A very wide range of designs can be accommodated from small transceivers or power management type devices through to large scale chips with a wide range of different pre-designed processor cores on board. Devices can be fabricated to suit many types of packaging and I/O as well. The complexity of the design and the technology will favour either a full ASIC implementation or a FPGA, the latter getting ever more sophisticated. Nowadays numerous devices, able to accommodate many different and varied functions and complexity, are readily available from firms such as Xilinx or Altera.

For most products requiring volume production it make sense to implement an ASIC/FPGA design, however this isn’t always the case. Furthermore, although it’s possible to effect designs at µm geometries that may be more difficult at larger scales, an ASIC isn’t a cure for poor initial concept, technology choice or fundamental design.

The decision to opt for a specific chip design will not only be based on projected production volume against total production costs but also on design longevity. That is, how long the design is likely to survive, and be useful, with current advances in technology, or until some other kind of hardware change is forced. 

Other factors that need to be carefully considered are decisions about what software and hardware boundaries are to exist, for example, we might ask are there any existing soft functions that it would make sense to move in to the hardware?

Other things to consider

Producing your own chip is an expensive business and it’s even more expensive if the prototype or pre-production devices don’t work. This would include ensuring that a new chip design incorporated in an end product would meet any approvals or certification required – particularly important for mixed-signal devices where isolation or higher frequencies are involved.

Nevertheless, custom chips, or Application Specific ICs (ASICs), can yield huge savings in production and in many cases deliver better performance as well as higher product reliability.  In many situations production costs, of any given product, may well be too high to contemplate without a VLSIC implementation.