Rich Q raised an interesting point in his comment, "Analog front ends," in response to my post, Superiority of Analog Integration Depends on Application. His question about AFEs (analog front ends) suggests an issue that design engineers always face when considering alternative product-design implementations: Is it better to take advantage of a highly integrated functional block or take on the additional design-cycle time and potential risk of a less integrated but more customizable design? The former locks your productís performance to an IC vendor's parametric choices and priorities but brings benefits in both time to market and design risk.
Engineering combines the art of balancing opposing forces, behaviors, and requirements with the science of fixing a given balance reliably and reproducibly through mass-manufacturing processes. In the current case, the trait opposing integration is segmentation: A decision about one is a commitment to the other.
Although it may seem to be an issue peculiar to IC technologies, this need to balance integration and segmentation predates monolithic circuits. Indeed, the concept dates back to the earliest complex electronic systems (if for no other reason than it allows designers to abstract a system as a collection of interconnected but otherwise approximately orthogonal functions).
So, at base level, segmentation can serve as an organizing principal allowing designers to manage project complexity and take advantage of a design teamís collective skill set and of vendorsí relevant technologies. But segmentation decisions do not only affect the design process. There are, for example, considerations for the competitive analysis designers and marketers must execute to justify a new-product development program before it starts: Market demand for parametric performance, cost, size, and power dissipation combine with time-to-market pressures and in-house expertise to inform make-or-buy decisions for various functional blocks.
In extremis, we have SOCs (systems on chip) -- an unfortunate marketing designation that always overstates the fact in an attempt to denote the current state-of-the-art in integration for a given application. These devices can subsume whole sections of a productís block diagram and can represent excellent value if and only if it leaves sufficient opportunity for the product developers to add differentiating value thatís meaningful to the marketplace.
And that, to my sensibilities, is where consideration of integration and segmentation begin: with an assessment of a product-development teamís value add in the context of market demands. After that the goal is often finding a balance in which highly integrated functions serve as problem-solving resources without their very solid-stateness imposing a brittleness on a design.