OE-9.1 Introduction to Embedded Systems

What this is and why it exists

An embedded system is defined by its constraints, not by its capabilities. That inversion is the whole of this topic.

Fixed function. Limited power. Real-time deadlines. A cost target that rules out generosity. A general-purpose computer has none of those, which is why the two are designed by different reasoning.

So processor selection is an engineering negotiation rather than a search for the fastest part. The fastest part is usually the wrong answer, and knowing why is the point.

The vocabulary

• Embedded system — a computer built into something that is not a computer.
• Fixed function — it does one job for its whole life.
• Real-time — correctness includes being on time.
• Hard deadline — missing it is a failure.
• Soft deadline — missing it degrades quality.
• Design metric — a measurable quantity a design is judged on.
• Unit cost — what one more of them costs to make.
• Non-recurring cost — what the design itself cost, once.

The mental model

A general-purpose computer is built to run programs nobody has written yet. An embedded system runs one program, forever, and everything about its design can be specialised to that. That is a freedom, not a limitation — you can drop everything the job does not need.

What replaces flexibility is a set of hard constraints. Power, because it may run on a battery for a year. Cost, because it may be made in millions and a rupee each matters. Size and weight, because it has to fit. And time, because in many of these systems being late is being wrong.

That last one deserves its own emphasis. In a real-time system, a correct answer delivered late is not a correct answer. An airbag controller that responds after the crash has failed completely. A hard deadline is one where missing it is failure; a soft deadline is one where missing it degrades quality, like a dropped video frame. Which kind you have changes the entire architecture.

Design metrics are the vocabulary for the negotiation: performance, power, unit cost, non-recurring cost, size, time to market, flexibility, maintainability. Learn them as a set that trades against one another, because that is how they behave. Improving one usually worsens another, and a design is a chosen position rather than an optimum.

Processor selection follows directly. You are not asking which is fastest. You are asking which one meets the deadlines with margin, within the power budget. And at a unit cost the product can carry, with the peripherals the job needs already on the chip. A part that satisfies all of those and is unremarkable on paper is the right answer.

What you should now be able to explain or do

Define an embedded system by its constraints. Say what specialising to one program buys. List the constraints that replace flexibility. Distinguish hard from soft deadlines and say why the distinction changes the architecture. Name the design metrics and say why they trade. Choose a processor by negotiation across those metrics rather than by speed.

Check yourself