Miniaturized packaging method for chips

By using adapter board and neural network model in chip package to optimize lead layout, the problem of unreasonable and excessive length of lead layout is solved, the quality and efficiency of chip package is improved, the risk of wire punching is reduced, and higher signals are achieved Integrity and stability.

CN118763010BActive Publication Date: 2024-11-15BEIJING HUACHUANG QIXING MICROELECTRONICS CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411224793.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-09-03
Publication Date
2024-11-15
Estimated Expiration
2044-09-03

AI Technical Summary

Technical Problem

Unreasonable lead layout and too long leads leads to low packaging quality in chip miniaturized packaging. Especially in three-dimensional stacking technology, there are problems such as too long lead spans of the top and bottom chips, reduced insulation resistance during welding and increased short circuit risk. .

Method used

By adding adapter boards during chip packaging, and using three-dimensional point cloud data and neural network models to analyze pin marking points and distribution angles, constructing distribution coefficients and differences matrix, optimizing lead layout to reduce lead collapse deformation and wire punching risks, Improve wire bonding strength and reduce yield loss of direct bonding between chips.

Benefits of technology

It significantly improves the strength of wire bonding, reduces the risk of wire punching, improves the efficiency and quality of chip packaging, achieves higher signal integrity and stability, and avoids performance degradation caused by differences in packaging space characteristics.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118763010B_ABST
    Figure CN118763010B_ABST
Patent Text Reader

Abstract

The present invention relates to the field of chip packaging technology, and proposes a miniaturized packaging method for chips, including: raw material preparation, mounting the bottom chip on the substrate, and then mounting the adapter board and the top chip; after mounting, wire bonding and welding are performed to complete the miniaturized packaging of the chip, and after packaging, the three-dimensional point cloud data of the entire packaged assembly device is collected, and the stacking structure of the package in the chip miniaturized packaging process is adaptively adjusted by analyzing the structural deviation characteristics of the three-dimensional point cloud data of the miniaturized packaging assembly device of the produced chip. The present application improves the quality of chip miniaturized packaging by adding an adapter board to the miniaturized packaging of the chip and adaptively adjusting the process of stacking and packaging the chip.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the technical field of chip packaging, and in particular to a miniaturized packaging method for chips. Background Art

[0002] Chip miniaturization packaging has multiple advantages, such as reduced size and weight, high integration, improved electrical performance, optimized heat dissipation and power consumption management, and increased cost and production efficiency. This makes miniaturization packaging an indispensable technology and means in the design and manufacturing of modern electronic products. As the size of transistors gradually approaches the atomic scale, the technical difficulty and cost of maintaining Moore's Law increase exponentially, and it is becoming increasingly uneconomical to achieve miniaturization of chips and devices by reducing the size of transistors. By stacking chips and devices in three dimensions in the height direction, device miniaturization can be achieved, which is less technically difficult and less costly.

[0003] Common three-dimensional stacking technologies include chip stacking bonding technology, package stacking technology, buried substrate technology, 2.5D wafer-level / board-level fan-out packaging technology, TSV technology, etc. Chip stacking bonding technology can achieve chip stacking through micro-assembly technologies such as bonding and bonding. It has low technical difficulty, small equipment requirements and low cost. From a structural perspective, it can generally be divided into three types: I-shaped, pyramid-shaped and stepped. The pyramid shape requires a larger size between chips. The I-shaped shape requires adding silicon wafers between chips to increase the total height of the package. The stepped shape requires that all chip pads are distributed on one side of the chip. These structures have strict requirements for wire-bonded chips and generally cannot achieve stacking of passive devices.

[0004] The packaging stacking technology can easily realize the stacking of chips and passive devices at the same time, but the interlayer interconnection channels such as solder balls / copper pillars occupy a large area of ​​the upper and lower substrates, and the size of the package is limited. In addition, the different warping sizes and directions of the upper and lower packages also significantly affect the yield of the finished product; the electrical characteristics of the chip package are crucial to the chip performance. For example, the layout of the chip pins, the length of the connection wires, and the impedance matching of the signal transmission will affect the signal integrity, speed and stability of the chip. In the packaging process, factors such as the length of the signal line, the routing method, and the packaging material will affect the integrity of the signal. Therefore, in the process of miniaturizing the chip packaging, unreasonable lead layout and excessive lead length will reduce the quality of the chip after packaging. Summary of the invention

[0005] The present application provides a miniaturized packaging method for a chip to solve the problem of unreasonable layout of chip miniaturized packaging leads and excessively long leads leading to low packaging quality. The technical solution adopted is as follows:

[0006] An embodiment of the present application provides a miniaturized packaging method for a chip, the method comprising the following steps:

[0007] Raw material preparation: Raw materials include flip chips, passive components, wire-bonded chips and adapter boards, where the adapter board consists of a main layer, an insulating layer, a circuit layer and a dielectric layer;

[0008] Mounting the bottom chip: fix the bottom chip and substrate;

[0009] Mounting the adapter board and top chip: fix the passive components, adapter board and top chip;

[0010] Wire bonding: Attach the adapter board with passive components, flip-chip chips and wire-bonded chips to the bottom chip for assembly, and finally bond the top chip, adapter board and bottom chip layer by layer;

[0011] Collect the 3D point cloud data of the packaged device and the plane coordinate data in the top view during the chip miniaturization packaging process to form a chip packaging layout sample;

[0012] The three-dimensional point cloud data is mapped to the XY plane in the spatial rectangular coordinate system to obtain the plane point cloud data, and the points in the plane point cloud data located on the plane coordinate data are used as pin marking points; the center point of the plane point cloud data is used as the segmentation center, and sections parallel to the XZ plane and the YZ plane in the spatial rectangular coordinate system are made to segment the three-dimensional point cloud data into four subspaces;

[0013] Determine the distribution coefficient of each subspace based on the spatial position distribution of the pin marking points in the subspace and the angle distribution of the line between the pin marking points and the central marking point;

[0014] The distribution coefficients of all subspaces are combined into distribution vectors. Based on the similarity of distribution vectors between chip package layout samples and the difference in packaging time, a difference matrix is ​​constructed. A neural network model is used to obtain the difference results of lead layout of chip miniaturization package, and the lead layout of chip miniaturization package is re-performed.

[0015] Preferably, the material of the circuit layer is copper, and the surface of the pad exposed to the dielectric layer is subjected to chemical nickel-palladium-gold plating, chemical nickel-gold plating, electrolytic nickel-gold plating and electrolytic nickel-silver plating.

[0016] Preferably, the substrate is one of a plastic package substrate, a lead frame, a ceramic substrate and a microcrystalline glass substrate.

[0017] Preferably, the method of fixing to the substrate is one of wire bonding and flip chip bonding.

[0018] Preferably, the assembling method is one of bonding and welding processes.

[0019] Preferably, the distribution coefficient of each subspace is determined by:

[0020] For each subspace, the positive direction of the X-axis in the spatial rectangular coordinate system is taken as the starting direction, and the pin marking points of each layer in the subspace are arranged in the order of rotating counterclockwise from the starting direction to form a distribution marking sequence of each layer;

[0021] The angle between the line connecting the pin mark point corresponding to each element in the distribution mark sequence and the center point of the plane point cloud data and the starting direction is used as the distribution mark angle of each element, and the distribution mark angles of all elements in the distribution mark sequence constitute a distribution mark angle sequence;

[0022] According to the difference between the distribution marking angle sequences of two adjacent layers in each subspace, the distribution coefficient of each subspace is obtained by combining the spatial distance of adjacent elements in the distribution marking sequences of each layer and the difference in the distribution marking angles of adjacent elements.

[0023] Preferably, the expression for further obtaining the distribution coefficient is: ; In the formula, Represents the 3D point cloud data The distribution coefficient of the subspace; Represents the 3D point cloud data In the subspace The layer's package layout factor; Represents the 3D point cloud data In the subspace The layer's package layout factor; and Respectively represent the first In the subspace Layer and The distribution label angle sequence corresponding to the layer, represents the Manhattan distance function; Represents the 3D point cloud data The number of levels of subspace division.

[0024] Preferably, the three-dimensional point cloud data In the subspace The expression for the layer's encapsulation layout factor is: ; In the formula, Represents the 3D point cloud data In the subspace The layer's package layout factor; Represents the 3D point cloud data In the subspace The number of pin marker points of the layer, Represents the 3D point cloud data The number of pin marking points of the layer; Represents the 3D point cloud data In the subspace The distribution label sequence corresponding to the layer elements and The Euclidean distance of elements in space; and Respectively represent the first In the subspace The distribution label sequence corresponding to the layer elements and The distribution of the elements is marked with angles, and cos represents the cosine function; Represents the 3D point cloud data In the subspace The number of elements in the distribution label sequence corresponding to the layer.

[0025] Preferably, the method for constructing the difference matrix includes:

[0026] The similarity between the distribution vectors corresponding to the chip package layout samples is used as the clustering distance, and a clustering algorithm is used to obtain the clustering results of all chip package layout samples;

[0027] For each cluster, a sequence of chip package layout samples in the cluster in the order of corresponding production time is used as a chip package layout sequence; the difference of packaging time corresponding to adjacent elements in the chip package layout sequence of the cluster is calculated, and the mean of all the differences in the chip package layout sequence is used as the packaging time difference factor of the cluster;

[0028] According to the differences in the encapsulation time difference factor, the number of elements and the distribution vector between the clusters, the difference factor between the clusters is obtained;

[0029] The difference factors between each cluster and all other clusters are composed into a sequence as the row vectors of the difference matrix.

[0030] Preferably, the calculation expression of the difference factor is: ; In the formula, Indicates Clusters and The difference factor between clusters; and Respectively represent Clusters and The encapsulation time difference factor of the clusters; Indicates The first distribution vector, Indicates The first distribution vector, express and The cosine similarity between ; and Respectively represent Clusters and The number of distribution vectors in a cluster.

[0031] The beneficial effect of the present application is as follows: by analyzing the problem of the lead process in the chip packaging process, an adapter plate is added in the chip miniaturization packaging process, which has the beneficial effect of overcoming the problem of lead collapse and deformation caused by the long lead span of the top chip, the punching caused by the plastic packaging process, and the insulation resistance is reduced or even short-circuited when the top and bottom chips with large stacking differences are stacked using the traditional pyramid chip stacking technology, and the strength of the lead bonding is significantly improved, and the risk of wire bonding such as punching in the subsequent plastic packaging process is reduced. In addition, the adapter plate with circuits can be used for the integration of passive devices, while reducing the problem of large yield loss due to direct bonding between the bottom chip and the top chip. Furthermore, according to the differences in the stacking space structure and the lead layout in the chip packaging production process after adding the adapter plate, the distribution coefficient and the difference factor are successively constructed, and the difference matrix is ​​constructed according to the difference factor. The stacking space deviation and the lead layout deviation in the chip packaging process are reflected through the difference matrix. The beneficial effect is that the spatial characteristic changes of the chip packaging production are considered to avoid reducing the performance and quality of the chip packaging due to the difference in the packaging space characteristics, thereby improving the efficiency and quality of the miniaturized packaging of the chip, and completing the miniaturized packaging of the chip. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0033] Figure 1 A schematic diagram of a process flow of a miniaturized packaging method for a chip provided by an embodiment of the present application;

[0034] Figure 2 A schematic diagram of an adapter plate structure provided in one embodiment of the present application;

[0035] Figure 3 A schematic diagram of mounting a bottom chip provided by an embodiment of the present application;

[0036] Figure 4 A schematic diagram of a mounting adapter board and a top-layer device provided in one embodiment of the present application;

[0037] Figure 5 A schematic diagram of wire bonding provided in accordance with an embodiment of the present application. DETAILED DESCRIPTION

[0038] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0039] See also Figure 1 , which shows a flow chart of a method for miniaturizing packaging of a chip provided by an embodiment of the present application, the method comprising the following steps:

[0040] Step S001, obtaining a chip package layout sample.

[0041] Using chip stacking technology to miniaturize chip packaging can effectively solve the problem of too long leads in the packaging process. However, chip stacking technology has limitations on chip size and pad distribution, poor applicability, and inability to integrate passive devices. Therefore, in order to address the above problems, the lead layout is adjusted in real time during the chip miniaturization packaging process. At the same time, an adapter board is added between the bottom chip and the top chip to reduce the problem of low chip miniaturization packaging quality caused by the lead layout during the chip miniaturization packaging process.

[0042] Furthermore, the process of chip miniaturization packaging includes: raw material preparation, mounting the bottom chip, mounting the adapter board and the top device, and wire bonding. The specific packaging process is as follows:

[0043] Raw material preparation: Raw materials include one or more of flip chips, passive components, wire bonding chips and adapter boards, among which the adapter boards are Figure 2 As shown: the main material 1 is silicon, which is closest to the thermal expansion coefficient of the chip, and the thermal stress inside the device is minimal. The insulating layer 2 is an optional structure, and its main purpose is to increase the insulation resistance between the circuit layer 3 and the main material 1. The material of the insulating layer is silicon dioxide, silicon nitride and a mixture thereof. The number of layers of the circuit layer 3 is generally within 4, and its material is copper Cu. The surface of the pad exposed to the dielectric layer 4 is treated with chemical plating of nickel palladium gold, chemical plating of nickel gold, electroplating of nickel silver and electroplating of nickel gold, etc., to facilitate subsequent wire bonding and prevent oxidation during bonding and welding. The dielectric layer 4 plays the role of insulation between circuit layers and protection of the top metal, and its material is silicon dioxide. The back film 5 is an optional structure, which plays the role of back protection and reducing the warping of the adapter board. After the multi-layer circuit layer and dielectric layer are made, the inside of the adapter board will warp due to stress. The back film applies reverse stress to it to balance the warping, and is generally an organic material.

[0044] Mounting the bottom chip: The substrate can be selected from a plastic substrate, a lead frame, a microcrystalline glass substrate and a ceramic substrate according to the number of pins, the pin density and the package type. The bottom chip can be interconnected with the substrate by wire bonding or flip-chip bonding. For example, the bottom chip 102 is fixed to the substrate 101 by bonding or welding by wire bonding. The bonding method can be liquid conductive glue or conductive dry film. The specific mounting schematic diagram is shown in the figure. Figure 3 shown.

[0045] Mounting the adapter board and top-layer components: The flip-chip chip 104 and the passive components 105 are mounted on the adapter board 103. Generally, welding is used for assembly. After welding, necessary cleaning is performed. To improve reliability, the flip-chip chip 104 can be selectively bottom-filled. Subsequently, the wire-bonded chip 106 is mounted, such as Figure 4 As shown, the wire-bonded chip 106 is fixed to the transfer plate 103 by bonding or welding, and bonding is generally used. If the cleaning liquid does little damage to the chip pad during cleaning after welding, and the adhesive of the wire-bonded chip 106 has a high temperature resistance, it can also be done first. Figure 5 The wire bonding chip is mounted, and then the passive components and flip chip are welded. The schematic diagram of the specific adapter board and top-level components is as follows Figure 4 shown.

[0046] Wire bonding: The adapter board 103 with the passive components 105, flip chip 104 and wire bonding chip 106 is mounted on the bottom chip 102. The assembly is generally performed by bonding, but welding can also be used. Finally, the top chip, adapter board and bottom chip are bonded layer by layer. The specific bonding diagram is shown in the figure. Figure 5 shown.

[0047] Among them, after the bottom chip, the adapter board and the top device are bonded, a CCD industrial camera is used to collect a top view of the bonded overall device directly above the bonded overall device, and the top view is used as input to obtain the plane coordinate data of the pins in the bonded overall device using a recurrent neural network model. The specific training process of the recurrent neural network model is a well-known technology and will not be repeated here. In other embodiments of the present application, a convolutional neural network can also be used to obtain the plane coordinate data of the pins in the bonded overall device.

[0048] A 3D scanner is used to collect the 3D point cloud data of the entire device after bonding. Since the collection environment is the miniaturized chip packaging workshop, the environment has a greater impact on the accuracy of the point cloud data. Therefore, the input is the collected 3D point cloud data of the entire chip after bonding, and the point cloud radius filtering algorithm is used to obtain the pre-processed 3D point cloud data.

[0049] Furthermore, during the production process n (The value in this embodiment is 100) 3D point cloud data of the chip miniaturization packaging process and the top view of the overall device after bonding, the pre-processed 3D point cloud data of each chip miniaturization packaging process and the plane coordinate data of the pins in the overall device after bonding are used as a chip packaging lead layout sample, n A set of chip package lead layout samples is taken as the chip package lead layout set. The specific calculation process of the point cloud radius filtering algorithm is a well-known technology and will not be described in detail.

[0050] Step S002, constructing a distribution coefficient by analyzing the distribution characteristics of point cloud data in the chip packaging process.

[0051] The combination of 3D point cloud technology in the miniaturized packaging process of chips can be used for high-precision 3D inspection and measurement to ensure the quality and performance of the packaging. In this regard, 3D point cloud technology can provide high-resolution and high-precision point cloud data to help achieve comprehensive inspection and analysis of chips. Since the miniaturized packaging process of chips often faces reliability issues such as warping, delamination, and lead offset, a reasonable 3D packaging structure is an effective means to improve device reliability from the perspective of production technology. The 3D packaging structure directly affects the layout of the leads, the packaging quality, and the performance of the packaged products.

[0052] The layout of the pins after chip bonding is analyzed by collecting the three-dimensional point cloud data in the chip packaging process, and the lead layout in the chip miniaturization packaging process is adjusted according to the analysis results. Specifically, taking a chip packaging layout sample in the chip packaging lead layout set as an example, the three-dimensional point cloud data of the chip packaging layout sample is mapped to the XY plane in the spatial rectangular coordinate system, and the mapped three-dimensional point cloud data is used as the plane point cloud data, and the points in the plane point cloud data located on the plane coordinate data of the pins in the overall device after the chip packaging layout is bonded are used as pin marking points.

[0053] Furthermore, since the top view of the chip after miniaturization packaging has a rectangular edge shape, the center point of the plane point cloud data of the chip packaging layout sample is used as the center marking point, and the center marking point is used as the segmentation center to make sections parallel to the XZ plane and the YZ plane in the spatial rectangular coordinate system, and the three-dimensional point cloud data of the chip packaging layout sample is divided into points in four subspaces; since the chip is divided into four layers after bonding: the first layer substrate, the second layer bottom chip, the third layer adapter board, and the fourth layer top chip, each layer has a corresponding pin marking point.

[0054] For each subspace of the three-dimensional point cloud data of the chip package layout sample, the positive direction of the X-axis in the spatial rectangular coordinate system is taken as the starting direction, and the sequence of the pin marking points of each layer in the subspace is arranged in the order of rotating counterclockwise from the starting direction as the distribution marking sequence, and the angle between the connecting line of the pin marking point and the center marking point corresponding to each element in the distribution marking sequence and the starting direction is taken as the distribution marking angle of each element, and the sequence composed of the distribution marking angles of all elements in the distribution marking sequence is taken as the distribution marking angle sequence, and the distribution coefficient is calculated according to the spatial distribution characteristics of the pin marking points. The specific calculation process is as follows:

[0055] ,

[0056] ,

[0057] In the formula, Represents the 3D point cloud data In the subspace The layer's package layout factor; Represents the 3D point cloud data In the subspace The number of pin marker points of the layer, Represents the 3D point cloud data The number of pin marking points of the layer; Represents the 3D point cloud data In the subspace The distribution label sequence corresponding to the layer elements and The Euclidean distance of elements in space; and Respectively represent the first In the subspace The distribution label sequence corresponding to the layer elements and The distribution of elements is marked with angles, and cos represents the cosine function express The cosine of ; Represents the 3D point cloud data In the subspace The number of elements in the distribution tag sequence corresponding to the layer;

[0058] Represents the 3D point cloud data The distribution coefficient of the subspace; Represents the 3D point cloud data In the subspace The layer's package layout factor; and Respectively represent the first In the subspace Layer and The distribution label angle sequence corresponding to the layer, express and The Manhattan distance between Represents the 3D point cloud data The number of levels of subspace division.

[0059] If the 3D point cloud data In the subspace The number of pin marking points on the layer is large, so the calculated The larger the value, the greater the In the subspace The pin marking points of the layer are arranged relatively far apart, but the angles between the pin marking points relative to the center marking point are close, so the calculated The larger the value of The smaller the value of The larger the value, the larger the 3D point cloud data In the subspace Packing layout factor of the layer The larger the value, the higher the 3D point cloud data In the subspace The pins of the layers are relatively densely distributed, and the leads are closely distributed during wire bonding, which affects the electrical performance and quality of the chip.

[0060] Furthermore, if the 3D point cloud data The larger the encapsulation layout factor between different layers in a subspace, the larger the calculated The larger the value, the higher the 3D point cloud data The distribution of pin marking points in each layer in the subspace may cause the distribution of leads to overlap. At the same time, the angle distribution difference of pin marking points in different layers relative to the central marking point is small, so the calculation results are The smaller the value, the more dense the pins on different layers are and the closer the pins are to each other, which makes it more likely that the pins will be similarly distributed during wire bonding. The distribution coefficient of the subspace The larger the value, the higher the 3D point cloud data The distribution of pins within a subspace causes the leads in the lead layout to be close or even overlapped, which in turn affects the chip packaging performance and quality.

[0061] So far, the distribution coefficient can be obtained according to the above process of this embodiment.

[0062] Step S003, calculating the difference factor according to the distribution of the distribution coefficients, and constructing a difference matrix.

[0063] A vector composed of distribution coefficients of all subspaces of the three-dimensional point cloud data of each chip package layout sample in the chip package lead layout set arranged in a counterclockwise order in the starting direction is used as a distribution vector, and the differences in package lead layouts in the automated packaging process are analyzed based on the distribution vectors corresponding to all chip package layout samples; specifically, the input is a chip package lead layout set, and an agglomerative hierarchical clustering algorithm is used to obtain the clustering results of the chip package lead layout set, and the cosine similarity between the distribution vectors corresponding to the package layout samples is used as a measurement result of the similarity between the package layout samples in the clustering process. The specific implementation process of the agglomerative hierarchical clustering algorithm is a well-known technology and will not be repeated here.

[0064] Furthermore, for the clustering results of the chip package lead layout set, the sequence of chip package layout samples of each cluster in the clustering results in the order of corresponding production time is used as the chip package layout sequence, and the difference factor is calculated according to the distribution vector difference between different clusters in the clustering results. The difference factor reflects the spatial structure difference and lead layout difference in the chip packaging process as the chip packaging production proceeds. The specific calculation process is as follows:

[0065] ,

[0066] ,

[0067] In the formula, Indicates The encapsulation time difference factor of the clusters; and Respectively represent The first cluster of chip package layout sequence and The encapsulation moment corresponding to each element; Indicates The number of elements in the chip package layout sequence of the cluster;

[0068] Indicates Clusters and The difference factor between clusters; and Respectively represent Clusters and The encapsulation time difference factor of the clusters; Indicates The first distribution vector, Indicates The first distribution vector, express and The cosine similarity between ; and Respectively represent Clusters and The number of distribution vectors in a cluster.

[0069] Jordi If the production time difference between the chip package layout samples in the clusters is large, the calculated Packing time difference factor of clusters The larger the value of , the more discontinuous deviation characteristics appear in the production process of chip packaging, and the greater the possibility of deviation in the spatial structure and lead layout of chip packaging. Furthermore, if both clusters show large discontinuous deviation characteristics, that is, The larger the value of is, the closer the number of distribution vectors in the two clusters is, and the similarity of the distribution vectors between the two clusters is relatively large, then the calculated Clusters and The difference factor between clusters The larger the value of, the greater the possibility of lead layout structure differences in the chip packaging process.

[0070] Furthermore, for each cluster in the clustering results of the chip package lead layout set, a sequence composed of difference factors between the cluster and other clusters in the clustering results is used as the chip miniaturization packaging performance difference sequence of the cluster, and a row of elements in the chip miniaturization packaging performance difference sequence matrix of each cluster, and a matrix composed of the chip miniaturization packaging performance difference sequences of all clusters is used as a difference matrix.

[0071] Step S004, adjusting the lead layout in the miniaturized packaging process of the chip according to the difference matrix, and packaging the chip based on the adjustment result.

[0072] Furthermore, the difference matrix is ​​used as input, and a CNN (Convolutional Neural Networks) convolutional neural network model is used to obtain the difference results of the lead layout of the chip miniaturization package. The output results include 1 or 0. If the output result is 0, there is no deviation in the lead layout of the current chip miniaturization package, and there is no need to adjust the layout of the chip miniaturization package; if the output result is 1, an adjustment instruction is sent to the chip miniaturization package control system, and the chip miniaturization package control system re-layouts the lead layout of the chip miniaturization package according to the received instruction, and completes the adjustment of the stacking between devices during the chip miniaturization packaging process. It should be noted that the loss function of the CNN convolutional neural network training adopts the cross loss function, and the optimizer is the Adam optimizer. The specific training process of the CNN convolutional neural network model is a well-known technology and will not be repeated.

[0073] In other embodiments of the present application, a recursive neural network model may also be used to obtain lead layout difference results of chip miniaturization packaging.

[0074] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the principles of the present application shall be included in the protection scope of the present application.

Claims

1. A miniaturized packaging method for a chip, characterized in that: The method comprises the following steps: Raw material preparation: Raw materials include flip chips, passive components, wire-bonded chips and adapter boards, where the adapter board consists of a main layer, an insulating layer, a circuit layer and a dielectric layer; Mounting the bottom chip: fix the bottom chip and substrate; Mounting the adapter board and top chip: fix the passive components, adapter board and top chip; Wire bonding: Attach the adapter board with passive components, flip-chip chips and wire-bonded chips to the bottom chip for assembly, and finally bond the top chip, adapter board and bottom chip layer by layer; Collect the 3D point cloud data of the packaged device and the plane coordinate data in the top view during the chip miniaturization packaging process to form a chip packaging layout sample; The three-dimensional point cloud data is mapped to the XY plane in the spatial rectangular coordinate system to obtain the plane point cloud data, and the points in the plane point cloud data located on the plane coordinate data are used as pin marking points; the center point of the plane point cloud data is used as the segmentation center, and sections parallel to the XZ plane and the YZ plane in the spatial rectangular coordinate system are made to segment the three-dimensional point cloud data into four subspaces; Determine the distribution coefficient of each subspace based on the spatial position distribution of the pin marking points in the subspace and the angle distribution of the line between the pin marking points and the central marking point; The distribution coefficients of all subspaces are combined into a distribution vector; based on the distribution vector similarity between chip package layout samples and the packaging time difference, a difference matrix is ​​constructed, and a neural network model is used to obtain the lead layout difference results of the chip miniaturization package, and the lead layout of the chip miniaturization package is re-performed; The expression of the distribution coefficient is: ; In the formula, Represents the 3D point cloud data The distribution coefficient of the subspace; Represents the 3D point cloud data In the subspace The layer's package layout factor; Represents the 3D point cloud data In the subspace The layer's package layout factor; and Respectively represent the first In the subspace Layer and The distribution label angle sequence corresponding to the layer, represents the Manhattan distance function; Represents the 3D point cloud data The number of levels of subspace division; The three-dimensional point cloud data In the subspace The expression for the layer's encapsulation layout factor is: ; In the formula, Represents the 3D point cloud data In the subspace The layer's package layout factor; Represents the 3D point cloud data In the subspace The number of pin marker points of the layer, Represents the 3D point cloud data The number of pin marking points of the layer; Represents the 3D point cloud data In the subspace The distribution label sequence corresponding to the layer elements and The Euclidean distance of elements in space; and Respectively represent the first In the subspace The distribution label sequence corresponding to the layer elements and The distribution of the elements is marked with angles, and cos represents the cosine function; Represents the 3D point cloud data In the subspace The number of elements in the distribution tag sequence corresponding to the layer; The method for constructing the difference matrix includes: The similarity between the distribution vectors corresponding to the chip package layout samples is used as the clustering distance, and a clustering algorithm is used to obtain the clustering results of all chip package layout samples; For each cluster, a sequence of chip package layout samples in the cluster in the order of corresponding production time is used as a chip package layout sequence; the difference of packaging time corresponding to adjacent elements in the chip package layout sequence of the cluster is calculated, and the mean of all the differences in the chip package layout sequence is used as the packaging time difference factor of the cluster; According to the differences in the encapsulation time difference factor, the number of elements and the distribution vector between the clusters, the difference factor between the clusters is obtained; The difference factors between each cluster and all other clusters are composed into a sequence as the row vectors of the difference matrix.

2. The miniaturized packaging method for a chip according to claim 1, characterized in that: The material of the circuit layer is copper, and the surface of the pad exposed to the dielectric layer is subjected to chemical nickel-palladium-gold plating, chemical nickel-gold plating, electrolytic nickel-gold plating and electrolytic nickel-silver plating.

3. The miniaturized packaging method for a chip according to claim 1, characterized in that: The substrate is one of a plastic package substrate, a lead frame, a ceramic substrate and a microcrystalline glass substrate.

4. The miniaturized packaging method for a chip according to claim 1, characterized in that: The method of fixing to the substrate is one of wire bonding and flip chip bonding.

5. The miniaturized packaging method for a chip according to claim 1, characterized in that: The assembling method is one of bonding and welding processes.

6. The miniaturized packaging method for a chip according to claim 1, characterized in that: The method for determining the distribution coefficient of each subspace is: For each subspace, the positive direction of the X-axis in the spatial rectangular coordinate system is taken as the starting direction, and the pin marking points of each layer in the subspace are arranged in the order of rotating counterclockwise from the starting direction to form a distribution marking sequence of each layer; The angle between the line connecting the pin marking point corresponding to each element in the distribution marking sequence and the center point of the plane point cloud data and the starting direction is used as the distribution marking angle of each element, and the distribution marking angles of all elements in the distribution marking sequence constitute a distribution marking angle sequence; According to the difference between the distribution marking angle sequences of two adjacent layers in each subspace, the distribution coefficient of each subspace is obtained by combining the spatial distance of adjacent elements in the distribution marking sequences of each layer and the difference in the distribution marking angles of adjacent elements.

7. The miniaturized packaging method for a chip according to claim 1, characterized in that: The calculation expression of the difference factor is: ; In the formula, Indicates Clusters and The difference factor between clusters; and Respectively represent Clusters and The package delay factor of the clusters; Indicates The first distribution vector, Indicates The first distribution vector, express and The cosine similarity between ; and Respectively represent Clusters and The number of distribution vectors in a cluster.