Nozzle (showerhead) design for a dotted epoxy pattern for a 8mm by 11mm die
using 0.9mm diameter pipes (needles); system's partial output example
Epoxy voids (photograph by Phoenix X-Ray)
A friend working with the semiconductor industry (Esec) visited me in December 1991 and told me about a problem they had: To bond dies (chips) to substrates they inject epoxy on the substrate through a nozzle (showerhead) with parallel, small diameter round pipes (needles), to generate a certain pattern of identical epoxy droplets at a measured epoxy quantity. During the bonding process the chip is pressed on the epoxy injected substrate. Both epoxy and air flow to the chip 's sides, and the epoxy pattern should ensure that no epoxy voids and air bubbles are left, and the spreading epoxy generates a uniform sealing wrap (fillet) with a desired hight around the chip's edges. The droplets provide initial conditions for the flow, and with a calculated epoxy quantity the flow stops when the die pressing is stopped at a desired distance from the substrate. It stops due to viscosity and surface tensions. At this point the epoxy is void-free and bubble-free, with approximately an ideal fillet of a desired hight. An expert engineer was working on each nozzle design "manually" for several hours at least, depending on size and number of pipes, using an expensive graphical workstation. Designs were evaluated with a program that checked against a long list of design rules (constraints) to be met, at a trial and error process. To automate the design process itself researches have been working for two years with no good solution. By the end of December I finished programming a design algorithm which incorporated the design rules, and sent to my friend solutions for design requests he had left. They were good. I designed a simple nozzle design interactive system around the algorithm, with a menu-driven user interface of several screens. The system also managed a repository of selected nozzle designs. My son Roii, at high-school at that time, programmed the system using Basic on an IBM AT PC with DOS (Before Windows and mouse!). Roii surprised me writing in January a bug-free system, with user messages that exceeded my design. The system finds all solutions for a given set of parameters, using common defaults (that can be overridden) for convenience. The output is both graphical (for visual overview; see left picture above) and numerical (for constructing the nozzle). It allows a full nozzle design in few minutes by choosing the best solution to given chip and parameters (if a solution exists). We sent them a floppy-disk with the system in February. The next day they wanted it with minor cosmetic changes, and started using it.
The dotted epoxy patterns generated by the design system can be used also for epoxy dispensing methods other than multi-pipe nozzles, with similar results. Descriptions of computing methods with very similar patterns started to appear in professional magazines several years later.
The self explanatory system can be downloaded and tried here (Link disabled. Contact Yoav Raz)
Originally a nozzle calculations lasted seconds to minutes (with "wait" messages), depending on die size. On today's PCs they end in a fraction of a second for a maximal die.