Solder joint inspection using laser doppler vibrometry - includes related article on Hewlett-Packard's laser Doppler vibrometer system design

Hewlett-Packard Journal, Oct, 1989 by Catherine A. Keely

Solder Joint Inspection Using Laser Doppler Vibrometry

LIKE MOST ELECTRONICS COMPANIES, Hewlett-Packard has been striving to fit more functionality into smaller packages. To achieve this, many of HP's products now use surface mount technology (SMT) and other technologies in their printed circuit assembly production. SMT involves the use of packages significantly smaller than standard through-hole packages. The package leads are soldered directly to pads on the surface of the board instead of passing through the board. SMT components have lead widths of 0.025 inch (0.625 mm) or less, with lead pitch double the lead width. As one can imagine, when a printed circuit board is packed with SMT components, inspection of the solder joints by visual means becomes a difficult task. This is especially true with certain lead shapes, such as J leads, where most of the solder joint is under the component. Sophisticted electrical testers can catch many solder joint errors, such as shorts or solder bridges. They can sometimes find unsoldered lead problems, but frequently these faults go unnoticed because the tester forces mechanical contact between the lead and the pad, making the missing joint undetectable.

To address this problem, the computer vision project of HP Laboratories has investigated alternative methods of inspecting SMT solder joints. The subject of this paper is one method that has proven successful at detecting one major joint error type: the unsoldered joint. The method, which does not actually involve computer vision although it is an optical method, is based on the vibration characteristics of an unsoldered lead. The idea is that an unsoldered lead, when stimulated, will vibrate at its resonant frequencies, which depend on its material and geometry. A soldered lead, under the same stimulation, will have different vibrational characteristics because it has different geometrical constraints. A laser Doppler vibrometer or velocimeter (LDV, see box, page 82) is used to measure the velocity of a vibrating lead, and as this paper describes, the peaks in the frequency spectrum of the vibration indicate whether the joint is soldered or not.

The SMT Application

The most common surface mount components fall into three general categories (see Fig. 1): J leads, which are found on PLCCs (plastic leaded chip carriers), gull wings, found on SOICs (small-outline integrated circuits) and PQFPs (plastic quad flat packs), and brick-shaped passive components. The gull wings can be further broken down into different shape standards: narrow SOICs, wide SOICs, and fine-pitch PQFPs (0.025 in./0.625 mm pitch or less). The passives can alse be broken down into the many available shapes and sizes, since the geometry determines the natural frequency, but this paper does not distinguish between the different passives.

There are many common joint error types, including shorts or bridges, solder balls or splashes, cold joints, insufficient solder, excess solder, dewetting or wicking, nonwetting, and lifted leads. In the first five error types, some sort of joint is formed, so examining the vibration should only indicate that there is a joint. For the last three, which form a significant portion of the errors, there is no mechanical joint, and thus they are considered open or unsoldered joints for the purposes of vibration analysis. Cracked joints fall somewhere in between, depending on how freely the lead can vibrate.

Before building the LDV system to measure the vibration, finite element models of the various leads were made and analyzed for their vibration characteristics. This was done to determine analytically the expected resonant frequencies of soldered and unsoldered leads, to verify that the expected resonant frequencies for good and bad joints are distinct, and to make sure that the frequencies are in the range detectable by the proposed LDV setup, that is, below 200 kHz. The results of the finite element analysis gave the following resonances under 100 kHz for unmounted leads: about 29 kHz for wide SOICs, 51 kHz for narrow SOICs, 14 kHz (first mode) and 40 kHz (second mode) for J leads (PLCCs), and 20 kHz for fine-pitch PQFPs. (1) No modeling was done for the leadless passive components. First-mode resonant frequencies of properly soldered leads are at least five times those of unsoldered leads, and are always above 85 kHz, with the specific frequencies depending on the lead type and thickness. The frequencies increase slowly as the lead thicknesses increase, with slopes of 50 to 400 Hz/[micrometer]. These results indicate that in theory the vibration characteristics should distinguish soldered leads from unsoldered leads.

To measure the vibration, the leads first have to be stimulated to vibrate. This can be done in several ways, such as using an impulse force, shaking the board, or sweeping a stimulus frequency, but the method we implemented is an air jet (at a pressure of about 35 kPa or 5 psgi) aimed near the joint, as shown in Fig. 2. (2) The air jet is a source of acoustic white noise, which sets up resonance vibrations in the leads. It is simple and flexible, and it works.