In an ideal world the prudent process engineer would carry out inspection or test after each step in the assembly process ie after solder paste printing, component placement and reflow soldering. In practice, however, one is forced to choose between two strategies for automatic optical inspection(AOI). AOI is either carried out directly after screen printing and component placement or only reflow soldering. We examine the criteria for the decision on which strategy is adopted.
Most defects that occur in the manufacture of SMT modules or assembled printed circuit boards (PCBs) have their causes in one of the three critical process steps - screen printing of solder paste, component placement and reflow soldering. It would be preferable to carry out some kind of inspection or test after each of these steps, in order to identify and eliminate the root cause of any defects. However, in practice this strategy is not possible as the available resources, whether financial, human or spatial, are always limited. The responsible process engineer is forced to use the resources that are actually available to achieve the highest return.
Automatic optical inspection (AOI) offers two different strategies: Strategy 1 - Detection of defects and Strategy 2 - prevention of defects. The decision between these two possibilities depends on the quality philosophy of the company, the type of failures that are encountered and the stability of the assembly process. In the sections below we examine the criteria for this decision.
Strategy 1 - Detection of defects
The most common application of AOI in the SMT assembly industry is post-reflow soldering and is, in fact, the only type of inspection that is offered by most suppliers of AOI equipment. As a result many users only ever consider inspecting their boards after soldering.
The driving force behind inspection of soldered assemblies is clear. It is intuitively obvious that inspection after the last step of a production process should ensure that no bad product is shipped from the factory. In the end-customer's drive to zero PPM failure rate many companies supplying for example, the automotive industry have felt themselves forced to perform 100% inspection at the end of the SMT line. With ever decreasing component size and increasing population densities only an automatic system can be considered.
In the case of SMT assembly AOI is capable of detecting a broad spectrum of defect types regardless of their origin - whether due to the process or the material (ie the components themselves). Defects that are caused at any stage of the process including the oven are detected. Furthermore, defects whose root cause is due to a combination of factors at different stages of the assembly process should also not pass the inspection. If the false accept rate (real defects that are not detected) is kept low, one can be confident of achieving a high first-pass yield at front-end testing.
But therein lies a problem. In practice optical or X-ray solder joint inspection (SJI) is a very difficult task. The demands (eg detection of excess or insufficient solder, solder balls, bridging and bent pins in both the horizontal and vertical directions) mean that the technology is extremely complicated with multiple cameras and lighting systems (often combined with mirrors) being the norm.
Given the huge number of variables that the inspection software has to deal with, it is not surprising that achieving a low false accept rate has in general only been possible by allowing high false fail rate (components or joints that are detected by the system as being defective when, in fact, they are good). One is faced with the Hobson's choice of either a high false accept or false fail rate. Most manufacturers choose the latter in order to prevent bad product being shipped. In this case a visual re-inspection of the automatically detected 'defects' by a trained operator is usually carried out. For some producers it is the norm that the majority of detected defects are, in fact, false fails. This has the consequence that the operator, who is expecting repeated false fails, can, through boredom, miss the real defects!
The second major disadvantage of this strategy is that it only facilitates the detection of defects but does not identify the root cause.
A given defect might have its cause in any of the three processes that preceded the inspection. There is no direct information, for example, if a tombstone was caused by incorrect solder paste printing, placement of the component offset relative to the paste/pad or nonsimultaneous melting of the paste during soldering. Only with a lot of experience can the process engineer say, in a few cases, which of the processes was actually at fault.
The inspection is slow to react to problems early in the manufacturing process. Trends or drifts in a process are not detectable until defects are actually being produced and when these are detected the line may be filled with partly manufactured product - all with the same defect!
Strategy 2 - Prevention of defects
There is another strategy, which is being followed by an ever-increasing number of SMT assemblers. They believe that the key to reducing the number of failures produced and in improving the process itself, is in checking the product as early as possible after each of the critical process steps of screen printing and placement of the surface mount devices (SMDs). AOI is essentially the only tool available, as no form of testing can be carried out on presoldered PCBs once paste has been applied.
This strategy calls for the rigorous implementation of the principles of statistical process control (SPC) in which critical parameters are measured and monitored over time, in order to detect trends that would lead to out-of-control conditions and ultimately to defective product being manufactured.
This strategy puts some demands on the AOI system which might explain why some system suppliers have shied away from it. The lifeblood of SPC is measurement data and so the AOI system must not only be able to locate the object (either solder paste deposit or SMD). It must also be capable (as defined by SPC) of measuring accurately and reliably the position and skew (rotational angle) of the object as well as deposit height and volume in the case of solder paste inspection. These are the data, which analysed according to SPC, will produce the information that will allow the process engineer to predict the onset of defects and take preventative action - before the defects actually occur.
In component placement inspection (CPI) for example, a trend analysis of the mean placement position by one of the placement machines in a production line might show a pronounced trend. This would indicate a shift in one direction of the position of all SMDs placed by this machine. If not counteracted this shift or drift would eventually lead to solder problems eg tombstones of chip components aligned in the direction of the shift. An upward trend in the standard deviation graph would indicate increasing scatter of the placed SMDs. Again this can be corrected by preventative maintenance before the onset of defects.
In the case of solder paste inspection (SPI), a reduction in the CpK value for the measured paste volume would arise given, for example, either of two causes. A change in the paste rheology can lead to a change in the amount of paste deposited over the whole board (shift in mean volume) or a variation in the squeegee force along its length can result in differing amounts of paste being deposited across the PCB (increase in standard deviation of volume of the deposits). In either case, the deterioration of the CpK value alerts the process engineer, who can then analyse the mean and standard deviation data further to identify the cause of the problem.
As well as providing absolute measurement data at each stage in the assembly process, Strategy 2 speeds up the reaction to encountered defects or process trends. Since the inspection is carried out at the earliest possible opportunity there is virtually no time lag between process step and AOI. In critical situations production can even be stopped until the cause of a defect is eliminated.
Compare this to the post-reflow inspection situation where, for example, valuable components may be placed on many PCBs, before an end-of-line inspection detects an expensive IC that has been placed in the wrong orientation! In this example the manufacturer would save a huge amount of rework or scrap, if pre-reflow AOI were available.
There are, of course, disadvantages with this approach. Problems in the oven may cause tombstoning although the paste was correctly printed and the component correctly placed. However, as oven technology has matured, not many assemblers now encounter serious problems during the actual soldering process.
When applied in high or medium volume lines, SPC allows the engineer to statistically predict the onset of and reduce the level of defects in a particular product whereas in a low volume line with frequent changeovers only statistics for the line itself are produced. It is imperative, in either case, that due to the large amount of data generated, that the analysis is carried out using an automatic SPC software package.
Conclusions
The two AOI strategies available to the process engineer reflect the different philosophies of quality control or assurance and quality management. In the first strategy the emphasis is on finding all defects at almost any cost. The task of finding and eliminating the root causes is secondary because of the pressure to produce 100% good product.
The second strategy demands that the whole process is brought and kept under control and relies on similar strategies in other parts of the business eg from the component suppliers. It requires a higher level of process engineering but delivers the benefits of reduced failures, reduced scrap and rework and higher first pass yield. And like all good investments it delivers higher profit margins at the end of the day.
| Tel: | +27 11 609 1244 |
| Email: | [email protected] |
| www: | www.zetech.co.za |
| Articles: | More information and articles about ZETECH ONE |
© Technews Publishing (Pty) Ltd | All Rights Reserved
printer friendly version