PCB Verification

In most mass-production manufacturing facilities an attempt is made to achieve 100% quality assurance of the product.

Abstract
In most mass-production manufacturing facilities an attempt is made to achieve 100% quality assurance of the product. One of the most difficult tasks in this process is an inspection that seeks to identify both functional and cosmetic defects. PCB’s are inspected extensively before the insertion of components and the soldering process to isolate defects. Studies shows that open / partial open, short, pinhole, breakout, overetch, underetch are the most frequent defects that occur.

In this project, we focused on six kinds of defects that can be classified into two general groups:

  • Lack of conductor – open, pinhole, mousebite
  • Addition of conductor – short, spur, dirt

In general, PCB inspection falls into one of three categories :

  • Reference comparison
  • Non-referential approaches
  • Hybrid approaches (combination of these methods)

In this project, we chose the reference comparison category, using the `Phase Only Transform` method.

The Problem
Finding an algorithm for checking PCBs. The algorithm compares a PCB to its mask in order to find defects and then classify them.

The Algorithm
1

1. Phase Only Transform :

  •  Divide PCB & MASK into blocks (e.g. 256×256) enables to find small defects reduces computational time
  • Phase Only Transform to each [PCB,MASK] block
  • Output : Errors Image

2. Binarization :

  • Image processing and analysis requires binary image
  • Binarization to errors image using empirical global threshold
  • Noise removal using minimum area threshold
  • Output : Binary Errors Image

3. Convex Hull Computation :

  • In order to improve defects display on PCB we use the image featuring : ConvexImage
  • It creates a binary convex image per “error object” according to the size of the object
  • Output : PCB with marked defects

4. Defects Classification :

  • Defects are divided to two general groups : addition of conductor & lack of conductor, by comparing the mean of each “error object” to a threshold
  • In a specific group, the classification is based on morphological methods and image featuring

The process
2

Results
3    4
PCB                                                                             MASK
5    6
PCB with classified defects

Conclusions
1. Phase Only method turned to be the best way for PCB inspection, regarding :

  • Computation time
  • Identification of defects

2. In comparison to other methods tried before, based on :

  • Morphological operations
  • Image featuring

3. The algorithm identifies quite well an addition/lack of conductor. There is some difficulty to determine whether an addition of conductor is spur or dirt because of shadows around the object
4. The algorithm performance depends on two parameters definitions:

  • Binarization global threshold
  • Size of blocks

Acknowledgment
We are grateful to our project supervisor Alex and Michael Bronstein throughout this work.We would also like to thank Johanan Erez and the laboratory stuff for his help and guidance.
We are also grateful to the Ollendorff Minerva Center Fund for supporting this project.