Satellite-borne computing force load current monitoring method

Abstract

The invention discloses a satellite-borne computing force load current monitoring method, and relates to the technical field of satellite-borne power supplies and measurement and control. A sampling point is arranged on the input side of a + 28V primary power supply passing through an isolated DCDC; four 100 m omega resistors are connected in series to form a sampling matrix, three acquisition channels are led out, the equivalent resistance values are 0.4 omega, 0.2 omega and 0.1 omega, and 0-4A segmented measurement is performed correspondingly; a channel signal enters the 0-12V ADC through the current detection amplifier with the gain of 20, the MCU executes 10-point extreme value removal mean filtering, and a second channel current is used as a criterion: lt; a channel 1 is selected for 1A, and a channel 2 and gt are selected for 1-2A; in order to avoid a single-point fault, the 2A selective channel 3 is connected in parallel with a direct-through channel without a sampling resistor at a + 28V side, the on-off of the direct-through channel is controlled by an MOS (Metal Oxide Semiconductor) and an MCU (Microprogrammed Control Unit), and surge suppression and current limiting are included; and when the acquisition channel is abnormal, switching straight-through and reporting an abnormal state, and meanwhile, recording abandoned channel data to support rollback / reselection. According to the invention, full-scale precision and power supply stability are both considered, + 12V ripples are avoided, and the circuit is suitable for loads such as satellite-borne GPU / NPU and the like.

Description

Technical Field

[0001] The present invention relates to the field of satellite-borne power supply and measurement and control technology, and in particular to a method for monitoring satellite-borne computing power load current. Background Art

[0002] Onboard computing payloads are typically powered by a +28V primary power supply that is stepped down to +12V via a DC-DC converter. Connecting a sampling resistor in series with the +12V side not only introduces additional voltage drop and power consumption, but also couples sampling ripple to the load, impacting GPU / NPU stability. Furthermore, the computing payload current varies over a wide dynamic range (approximately 0 to 4A). A single sampling resistor cannot simultaneously achieve both low-current resolution and high-current range, easily leading to over-range or reduced accuracy. Furthermore, a failure in the traditional monitoring link could potentially interrupt power supply, creating a single point of failure risk. Summary of the Invention

[0003] In view of the above existing problems, the present invention is proposed.

[0004] The present invention provides a satellite-borne computing power load current monitoring method to solve the problem of achieving high-precision monitoring and power supply stability under large dynamic current conditions and avoiding single point failures.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions: An embodiment of the present invention provides a method for monitoring the current of a satellite-borne computing power load, which includes the following steps: Step S1: A +28V primary power supply provides +12V power to the computing load GPU / NPU via a DCDC, and a current sampling point is set on the +28V line on the DCDC input side; Step S2: constructing a sampling matrix on the +28V line to form three independent acquisition channels, wherein the three acquisition channels correspond to different equivalent sampling resistors to cover a current dynamic range of 0 to 4A; Step S3: The three acquisition channels respectively amplify their respective pressure difference signals through current detection amplifiers and send them to ADCs, and the MCU reads the three-channel digital values; In step S4, the MCU performs mean filtering on the continuous sampling data of each channel, removes abnormal points, and averages the valid samples to obtain a single sampling result of the channel; In step S5, the MCU selects a channel based on the threshold value: it reads the current value of the second channel. When the current value is greater than 2A, the result of the third channel is selected as the final monitoring value; when the current value is less than 1A, the result of the first channel is selected; and when the current value is between 1A and 2A, the result of the second channel is selected. Step S6, a straight-through channel with no sampling resistance is connected in parallel on the +28V line, and the on-off of the straight-through channel is controlled by the MCU, and the power-on default is off; when any acquisition channel appears abnormal, switch to the straight-through channel to maintain power supply and report the abnormal state, the straight-through channel includes a surge suppression circuit and a current limiting resistor.

[0006] As a preferred scheme of the satellite computing power load current monitoring method, wherein: the equivalent sampling resistances of the three acquisition channels are 0.4Ω, 0.2Ω and 0.1Ω respectively, and the corresponding covered current sections are I<1A, 1A≤I<2A and 2A≤I≤4A respectively.

[0007] As a preferred scheme of the satellite computing power load current monitoring method, wherein: the sampling matrix is composed of four sampling resistances R4, R5, R6 and R7 connected in series, and the resistance values of the four are all 100mΩ, wherein: the first channel is connected across R4+R5+R6+R7 to obtain an equivalent resistance value of 0.4Ω, the second channel is connected across R4+R5 to obtain an equivalent resistance value of 0.2Ω, and the third channel is connected across R7 to obtain an equivalent resistance value of 0.1Ω.

[0008] As a preferred scheme of the satellite computing power load current monitoring method, wherein: the gain of the current detection amplifier is 20, the ADC input dynamic range is 0~12V, and the MCU calculates the current value of each channel according to I=UADC / (20×Req), wherein Req is the equivalent sampling resistance of the current channel.

[0009] As a preferred scheme of the satellite computing power load current monitoring method, wherein: the surge suppression circuit of the straight-through channel is composed of a capacitor C1 and resistors R1 and R3, and the current limiting resistor is R2, which is used to suppress the impact current and surge during the on-off process of the MOS tube.

[0010] As a preferred scheme of the satellite computing power load current monitoring method, wherein: when powered on, the MCU defaults to close the straight-through channel and only enables the sampling matrix for monitoring; when detecting an open circuit of the acquisition channel, saturation of the amplifier or ADC, out-of-range sampling value and other abnormalities, immediately switch to the straight-through channel and report the abnormal state.

[0011] As a preferred scheme of the satellite computing power load current monitoring method, wherein: the DCDC is an isolated power supply with a large input voltage range, and because the sampling point is set at the +28V side, the power supply at the +12V side does not generate ripple due to sampling voltage drop, thereby not affecting the power supply stability of the GPU / NPU.

[0012] As a preferred scheme of the satellite computing power load current monitoring method, the sampling resistance single power consumption is selected as 2W, the single sampling resistance value is determined as 100mΩ by combining the amplifier gain and the ADC range.

[0013] As a preferred scheme of the satellite computing power load current monitoring method, the mean filtering is specifically that the latest 10 sampling points of each channel are cached in time sequence, the maximum value and the minimum value are removed, the remaining 8 samples are averaged to be the sampling result of the channel and participate in the channel selection.

[0014] As a preferred scheme of the satellite computing power load current monitoring method, the MCU records the current value of the discarded channel, the current selected channel number and the threshold judgment result at the same time of outputting the final current value, so as to back up or switch to other channels for continuous monitoring when the fault occurs, and realize fault tolerance without single point fault.

[0015] The present application has the following advantages: The sampling point is located at the +28V side, avoiding introducing additional voltage drop and ripple at the +12V load side, and ensuring the steady-state and transient power supply quality of the GPU / NPU.

[0016] The present application adopts three-channel segmented measurement and threshold selection, and considers small current resolution and large current range in the range of 0-4A.

[0017] The mean filtering of the present application suppresses transient interference and improves measurement stability.

[0018] The present application sets a pass-through channel controlled by the MCU, maintains power supply and reports the exception when the acquisition link is abnormal, and realizes no single point fault.

[0019] The present application has simple structure, low power consumption and is convenient for satellite power system integration. BRIEF DESCRIPTION OF DRAWINGS

[0020] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0021] Figure 1 The hardware structure diagram of the satellite computing power load current monitoring method in embodiment 1.

[0022] Figure 2 The MCU software flow diagram of the satellite computing power load current monitoring method in embodiment 1. DETAILED DESCRIPTION

[0023] In order to make the above objectives, features and advantages of the present application more apparent, specific embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0024] In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present application. However, it will be apparent to one skilled in the art that the present application can be practiced without the specific details given herein, that the present application can be practiced with other than the described embodiments, and that variations from the particular embodiments described herein can be made and still be within the scope of the present application.

[0025] Secondly, the "one embodiment" or "embodiment" referred to herein means that the specific features, structures or characteristics can be included in at least one implementation of the present application. "In one embodiment" appearing in different places in the specification does not mean the same embodiment, nor is it an embodiment that is independent or alternative to other embodiments.

[0026] With the continuous improvement of satellite earth observation resolution, the amount of data generated has increased explosively. Transmitting all the data back to the ground for processing is faced with the dilemma of limited satellite and ground communication bandwidth, resulting in low transmission efficiency, high cost, and greatly discounted timeliness of data application. On the other hand, in many application scenarios such as forest fire prevention and emergency rescue, high requirements are put forward for the real-time and autonomy of satellite data processing. Therefore, sending computing power to space and constructing a space-based computing constellation-based space-based computing facility to reduce data transmission and processing delay have become the trend.

[0027] In the process of sending computing power load to space, there is an urgent need for GPU / NPU current monitoring of computing power load. The space environment in which the satellite operates is extremely complex and harsh, and there are many high-intensity radiations such as galactic cosmic radiation, geomagnetic trapped radiation, and solar particle events. These radiations can pose a serious threat to the chip, such as total dose effect, which can cause the chip performance to continuously decline until it is damaged, and high-energy particle impact on the chip can cause internal technology damage, resulting in sudden increase in current and even chip burning. As the core component of the computing power load, the stable operation of the GPU / NPU is directly related to the reliability of the entire satellite computing power system. By monitoring the current of the GPU / NPU, abnormal changes in the working state of the chip can be detected in time, and when the current abnormally increases, a warning can be quickly issued so that appropriate protection measures can be taken to avoid damage to the chip due to abnormal current and to ensure the continuous and stable operation of the satellite computing power system in a complex space environment, providing a solid guarantee for efficient "space computing".

[0028] The power consumption dynamic range of GPU / NPU is large. For example, the most commonly used GPU / NPU, the static power consumption is only a few watts, and the running power consumption can reach dozens of watts or even hundreds of watts. The general GPU power supply voltage is +12V. Assuming that the maximum power consumption is 100W, the corresponding current dynamic range is 0-8.3A. How to accurately collect the current without affecting the normal power supply is a key problem.

[0029] The existing general system scheme is that a +28V primary power supply is converted into +12V by a DCDC to supply power to the GPI / NPU. The power supply current collection method is to connect a resistor in series on the +12V power supply line, amplify the sampling signal through a current detection amplifier, and then collect it to the ADC controlled by the MCU.

[0030] This method has the following problems: 1. A single series resistor can introduce a single point of failure. When this resistor breaks down due to any problem, the entire computing power load will not work.

[0031] 2. It is difficult to ensure the accuracy of current collection in the dynamic range. In order to ensure the normality of the power supply voltage, the voltage drop on the sampling resistor must be within the specified range. In the 0-8.3A large current dynamic range, the sampling resistor can only be selected according to the maximum current (8.3A). Assuming that the maximum voltage drop allowed by the sampling resistor is 100mV, the sampling resistor is 100mV / 8.3A=12mΩ. The sampling value of this resistor is 6mV when the current is 0.5A. The voltage is too small, and it is difficult to ensure the sampling accuracy due to noise.

[0032] The artificial introduction of 100mV ripple to the GPU / NPU +12V power supply may cause system instability in extreme cases.

[0033] Embodiment 1, refer to Figure 1 and Figure 2 The first embodiment of the present application provides a satellite computing power load current monitoring method. A +28V primary power supply provides +12V for the computing power load through an isolated DCDC. The current sampling point is set on the +28V line at the input side of the DCDC. The sampling matrix is composed of four 100mΩ resistors R4, R5, R6 and R7 connected in series, and three acquisition channels are respectively led out: Channel 1 is connected across R4+R5+R6+R7 (0.4Ω), which is used for small current section (I<1A); Channel 2 is connected across R4+R5 (0.2Ω), which is used for medium current section (1–2A); Channel 3 is connected across R7 (0.1Ω), which is used for large current section (2–4A).

[0034] Three-channel differential pressure signal is amplified by current detection amplifier with a gain of 20 and then enters the ADC (range 0-12 V). The MCU calculates the current value according to the following figure description.

[0035] The MCU continuously executes the following: 1. The average value filtering of the last 10 samples of each channel is performed by "8-point average after extreme value removal", and the channel sampling result is obtained. 2. The current value of the second channel is read and threshold determination is performed: If the value > 2 A, the result of channel 3 is selected as the final monitoring value; If the value < 1 A, the result of channel 1 is selected; If the value is between 1-2 A, the result of channel 2 is selected; 3. The final monitoring value is output to the upper computer / telemetry, and the current value of the discarded channel, the current channel number, and the threshold determination result are recorded for abnormal backtracking or reselection; 4. Abnormal monitoring: when any acquisition channel is detected to be open, the operational amplifier / ADC is saturated, or the sample is out of range, the MCU controls the MOS to be turned on, switches to the straight-through channel to maintain power supply, and reports the abnormal state.

[0036] The straight-through channel includes a surge suppression circuit composed of C1, R1, and R3, and a current limiting resistor R2, which are used to suppress impact current and surge during MOS on-off.

[0037] Figure 1 For the hardware design in Example 1, it includes a straight-through channel, an acquisition channel, and a DCDC. The straight-through channel is composed of a MOS tube, C1, R1, R2, and R3. The MOS tube is selected as a large-current MOS tube. C1, R1, and R3 form a surge suppression circuit. R2 is a current limiting resistor.

[0038] The acquisition channel includes acquisition channels 1, 2, and 3, each of which includes an amplifier and an ADC, and shares one MCU.

[0039] R4+R5+R6+ R7 constitute the sampling resistance of channel 1. R4+R5 constitute the sampling resistance of channel 2. R7 constitutes the sampling resistance of channel 3.

[0040] The DCDC selects an isolated power supply with a large input voltage range.

[0041] Figure 2 For the MCU software in Example 1, it includes a calculation program, average value filtering, and channel selection program.

[0042] 1. The calculation formula of the MCU channel sampling value is: Channel 1: Current = Sampling value / 20 / 0.4, unit A; Channel 2: Current = sample value / 20 / 0.2, unit A; Channel 3: Current = sample value / 20 / 0.1, unit A; 2, After the MCU collects the signal, it needs to be filtered to filter out outliers. Adopt mean filtering, take 10 sampling points, remove the maximum and minimum values, take the mean of the remaining 8 sampling points as the sampling result, and use it for subsequent judgment.

[0043] 3, MCU channel selection scheme: (1) When powered on, the MCU program defaults to turn off the pass-through channel; (2) Determine the calculation result of sampling channel 2, greater than 2A, select the current of channel 3 as the final value. Less than 2A, continue to judge and select the current of channel 1 as the final value. Otherwise, select the current of channel 2 as the final value.

[0044] In the drawing: +28V - primary power bus; DCDC - isolated step-down power supply; GPU / NPU - computing load; R4, R5, R6, R7 - sampling resistors (each 100 mΩ); C1, R1, R3 - surge suppression circuit elements; R2 - current limiting resistor; op-amp - current detection amplifier; ADC - analog-to-digital converter; MCU - controller; MOS - pass-through channel power switch.

[0045] Specifically, the starboard computing load GPU / NPU current monitoring method using a sampling matrix proposed in this embodiment can realize no single point failure, and in a large dynamic current range, it can ensure normal power supply voltage and current sampling accuracy, and solve the following problems: 1) One DCDC is used for GPU / NPU, and the sampling point is moved to the left side of DCDC on the +28V line. This will bring three benefits: First, the sampling current range becomes smaller. With a 100W power consumption, the maximum current on the +28V line is 3.6A, and the DCDC efficiency is 90%, the maximum current is 4A, which is much smaller than the 8.3A on the +12V line.

[0046] Second, it will not cause +12V ripple. The sampling resistor is placed at the +28V end, and the voltage drop formed after passing through the DCDC will not cause fluctuations in the rear +12V.

[0047] Third, the allowed voltage drop will be larger. The input dynamic range of the DCDC is very large, so the selection of the sampling resistor is no longer limited by the maximum voltage drop.

[0048] 2) Calculate the value of the sampling resistor: The selection of the sampling resistor is limited by: power consumption and amplifier multiplication.

[0049] The power consumption is large, and the package will be large accordingly, occupying the PCB area. Moreover, the resistance with too large power consumption cannot meet the precision at the same time. If the resistance for sampling is selected to be 2W, the maximum sampling resistance value is 125mΩ.

[0050] Supposing that the amplifier multiple is 20 and the ADC dynamic range is 0~12V, the maximum sampling voltage is 12V / 20=600mV, and the corresponding sampling resistance is 600mV / 4A=150mΩ.

[0051] Comprehensively considering the two parameters, the sampling resistance is selected to be 100mΩ.

[0052] 3) However, direct single resistance collection will bring another problem: it is impossible to ensure the precision in the whole current dynamic range.

[0053] For example, when the current is 4A, the sampling voltage is 500mV, and when the current is 0.5A, the sampling voltage is 50mV, which is easily affected by noise. Here, the current range of 0~4A is divided into three ranges, i.e., below 1A, 1A~2A and 2A~4A, and 400mΩ, 200mΩ and 100mΩ sampling resistances are respectively used for collection. In this way, when the current is 0.5A, the sampling value is 0.5A*400mΩ=200mV, which can still be normally collected.

[0054] 4) In order to prevent single-point failure, a channel without sampling resistance needs to be added, which is controlled by the MCU through a large-current MOS tube. When the collection channel has a problem, it is switched to the straight-through channel to avoid single-point failure.

[0055] 5) After the signal is collected by the MCU, it needs to be filtered to remove outliers. The mean value filtering is adopted, 10 sampling points are taken, the maximum value and the minimum value are removed, the remaining 8 sampling points are taken to get the mean value, which is used as the sampling result for subsequent judgment.

[0056] 6) The sampling value calculation formula of the MCU channel is as follows: Channel 1: current= sampling value / 20 / 0.4, unit A; Channel 2: current= sampling value / 20 / 0.2, unit A; Channel 3: current= sampling value / 20 / 0.1, unit A; The MCU channel selection scheme is as follows: (a) When powered on, the MCU program defaults to close the straight-through channel; (b) The calculation result of the sampling channel 2 is judged, and if it is greater than 2A, the current of the channel 3 is selected as the final value. If it is less than 2A, the current of the channel 1 is selected as the final value. Otherwise, the current of the channel 2 is selected as the final value.

[0057] Embodiment 2, as a second embodiment of the present application, can be equivalently replaced as follows without changing the method flow S1-S6: a) The threshold value can be set to 0.8 A and 1.8 A according to the load characteristics; b) The current detection amplifier gain can be 15-25; c) The sampling matrix can be realized by parallel connection of 50 mΩ resistors to achieve equivalent 100 mΩ to reduce temperature rise; d) The sample number of mean filtering can be 8-16, still retaining the de-extreme value strategy; e) The abnormal criterion can increase the ADC code type jump / no refresh timeout software alarm items.

[0058] The above replacements do not change the essential characteristics of the sampling point setting on the +28V side, three-channel segmented measurement and threshold selection, and straight-through channel fault tolerance.

[0059] The present method is suitable for star-borne GPU / NPU, FPGA board card, software radio and other load scenarios with high requirements for power supply stability and current monitoring accuracy and large current dynamic range, is easy to integrate with existing star-borne power supply architecture, and has good engineering implementability and popularization value.

[0060] In summary, the embodiment can bring the following effects: Outside the sampling circuit, a backup straight-through path is added to avoid single-point failure.

[0061] The accuracy of current acquisition in the full range can be ensured.

[0062] The acquisition will not affect the GPU / NPU power supply voltage.

[0063] The MCU internally filters the data to avoid wild values.

[0064] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and not to limit, although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced equivalently without departing from the spirit and scope of the technical solutions of the present application, which should be covered in the scope of the claims of the present application.

Claims

1. A method for monitoring the current of a satellite-borne computing load, characterized in that: include: Step S1: A +28V primary power supply provides +12V power to the computing load GPU / NPU via a DCDC, and a current sampling point is set on the +28V line on the DCDC input side; Step S2: constructing a sampling matrix on the +28V line to form three independent acquisition channels, wherein the three acquisition channels correspond to different equivalent sampling resistors to cover a current dynamic range of 0 to 4A; Step S3: The three acquisition channels respectively amplify their respective pressure difference signals through current detection amplifiers and send them to ADCs, and the MCU reads the three-channel digital values; In step S4, the MCU performs mean filtering on the continuous sampling data of each channel, removes abnormal points, and averages the valid samples to obtain a single sampling result of the channel; In step S5, the MCU selects a channel based on the threshold value: it reads the current value of the second channel. When the value is greater than 2A, the result of the third channel is selected as the final monitoring value; when the value is less than 1A, the result of the first channel is selected; and when the value is between 1A and 2A, the result of the second channel is selected. Step S6: Connect a through channel without a sampling resistor in parallel to the +28V line. The through channel is controlled by the MCU and is closed by default upon power-on. When an abnormality occurs in any acquisition channel, the through channel is switched to maintain power supply and report the abnormal state. The through channel includes a surge suppression circuit and a current limiting resistor.

2. A method for monitoring satellite-borne computing power load current according to claim 1, characterized in that: The equivalent sampling resistances of the three acquisition channels are 0.4Ω, 0.2Ω and 0.1Ω respectively, and the corresponding current segments covered are I<1A, 1A≤I<2A, and 2A≤I≤4A respectively.

3. A method for monitoring satellite-borne computing power load current according to any one of claims 1 and 2, characterized in that: The sampling matrix consists of four sampling resistors R4, R5, R6, and R7 connected in series, all of which have a resistance of 100 mΩ. The first channel is connected across R4+R5+R6+R7 to obtain an equivalent resistance of 0.4Ω, the second channel is connected across R4+R5 to obtain an equivalent resistance of 0.2Ω, and the third channel is connected across R7 to obtain an equivalent resistance of 0.1Ω.

4. A method for monitoring satellite-borne computing power load current according to any one of claims 1 to 3, characterized in that: The gain of the current detection amplifier is 20, the ADC input dynamic range is 0-12V, and the MCU calculates the current value of each channel according to I=UADC / (20×Req), where Req is the equivalent sampling resistance of the current channel.

5. The method for monitoring the current of a satellite-borne computing power load according to claim 1, wherein: The surge suppression circuit of the through channel is composed of a capacitor C1 and resistors R1 and R3. The current limiting resistor is R2, which is used to suppress impact current and surge during the switching process of the MOS tube.

6. The method for monitoring the current of a satellite-borne computing power load according to claim 1, wherein: When the MCU is powered on, the pass-through channel is disabled by default and only the sampling matrix is ​​enabled for monitoring. When an abnormality is detected, such as an open circuit in the acquisition channel, saturation of the amplifier or ADC, or an out-of-range sampling value, the MCU immediately switches to the pass-through channel and reports the abnormal status.

7. The method for monitoring the current of a satellite-borne computing power load according to claim 1, wherein: The DCDC is an isolated power supply with a large input voltage range. Since the sampling point is set on the +28V side, the +12V side power supply does not generate ripples due to the sampling voltage drop.

8. A method for monitoring satellite-borne computing power load current according to any one of claims 1 to 4, characterized in that: The sampling resistor is selected based on a single power consumption of 2W, and the resistance of a single sampling resistor is determined to be 100mΩ in combination with the amplifier gain and the ADC range.

9. The method for monitoring the current of a satellite-borne computing power load according to claim 1, wherein: The mean filtering is specifically as follows: the 10 most recent sampling points of each channel are cached in chronological order, the maximum and minimum values ​​are removed, and the remaining 8 samples are averaged as a sampling result of the channel and used in channel selection.

10. The method for monitoring the current of a satellite-borne computing power load according to claim 1, wherein: While outputting the final current value, the MCU records the current value of the discarded channel, the currently selected channel number, and the threshold judgment result.

Images