AIF Roadmap

January 2003

AIF I was released in 1998 and was sufficient only to describe the die pad number, coordinates, net name and the ball pad that the die pad should be connected to. It was primarily an attempt to get semiconductor designers to consistently describe the die surface. Artwork developed this format together with Amkor who had a vested interest in reducing the time and error it took to get the die pad coordinates into a package design tool.

The AIF I spec was extremely simple and anyone with a spreadsheet could create a AIF I file in a few minutes. The cost of this simplicity was that many things could not be described. For example, only square die pads were supported and all die pads had to be the same size. The die outline was assumed to be symmetrically arranged around the die pads. There was no way to describe bond fingers, rings or wires.

AIF II

Not long after completing the AIF I spec and writing some CAD tools that would import it into package software such as Encore we started on the next generation of AIF -- one that could describe the entire bond shell - die pads, fingers, rings and wires.

We kept the same column format in order to maintain as much compatibility with spreadsheets (something that the package design world adopted and continues to use to this day) as possible. We had to add some addtional concepts: pad stacks, bond fingers with rotation, metal areas and traces (for the rings) and a BGA outline description.

While this new format was more complex, it was also much more useful and we wrote quite a few CAD tools based on AIF II. We also wrote import and export tools for both Encore BGA (from Xynetix) and Cadence APD which were the two dominant design tools for BGA packages.

The AIF II format is in wide use today primarily as a way of importing/exporting die and package data to Encore and APD. It is also used for producing wirebond documentation. Artwork released a free Windows viewer for AIF which aided in adoption of the format.

More Capability Requested

While AIF II was very useful it too made many assumptions to simplify reading and writing the format. These included:

  • One Die - AIF II could not describe multiple die in a single package. This became a problem with the advent of stacked die and MCMs. In addition there was no way to specify a wire from die pad to die pad.

  • No Vertical Data - AIF II was essentially layerless. It pretty much assumed that everything that happened on the surface was at one level. Again, this would be a problem for stacked die.

  • No Support for Routes/Vias - users wanted to get a more complete description of their packages including the routes from bond finger to ball and the vias that interconnected the various layers.