What's changed: Initial version
1.1Embedded processors and MCU/DSP
Covers the roles of the MCU (microcontroller), the DSP, the general-purpose CPU, and the SoC; the design-philosophy difference between RISC and CISC; the use-case lineup of the ARM Cortex-M/R/A series; and the judgment skill of choosing the optimal processor configuration from an embedded device's requirements (real-time behavior, power consumption, cost, and computational characteristics).
ES's exam A-2 does not test rote definitions such as "what is an MCU"; it tests the ability to judge, from the requirements of the embedded device in front of you (real-time control, signal processing, power consumption, cost), which processor configuration (MCU/DSP/general-purpose CPU/SoC, RISC/CISC, Cortex-M/R/A) to choose. Even for the same requirement of "processing a temperature sensor reading," a simple threshold check is fine on a cheap MCU, but real-time digital signal processing (filtering, FFT) calls for a DSP, and running a rich UI on top of an OS calls for an application processor (the Cortex-A family).
1.1.1Roles of the MCU, DSP, general-purpose CPU, and SoC
- A MCU (microcontroller) is an embedded-oriented processor that integrates a CPU core together with peripheral functions—ROM/RAM, timers, GPIO, an A/D converter—onto a single chip. Low cost and low power consumption make it well suited to executing simple control logic in real time, as in home appliances, sensors, and motor control.
- A DSP (digital signal processor) is specialized for rapidly repeating multiply-accumulate operations (multiply plus add) in a single cycle. It is well suited to processing large volumes of continuous numerical computation in real time, such as audio/image filtering or an FFT (fast Fourier transform). A general-purpose CPU can perform the same computation, but a DSP's dedicated hardware multiplier and specialized addressing modes let it process faster at comparable power consumption.
- A general-purpose CPU is a high-performance processor whose primary purpose is running diverse applications on top of an OS. An SoC (System on Chip) integrates a CPU core along with a GPU, memory controller, wireless communication module, and more onto a single chip, and is used in smartphones and embedded devices with a rich UI (such as industrial tablets).
1.1.2RISC/CISC and the ARM Cortex series
- RISC (Reduced Instruction Set Computer) is a design philosophy that simplifies the instruction set into fixed-length instructions aiming for close to one cycle per instruction. CISC (Complex Instruction Set Computer) is a design philosophy with a rich instruction set where a single instruction can perform complex processing, at the cost of variable instruction length and cycle count. Because the embedded field demands low power consumption and high clock efficiency, RISC-based designs (such as ARM) are mainstream.
- The ARM Cortex series has three families by use case. Cortex-M is an MCU-oriented core with low cost, low power consumption, and high interrupt responsiveness (home appliances, sensor control). Cortex-R is an embedded-control-oriented core emphasizing hard real-time behavior and functional safety (redundant configurations such as lockstep), used in automotive ECUs and storage controllers. Cortex-A is a high-performance application-processor-oriented core for running an OS (Linux/Android, etc.), used in smartphones and industrial tablets.
Most-tested: a DSP is specialized for multiply-accumulate operations and suits filtering/FFT; Cortex-M is MCU-oriented, Cortex-R emphasizes hard real-time behavior and functional safety, and Cortex-A is a high-performance core for running an OS; and embedded systems are predominantly RISC-based. Practice being able to work backward from requirements (real-time behavior, computational characteristics, power consumption, whether an OS is needed) to the processor configuration.
Suppose you are the embedded designer at a consumer-electronics maker, and a portable audio device is given the requirements of "real-time noise-canceling processing" and "long battery-powered operation." The first candidate, a general-purpose MCU (Cortex-M family), is low-power and low-cost, but its computational efficiency falls short for the use case of executing large volumes of continuous multiply-accumulate operations in real time (digital filtering such as an FIR filter) needed for noise cancellation; achieving the same processing would require raising the clock speed, increasing power consumption and failing to meet the battery-life requirement. Considering a DSP next, its dedicated hardware multiplier and architecture optimized for multiply-accumulate operations let it complete the filtering faster at power consumption comparable to the MCU; finishing sooner frees up more time to return the processor to sleep, which in turn also satisfies the battery-life requirement. If the device then acquires the additional requirement of "support music playback over Bluetooth plus a touch-panel UI and app installation," that means the need has shifted from real-time signal processing alone to running diverse applications on top of an OS—at which point the choice should shift to an application processor in the Cortex-A family (an SoC). The core of embedded processor selection is the procedure of "first identify the essential nature of the processing—simple control, continuous signal processing, or rich OS execution—then choose the optimal configuration under the power and cost constraints."
| Processor type | Well-suited processing | Typical use case | Power/cost tendency |
|---|---|---|---|
| MCU (Cortex-M) | Simple control logic, GPIO/interrupt handling | Home appliances, sensors, motor control | Low power, low cost |
| DSP | Multiply-accumulate, filtering, FFT | Audio/image processing, communication modem | High efficiency per computation |
| Cortex-R | Hard real-time control, redundant configuration | Automotive ECU, storage controller | Moderate cost, high reliability |
| Cortex-A (SoC) | Running an OS, rich UI, diverse apps | Smartphones, industrial tablets | Higher power, higher cost |
Trap: "A DSP outperforms a general-purpose CPU, so a DSP should always be chosen" is wrong—a DSP's advantage is limited to continuous signal processing centered on multiply-accumulate operations; for simple control logic or running an OS, an MCU/Cortex-A is more appropriate. Also wrong: "the Cortex-M series has low performance, so it cannot be used for functional-safety applications"—functional safety (hard real time, redundancy) calls for Cortex-R, a different design philosophy, rather than Cortex-M; it is not a matter of relative performance.
1.1.3Section summary
- An MCU is low-cost, low-power for simple control; a DSP specializes in multiply-accumulate for filtering/FFT; an SoC is a high-performance configuration for running an OS
- ARM Cortex is chosen by use case: M for MCU-oriented work, R for hard real time/functional safety, A for high-performance OS execution
- Processor selection is a judgment exercise: identify the essential nature of the processing (control/signal processing/OS execution) and choose the optimal configuration under power and cost constraints
Sign in to track progress — Log in.
Quick check
(just a quick review)Q1. A portable audio device must combine real-time noise-canceling processing (continuous digital filter computation) with long battery-powered operation. Which processor best fits this requirement?
Q2. A designer of an automotive ECU wants to select a processor core that satisfies both hard real-time control and functional safety (malfunction detection via a redundant configuration such as lockstep). Which choice is most appropriate?
Q3. Which explanation is most appropriate for why RISC, as a processor design philosophy, has become mainstream over CISC in the embedded field?