A transient oscillation bidirectional compensation circuit for XPU power supply

Abstract

The application belongs to the technical field of power electronics, and provides a transient oscillation bidirectional compensation circuit for an XPU power supply, comprising a high-pass sampling module, a dynamic compensation analog driving module, a push-pull signal amplification module and a power stage push-pull compensation module; the high-pass sampling module is connected with the output end of a voltage regulator module; the dynamic compensation analog driving module is connected with the high-pass sampling module; the push-pull signal amplification module is connected with the dynamic compensation analog driving module; the power stage push-pull compensation module is connected with the push-pull signal amplification module, a PDN and a load module respectively; and the power stage push-pull compensation module is used for converting a driving signal into a compensation current so as to inject the output end of the voltage regulator module. The application can effectively reduce voltage overshoot and drop, and has the advantages of small volume, low cost and simple structure.

Description

Technical Field

[0001] This invention belongs to the field of power electronics technology and provides a bidirectional compensation circuit for transient oscillations in XPU power supplies. Background Technology

[0002] With the rapid development of big data and artificial intelligence technologies, the performance requirements for high-performance computing chips are becoming increasingly stringent. However, the application of high-performance computing chips faces severe challenges, one of the most significant being chip power supply. Compared to the traditional 12V power distribution architecture, the 48V power distribution architecture has gradually become the mainstream solution due to its lower power distribution losses. Currently, the power of a single microprocessor module exceeds 750W, and the current reaches 1000A or even higher. Simultaneously, with the increase in processor clock frequency, the load change rate will reach over 1000A / µs. Extensive research has been conducted on the architecture analysis of 48V-VRM, which can be broadly categorized into single-stage and two-stage solutions. The two-stage solution can be further divided into the Factorized Power Architecture (FPA) architecture, represented by Vicor, and the Intermediate Bus Converter (IBC) architecture. The latter is more widely used in the industry.

[0003] For 48V IBC architecture power supply schemes, the front-end typically uses a fixed-ratio DC-X transformer (DCX) without voltage adjustment to reduce the 48V bus voltage to an intermediate bus voltage of 12V or 6V, and then uses an adjustable voltage module for voltage regulation. There are two main topologies for the front-end DCX module. One is the LLC topology, which achieves extremely high efficiency by operating at the resonant frequency. The introduction of matrix transformers and fractional-turn transformers has further improved the power density and efficiency of LLC-DCX. The other DCX topology is the switched-capacitor topology, which uses higher-density capacitors as energy storage elements in the converter to achieve higher power density. The switched-capacitor topology is more suitable for conversions with smaller voltage ratios, such as 2:1 and 4:1. As the ratio increases, its efficiency drops sharply.

[0004] In low-voltage and high-current applications, Buck converters are widely used in the industry due to their simple and scalable structure and excellent dynamic performance. Buck converters have many control methods. Voltage-mode control is the most basic control method for VRMs. This method has a simple loop control and good disturbance rejection, but because the LC filter at the output adds two poles to the control system, it leads to complex phase compensation in the control loop, making it difficult to improve the closed-loop control bandwidth, thus resulting in relatively poor dynamic response. Current-mode control is also a widely used control strategy. This method directly controls the inductor current signal, and the filter introduces only one pole in the control loop, greatly improving the loop gain and bandwidth. Compared with voltage-mode control, it shows a significant advantage in improving the dynamic performance of the VRM. Current-mode control strategies can be further summarized into three categories: peak current control strategy, valley current control strategy, and average current control strategy. Among them, peak current control and valley current control have better transient response capabilities, but their loop noise immunity is poor and they are very prone to subharmonic oscillations. Therefore, additional ramp compensation is needed to improve system stability, but ramp compensation also has an adverse effect on transient performance. Hysteresis control has an extremely simple loop, and it can achieve a very large loop bandwidth without considering compensation issues and error amplifiers. Its transient response speed is extremely fast. A highly representative control strategy is constant-time on-state control.

[0005] The principle of constant-time turn-on (COT) control is that when the output voltage is lower than the reference voltage, a constant-time pulse is generated to control the turn-on of the Buck converter's active transistor. However, COT control also faces system stability issues. Sufficient system stability can only be guaranteed when the output voltage ripple is large enough, requiring the VRM's output capacitor to have a sufficiently large equivalent series capacitance (ESR). Traditional electrolytic capacitors have a large ESR, but their large package size cannot meet the high power density requirements of the XPU for the VRM. Therefore, electrolytic capacitors are replaced by multilayer ceramic capacitors (MLCCs) with smaller packages. MLCCs have a smaller ESR, so COT control also needs to introduce ramp compensation to change the output voltage ripple. The presence of ramp compensation also affects the system's transient performance, so a suitable ramp slope needs to be selected by comprehensively considering system stability and dynamic performance. As a typical control strategy for Buck converters, COT control has a very large loop bandwidth and extremely fast transient response speed, offering significant advantages over traditional voltage and current control modes, and is therefore widely used in VRM scenarios.

[0006] However, Buck converters exhibit output voltage dips during load loading and overshoots during load unloading, which can negatively impact the chip's core operation. The reasons are as follows: During load loading, the load current suddenly increases. Because the inductor current is limited by the loop bandwidth and cannot keep up with the load current, the output capacitor must provide instantaneous current. This current flows in reverse through the output capacitor, causing a drop in the output capacitor voltage, resulting in the output voltage dip. During load unloading, the load current suddenly decreases. The inductor current response is affected by the filter inductor, so the rate of change of the inductor current is less than the rate of change of the load current. At this time, the output capacitor current flows in the forward direction, and its magnitude is the difference between the load current and the inductor current. This causes the output capacitor voltage to gradually increase, resulting in an output voltage overshoot. Output voltage spikes may exceed the chip's safe operating range, causing the chip to malfunction. Output voltage dips may cause the chip's supply voltage to become too low, leading to system crashes.

[0007] To address the aforementioned issues, the academic community has proposed numerous solutions. The simplest measure to reduce output voltage sag and overshoot is to increase the output capacitor, but this often comes with cost and size limitations. Using a negatively coupled inductor or TLVR topology can improve the transient response of the VRM by reducing the equivalent transient inductance, but its response speed is still limited by the closed-loop control bandwidth. The idea behind active PDN is to connect a current injection or absorption circuit in parallel across the PDN decoupling capacitor. When the load becomes dynamic, this circuit directly exchanges energy with the load, reducing charge changes in the decoupling capacitor and thus reducing voltage fluctuations, i.e., reducing processor voltage spikes. Using an active PDN architecture allows for rapid response to changes in chip load current, seamlessly switching between steady-state and transient modes during dynamic events, thereby improving the dynamic performance of the VRM.

[0008] Based on the location of the input ports of the compensation circuit used in the active PDN architecture, it can be divided into full-bridge compensation circuits, indirect-bridge compensation circuits, and direct-parallel compensation circuits. Full-bridge compensation circuits typically use switching converters as auxiliary circuits. Their essence is the same as indirect-bridge compensation circuits: improving dynamic response speed by reducing the inductance value during transient response. However, they usually require complex circuit topologies and additional magnetic components, resulting in a larger board area. Therefore, using the simpler direct-parallel compensation circuit allows for a smaller size.

[0009] Direct parallel compensation circuits can be classified into transistor-type linear compensation circuits and MOS-type linear compensation circuits based on the power devices used. Their working principles are basically similar, but transistors, as power devices in compensation circuits, have insufficient output capability. Therefore, metal-oxide-semiconductor field-effect transistors (MOSFETs) can be used to replace transistors to achieve a greater current output capability in the compensation circuit.

[0010] Currently, traditional direct parallel compensation circuits detect the transient AC voltage component at the output of the voltage regulator module, amplify it using an operational amplifier to obtain an analog drive signal, and finally use the drive signal to drive a power transistor to generate a compensation current injected into the output of the voltage regulator module. Both the detection point and the injection point are located at the output of the voltage regulator module.

[0011] However, existing direct parallel compensation circuits have the following drawbacks: Existing direct parallel compensation circuits require different compensation circuits to be designed for VRM loading and unloading scenarios, which leads to the use of too many components and makes it difficult to further reduce costs and size.

[0012] Existing active PDN architectures often focus on mitigating voltage spikes that occur when the VRM is unloaded. While they do attenuate voltage drops that occur during loading to some extent, the attenuation is very low. Nowadays, chips are constantly improving their tolerance to voltage spikes, and voltage drops can directly lead to insufficient chip power supply voltage, resulting in shutdown. Therefore, low voltage drop attenuation capability will prevent the chip from working properly. Summary of the Invention

[0013] To address the aforementioned technical problems, this invention provides a bidirectional transient oscillation compensation circuit for XPU power supplies, which can effectively reduce voltage overshoot and drop, and is a small, low-cost, and simple active transient compensation circuit.

[0014] The transient oscillation bidirectional compensation circuit of the present invention is connected to the voltage regulator module, the PDN, and the load module, respectively, and includes: A high-pass sampling module is connected to the output terminal of the voltage regulator module. The high-pass sampling module is used to acquire the transient voltage signal output by the voltage regulator module during the transient period.

[0015] A dynamic compensation analog drive module is connected to the high-pass sampling module. The dynamic compensation analog drive module is used to amplify the transient voltage signal acquired by the high-pass sampling module in reverse and provide a DC bias to obtain an analog drive signal.

[0016] A push-pull signal amplification module is connected to the dynamic compensation analog drive module. The push-pull signal amplification module is used to amplify the analog drive signal to obtain a drive signal.

[0017] The power stage push-pull compensation module is connected to the push-pull signal amplification module, the PDN, and the load module, respectively. The power stage push-pull compensation module is used to convert the drive signal into a compensation current and inject it into the output of the voltage regulator module.

[0018] Furthermore, the high-pass sampling module includes: sampling capacitor One end of it is connected to the output of the voltage regulator module, and the other end is connected to the dynamic compensation analog drive module.

[0019] sampling resistor One end of it is connected to the sampling capacitor. One end is connected, and the other end is grounded.

[0020] Furthermore, the dynamic compensation simulation drive module includes: negative input resistance One end of it is connected to the sampling resistor connect.

[0021] An operational amplifier, whose inverting input is connected to the negative input resistor. The other end is connected.

[0022] Positive input large voltage divider resistor One end of it is connected to the non-inverting input terminal of the operational amplifier, and the other end is connected to the negative terminal of the first power supply. connect.

[0023] Positive input small voltage divider resistor One end of it is connected to the non-inverting input of the operational amplifier, and the other end is grounded.

[0024] Positive input decoupling capacitor With the positive input small voltage divider resistor in parallel.

[0025] Feedback resistor It is connected across the inverting input and output of the operational amplifier.

[0026] Feedback decoupling capacitor , and the feedback resistor in parallel.

[0027] Furthermore, the push-pull signal amplification module includes: Push-pull half-bridge structure, including NPN transistors connected in series. and PNP type transistors The NPN transistor The input terminal of the PNP transistor is connected to the positive terminal of the second power supply. The output terminal is connected to the negative terminal of the second power supply.

[0028] upper diode One end of it is connected to an NPN transistor. The base of one end is connected, and the other end is connected to the output of the operational amplifier.

[0029] lower diode One end of it is connected to a PNP transistor. The base of one end is connected, and the other end is connected to the output of the operational amplifier.

[0030] Pull-up resistor One end of it is connected to the upper diode. Connect one end to the other end of the power supply, and the other end to the positive terminal of the second power supply. connect.

[0031] Pull-down resistor One end of it is connected to the lower diode. Connect one end to the other end, and connect the other end to the negative terminal of the second power supply. connect.

[0032] Furthermore, the power stage push-pull compensation module includes: N-channel field-effect transistor Its first end is connected to the positive terminal of the auxiliary power supply. The second terminal is connected to the voltage regulator module, the DN, and the load module, and the third terminal is connected to the NPN transistor. Output terminal and PNP transistor Input end.

[0033] P-channel field-effect transistor Its first terminal is grounded, the second terminal is connected to the voltage regulator module, DN and load module, and the third terminal is connected to the NPN transistor. Output terminal and PNP transistor Input end.

[0034] Furthermore, the positive power supply terminal of the operational amplifier is connected to the positive terminal of the first power supply. Connect the negative power supply terminal to the negative terminal of the first power supply. The connection is made so that the voltage of the third power supply is 30V, and its positive terminal is... +15V, negative terminal It is -15V.

[0035] Furthermore, the analog drive signal The calculation formula is: ; In the formula, This represents the output signal of the operational amplifier. This indicates the output signal of the push-pull signal amplification module. and All are analog drive signals. Indicates sampling resistor The voltage, i.e., the transient voltage signal. Indicates the negative terminal of the first power supply. Voltage.

[0036] Furthermore, the formula for calculating the compensation current is as follows: ; In the formula, To compensate for the current, This indicates the moment when the transient oscillation bidirectional compensation circuit starts working, i.e., the moment when the compensation current begins to increase from 0. This indicates the moment when the compensation current reaches its peak value. This indicates the moment when the transient oscillation bidirectional compensation circuit stops working, that is, when the compensation current returns to 0. Indicates the compensation current in [ , The slope of the rise over a given time period. Indicates the compensation current in [ , The slope of the descent.

[0037] Furthermore, the power stage push-pull compensation module satisfies: ; In the formula, To compensate for the current, For inductor current, For load current, This represents the capacitor current.

[0038] Furthermore, the losses of the transient oscillation bidirectional compensation circuit for: ; In the formula, The period of load current variation. For auxiliary power supply voltage, Output voltage for the voltage regulator module. N-channel field-effect transistor The current, P-channel field-effect transistor The current.

[0039] The technical solution provided by this invention has the following advantages compared with the prior art: This invention acquires the transient voltage signal output by the voltage regulator module during its transient period using a high-pass sampling module. A dynamic compensation analog drive module amplifies the transient voltage signal acquired by the high-pass sampling module in reverse and provides a DC bias to obtain an analog drive signal. A push-pull signal amplification module amplifies the analog drive signal to obtain a drive signal. A power stage push-pull compensation module converts the drive signal into a compensation current, which is then injected into the output of the voltage regulator module. Compared to existing technologies, this invention's dynamic compensation analog drive module can simultaneously compensate for output voltage dips and overshoots, while significantly improving the ability to compensate for output voltage dips.

[0040] Other advantages, objectives and features of the present invention will become apparent in part from the following description, and in part from those skilled in the art through study and practice of the invention. Attached Figure Description

[0041] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0042] Figure 1 This is a circuit topology diagram of a transient oscillation bidirectional compensation circuit for XPU power supply according to the present invention; Figure 2 This is the circuit timing diagram of the present invention.

[0043] Figure 3 This is a circuit diagram of the present invention.

[0044] Figure 4 This is a test scenario diagram for testing the dynamic output voltage of this invention.

[0045] Figure 5 This is a waveform diagram of the load current under the test environment of this invention.

[0046] Figure 6 This is a waveform diagram of the load output voltage drop under the test environment of this invention without using the transient oscillation bidirectional compensation circuit.

[0047] Figure 7 This is a waveform diagram of the overshoot of the unloaded output voltage under the test environment of the present invention without using the transient oscillation bidirectional compensation circuit.

[0048] Figure 8 This is a waveform diagram of the load output voltage drop when using the transient oscillation bidirectional compensation circuit under the test environment of this invention.

[0049] Figure 9 This is a waveform diagram of the overshoot of the output voltage under load reduction using the transient oscillation bidirectional compensation circuit in the test environment of this invention.

[0050] Figure label: 101. Voltage Regulator Module; 102. Sampling Capacitor; 103. Sampling Resistor; 104. Negative Input Resistor; 105. Positive Input Voltage Divider Large Resistor; 106. Positive Input Voltage Divider Small Resistor; 107. Positive Input Decoupling Capacitor; 108. Feedback Resistor; 109. Feedback Decoupling Capacitor; 110. Operational Amplifier; 111. Pull-up Resistor; 112. Upper Diode; 113. Lower Diode; 114. Pull-down Resistor; 115. NPN Transistor; 116. PNP Transistor; 117. N-channel Field-Effect Transistor; 118. P-channel Field-Effect Transistor; 119. PDN and Load Module. Detailed Implementation

[0051] The following detailed description of a specific embodiment of the present invention is provided in conjunction with the accompanying drawings. However, it should be understood that the scope of protection of the present invention is not limited to the specific embodiment.

[0052] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "axial," "radial," and "circumferential" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing the technical solution of this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0053] In the description of the embodiments of the present invention, unless otherwise stated, "a plurality of" means two or more.

[0054] like Figure 1 As shown, this invention provides a transient oscillation bidirectional compensation circuit (hereinafter referred to as DATCC) for XPU power supply, which is connected to the voltage regulator module, PDN and load module respectively, and includes: A high-pass sampling module is connected to the output terminal of the voltage regulator module. The high-pass sampling module is used to acquire the transient voltage signal output by the voltage regulator module during the transient period.

[0055] A dynamic compensation analog drive module is connected to the high-pass sampling module. The dynamic compensation analog drive module is used to amplify the transient voltage signal acquired by the high-pass sampling module in reverse and provide a DC bias to obtain an analog drive signal.

[0056] A push-pull signal amplification module is connected to the dynamic compensation analog drive module. The push-pull signal amplification module is used to amplify the analog drive signal to obtain a drive signal.

[0057] The power stage push-pull compensation module is connected to the push-pull signal amplification module, the PDN, and the load module, respectively. The power stage push-pull compensation module is used to convert the drive signal into a compensation current and inject it into the output of the voltage regulator module.

[0058] The voltage regulator module is existing technology, and its specific structure will not be described in detail. It adopts a 10-phase interleaved parallel Buck converter topology and uses a current-mode constant-time conduction control strategy (CMCOT). The input voltage is 12V, the output voltage is 0.9V, the rated output current is 300A, the switching frequency is 600kHz, the inductor is 150nH, and the output capacitor is 7.52mF.

[0059] In the embodiments provided by the present invention, the high-pass sampling module includes: sampling capacitor One end of it is connected to the output of the voltage regulator module, and the other end is connected to the dynamic compensation analog drive module. Sampling capacitor. These are μF level capacitors, with capacitance values ​​ranging from 10μF to 22μF.

[0060] sampling resistor One end of it is connected to the sampling capacitor. One end is connected, and the other end is grounded. Sampling resistor. The resistance is 1Ω to 10Ω.

[0061] In the embodiments provided by the present invention, the dynamic compensation simulation driving module includes: negative input resistance One end of it is connected to the sampling resistor connect.

[0062] An operational amplifier, whose inverting input is connected to the negative input resistor. The other end is connected.

[0063] Positive input large voltage divider resistor One end of it is connected to the non-inverting input terminal of the operational amplifier, and the other end is connected to the negative terminal of the first power supply. connect.

[0064] Positive input small voltage divider resistor One end of it is connected to the non-inverting input of the operational amplifier, and the other end is grounded.

[0065] Positive input decoupling capacitor With the positive input small voltage divider resistor in parallel.

[0066] Feedback resistor It is connected across the inverting input and output of the operational amplifier.

[0067] Feedback decoupling capacitor , and the feedback resistor in parallel.

[0068] In the embodiments provided by the present invention, the push-pull signal amplification module includes: Push-pull half-bridge structure, including NPN transistors connected in series. and PNP type transistors The NPN transistor The input terminal of the PNP transistor is connected to the positive terminal of the second power supply. The output terminal is connected to the negative terminal of the second power supply.

[0069] upper diode One end of it is connected to an NPN transistor. The base of one end is connected, and the other end is connected to the output of the operational amplifier.

[0070] lower diode One end of it is connected to a PNP transistor. The base of one end is connected, and the other end is connected to the output of the operational amplifier.

[0071] Pull-up resistor One end of it is connected to the upper diode. Connect one end to the other end of the power supply, and the other end to the positive terminal of the second power supply. connect.

[0072] Pull-down resistor One end of it is connected to the lower diode. Connect one end to the other end, and connect the other end to the negative terminal of the second power supply. connect.

[0073] In the embodiments provided by the present invention, the power stage push-pull compensation module includes: N-channel field-effect transistor Its first end is connected to the positive terminal of the auxiliary power supply. The second terminal is connected to the voltage regulator module, the DN, and the load module, and the third terminal is connected to the NPN transistor. Output terminal and PNP transistor Input end.

[0074] P-channel field-effect transistor Its first terminal is grounded, the second terminal is connected to the voltage regulator module, DN and load module, and the third terminal is connected to the NPN transistor. Output terminal and PNP transistor Input end.

[0075] Specifically, the Qualcomm sampling module uses a sampling capacitor. Series sampling resistor The AC component of the transient output voltage of the voltage regulator module VRM, i.e., the transient voltage signal, is formed by the RC high-pass filter circuit; the operational amplifier of the dynamic compensation analog drive module is connected to the negative terminal of the first power supply. For reference, the transient voltage signal detected by the high-pass sampling module is inverted and amplified, and a certain DC bias is provided to obtain an analog drive signal; the push-pull signal amplification module follows and amplifies the analog drive signal output by the operational amplifier, so as to drive the power device; the power transistor in the power stage push-pull compensation module is driven by the amplified analog drive signal, generating a compensation current injected into the output of the VRM.

[0076] In the embodiments provided by the present invention, the positive power supply terminal of the operational amplifier is connected to the positive terminal of the first power supply. Connect the negative power supply terminal to the negative terminal of the first power supply. The connection is made so that the voltage of the third power supply is 30V, and its positive terminal is... +15V, negative terminal It is -15V.

[0077] In other words, the operational amplifier in the dynamic compensation analog drive circuit is referenced to the negative terminal (-15V) of the first power supply, rather than to the ground GND of the voltage regulator module, PDN, and load module.

[0078] In the embodiments provided by the present invention, the analog drive signal The calculation formula is: ; In the formula, This represents the output signal of the operational amplifier. This indicates the output signal of the push-pull signal amplification module. and All are analog drive signals. Indicates sampling resistor The voltage, i.e., the transient voltage signal. Indicates the negative terminal of the first power supply. Voltage.

[0079] In the embodiments provided by the present invention, the formula for calculating the compensation current is: ; In the formula, To compensate for the current, This indicates the moment when the transient oscillation bidirectional compensation circuit starts working, i.e., the moment when the compensation current begins to increase from 0. This indicates the moment when the compensation current reaches its peak value. This indicates the moment when the transient oscillation bidirectional compensation circuit stops working, that is, when the compensation current returns to 0. Indicates the compensation current in [ , The slope of the rise over a given time period. Indicates the compensation current in [ , The slope of the descent.

[0080] In the embodiments provided by the present invention, the power stage push-pull compensation module satisfies: ; In the formula, To compensate for the current, For inductor current, For load current, This represents the capacitor current.

[0081] Furthermore, the losses of the transient oscillation bidirectional compensation circuit for: ; In the formula, The period of load current variation. For auxiliary power supply voltage, Output voltage for the voltage regulator module. N-channel field-effect transistor The current, P-channel field-effect transistor The current.

[0082] The voltage regulator module used to verify the DATCC function employs a 10-phase interleaved parallel Buck converter topology. A power module (DrMos) with integrated drivers and metal-oxide-semiconductor field-effect transistors can achieve extremely high power density. The selected DrMos is the MP86962, with a maximum continuous output current of 80A and an input voltage range of 3V to 12V. The VRM uses a CMCOT control strategy, with the selected controller model being the MP2882. Its switching frequency range is 200kHz to 3MHz, supporting dual outputs and configurable up to 16 phases.

[0083] Figure 2for Figure 1 The timing diagram of the DATCC circuit is shown, and the load current is plotted. VRM output voltage Transient voltage AC component Analog drive signals and compensation current When the load current When the load increases suddenly, a voltage drop occurs, which causes the output voltage to drop. The magnitude of this drop is measured by a high-pass sampling module. The acquired voltage signal is amplified by the inverting amplifier and the power amplification module of the push-pull signal amplifier to obtain a positive pulse analog drive signal. ,when The voltage drops to the compensation circuit load function start voltage. hour, Just rose to Turn-on threshold voltage of a channel field-effect transistor The channel field-effect transistor begins to conduct, and as... Continue to increase, compensation current It also gradually increased, and then and After reaching their respective peaks, they begin to decline until... Return to The compensation circuit loading function is off at this time. Return to The channel field-effect transistor is off. Under load reduction, the DATCC operates similarly, except that at this time... It is a positive overshoot voltage signal. It is a negative pulse; the voltage at which the load shearing function of the compensation circuit is activated is... The turn-on threshold voltage of a P-channel field-effect transistor It is a negative value.

[0084] The parameters for DATCC are designed by first designing the sampling capacitor and sampling resistor of the high-pass sampling module. The time constant of the high-pass sampling module is: ; A larger time constant results in higher DATCC sensitivity and less response to transient fluctuations in the output voltage; however, this comes at the cost of a longer transient response time. Conversely, a smaller time constant results in lower sensitivity and a shorter transient response time. Therefore, a compromise is reached by selecting the cutoff frequency f. c Close to the VRM closed-loop cutoff frequency f c_VR ,Pick =22μF, =5.1Ω, cutoff frequency f c =1.42kHz.

[0085] For the power stage push-pull compensation module, the peak compensation current required for both loading and unloading is set to 280A. Due to the high requirements for pulse current of the power devices, the EPC2023 is selected as the N-channel MOSFET, with a gate threshold voltage of 1.4V, a maximum pulse drain current of 590A, and a gate-source drive voltage of 3.01V required for a drain current of 280A. The BSZ086P03NS3G is selected as the P-channel MOSFET, with a gate threshold voltage of -2.5V, a maximum pulse drain current of 160A, and a gate-source drive voltage of -6.81V required for a drain current of 280A. To allow for margin, three BSZ086P03NS3G transistors are connected in parallel.

[0086] For the push-pull signal amplification module, the drive voltage signal required for loading and unloading is: ; ; This is the gate-source drive voltage required by the EPC2023 when the drain current is 280A. This is the gate-source drive voltage required by the BSZ086P03NS3G when the drain current is 280A. This is the steady-state output voltage of the VRM. This refers to the voltage drop of the VRM output voltage during loading. The overshoot amplitude of the VRM output voltage during load reduction is given. From this, the required drive voltage signal range can be calculated to be -5.95V to 3.95V, and the static operating point is selected as -1.00V.

[0087] For the dynamic compensation analog drive module, due to the need for operational amplifiers with sufficient bandwidth, slew rate, and power supply range, the AD829 model was selected. It has a bandwidth of 120MHz, a slew rate of 230V / μs, and a power supply voltage of ±5V to ±15V. The operational amplifier is powered by ±15V. The analog drive signal output by the operational amplifier is:

[0088] ; It can be approximated that the output voltage fluctuation reaches its maximum value when the compensation current reaches its maximum value. Here, the threshold voltage for enabling the compensation circuit loading function is taken as -20mV. At this time, the analog drive signal just causes the gate-source voltage of the N-channel field-effect transistor to reach the gate threshold voltage. The load shearing function is enabled at a threshold voltage of 20mV. At this time, the analog drive signal just causes the gate-source voltage of the P-channel field-effect transistor to reach the gate threshold voltage V. th2The maximum output voltage fluctuation limit is 40mV. When the output voltage drops to 40mV, the gate-source voltage of the N-channel MOSFET reaches the transient operating point under load. When the output voltage overshoot reaches 40mV, the gate-source voltage of the P-channel MOSFET reaches the load shear transient operating point. Therefore, we can calculate that R1=200Ω, R2=18.7kΩ, R3=10Ω, R4=24.7kΩ, C1=2.2μF, and C2=10pF.

[0089] Figure 3 Displayed based on Figure 1 The actual circuit fabrication shows that the DATCC occupies a very small area on the board. If future power transistors capable of supporting larger pulse drain currents and smaller sizes become available, the overall size can be further reduced. The parameters and models of the components in the actual DATCC are shown in Table 1.

[0090] Table 1: Parameters and Models of Devices in DATCC Physical Samples The results of the 10-phase Buck converter prototype verification under standard conditions are presented here. The standard test environment is as follows: Figure 4 As shown, a DC power supply provides a 12V DC input voltage. Two isolated channels of an auxiliary power supply provide ±15V power to the dynamic compensation analog drive module and the push-pull signal amplification module. Six dynamic load boards provide 300A of short-time high dynamic current to simulate the operation of the chip. Another auxiliary power supply provides 5V power to the drive chip on the dynamic load board. A signal generator sends a trigger signal to control the dynamic load to conduct a large current. Then, an oscilloscope is used to observe the output voltage of the 10-phase Buck converter, the analog drive signal of the transient oscillation bidirectional compensation circuit, and the dynamic load control signal.

[0091] Test results are as follows Figures 5-9 As shown in the figure, after adding DATCC, the maximum output voltage drop of the 10-phase Buck converter when the load is increased from 0 to 300A is reduced from 144mV to 86mV, a year-on-year reduction of 40.278%; and the maximum output voltage overshoot when the load is reduced from 300A to 0 is reduced from 194mV to 100mV, a year-on-year reduction of 48.453%.

[0092] In summary, this invention utilizes dual power supplies and two types of transistors and field-effect transistors. This architecture uses only a single circuit to simultaneously compensate for VRM output voltage dips under load and output voltage overshoots under load, thereby improving the transient performance of the VRM and reducing the number of output capacitors. Therefore, this invention reduces the size of the VRM, simplifies design and use, significantly reduces design costs, and effectively suppresses both output voltage dips and overshoots. It should be noted that any parts not disclosed or specifically described in this invention are existing technology or conventional configurations, and their specific structures and working principles will not be elaborated further. In this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0093] Although embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the specification and embodiments. It can be applied to various fields suitable for the present invention. Other modifications can be readily implemented by those skilled in the art. Therefore, without departing from the general concept defined by the claims and their equivalents, the present invention is not limited to the specific details and examples shown and described herein.

Claims

1. A transient oscillation bidirectional compensation circuit for XPU power supply, connected to a voltage regulator module, a PDN, and a load module respectively, characterized in that, The transient oscillation bidirectional compensation circuit includes: A high-pass sampling module is connected to the output terminal of the voltage regulator module. The high-pass sampling module is used to acquire the transient voltage signal output by the voltage regulator module during the transient period. A dynamic compensation analog drive module is connected to the high-pass sampling module. The dynamic compensation analog drive module is used to amplify the transient voltage signal acquired by the high-pass sampling module in reverse and provide a DC bias to obtain an analog drive signal. A push-pull signal amplification module is connected to the dynamic compensation analog drive module. The push-pull signal amplification module is used to amplify the analog drive signal to obtain a drive signal. The power stage push-pull compensation module is connected to the push-pull signal amplification module, the PDN, and the load module, respectively. The power stage push-pull compensation module is used to convert the drive signal into a compensation current and inject it into the output of the voltage regulator module.

2. The transient oscillation bidirectional compensation circuit for XPU power supply as described in claim 1, characterized in that, The high-pass sampling module includes: sampling capacitor One end of it is connected to the output of the voltage regulator module, and the other end is connected to the dynamic compensation analog drive module; sampling resistor One end of it is connected to the sampling capacitor. One end is connected, and the other end is grounded.

3. The transient oscillation bidirectional compensation circuit for XPU power supply as described in claim 2, characterized in that, The dynamic compensation simulation drive module includes: negative input resistance One end of it is connected to the sampling resistor. connect; An operational amplifier, whose inverting input is connected to the negative input resistor. The other end is connected; Positive input large voltage divider resistor One end of it is connected to the non-inverting input terminal of the operational amplifier, and the other end is connected to the negative terminal of the first power supply. connect; Positive input small voltage divider resistor One end of it is connected to the non-inverting input of the operational amplifier, and the other end is grounded; Positive input decoupling capacitor With the positive input small voltage divider resistor in parallel; Feedback resistor It is connected across the inverting input and output terminals of the operational amplifier; Feedback decoupling capacitor , and the feedback resistor in parallel.

4. The transient oscillation bidirectional compensation circuit for XPU power supply as described in claim 3, characterized in that, The push-pull signal amplification module includes: Push-pull half-bridge structure, including NPN transistors connected in series. and PNP type transistors The NPN transistor The input terminal of the PNP transistor is connected to the positive terminal of the second power supply. The output terminal is connected to the negative terminal of the second power supply; upper diode One end of it is connected to an NPN transistor. The base is connected, and the other end is connected to the output of the operational amplifier; lower diode One end of it is connected to a PNP transistor. The base is connected, and the other end is connected to the output of the operational amplifier; Pull-up resistor One end of it is connected to the upper diode. Connect one end to the other end of the power supply, and the other end to the positive terminal of the second power supply. connect; Pull-down resistor One end of it is connected to the lower diode. Connect one end to the other end, and connect the other end to the negative terminal of the second power supply. connect.

5. The transient oscillation bidirectional compensation circuit for XPU power supply as described in claim 4, characterized in that, The power stage push-pull compensation module includes: N-channel field-effect transistor Its first end is connected to the positive terminal of the auxiliary power supply. The second terminal is connected to the voltage regulator module, the DN, and the load module, and the third terminal is connected to the NPN transistor. Output terminal and PNP transistor Input terminal; P-channel field-effect transistor Its first terminal is grounded, the second terminal is connected to the voltage regulator module, DN and load module, and the third terminal is connected to the NPN transistor. Output terminal and PNP transistor Input end.

6. The transient oscillation bidirectional compensation circuit for XPU power supply as described in claim 3, characterized in that, The positive power supply terminal of the operational amplifier is connected to the positive terminal of the first power supply. Connect the negative power supply terminal to the negative terminal of the first power supply. The connection is made so that the voltage of the third power supply is 30V, and its positive terminal is... +15V, negative terminal It is -15V.

7. The transient oscillation bidirectional compensation circuit for XPU power supply as described in claim 5, characterized in that, The analog drive signal The calculation formula is: ； In the formula, This represents the output signal of the operational amplifier. This indicates the output signal of the push-pull signal amplification module. and All are analog drive signals. Indicates sampling resistor The voltage, i.e., the transient voltage signal. Indicates the negative terminal of the first power supply. Voltage.

8. The transient oscillation bidirectional compensation circuit for XPU power supply as described in claim 7, characterized in that, The formula for calculating the compensation current is: ； In the formula, To compensate for the current, This indicates the moment when the transient oscillation bidirectional compensation circuit starts working, that is, the moment when the compensation current begins to increase from 0. This indicates the moment when the compensation current reaches its peak value. This indicates the moment when the transient oscillation bidirectional compensation circuit stops working, that is, when the compensation current returns to 0. Indicates the compensation current in [ , The slope of the rise over a given time period. Indicates the compensation current in [ , The slope of the descent.

9. The transient oscillation bidirectional compensation circuit for XPU power supply as described in claim 8, characterized in that, The power stage push-pull compensation module satisfies: ； In the formula, To compensate for the current, For inductor current, For load current, This represents the capacitor current.

10. A transient oscillation bidirectional compensation circuit for XPU power supply as described in claim 5, characterized in that, The loss of the transient oscillation bidirectional compensation circuit for: ； In the formula, The period of load current variation. For auxiliary power supply voltage, Output voltage for the voltage regulator module. N-channel field-effect transistor The current, P-channel field-effect transistor The current.

Images