Multi-integrated environment sensor chip based on end-side AI chip

Abstract

The invention discloses a multi-integrated environment sensor chip based on an end-side AI chip, which comprises a chip substrate, and is characterized in that the end-side AI SoC chip, and a temperature and humidity sensor chip, an illumination sensor chip and a peripheral circuit which are electrically connected with the AI SoC chip are integrated on the chip substrate through a SiP packaging technology; the end side AI SoC chip is internally provided with an algorithm model used for calibrating the collected temperature, humidity and illumination data. According to the technical scheme, the temperature and humidity sensor chip, the illumination sensor chip, the end-side AI SoC chip and peripheral circuits thereof are integrated on the same chip, so that the packaging space of a plurality of chips can be reduced, the design integration level of a circuit system is improved, the power consumption of the system can be reduced, and the reliability is improved.

Description

Technical Field

[0001] The present invention relates to the field of environmental sensor technology, and in particular to a multi-integrated environmental sensor chip based on an end-side AI chip. Background Art

[0002] Monitoring of temperature, humidity and light intensity is one of the most common, widely used and effective monitoring methods in agricultural environmental monitoring. Accurate and real-time environmental monitoring of temperature, humidity and light intensity is not only a key indicator for evaluating the quality of crop growth environment, but also an important means to guide agricultural production and improve crop yield and quality, providing agricultural producers with a vital basis for decision-making. Through the monitoring of temperature, humidity and light intensity, agricultural producers can timely understand the temperature and humidity conditions of the field environment and take necessary control measures, such as opening sunshade nets, irrigation and cooling, covering and heat preservation, and pest and disease control, to create the most conducive environmental conditions for crop growth. The demand for temperature, humidity and light sensors in the field of agricultural environmental monitoring is increasing in terms of high precision and stability, intelligence and integration, cost-effectiveness and convenience.

[0003] At present, there are mature independent sensor chip devices for temperature and humidity sensors and light sensors at home and abroad, which can independently measure temperature, humidity and light intensity. However, the discrete use of temperature, humidity and light sensors will lead to low integration of circuit system design, high complexity of secondary development, large volume and high power consumption, and high application cost. At the same time, due to the lack of edge computing capabilities to coordinate the processing of temperature, humidity and light measurement data, it is also difficult to achieve methods such as calibration of interference between multiple measurement parameters to improve measurement accuracy.

[0004] Therefore, the existing discrete measurement solutions of temperature, humidity and light sensors have gradually become unable to meet the development needs of the agricultural environment monitoring field. Summary of the invention

[0005] The purpose of the present invention is to propose a multi-integrated environmental sensor chip based on an end-side AI chip, which can realize integrated measurement of temperature, humidity, and light intensity and intelligent calibration based on the end-side AI chip.

[0006] To achieve the above technical objectives, the technical solution of the present invention is implemented as follows:

[0007] A multi-integrated environmental sensor chip based on an edge AI chip includes a chip substrate, on which an edge AI SoC chip and a temperature and humidity sensor chip electrically connected to the AI ​​SoC chip, a light sensor chip, and peripheral circuits are integrated through SiP packaging technology. The edge AI SoC chip is provided with an algorithm model for calibrating the collected temperature, humidity, and light data.

[0008] Furthermore, the end-side AI SoC chip includes an NPU acceleration module and a CPU module. The AI ​​SoC chip is electrically connected to the temperature and humidity sensor chip and the light sensor chip through the I2C bus. The peripheral circuit includes a power management module composed of an MCU power filter capacitor and an I2C pull-up resistor.

[0009] Furthermore, a window design is adopted above the measurement holes of the temperature and humidity sensor chip and the light sensor chip.

[0010] Furthermore, the chip substrate is made of BT and has a 2-4 layer stacked structure.

[0011] Beneficial effects of the present invention: The present invention adopts the technical solution of SiP integrated chip to integrate the temperature and humidity sensor chip, the light sensor chip, the end-side AI SoC chip and its peripheral circuits on the same chip, which can not only reduce the packaging space of multiple chips and improve the integration of circuit system design, but also reduce system power consumption and improve reliability;

[0012] In response to the problem that environmental interference between multiple sensors affects measurement accuracy, the solution of the present invention can provide edge computing capabilities for collaborative processing of multi-parameter data by integrating an AI SoC chip, and can realize the calibration of mutual interference between multiple measurement parameters such as temperature, humidity, and light intensity, thereby improving the measurement accuracy and anti-interference ability of multiple integrated environmental sensor chips, and has high engineering application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 It is a basic circuit block diagram of a multi-integrated environmental sensor chip based on an end-side AI chip according to an embodiment of the present invention;

[0014] Figure 2 Schematic diagram of the top package layout and window opening of the environmental sensor chip according to an embodiment of the present invention;

[0015] Figure 3 This is a schematic diagram of the bottom package layout of the environmental sensor chip according to an embodiment of the present invention;

[0016] Figure 4 is a schematic diagram of chip packaging according to an embodiment of the present invention;

[0017] Figure 5 is a side view of the finished structure of the environmental sensor package according to an embodiment of the present invention;

[0018] Figure 6 is a top view of the finished structure of the environmental sensor package according to an embodiment of the present invention;

[0019] Figure 7 This is a network structure diagram of a temperature prediction algorithm according to an embodiment of the present invention;

[0020] Figure 8 This is the temperature prediction algorithm test result 1;

[0021] Fig. 9 This is the second test result of the temperature prediction algorithm. DETAILED DESCRIPTION

[0022] The technical solutions in the embodiments of the present invention are described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0023] A multi-integrated environmental sensor chip based on an edge-side AI chip described in an embodiment of the present invention includes a chip substrate, on which an edge-side AI SoC chip and a temperature and humidity sensor chip electrically connected to the AI ​​SoC chip, a light sensor chip, and peripheral circuits are integrated through SiP packaging technology; the edge-side AI SoC chip is provided with an algorithm model for calibrating the collected temperature, humidity, and light data.

[0024] Specifically, Figure 1 As shown, the circuit design scheme of the multi-integrated environmental sensor chip based on the end-side AI chip of the present invention mainly includes environmental sensors (temperature and humidity sensors, light sensors), three core chips of the end-side AI SoC and their peripheral circuits, and its main function is that the AI ​​SoC is used as the main control and computing unit to collect data from the temperature and humidity sensors and the light sensors through the I2C interface for calculation. The AI ​​SoC provides a variety of standard communication interfaces and multiple GPIO interfaces, which are convenient for interaction with other storage, communication and other modules in the circuit to complete data storage, caching, communication and other functions. The AI ​​SoC is interconnected with the two sensors through the I2C bus. The power supply of the three functional chips is provided by the power management module.

[0025] Based on the above circuit design scheme, this scheme uses SiP (System-in-Package) system integration technology to integrate the above chips, devices and circuits into a package. Figure 2-3 As shown, the specific technical solution is as follows:

[0026] 1. The specific packaging solution includes 1 temperature and humidity sensor chip, 1 light sensor chip, 1 AI SoC chip, 4 MCU power filter capacitors, and 2 I2C pull-up resistors;

[0027] 2. The temperature and humidity sensor chip and the light sensor chip have windows above their measurement holes to ensure environmental data collection and avoid temperature interference between the two sensors.

[0028] 3. Adopt organic substrate packaging, the organic substrate stacking structure adopts 2-4 layers, and the packaging substrate adopts BT material.

[0029] When the multi-integrated environmental sensor chip designed by the present invention is used to collect temperature, humidity, and light data, the collected data can be intelligently calibrated through the end-side AI SoC. The specific technical solution is as follows.

[0030] 1. The external system issues data collection instructions through multiple standard communication interfaces or multiple GPIO interfaces provided by the AI ​​SoC chip;

[0031] 2. When the AI ​​SoC chip receives the data collection instruction, it collects and stores the measurement data of the temperature and humidity sensor and the light sensor through the I2C interface;

[0032] 3. The AI ​​SoC chip collaboratively calibrates the temperature, humidity, and light data through a locally stored calibration algorithm model based on a deep neural network. The model input parameters can be the historical trend data and current data of temperature, humidity, and light stored locally on the AI ​​SoC chip, and the output is the calculated model prediction data of temperature, humidity, and light.

[0033] 4. The AI ​​SoC chip outputs the final model prediction data of temperature, humidity, and light through a variety of standard communication interfaces or multiple GPIO interfaces.

[0034] The calibration scheme proposed in the present invention can realize the calibration of mutual interference among multiple measurement parameters such as temperature, humidity and light intensity, and also supports trend predictive calibration of current data through historical trend data, thereby improving the measurement accuracy and anti-interference ability of multi-integrated environmental sensor chips.

[0035] The embodiment of the present invention uses three core chips with the following characteristics: a temperature and humidity sensor, a light sensor, and an end-side AI SoC for SiP packaging integration. The specific implementation plan is as follows.

[0036] The temperature and humidity sensor consists of three parts: a temperature-sensitive element, a humidity-sensitive element, and a digital control circuit. Among them, temperature acquisition mainly relies on the characteristic that thermistor value changes with temperature to measure temperature; humidity acquisition mainly relies on the characteristic that the capacitance of humidity-sensitive capacitors made of polymer film capacitors changes with ambient humidity to measure humidity. The analog front end of the sensor is converted into an electrical signal through a humidity-sensitive element, and then converted into a digital signal through an analog-to-digital conversion module. The digital signal is linearized and consistent, and then output to the AI ​​SoC chip. The communication interface of the chip uses I2C interface communication.

[0037] The light sensor mainly includes a photosensitive element, a signal processing circuit and an output circuit. The working principle of the light sensor is mainly based on the photoelectric conversion characteristics of the photosensitive element. When the light intensity changes, the photosensitive element will sense this change and convert it into a corresponding electrical signal. The signal processing circuit is used to amplify, filter and linearize the electrical signal output by the photosensitive element to improve the measurement accuracy and stability of the sensor. The output circuit is responsible for converting the processed electrical signal into a standard signal output, such as analog voltage, analog current or digital signal, to facilitate subsequent data collection and processing. The communication interface of the chip uses I2C interface communication.

[0038] The AI ​​SoC chip mainly uses the in-memory computing NPU acceleration module as the main computing unit, and has a built-in 32-bit RISC-V architecture CPU as the main controller. In addition, its peripheral circuits include access modules such as ADC, AGC, FIFO, and digital filters for collecting various time series data; in terms of digital interfaces, it mainly includes UART, I2C, SPI, I2S, and standard GPIO and other interface forms. In addition, the chip architecture also includes modules such as built-in crystal oscillator, 320KB SRAM memory, and power management unit to ensure the normal operation of the chip. The interface between the AI ​​SoC chip and the temperature and humidity sensor and the light sensor is interconnected through the I2C bus.

[0039] This solution uses SiP system integration technology to integrate three core chips, temperature and humidity sensor, light sensor, and end-side AI SoC, as well as peripheral circuits, into one package. The package form is FCBGA, the package substrate area size is 10*10mm, the BGA ball pitch is 1mm, and the 9*9 array has a total of 81 balls. The organic substrate stacking structure uses 2-4 layers, the package substrate uses BT material, and the board thickness is 570um. The chip package diagram is shown in the figure. Figure 4 shown.

[0040] The side view of the finished sensor package structure is as follows: Figure 5 As shown, the BGA height (A1) is 0.6 mm, the substrate thickness (A2) is 0.57 mm, the total thickness of the metal cover (A3) is 1.04 mm, the thickness of the upper metal layer of the metal cover is 0.5 mm, and the height of the metal cover cavity is 0.54 mm.

[0041] The top view of the finished sensor packaging structure is as follows Figure 6As shown, the substrate size is 10*10mm, the metal cover size is 9.6*9.6mm, the metal cover cavity size is 6*6mm, and the metal cover foot width is 1.8mm. The metal cover is processed by CNC, and the visible light holes and air holes are reserved for the two sensors. Both are circular holes with diameters of 0.8mm and 0.6mm respectively. The embodiment of this solution uses the following characteristics of temperature and humidity sensor, light sensor, and end-side AI SoC three core chips for SiP packaging integration. The specific implementation plan is as follows.

[0042] The present invention can intelligently calibrate the collected temperature, humidity, and light data through the end-side AI SoC, and the AI ​​SoC can support the deployment of multiple calibration algorithm models. This solution takes a temperature data prediction algorithm based on deep learning as an example, and the specific implementation plan is as follows:

[0043] In this implementation, a temperature data prediction algorithm based on deep learning is used. The algorithm adopts a 6-layer structure, and its network structure diagram is as follows: Figure 7 As shown in the figure. The main computing operators of the algorithm include CNN, GeMM, etc. The input data of the algorithm model is the processed time series temperature data, and the numerical value of the next temperature data is predicted. After quantization, pruning, and retraining optimization, the algorithm model is deployed to the AI ​​SoC for calculation, and finally the data is read out through Matlab for accuracy comparison.

[0044] The algorithm test results are as follows Figure 8 and Fig. 9 The blue line represents the original test data, the red line represents the floating-point model inference result before quantization, the green line represents the quantized model inference result, the yellow line represents the simulation hardware inference result, and the purple line represents the hardware measurement result of the multi-integrated environmental sensor chip. According to the test results, it can be verified that the multi-integrated environmental sensor chip can meet the algorithm model deployment requirements and realize the intelligent calibration solution based on the end-side AI chip.

[0045] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A multi-integrated environmental sensor chip based on an end-side AI chip, comprising a chip substrate, characterized in that: The chip substrate is integrated with an end-side AI SoC chip, a temperature and humidity sensor chip electrically connected to the AI ​​SoC chip, a light sensor chip, and peripheral circuits through SiP packaging technology. The end-side AI SoC chip is equipped with an algorithm model for calibrating the collected temperature, humidity, and light data.

2. The multi-integrated environmental sensor chip based on the end-side AI chip according to claim 1, characterized in that: The end-side AI SoC chip includes an NPU acceleration module and a CPU module. The AI ​​SoC chip is electrically connected to the temperature and humidity sensor chip and the light sensor chip through the I2C bus. The peripheral circuit includes a power management module composed of an MCU power filter capacitor and an I2C pull-up resistor.

3. The multi-integrated environmental sensor chip based on the end-side AI chip according to claim 2, characterized in that: A window design is adopted above the measurement holes of the temperature and humidity sensor chip and the light sensor chip.

4. The multi-integrated environmental sensor chip based on the end-side AI chip according to claim 3 is characterized in that: The chip substrate is made of BT and has a 2-4 layer stacked structure.

Images