QFD-ID: QFD-Forum Ausgabe 23

The manufacture of modern, multilayer printed circuit boards is a highly complicated process. The overall process involves a sequence of individual, very complex sub-processes, some of which are carried out many times. The result is that the manufacture of a multilayer printed circuit board can consist of a sequence of 40, 60 or even more individual processes. This involves a wide range of different processes and processes of widely differing complexity. For example, mechanical tasks (drilling, milling, scoring, brushing, pressing), chemical/electrochemical steps (cleaning, galvanising, coating, etching etc.), photo-technology (board patterns, solder resist masks, etc.), printing, optics, etc.. Modern printed circuit boards are highly sensitive and have precision structures; track widths of 100 m and smaller are standard today; 4, 8, 10 right up to 14 and sometimes even more layers are pressed onto each other in order to manufacture a printed circuit board that provides a high packing density. The position of the individual layers with respect to each other is very important here as metallic feed-throughs have to be manufactured between the layers. Quite often position tolerances of less than 100 m have to be maintained.

I think it is patently clear that at the end of the whole process chain there can only be a satisfactory number of good printed circuit boards if each individual process can be run with an extremely high yield. For example, error propagation shows that in order to obtain a final yield of 90% with a sequence of 50 individual processes, on average each process must be operated with a yield of about 99.8%. This corresponds fairly closely to a 3 sigma process.

This requirement, which is often taken as standard today, can only be achieved by having an excellent knowledge and control of each individual process. The magnitude of the challenge becomes apparent when one realises that we are concerned here not just with processes which are influenced by a few parameters (up to 10) but also some processes with 100 and more influencing factors.

In order to demonstrate this complexity, a QFD-approach is used to describe and classify the process parameters. Here, a team works together, comprising the following participants: process owner and specialists in the process to be described, process owner and specialists in the subsequent process (internal client), quality experts, and possibly also specialist personnel of the manufacturer, developers, etc. of the process.

For this, the central matrix of a HoQ (House of Quality) is used in the following way:

The team defines the essential output (process results) of the relevant process (that is why the internal client is also very important). This output, which consists of the individual requirements (the output parameters), is put into the rows of the matrix and can, if necessary and desired, also be weighted (e.g. pair-wise comparison). All the parameters which control or influence the process, namely the process parameters, are then put into the columns of the matrix. When doing this, it is beneficial to work through the process from the start to the end, from module to module.... Attention to detail is an advantage. Indeed it is recommended to pay very close attention to completeness and to include all possible parameters whose importance will be determined later.

As in a classical QFD, the relationships between the individual process parameters and the output parameters are defined following standard conventions: the relevant process parameter has either a large (9), medium (3), small (1) or no (0) effect on the relevant output parameter. Here, the standard QFD symbols are used. After making this link between the output parameters and the process parameters, the sequence of importance of the process parameters is calculated, once again following standard QFD conventions. What one finally obtains is "an order", a sequence of the process parameters according to their importance, i.e. their effect on the process and hence on the output parameters.

This represents a key step in understanding a process: the definition of the process results (output parameters), their rank and the process parameters which control the process, and affect the results - ordered according to their importance for the process.

In the following two steps the matrix is extended to the right and downwards. The reference value and tolerances of each output parameter, and how they are measured, are recorded to the right, in the rows of the output parameters.

The matrix is extended downwards with the reference values and the upper and lower limits of the process parameters and how they are measured, controlled or monitored.

One now has a matrix which provides a virtually complete representation and description of the process. From this, key properties of the process can be gleaned, including whether and how all output parameters and process parameters are defined, quantitatively recorded and measurable or whether there are still gaps, even perhaps for the key parameters.

This matrix cannot only be used to identify the key process parameters but can also be used to promote thought about how they can be measured and controlled. It can also be used for training purposes, to draw up maintenance plans, etc..

In addition, a variable comparison can be carried out, at least for processes having process parameters which are relatively east to vary. To do this, one runs the 5 - 20 main process parameters at level 1, which roughly corresponds to the standard process, and afterwards at level 2, at which poorer process results are expected. If one has recorded the key process parameters in their sequence when drawing up the matrix, the process results should significantly change when this relatively simple experiment is undertaken. If this is not the case, the most important process parameter(s) has/have still not been recognised.

Author

Dr. Klaus Bischoff
Studied Physics and Mathematics at the University of Karlsruhe Worked 5 years as Vendor Engineer for Digital Equipment, at that time (1985 - 1990) one of the biggest computer manufacturer of the world. From 1990 to 2000 quality manager at Endress + Hauser Maulburg: Endress + Hauser is a manifacturer of equipment for measuring and automation with about 5.500 employees world wide. Since 2000 quality manager of the photo print electronics a company of the Endress + Hauser group that produces printed circiut boards.

Process QFD, a method for process control

Abstract

Author