Design metrics of embedded system

A design metric is a measurable feature of a system's implementation. Designers must optimise several metrics at once, and these metrics usually compete with each other (improving one worsens another). This is the design-metric (optimisation) challenge.

Important design metrics

• NRE cost (non-recurring engineering): the one-time cost of designing the system. It is paid once, however many units are made.
• Unit cost: the cost of manufacturing each copy (excluding NRE). Total cost = NRE + unit cost × number of units.
• Size: physical space or memory, measured in bytes (software), gates (hardware) or cm³.
• Performance: execution time. It includes latency (time for one task) and throughput (tasks per second).
• Power: the power consumed; it decides battery life, heat and cooling.
• Flexibility: the ability to change functionality without large extra NRE.
• Time-to-prototype: the time needed to build a working version.
• Time-to-market: the time until the product can be sold. Late entry loses revenue.
• Maintainability: how easily the system can be modified after release.
• Correctness: confidence that the system implements its functionality correctly.
• Safety: the probability that the system will not cause harm.
• Reliability is also commonly listed with these.

Competing example

Using a faster processor improves performance but raises unit cost and power. A custom single-purpose chip gives high performance and low power but has high NRE and low flexibility.