README.md

metadata

license: mit
language:
  - en
pretty_name: iDATA-Capcitance Dataset
size_categories:
  - 100M<n<1B

3D Capacitance Dataset

This project serves as the submodule of PC-Cap.

The dataset generation is briefly shown in Fig. 1.

Dataset Description

Directory Descriptions

The complete directory structure is as follows:

dataset
├── pc
├── pc_annotation_coupling_test_clean.csv
├── pc_annotation_coupling_test.csv
├── pc_annotation_coupling_train_clean.csv
├── pc_annotation_coupling_train.csv
├── pc_annotation.csv
├── pc_annotation_total_test_clean.csv
├── pc_annotation_total_test.csv
├── pc_annotation_total_train_clean.csv
└── pc_annotation_total_train.csv

The directory dataset/pc contains all information of the point cloud.

File descriptions

• pc_annotation.csv - records some information from the original 3D Capacitance Sample of the point cloud

  column name	description
  point_cloud_path	point cloud file path
  conductor_num	the number of conductors that the original sample contains
  point_num	the size of point cloud
  electrode1_point_num	points generated based on the master conductor
  electrode2_point_num	points generated based on the target environment conductor
  cap_type	total or coupling capacitance
  cap	reference capacitance value, the label
  v1_x	the absolute coordinate-low-x of the original sample in the layout
  v1_y	the absolute coordinate-low-y of the original sample in the layout
  v1_z	the absolute coordinate-low-z of the original sample in the layout
  dx	the size of the original sample along direction-x
  dy	the size of the original sample along direction-y
  dz	the size of the original sample along direction-z
  timestamp	timestamp of the point cloud file generation
  solver_time	the calculation time that the solver takes to get the capacitance result

• pc_annotation_xxx.csv - part of records from pc_annotation.csv
  The details of dataset partition and data cleaning are available at PC-Cap repository

  sub-dataset used in our research	descriptions
  pc_annotation_coupling_train_clean.csv	the training set for coupling capacitance task
  pc_annotation_coupling_test_clean.csv	the testing set for coupling capacitance task
  pc_annotation_total_train_clean.csv	the training set for total capacitance task
  pc_annotation_total_test_clean.csv	the testing set for total capacitance task

• w500l2_1_C_0_0.txt - example of a point cloud file

  • file name format

    • w500l2:
      The 3D sampling window is height-adaptive and can accommodate 2 layers above and 2 layers below.
      5 metal layers in maximum: Routing-Via-Routing-Via-Routing or Via-Routing-Via-Routing-Via.
      The 3D sampling window project to XOY-plane with an area sized $500 nm \times 500nm$.
    • 1:
      the sample id when generation
    • C_0_0:
      the position in the sample capacitance matrix

  • content format: point cloud file format that CloudCompare can read directly.

    symbol	descriptions
    $x$	coordinate-x
    $y$	coordinate-y
    $z$	coordinate-z
    $n_x$	component-x of normal vector
    $n_y$	component-y of normal vector
    $n_z$	component-z of normal vector
    $\varepsilon_r$	the dielectric channel, relative permittivity
    class	a point belongs to a class that relates to its generation
    $\Phi$	the selecting channel

  • line 1 : 432.715, 171.114, 0.5879, 1, 0, 0, 4.06, 1, 0
    A point with coordinates (432.715, 171.114, 0.5879) in layout.
    This point is generated on the surface of a region whose local normal vector is (1, 0, 0).
    This local region is full of dielectric with a relative permittivity of 4.06, except the region inner the conductor.
    The id of this region equals 1.
    The selecting channel value of this point is 0, such as non-target environment conductors. The details of the setting are related to the Gauss Law-based Point Cloud feature.