically:

[0067] Primary winding formation: Current flows in from the inner end (near the center of the substrate 30) of the bottommost primary spiral coil 33 (e.g., L1 layer), travels along the spiral to the outer end (outer disk of the substrate 30), and then jumps vertically upward through the second through-hole 31 to the outer end of the next layer's primary spiral coil 33 (e.g., L3 layer), then travels along the spiral inward. This process is repeated. Through the first through-hole 32, the inner end of L3 layer connects to the inner end of L5 layer; through the second through-hole 31, the outer end of L5 layer connects to the outer end of L7 layer; through the first through-hole 32, the inner end of L7 layer connects to the inner end of L9 layer; and through the second through-hole 31, the outer end of L9 layer connects to the outer end of L11 layer, thus connecting all primary spiral coils 33 in series to form the primary winding. The primary winding is connected to the input terminal through the inner ends of L1 layer and L11 layer.

[0068] Secondary winding formation: Current flows in from the inner end (near the center of the substrate 30) of the top-level spiral coil 34 (e.g., L12 layer), travels along the spiral to the outer end (outer disk of the substrate 30) of the bottom-level spiral coil 34, and then jumps vertically upward through the second through-hole 31 to the outer end of the next-level spiral coil 34 (e.g., L10 layer), and then travels along the spiral inward. This process is repeated. Through the first through-hole 32, the inner end of L10 layer is connected to the inner end of L8 layer; through the second through-hole 31, the outer end of L8 layer is connected to the outer end of L6 layer; through the first through-hole 32, the inner end of L6 layer is connected to the inner end of L4 layer; and through the second through-hole 31, the outer end of L4 layer is connected to the outer end of L2 layer, thus connecting all secondary spiral coils 34 in series to form the secondary winding. The secondary winding is connected to the output terminal through the inner ends of L2 layer and L12 layer.

[0069] In practical implementation, the magnetic field lines in this embodiment circulate circumferentially along the toroidal core, passing vertically through all winding layers to form a three-dimensional closed loop. This ensures uniform magnetic flux distribution and even transmission of magnetic field lines between core units 2, avoiding excessively high local magnetic flux density. Specifically: Starting from any core unit 2 (e.g., unit A1), magnetic field lines originate vertically downwards from the bottom of unit A1, pass through all helical coils, and enter the top of the opposite core unit 2 (e.g., unit A2). Subsequently, the magnetic field lines flow horizontally along the toroidal core from unit A2 to the adjacent core unit 2 (e.g., unit A3). Then, the magnetic field lines rise vertically upwards from the top of unit A3, pass through all helical coils, and return to the bottom of the core unit 2 (e.g., unit A4) directly opposite unit A3. This cycle repeats, forming a continuous, closed loop in three-dimensional space that travels circumferentially along the toroidal core and repeatedly passes vertically through the winding assembly 3 region.

[0070] In this embodiment, the cavity frame 1 has a ring-shaped symmetrical structure and is made of aluminum nitride ceramic with high thermal conductivity and low radio frequency loss, which is suitable for high frequency and high power application environments.

[0071] In application, the cavity frame 1 of this embodiment is a thick-walled circular ring (i.e., a thick-walled annular symmetrical structure), and the inner ring surface 12 and the outer ring surface 11 of the thick-walled circular ring are concentric cylinders. Further, refer to... Figure 4 Along the circumference of the thick-walled ring, the cavity frame 1 of this embodiment has a plurality of mounting through holes 13 (e.g., 6 to 8) spaced at equal intervals. The number of mounting through holes 13 corresponds one-to-one with the number of magnetic core units 2, and the cross-sectional shape of the mounting through holes 13 is the same as that of the magnetic core units 2. Each mounting through hole 13 is used to mount one magnetic core unit 2. Furthermore, the distance from the center point of each mounting through hole 13 to the center of the thick-walled ring is equal. Further, the thickness of the thick-walled ring is 5 mm to 15 mm, and its height is slightly higher than that of the magnetic core units 2. Preferably, the height of the mounting through holes 13 is equal to the height of the magnetic core units 2.

[0072] In practical applications, refer to Figure 4 Along the axial direction (e.g., the Z-axis) of the thick-walled annulus, a plurality of annular grooves 15 (e.g., 12) are equally spaced on the inner annular surface 12, with the centers of each annular groove 15 evenly distributed along the central axis of the inner annular surface 12. Further, the annular grooves 15 are used to mount the winding assembly 3. Specifically, a portion of the outer disk of the substrate 30, on which the spiral coil is printed, is embedded within the annular groove 15, so that the primary spiral coil 33 and the secondary spiral coil 34 are alternately stacked in the cavity in the middle of the thick-walled annulus, and the distance from each coil turn to each magnetic core unit 2 is equal. Further still, the cross-section of the annular groove 15 is rectangular, and the depth and width of the annular groove 15 are determined according to the current carrying capacity. Specifically: the width of the annular groove 15 is determined according to the ratio of the effective current value to the allowable current density. The depth of the annular groove 15 is determined according to the sum of the thickness of the substrate 30 and the thickness of the spiral coil.

[0073] In actual implementation, refer to Figure 4 The outer ring surface 11 of the thick-walled annulus is provided with a heat dissipation assembly, which includes multiple equally spaced heat dissipation fins 14 to enhance convective heat dissipation by increasing the surface area. A heat-perforation array is provided inside the thick-walled annulus. One side of the heat-perforation array is connected to a mounting hole 13, and the other side is connected to the root of the heat dissipation fins 14, so as to directionally conduct the heat generated by the magnetic core unit 2 to the external environment to eliminate local hot spots in the transformer. Furthermore, the heat-perforation array is filled with a highly thermally conductive material (such as silver paste or copper pillars).

[0074] Example 2: Based on Example 1, this example discloses a transformer.

[0075] In this embodiment, the transformer cavity frame 1 has multiple magnetic core units 2 and winding assemblies 3 embedded within it. The winding assembly 3 includes a primary winding and a secondary winding. The primary winding includes multiple primary spiral coils 33 connected in series, and the secondary winding includes multiple secondary spiral coils 34 connected in series. Furthermore, the primary spiral coils 33 and the secondary spiral coils 34 are staggered, with the central axes of all spiral coils collinear, and the distance from the central axis of each spiral coil to each magnetic core unit 2 is equal.

[0076] The cavity frame 1 of this embodiment has an electrical interface area, within which high-density metallized vias (formed using high-density interconnect technology) are provided for vertically leading the independent ends of the multi-layer spiral coils disposed inside the cavity frame 1 to the surface of the cavity frame 1. Furthermore, to support complex multi-phase parallel, interleaved parallel, or midpoint tap topologies, the electrical start and end points of each layer of spiral coil are independently led out to the electrical interface area on the surface of the cavity frame 1. For example, a 12-layer spiral coil can be configured to operate in parallel with 2, 3, 4, 6, or even 12 phases, thereby adapting to different power levels and dynamic response requirements without changing the hardware. Even further, on the surface of the electrical interface area, the ends led out by the high-density metallized vias form a high-density pad array.

[0077] In some embodiments, the primary side of the transformer is coupled to a first gallium nitride (GaN) power chip, and the secondary side of the transformer is coupled to a second GaN power chip to support MHz-level high-frequency switching and synchronous rectification, thereby improving overall efficiency and power density. Further, multiple first GaN power chips are connected to the primary side of the transformer, and multiple second GaN power chips are connected to the secondary side. The multiple first GaN power chips drive the primary winding of the transformer in a multi-phase interleaved manner, and the multiple second GaN power chips control the output of the secondary winding of the transformer in a synchronous rectification manner. Each first GaN power chip corresponds one-to-one with its corresponding second GaN power chip, and each first GaN power chip works in conjunction with its corresponding second GaN power chip to form a multi-phase power circuit.

[0078] In some embodiments, a magnetic flux sensor is provided in each mounting through hole 13. Furthermore, the magnetic flux sensor is attached to or embedded in the surface of each magnetic core unit 2 to collect the magnetic field state inside the magnetic core unit 2, while attenuating stray magnetic field interference (such as noise from adjacent windings or switching nodes) in the space to the greatest extent possible, ensuring the authenticity and accuracy of the collected signal.

[0079] In some embodiments, a temperature sensor is provided in each mounting through-hole 13. Furthermore, the temperature sensor is disposed on the surface of the magnetic core unit 2 for monitoring the operating temperature of the magnetic core unit 2.

[0080] Example 3: This example discloses a method for preventing transformer saturation.

[0081] The transformer in this embodiment can be the transformer described in Embodiment 1 or Embodiment 2. The anti-saturation method in this embodiment includes steps SS1 to SS3, as referred to... Figure 5 .

[0082] Step SS1: Obtain transformer parameters.

[0083] In application, the transformer parameters include the primary current signal of the primary winding, the magnetic state signal of the core of each phase, and the operating temperature of the core of each phase.

[0084] In practical applications, the primary current signal flowing through the primary winding of the transformer is collected in real time, and the magnetic state signal of the magnetic core unit 2 collected from each magnetic flux sensor and the operating temperature signal of the magnetic core collected from the temperature sensor are read simultaneously.

[0085] In actual implementation, a miniature Hall effect sensor is used as the flux sensor. Furthermore, in order to capture the instantaneous details of the current without distortion, a current sensor is connected in series at the input terminal of the primary winding of the transformer to collect the primary current signal, and the bandwidth of the current sensor is not less than 10 times the switching frequency.

[0086] Step SS2: Extract observation indicators based on the transformer parameters.

[0087] In application, the observation indicators include a first observation indicator and a second observation indicator. Further, the amplitude of the odd harmonics of the primary current signal is extracted, and the amplitude of the odd harmonics is used as the first observation indicator. The magnetic state imbalance of core unit 2 is calculated based on the magnetic state signals of each phase, and the magnetic state imbalance is used as the second observation indicator. The magnetic state signals can be magnetic flux, magnetic flux density, magnetic field strength, and induced voltage. Among them, the induced voltage reflects the rate of change of magnetic flux.

[0088] In practical applications, when the core unit 2 is in the linear region, the BH curve is a straight line, and the inductance L is constant. When a sinusoidal (or approximately sinusoidal) voltage is applied, the primary current signal is a sinusoidal wave with extremely low harmonic content. As the core unit 2 begins to approach saturation, the BH curve enters the nonlinear saturation region, and the inductance L begins to decrease. At this time, the current waveform of the primary current signal changes from a sinusoidal wave to a spike. Fourier analysis of the distorted primary current signal reveals that the amplitudes of odd harmonics (especially the third and fifth harmonics) increase significantly. The amplitudes of the third and fifth harmonics are very sensitive to small nonlinear changes and are early warning signals of saturation. Therefore, in this embodiment, the amplitude of the odd harmonics is the amplitude of the third harmonic and / or the amplitude of the fifth harmonic.

[0089] Furthermore, by monitoring the changes in the magnetic state imbalance of core unit 2, the trend of uneven magnetic state distribution can be identified, providing early warning before localized magnetic flux accumulation leads to saturation, thus enabling preventative control. Specifically, the magnetic state imbalance of core unit 2 includes the following formula:

[0090] ;

[0091] In the formula, The magnetic state imbalance of the core unit; The magnetic state signal of any magnetic core unit within one switching cycle; for The maximum value; for The minimum value; for The average value.

[0092] Step SS3: Determine whether the magnetic core unit 2 is about to enter the nonlinear saturation region based on the observed indicators. In response to the magnetic core unit 2 being about to enter the nonlinear saturation region, perform targeted control to keep the working position of the magnetic core unit 2 within the linear region of the BH curve and avoid magnetic saturation.

[0093] When applied, in response to the first observation index (amplitude of odd harmonics) being greater than the first preset threshold, it is determined that the magnetic core unit 2 is about to enter the nonlinear saturation region, and a feedforward correction command is generated for targeted control.

[0094] Furthermore, the saturation magnetic flux density and permeability of the core unit 2 are sensitive to temperature. Specifically, the saturation magnetic flux density decreases as the temperature increases. That is, under the same magnetic field strength, at high temperatures, the same current value will cause the core unit 2 to approach saturation. Therefore, in this embodiment, a first preset threshold is determined based on the core operating temperature of each phase.

[0095] Furthermore, to overcome the influence of temperature drift on the characteristics of the core unit 2 and improve the accuracy of early warning, the method for determining the first preset threshold in this embodiment includes: dynamically updating the first preset threshold based on the operating temperature of the core using a temperature threshold relationship table. Specifically, the temperature threshold relationship table includes the third harmonic amplitude of the power phase when the BH curve of the core unit 2 reaches the nonlinear inflection point at each calibration temperature.

[0096] In some embodiments, to issue an early warning before the core unit 2 actually enters the nonlinear saturation region, and to allow sufficient response time for control to achieve true feedforward control, this embodiment builds a test circuit on the transformer sample. At each calibration temperature, the excitation is gradually increased (e.g., by increasing the input voltage or duty cycle), while the primary current signal is monitored and the corresponding third harmonic amplitude is calculated. When the BH curve reaches the nonlinear inflection point (i.e., the point where the permeability drops to 80% of the initial value, or the point where the magnetic flux density reaches 90% of the saturation magnetic flux density), the corresponding third harmonic amplitude is recorded as the first preset threshold corresponding to that calibration temperature. A temperature threshold relationship table is constructed based on each calibration temperature and the corresponding first preset threshold.

[0097] In practical applications, the first preset threshold is read from the temperature threshold relationship table based on the core operating temperature of core unit 2. The amplitude of the odd harmonic is compared with the first preset threshold. When the amplitude of the odd harmonic is greater than the first preset threshold, it is determined that core unit 2 is about to enter the nonlinear saturation region. In response to core unit 2 entering the nonlinear saturation region, a feedforward correction command is generated for targeted control.

[0098] Furthermore, to be applicable to dynamic load scenarios and to identify abnormal acceleration trends in the magnetic state, this embodiment determines that the magnetic core unit 2 is about to enter the nonlinear saturation region if the growth rate of the second observed index exceeds a second preset threshold over multiple consecutive cycles. Even further, if the average growth rate of the second observed index exceeds the second preset threshold over N consecutive switching cycles (N≥3, e.g., 5), the magnetic core unit 2 is determined to be about to enter the nonlinear saturation region. Specifically, the second preset threshold includes the following formula:

[0099] ;

[0100] In the formula, For safety, the factor can be 3 to 5; The mean of the sample set of average growth rates of the second observed index over N consecutive switching cycles; Let be the standard deviation of the sample set of average growth rates of the second observed indicator over N consecutive switching cycles.

[0101] In CPU / GPU power supply scenarios, the power supply must respond to load jumps of hundreds of amperes within 1 microsecond. Traditional solutions, once triggering protection shutdown, even within just a few microseconds, can cause CPU voltage drops and calculation errors (blue screen / system crash). To maintain overall output performance and stability without reaching the nonlinear saturation region, this embodiment performs rapid and precise current limiting and global compensation before actual saturation occurs, keeping the total output power and voltage constant, thus rendering the load imperceptible.

[0102] In some embodiments, when the amplitude of the odd harmonic is greater than a first preset threshold, the targeted control includes step SS31. Step SS31: When the amplitude of the odd harmonic is greater than the first preset threshold, a first target control phase is determined based on the primary current signal of the primary winding, and the peak current reference voltage of the first target control phase, as well as the duty cycle of other power phases, are adjusted. Specifically, step SS31 includes steps SS311 to SS314.

[0103] Step SS311: Based on the first observation index (primary current signal of the primary winding), the power phase with the largest average current is selected as the first target control phase (e.g., phase m).

[0104] In application, based on the primary current signal of the primary winding, the first average current of each power phase is calculated. The power phase with the largest first average current is selected as the first target control phase to directly intervene in the power phase most prone to saturation. Specifically, to avoid misjudgments caused by instantaneous fluctuations and to ensure the accuracy of the selection of the first target control phase, the first average current includes the following formula:

[0105]

[0106] In the formula, For the number The power phase, for Integers in This represents the total number of power phases. For switching cycles; For the number The instantaneous current signal of the power phase; For the number The first average current of the power phase.

[0107] Step SS312: Query the overload index reduction coefficient mapping relationship table, and determine the current limit reduction coefficient based on the overload value of the first observed index.

[0108] In application, the overload index reduction coefficient mapping table includes the correspondence between overload values ​​and suggested reduction coefficients. Wherein, overload value = first observed index - first preset threshold. The suggested reduction coefficient corresponds to the current limit reduction coefficient.

[0109] Step SS313: Update the peak current reference voltage of the first target control phase according to the current limit reduction coefficient.

[0110] In application, excessive harmonics are often caused by excessive current. The saturation tendency can be suppressed by directly limiting the current amplitude of that phase by adjusting the peak current reference voltage. Furthermore, the peak current reference voltage of the second target control phase includes the following formula:

[0111] ;

[0112] In the formula, The peak current reference voltage; This is the current limit adjustment factor, typically ranging from 0.01 to 0.05, to achieve small, gradual adjustments and avoid output voltage oscillations caused by sudden changes.

[0113] In practical applications, the updated peak current reference voltage is written into the digital-to-analog converter register of the digital multiphase controller corresponding to the first target control, and the peak current limit of the first target control phase then decreases proportionally.

[0114] Step SS314: Determine the total compensation current of the transformer based on the original average current of the first target control phase and the current limit reduction coefficient, and update the duty cycle of other power phases.

[0115] In application, in order to maintain the overall stable output of the system, the total compensation current of the transformer is determined based on the first average current of the first target control phase and the current limit reduction coefficient. The duty cycle increment of other power phases is determined based on the total compensation current, and the duty cycle of the corresponding power phase is updated based on the duty cycle increment to ensure that the total output current of the system remains unchanged while reducing the load of the first target control phase.

[0116] Furthermore, the total compensation current includes the following formula:

[0117] ;

[0118] In the formula, This is the total compensation current; This is the current limit reduction factor, typically ranging from 0.01 to 0.05; The first average current of the first target control phase.

[0119] Furthermore, the duty cycle increments for other power phases include the following formula:

[0120] ;

[0121] ;

[0122] In the formula, This represents the total number of power phases. The numbering of the first target control phase; For the number The power phase, for Integers in, and ; For the number The current increment of the power phase; For the number The duty cycle increment of the power phase; For the number The current duty cycle conversion factor for the power phase is expressed in units of duty cycle per ampere. Furthermore, the current duty cycle conversion factor can be determined based on the inductance value, switching frequency, input voltage, and output voltage of each phase, or obtained through experimental calibration.

[0123] In some embodiments, the current duty cycle conversion factor may include the following formula:

[0124]

[0125] In the formula, For the number The output inductance value of the power phase; for The switching frequency of the phase power circuit; For the number The input voltage of the power phase; For the number The output voltage of the power phase.

[0126] In some embodiments, updating the duty cycle of the corresponding power phase based on the duty cycle increment includes the following formula:

[0127] ;

[0128] In the formula, For the number The duty cycle of the power phase.

[0129] In multiphase applications, even if the current in each phase is perfectly equal after adjustment by the algorithm, local hot spots may still occur in a certain core unit 2 due to millimeter-level asymmetry in PCB layout (i.e., winding assembly 3), slight differences in core parameters, or uneven heat dissipation conditions. This causes the core unit 2 to saturate before the others. However, this uneven current cannot be detected by traditional current sampling. In this embodiment, the magnetic field strength of each core unit 2 is directly measured, and targeted control is performed when the magnetic state imbalance exceeds a second preset threshold.

[0130] Furthermore, under ultrafast dynamic loads, even if the average current of each power phase is balanced, slight differences in parasitic parameters and drive delays at the switching instant can cause instantaneous magnetic flux superposition in the magnetic circuit due to the rising edges of the currents of adjacent power phases, forming a nanosecond-level dynamic congestion point and inducing local saturation spikes. Therefore, the higher the switching frequency, the more fatal the impact of parasitic parameters becomes, and the more prominent the dynamic congestion problem becomes. To break this harmful instantaneous synchronization, this embodiment adjusts the phase of the second target control phase, causing it to deviate from its adjacent power phase with a higher magnetic flux density in time, fundamentally avoiding the malignant superposition of pulses in time and space.

[0131] In some embodiments, when the magnetic state imbalance is greater than a second preset threshold, the targeted control includes step SS32. Step SS32: When the magnetic state imbalance is greater than the second preset threshold, a second target control phase is determined based on the magnetic core magnetic state signal, and the phase of the second target control phase is adjusted so that the second target control phase deviates from the adjacent power phase with a larger magnetic flux density in time. Specifically, step SS32 includes steps SS321 to SS324.

[0132] Step SS321: Based on the magnetic state signal of the magnetic core, the magnetic core unit 2 with the highest magnetic field strength is selected as the second target unit (e.g., the magnetic core unit 2 of phase m).

[0133] Step SS322: Query the phase magnetic circuit mapping table to determine one or more power phases (e.g., m phase) with the highest coupling weight to the second target unit, and identify them as the second target control phase.

[0134] In application, the total magnetic flux in any core unit 2 of the transformer in this embodiment is the vector sum of the magnetic fields generated by all power phase currents. Therefore, the phase magnetic circuit mapping table includes: the contribution ratio of the unit current (or unit ampere-turns) of the i-th phase power circuit to the magnetomotive force or magnetic flux density generated in the magnetic circuit of the j-th core unit 2. Where i and j are... In the integers, i can be equal to j or not equal to j.

[0135] In practical applications, each magnetic core unit 2 corresponds to one power phase.

[0136] Step SS323: Calculate the phase adjustment amount of the second target control phase based on the second observation index.

[0137] In application, in order to calculate the phase adjustment amount based on the magnetic state imbalance and its cumulative error, and to achieve fast response and zero steady-state error adjustment, the phase adjustment amount δ includes:

[0138] ;

[0139] ;

[0140] ;

[0141] In the formula, The second observation index is the magnetic state imbalance of core unit 2. This is the integral coefficient, which can take values ​​from 0.1 to 0.5 times the switching period, with a unit of 1. This is a proportionality coefficient, which can range from 0.01 to 0.1 times the switching frequency, and the unit is seconds (s). The phase-time adjustment amount for the second target control phase, in seconds; This is the initial amount for phase adjustment; To adjust the amplitude; The switching cycle.

[0142] Step S324: Update the phase angle of the second target control phase according to the phase adjustment amount, so that the switching event of the second target control phase moves in the opposite direction to the adjacent power phase with a larger magnetic flux density, so that the second target control phase deviates from the adjacent power phase with a larger magnetic flux density in time.

[0143] When applying the algorithm, updating the phase angle includes:

[0144]

[0145] In the formula, The phase angle of the second target phase is controlled.

[0146] In practical applications, the updated phase angle of the second target control phase is written into the phase offset register of the digital PWM generator of the corresponding digital multiphase controller.

[0147] In practical implementation, if the transformer includes four magnetic core units 2, comprising four power phases with staggered phases of 0°, 90°, 180°, and 270°, and the second target control phase is the 90° phase, with the preceding adjacent phase being the 0° phase and the following phase being the 180° phase, and if the 180° phase has a larger magnetic flux, then the second target control phase is finely adjusted from the 90° phase towards the 0° phase (at this time...). (To increase the phase angle). By fine-tuning the phase of the second target control phase, the magnetic flux distribution of each power phase is changed, thereby achieving magnetic flux density balance.

[0148] Example 4: This example also introduces a transformer-based anti-saturation system.

[0149] To achieve MHz-level power conversion, the anti-saturation system in this embodiment includes: a transformer, a digital multiphase controller, a first GaN power chip, and a synchronous rectifier. The transformer is the same as described in Embodiment 1 or Embodiment 2, and will not be repeated here. The digital multiphase controller is connected to the first GaN power chip and includes a digital PWM generator for executing the steps of the anti-saturation method described in Embodiment 3, which will not be repeated here.

[0150] In application, the workflow of this embodiment includes: power conversion, energy transmission and conversion, and output rectification.

[0151] (1) Power conversion: The input DC power (e.g., 48V DC) is connected to the first GaN power chip (as a primary-side switch) driven by the digital multiphase controller. The first GaN power chip converts the DC power into a high-frequency square wave voltage in the MHz range and applies it directly to the two ends of the primary winding of the transformer through the pad array.

[0152] (2) Energy transmission and conversion: High-frequency current generates a high-speed changing magnetic field in the primary winding of the transformer. The transformer’s symmetrical distributed magnetic circuit efficiently couples the high-speed changing magnetic field to the secondary winding.

[0153] (3) Output rectification: The high-frequency voltage induced by the secondary winding is rectified and filtered by the synchronous rectifier (i.e. the second GaN power chip) integrated in the same package, and finally outputs a stable, low-voltage, high-current DC power to directly power the load (such as CPU / GPU).

[0154] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A transformer, characterized in that, The system includes a cavity frame, within which multiple magnetic core units and a winding assembly are embedded. The winding assembly includes: Primary winding, including multiple series-connected primary helical coils; Secondary windings, including multiple series-connected secondary helical coils; The primary spiral coil and the secondary spiral coil are stacked alternately, with the central axes of each spiral coil being collinear, and the distance from the central axis of the spiral coil to each magnetic core unit is equal. The cavity skeleton is a thick-walled circular ring; Along the axial direction of the thick-walled ring, multiple annular grooves are formed on the inner ring surface for mounting the primary helical coil and the secondary helical coil; the centers of each annular groove are equally spaced on the central axis of the inner ring surface; Multiple mounting through holes are equally spaced along the circumference of the thick-walled ring for mounting the magnetic core unit; A heat dissipation hole array is provided inside the thick-walled circular ring. One side of the heat dissipation hole array is connected to the mounting hole, and the other side of the heat dissipation hole array is connected to the heat dissipation component. The heat dissipation component includes multiple heat dissipation fins, which are equally spaced on the outer ring surface of the thick-walled circular ring.

2. The transformer according to claim 1, characterized in that, The primary helical coil and the secondary helical coil are respectively distributed on the corresponding substrates, and their rotation directions are opposite and mirror symmetrical.

3. The transformer according to claim 1, characterized in that, The magnetic core unit has an air gap groove, the length of which is equal to the height of the magnetic core unit.

4. A method for preventing transformer saturation as described in any one of claims 1 to 3, characterized in that, Includes the following steps: Obtain transformer parameters; the transformer parameters include the primary current signal of the primary winding, the magnetic state signal of the core of each phase, and the operating temperature of the core of each phase; Based on the transformer parameters, observation indicators are extracted; the observation indicators include the amplitude of odd harmonics and the magnetic state imbalance. Based on the observed indicators, it is determined whether the core unit is about to enter the nonlinear saturation region. In response to the core unit entering the nonlinear saturation region, targeted control is performed to keep the operating position of the core unit within the linear region of the BH curve and avoid magnetic saturation.

5. The method for preventing transformer saturation according to claim 4, characterized in that, The magnetic state imbalance includes the following formula: ； In the formula, The magnetic state imbalance of the core unit; The magnetic state signal of any magnetic core unit within one switching cycle; for The maximum value; for The minimum value; for The average value.

6. The transformer anti-saturation method according to claim 4, characterized in that, The step of determining whether a magnetic core unit is about to enter the nonlinear saturation region based on observed indicators includes: The amplitude of the odd harmonic is compared with the first preset threshold. When the amplitude of the odd harmonic is greater than the first preset threshold, it is determined that the magnetic core unit is about to enter the nonlinear saturation region. The method for determining the first preset threshold includes: using a temperature threshold relationship table, dynamically updating the first preset threshold according to the core operating temperature; the temperature threshold relationship table includes the third harmonic amplitude of the power phase when the BH curve of the core unit reaches the nonlinear inflection point at each calibration temperature. Compare the magnetic state imbalance with the second preset threshold. When the magnetic state imbalance is greater than the second preset threshold, it is determined that the magnetic core unit is about to enter the nonlinear saturation region. The second preset threshold is determined based on the mean and standard deviation of the sample set of the average growth rate of the second observation index over multiple consecutive switching cycles.

7. The method for preventing transformer saturation according to claim 6, characterized in that, The targeted regulation includes: Step SS31: When the amplitude of the odd harmonic is greater than the first preset threshold, determine the first target control phase according to the primary current signal of the primary winding, and adjust the peak current reference voltage of the first target control phase, as well as the duty cycle of other power phases. Step SS32: When the magnetic state imbalance is greater than the second preset threshold, determine the second target control phase according to the magnetic core magnetic state signal, and adjust the phase of the second target control phase so that the second target control phase deviates from the adjacent power phase with a larger magnetic flux density in time.

Images