Embedded System Design Cycle Questions and Answers

Answer: A

If the driver reaches registers directly it cannot be built for the host at all, whereas a thin abstraction can be replaced by a stub that records what was written. Designing for that substitution is one of the practical benefits of a hardware abstraction layer.

Answer: B

Every fault fixed should leave behind a test that would have caught it, and running the whole set on every change is what stops the same fault returning. It is the accumulated set rather than any single test that gives the value.

Answer: C

Rework, recall, field visits and reputation all attach to a fault that has escaped, and the engineering to fix it is the smallest part of the bill. That escalation is the whole economic argument for review and test at every stage.

Answer: D

Reaching for the wrong instrument wastes hours: an analogue fault is invisible on an analyser and a protocol sequence is unreadable on two scope channels. Knowing which question is being asked is what selects the tool.

Answer: A

Correct answer: the timing relationship among many digital lines

A logic analyser samples many channels at once and displays them against a common time base, so it answers questions about sequence and relative timing - whether the chip select fell before the clock, whether the bus settled before the strobe. An oscilloscope shows the analogue detail of one or a few signals, which is what is wanted for rise times, amplitudes and noise.

Answer: B

Several hundred microseconds spent formatting and transmitting a line displaces the very event under study, and the fault often disappears. Toggling a spare pin and watching it on an analyser costs a few cycles and disturbs nothing.

Answer: C

Correct answer: costs a few cycles and can be watched externally

Setting a pin on entry and clearing it on exit costs almost nothing, does not stop the processor, and can be watched on a scope or analyser while the system runs at full speed. That makes it the technique of choice for anything timing-sensitive, where a breakpoint would change the behaviour being investigated and printing would be far too slow.

Answer: D

A fault that has already corrupted the stack leaves a breakpoint with nothing useful to say, whereas a trace buffer records how the program arrived there. It needs hardware support and a probe that can stream the data out.

Answer: A

Correct answer: checking mechanically for defects a reader would tire of looking for

A tool checks every path in every function without fatigue - uninitialised variables, unreachable code, missing break statements, possible overflows, violations of a coding standard. A reviewer is far better at judgement: whether the design is right, whether the code will be understood in a year. The two are complementary because each is weak where the other is strong.

Answer: B

Correct answer: code that is correct today but easy to break tomorrow

Testing can only observe behaviour, so code that passes every test may still be badly named, duplicated, tightly coupled or resting on an unstated assumption - and none of that shows until someone changes it. A reviewer reads the code rather than running it, and so sees the qualities that determine what happens next year rather than today.