Optical Inspection Module for the Printed Circuit Board Production

A prototype device for the optical inspection of printed circuit boards has been built up in Optical Measurement Laboratory of Kemi-Tornio Polytechnic. The device is intended for the inspection of the presence and pose of the surface mounted devices, but it can be altered for other PCB inspection tasks as well. The inspection device constructed can be integrated into a production line as an autonomous module. Module comprises a conveyor, a machine vision system, and a storage buffer for faulty boards.

Introduction

Production of the printed circuit boards (PCB) is nowadays highly and quite successfully automated. E.g. common day assembly machines of the surface mounted devices (SMD) are both fast, reliable, and accurate. However, some errors occur time to time and thus the inspection of the assembly quality is needed. The more the packing density and the production speed of the PCBs increase the more demanding, time-consuming and laborious inspection tasks are for a human operator. So, there is growing need for an automatic and cost-effective on-line inspection mod-ule.
A prototype inspection module for the automatic inspection of the presence and the pose (i.e. position and orientation) of the SMDs has been constructed in Optical Measurement Laboratory of Kemi-Tornio Polytechnic. The inspection module is realized for the inspection of 250 x 151 mm -sized PCB panels comprising 16 similar PCBs (size of a PCB 68 x 28 mm). In every PCB there is 35 various SMDs, whose presence and pose relative to two fiducial marks of the PCB are to be inspected. So, in one PCB panel there is totally 592 features to be detected. In current product the dimen-sions of the SMDs vary from 2.0 x 1.25 x 0.5 mm to 6.5 x 4.6 x 3.6 mm.

Structure and operation of the module

Main components of the inspection module are a PC controlled machine vision system, a motion mechanism, a storage buffer for the erroneous PCB panels, and equipment for the data transfer between different components of the module (Fig. 1). The motion mechanism comprises a step-motor based conveyor and an air operated translation mechanism for the machine vision system.

The operation of the inspection is essentially the following: PCB panel is fed into the inspection module from the previous section of the production line (e.g. from the SMD assembly machine). Machine vision system captures images needed for the inspection and inspects panel for defects. If any defect is detected, the panel is stored into the storage buffer, operator is informed of the defect (defect type and location are presented in a monitor), and the results of the inspection are saved e.g. for the statistical process control (SPC) -analysis. Approved PCB panels are fed forward to the next section of the production line.

Operation of the machine vision system

The machine vision system includes a led-based lighting system, a digital B/W CCD-camera (resolution 1300 x 1030 pix), and adequate optics, frame grabber and data cables. The field-of-view of the system is optimized for imaging one PCB at a time. In order to obtain images of every PCB, the PCB panel is either translated with the conveyor or the machine vision system is moved with the air-operated motion mechanism. During the image capturing both the PCB panel and the machine vision system motion are at rest.
Three images are captured from every PCB for the inspection, one by using red, one by using green, and one by using blue light. These three different grayscale images are needed in order to automatically distinguish every SMD from the image from its background, which could be the surface of the PCB, a glue spot, or a pad (Fig. 2).

Image processing (i.e. feature detection) is based on two types of a priori knowledge. Design information of the PCBs is used in order to restrict the area of the feature detection in the image. Informa-tion of the component reflectance properties (color tone) is used in the actual feature detection. This information is got from a semi-automatic teaching procedure performed before the actual inspec-tion.
In order to determine the actual poses of the SMDs the detected image coordinates of the SMDs are transformed from the image coordinate frame to the PCB coordinate frame. The lens distortions of the detected image coordinates are corrected and also the relief displacements are compensated in this transformation.

Present results 

The inspection module has been constructed and its functionality has been confirmed. The operation of the motion mechanisms and image capturing are synchronized. Image processing algorithms were preliminary tested by inspecting the poses of the SMDs from one PCB panel 10 times. The repeatability of the pose determina-tion was 0.013 mm (standard deviation). These preliminary tests indicate that the inspection including image formation, image processing, SMD pose determination, and defect detection can be done in 30 s.

Prospects

The inspection module introduced in this paper will be tested comprehensively first in the laboratory and after that on an actual PCB production line. Image processing algorithms will be devel-oped so that the same module will be suited, besides the inspection of the SMDs, also for the quality inspection of the soldering process. 

Contact information

Teuvo Heimonen, Heikki Juvalainen, Jani Sipola, Samuli Vähä
Kemi-Tornio Polytechnic
The Unit of Technical Education
Kiveliönkatu 36
Fin-94600 KEMI
Finland
e-mail: teuvo.heimonen@tokem.fi
tel. +358 40 5868108, fax. +358 16 258800
Internet: http://www.tokem.fi/teku/omlab/