|
|
|
|
T&M: Zero-defect IC inspection strategy with AOI
|
|
| by Peter Krippner, Viscom AG |
|
| 1 March 2006 |
|
| Many types of IC defects can be detected using either orthogonal or inclined cameras, both of which have their advantages and disadvantages. Other influencing factors such as camera resolution, AOI software and transport concepts are also discussed. |
|
|
During the automatic placement and soldering of electronic assemblies in the manufacture of printed-circuit boards, production errors cannot be completely avoided. In order to detect these errors and assure quality, automatic optical inspection systems (AOI) are used. These systems are often stationed after the soldering oven for post reflow inspection purposes, since the position at this gate results in the greatest error coverage.
One of the housing types with a broad spectrum of defects is that of ICs (QFP, TSOP, PLCC, SOIC, BGA, QFN, etc.), with a typical pitch of up to 0.5 mm, and sometimes even less. These very defects are often either very difficult or even impossible to detect in the subsequent electrical tests and function check.
Even when the inspection is carried out using an AOI system, IC solder joint defects and IC assembly defects make considerable demands on the performance of the inspection equipment and have a decisive influence on the depth of the AOI inspection. It is therefore necessary to differentiate between AOI systems with purely orthogonal camera views, and those which are also equipped with cameras which present inclined views.

Types of IC defects The types of defects typical of ICs may be classified according to the manufacturing phases involved, namely solder paste print, components assembly and soldering. And it is particularly during soldering that the component has an additional influence over its own (soldering) characteristics. The following defects can be assigned to the solder paste print or solder paste: • deficient tin or solder joints that are too thin • solder bridges • solder droplets The following defects can be assigned to component placement: • missing components • “stray” IC components • displaced components • twisted components • reverse-polarity components • wrong components • bent IC pins

The following defects can be assigned to soldering or the components themselves: • “lifted lead”, i.e. IC pin raised or bent upwards • coplanarity of an IC pin, e.g. due to lack of wet ability • unmelted solder paste
Depending on the type of IC housing, certain kinds of defect occur more or less frequently. Lifted leads and pin coplanarity occur more frequently with fine-pitch QFPs than with SOICs.

External factors influencing AOI inspection The objective of AOI inspection is to cover the types of defects described without exception, so that all defects which occur during production are actually detected. At the same, the number of pseudo-defects (false alarms, bogus defects) must be reduced to a minimum and the system throughput rate raised to its maximum.
When inspections of solder joints and components are being performed by an AOI system, certain external factors, i.e. factors not connected with the AOI, have to be taken into consideration, which can make achieving the objectives mentioned easier or less easy.
One of the most important influencing factors is that of pad design. A uniform pad design for various assemblies and types of housing has a positive effect over a period of time, since this also results in the appearance of the good solder joints being more uniform. The more uniform the appearance of the good solder joints is, the more certain it becomes that solder defects will be detected, and the lower the pseudo-defect rate will get. Pads which are too small, and which result in the solder meniscus not being visible, have a negative influence on the inspection process.
Another factor of major importance is that of component quality as far as good wetability of the connections and good dimensional stability is concerned. Good wetability results in a better, more constant impression of the solder joint, better dimensional stability, eg. of the pin length in QFPs on the other hand, can make programming easier.
Other influencing factors include the colour of the printed-circuit board and that of the solder mask lacquer, as well as the amount of printed-circuit board sag. Since it is never possible to influence all of the peripheral conditions favourably, an AOI system must be so configured (sensors, software, system set-up), as to be able to compensate for any negative influencing factors.

One of the most crucial characteristics when differentiating between the AOI systems currently on the market is that of the camera view. There are two distinct types, the orthogonal camera view (plan view) and the inclined camera view (slanted view).
Most types of defect can, in principal, be detected using the orthogonal camera view. The critical types of defect for the orthogonal camera view are coplanarity and lifted leads. These can only be found in favourable circumstances using the plan view. In such cases, advantage is often taken of the so-called capillary effect, by virtue of which a good solder joint attracts the solder tin to the IC pin. In this favourable pad design, there is no deposit of tin on the projecting edge of the pad, so that under suitable lighting conditions, the projecting edge of the pad appears light. If the pin has not been wetted (coplanarity), or if the pin has even been bent upwards (lifted lead), the solder tin is distributed evenly on the pad, which results in a clearly different impression (Figure 4). This even works in the case of a PLCC, where the solder joint itself is located underneath the component and is not even visible when viewed from above. As long as the pad projects sufficiently far beyond the contour of the component, the capillary effect can be utilized.
The uses of the approach mentioned above do, however, reach their limits when it comes to fine-pitch ICs without an appropriately adapted pads design, or in the case of micro-bridges. The difference between good and bad cases is hardly distinguishable, as shown in figures 5A and 5B.


Zero defect inspection is no longer possible. The maximum defect detection rate relative to the number actual defects is between 50% and 75%, and the pseudo-error rate is also dissatisfactory. Considerably better results are achieved with an inclined camera view, in which – similar to the method used in manual optical inspection – the printed-circuit board is viewed at an angle of 45° degrees to the horizontal. The more acute the angle of vision, the essinformation can be obtained using the inclined view. Figures 6A and 6B depict the same solder joints as in figures 5A and 5B using an inclined camera view.


The considerably improved optical differentiation between good and bad cases is obvious. One of the axioms of image processing is, the better one is able to differentiate between good and bad cases, the better error detection and pseudo error rate become. As in the plan view, switchable lighting permitting different programmable angles of illumination is the optimum. Fixed, non-switchable lighting systems don’t allow the flexibility required to obtain the best possible contrast for every situation.
As shown in figures 5/6, there are solder bridges situated on the inside between the pins which are in the shadow in the orthogonal view, which makes them virtually invisible.
Recognizability is considerably better in the inclined camera view.
The use of inclined cameras requires know-how in the fields of calibration and gray scale compensation. If the system is only equipped with an orthogonal camera, the AOI system itself does not necessarily have to have the capability of continually monitoring the calibration, as the camera is able to self-calibrate itself within certain limits. In the case of a multiple camera system, however, the software must enable calibration of the cameras relative to one another and ideally also monitor this beyond the limits of basic calibration. If the software does not permit simple, exact and rapid calibration, the AOI supplier will have problems implementing the integration of the inclined view.
A check on the results is always possible by carrying out a so-called machine capability investigation (MCI). During this procedure, suitable calibration patterns with, for example 50 runs, are carried out and calculated in accordance with prescribed set values and measuring tolerances (cm/cmk values), in which a Gaussian distribution of the measured values is assumed. Whereas cmk values of >1 have been accepted up to the present (3 sigma quality), the trend nowadays is towards cmk > 1.67 or even >2 (5 sigma or 6 sigma quality).
It is clear that in the light of these requirements, stable sensory systems are at an advantage. Moveable parts inside the sensor modules tend to degrade precision and a high degree of reproducibility.
A further skill which has to be mastered in the use of inclined cameras is that of compensating for printed-circuit board sag. This effect causes the solder joint being inspected to "slide" downwads in the field of view (refer to figure 7).
One possibility for compensating is to scan the sag at a sufficient number of points in order to calculate a model of the sag. This method, however, does necessitate addition inspection time. An integrated approach is more advantageous, in which a larger field of inspection is isolated by seeking significant reference points vertically using software. If appropriate software is available, this method provides reliable results without requiring additional time.
A further important criterion in differentiating between AOI systems is that of the camera pixel resolution. With fine-pitch ICs, picture elements of 0.5 mm are normal, as in many designs, the pad and the gaps between the pads each measure 250μm. If it is necessary to cover the width of the frontal meniscus with at least 15 pixels, the standard resolution of the inclined cameras ought to be well under 20μm per pixel in order to obtain images with optimum information content.
A very good camera resolution does of course mean that the system performance required be commensurately higher, so that the AOI does not become a bottleneck on the production line. This applies both to the software (automatic, optimized position generation, use of TSP (Travelling-Sales-Man-Problem) approaches), as well as the camera technology (use of megapixel cameras with a high frame rate) and the positioning technology (high-speed linear drives).
The biggest challenges which have to meet in this connection are to be found in the field of software. Whereas the orthogonal camera view is relatively easy to master, the inclined camera view makes considerably different and higher mathematical demands on the software development, due of the additional need to differentiate between the angles in the field of view (at least in the four cardinal directions).


Summary and perspective In the zero-defect IC inspection strategy, image capture, camera technology and AOI software are the deciding criterion of an inspection system. And it is the image capture which plays a major role, since only that which”is seen“is available for subsequent evaluation.
The orthogonal camera view permits an adequate inspection depth for ICs when the pad design is favourable, although residual escape cannot be completely eliminated, particularly in the case of coplanar IC pins. If the pad design is unfavourable, defect escape can increase to between 25 and 50%, i.e. defect detection of the critical types of defects is only 50 – 75%. This occurs often in practice because of the small pad sizes.
Inclined image acquisition, on the other hand, makes optimum results achievable – even in the case of critical defects. A high-resolution camera with about 15μm per pixel is well suited to this task and, in combination with AOI software which clearly contrasts defect characteristics, provides a solid basis for zero-defect IC inspection.
The future use of inclined camera views will continue to be an advantage. New components, such as QFNs, which have space-saving connections both at the side as well as beneath the component, cannot be adequately inspected orthogonally. Therefore, the use of inclined image acquisition is expected to increase.
|
|
|
|
|