Firmware

Question 1

A greenhouse sensor node prints every received byte with printf inside its UART (serial port) receive interrupt service routine (ISR). At higher data rates, bytes go missing. What is the most likely cause?

Accepted Answer

Slow printf keeps the ISR busy while new bytes overrun the receiver.

Question 2

As a general rule, what should an interrupt service routine (ISR) do?

Accepted Answer

Do the urgent minimum, then hand the rest to a task or the main loop.

Question 3

After the first coin is inserted, a parking meter freezes. The debugger shows the CPU entering the coin-sensor interrupt handler over and over, and the main loop never runs. What is the most likely cause?

Accepted Answer

The handler never clears the pending flag, so it fires again at once.

Question 4

On a pump motor controller, the encoder interrupt sometimes misses edges while a long analog-to-digital converter (ADC) interrupt is running at the same priority. What is the best fix?

Accepted Answer

Give the encoder a higher preemption priority and shorten the ADC handler.

Question 5

What does interrupt latency measure?

Accepted Answer

The time from the hardware event to the start of its handler.

Question 6

A weather station's main loop disables interrupts while it computes a checksum over a 4 KB buffer. The serial receiver starts losing bytes. What is the best fix?

Accepted Answer

Keep interrupts disabled only around the few lines that touch shared data.

Question 7

A formatting helper uses a static char buffer. It is called from the main loop and from a timer interrupt handler. Output sometimes looks scrambled. What is the problem?

Accepted Answer

The helper is not reentrant, so the interrupt overwrites the buffer mid-call.

Question 8

Beehive scale firmware waits with while (!sample_ready) {}. A timer interrupt sets sample_ready = 1. It works with optimization off but hangs at -O2. What is the most likely cause?

Accepted Answer

The flag is not volatile, so the compiler reads it once and loops forever.

Question 9

Which statement about the volatile keyword in C is true?

Accepted Answer

It forces every access to happen, but does not make read-modify-write atomic.

Question 10

On an 8-bit microcontroller, a timer interrupt increments a 32-bit millisecond counter. The main loop sometimes sees the count jump ahead by about 256 and then go back. What is the most likely cause?

Accepted Answer

Main reads it in several byte loads, and the interrupt updates it in between.

Question 11

On an 8-bit microcontroller, main loop code reads a 32-bit tick counter that a timer interrupt updates. Which approach gives a consistent value?

Accepted Answer

Briefly disable the timer interrupt, copy the counter, then re-enable it.

Question 12

On a bike-share dock, two tasks each run hits++ on a shared counter. Over a day, the total is lower than the real number of events. Why?

Accepted Answer

The increment is a load, add, and store, and a switch can land between.

Question 13

On a sprinkler board, main does PORTB |= LED_BIT, and an interrupt does PORTB &= ~VALVE_BIT on the same port. The valve sometimes turns back on. What is the best fix?

Accepted Answer

Use the port's atomic set and clear registers, or a critical section.

Question 14

A helper disables interrupts on entry and always re-enables them on exit. It is called from inside another critical section, and a rare data corruption appears. What is the fix?

Accepted Answer

Save the prior interrupt state on entry and restore that saved state on exit.

Question 15

A serial receive interrupt writes bytes into a ring buffer and advances head. The main loop reads bytes and advances tail. Why can this work without a lock on a single-core chip?

Accepted Answer

Each index has a single writer, and each index update is one atomic store.

Question 16

On a dual-core chip in a cold-storage logger, core A fills a buffer and then sets a volatile ready flag. Core B sees the flag but sometimes reads stale buffer data. What is the fix?

Accepted Answer

Use atomics with release and acquire ordering, or memory barriers.

Question 17

In an irrigation controller, a low-priority logging task holds a mutex on the flash chip. A high-priority valve task waits for it. A medium-priority network task then runs for seconds. What is this called?

Accepted Answer

Priority inversion

Question 18

A real-time operating system (RTOS) shows priority inversion: a high-priority task waits on a mutex held by a low-priority task that keeps getting preempted. What is the standard fix?

Accepted Answer

Use a mutex with priority inheritance so the holder is boosted.

Question 19

In a real-time operating system (RTOS), when should you use a mutex rather than a binary semaphore?

Accepted Answer

To guard a shared resource, since it has an owner and supports inheritance.

Question 20

In a vending machine, task A locks the serial port and then the I2C (inter-integrated circuit) bus. Task B locks the I2C bus and then the serial port. Both tasks sometimes freeze forever. What is the best fix?

Accepted Answer

Make every task take the two locks in the same fixed order.

Question 21

A fixed-priority scheduler runs three periodic tasks with periods of 1 ms, 10 ms, and 100 ms, each due by its next period. Under rate-monotonic scheduling, which task gets the highest priority?

Accepted Answer

The 1 ms task, because the shortest period gets the highest priority.

Question 22

In a smart thermostat, the highest-priority task polls a sensor in a tight loop with no delay. The display and network tasks never run. What is the best fix?

Accepted Answer

Have the task block on a timer or an event between polls.

Question 23

A data-logging task calls delay(10 ms) after its work, which takes 1 to 3 ms. Samples drift and are not spaced evenly. What is the better approach?

Accepted Answer

Use a delay-until call that wakes at fixed absolute times.

Question 24

A serial receive interrupt posts each byte to a queue, and sometimes the queue is full. What should the interrupt handler do?

Accepted Answer

Count the dropped byte and return at once, since it must not wait.

Question 25

A cooperative scheduler runs a handful of tasks in a loop. One task now does a 40 ms image compression in a single call, and button presses feel laggy. What is the core problem?

Accepted Answer

No task can run until the long task yields, so the long call delays everyone.

Question 26

A developer adds a 2 KB local array to a task that has a 1 KB stack. Other tasks start crashing at random. What is the most likely cause?

Accepted Answer

The task's stack overflows into memory used by something else.

Question 27

How can you measure the peak stack use of a task on a small microcontroller with no memory protection?

Accepted Answer

Fill the stack with a known pattern, then later see how much was overwritten.

Question 28

A cafe point-of-sale terminal runs for weeks, then malloc fails for a 512-byte request even though the heap report shows 3 KB free in total. What is the most likely cause?

Accepted Answer

Heap fragmentation, so no single free block is 512 bytes long.

Question 29

Long-running firmware suffers from heap fragmentation. Which change best removes the risk?

Accepted Answer

Use fixed-size block pools, or allocate everything once at startup.

Question 30

A sensor hub parses JSON from the network with a recursive parser. A message with deeply nested arrays crashes the device. What is the best fix?

Accepted Answer

Limit the nesting depth, or use a parser that does not recurse.

Question 31

On a typical 32-bit microcontroller, adding a 4 KB table of fixed values grows both flash and RAM use by 4 KB. What change most likely saves the RAM?

Accepted Answer

Declare the table const so the linker can keep it in flash only.

Question 32

On a bare-metal board with no memory protection, the heap grows up and the stack grows down in the same RAM region. What happens if they meet?

Accepted Answer

They overwrite each other's data, often with no immediate fault.

Question 33

A fan speed controller runs a filter in floating point on a microcontroller with no floating-point unit (FPU). The loop misses its 1 ms deadline. What is a common fix?

Accepted Answer

Rewrite the filter in fixed-point integer math.

Question 34

In Q15 fixed-point format, stored in a signed 16-bit integer, what range of values can be represented?

Accepted Answer

From -1 up to just under +1

Question 35

You multiply two Q16.16 fixed-point numbers held in 32-bit integers. What is the correct approach?

Accepted Answer

Multiply into a 64-bit result, then shift right by 16 bits.

Question 36

A chip has a single-precision floating-point unit. The line y = x * 0.3; with float x runs much slower than expected. What is the most likely cause?

Accepted Answer

0.3 is a double constant, so the math runs in software doubles.

Question 37

A soil moisture probe sums 1000 readings from a 12-bit analog-to-digital converter (ADC) to average them. The average is often far too low. What is wrong?

Accepted Answer

The sum is kept in a 16-bit variable, which overflows past 65535.

Question 38

A fixed-point audio filter on a door chime makes loud clicks on strong input. Wrapping overflow sends large positive values to large negative ones. What is the usual fix?

Accepted Answer

Use saturating arithmetic that clamps at the maximum and minimum values.

Question 39

An uptime counter on a ferry ticket kiosk adds 1 to a 32-bit float every millisecond. After a few hours, the value stops increasing. Why?

Accepted Answer

Past about 16.7 million, the float cannot hold every integer, so +1 is lost.

Question 40

A battery gauge converts a 12-bit reading to millivolts with mv = (raw / 4095) * 3300 using integer math. It nearly always shows 0. What is the fix?

Accepted Answer

Multiply before dividing: raw * 3300 / 4095 in a 32-bit type.

Question 41

A status register's flags are write-1-to-clear. Code clears one flag with STATUS |= FLAG_RX. Other pending flags sometimes vanish. Why?

Accepted Answer

The read returns all set flags, and writing them back clears them all.

Question 42

Which C statement sets bit 5 of a control register without changing the other bits?

Accepted Answer

CTRL |= (1u << 5);

Question 43

Why are memory-mapped hardware registers accessed through volatile pointers?

Accepted Answer

Hardware can change the value, and each access can have side effects.

Question 44

On a little-endian microcontroller, the 32-bit value 0x12345678 is stored at address 0x100. Which byte is at address 0x100?

Accepted Answer

0x78

Question 45

A little-endian tide gauge sends a struct of 32-bit integers straight from memory to a server that expects network byte order. The numbers arrive scrambled. What is the best fix?

Accepted Answer

Serialize each field byte by byte in the defined big-endian order.

Question 46

A struct holds a uint8_t followed by a uint32_t. On a typical 32-bit compiler, sizeof reports 8, not 5. Why?

Accepted Answer

The compiler pads so the 32-bit field sits on a 4-byte boundary.

Question 47

Parser code casts a byte buffer pointer at offset 1 to uint32_t * and reads through it. On some cores this causes a hard fault. What is a safe fix?

Accepted Answer

Copy the 4 bytes into a local uint32_t with memcpy before using it.

Question 48

On a new board for a tool rental locker, the I2C (inter-integrated circuit) bus reads all zeros or hangs, and the scope shows the lines never go high. What is the most likely cause?

Accepted Answer

The pull-up resistors on the clock and data lines are missing.

Question 49

An I2C (inter-integrated circuit) bus with 47 kilohm pull-ups works at 100 kHz but fails at 400 kHz. The scope shows slow, rounded rising edges. What is the best fix?

Accepted Answer

Use smaller pull-up resistors so the lines rise faster.

Question 50

A temperature sensor datasheet lists its write address as 0x90. The I2C (inter-integrated circuit) driver takes a 7-bit address, and every transfer gets a NACK (not acknowledged). What is the likely fix?

Accepted Answer

Pass 0x48, since 0x90 is the 7-bit address shifted left with the write bit.

Question 51

A display on SPI (serial peripheral interface) returns values that look shifted by one bit, or garbage, though the wiring is correct. What should you check first?

Accepted Answer

That the clock polarity and phase match what the device expects.

Question 52

A UART (serial port) link from a farm gate controller shows framing errors and random characters. Both ends are set to 8 data bits, no parity, 1 stop bit. What is the most likely cause?

Accepted Answer

The two sides are not running at the same baud rate.

Question 53

A UART (serial port) runs at 115200 baud with 8 data bits, no parity, and 1 stop bit. Roughly how many data bytes per second can it carry?

Accepted Answer

About 11,500

Question 54

A GPS module streams at high baud into a chip whose UART (serial port) has a 1-byte buffer. With a per-byte interrupt, bytes are lost when the CPU is busy. What is a good fix?

Accepted Answer

Receive with DMA (direct memory access) into a circular buffer.

Question 55

A 12-bit analog-to-digital converter (ADC) uses a 3.3 V reference. About how much input voltage does one step represent?

Accepted Answer

About 0.8 mV

Question 56

A vibration sensor is sampled at 1 kHz. The data shows a strong 100 Hz tone, but the real vibration is at 900 Hz. What explains it, and what fixes it?

Accepted Answer

The 900 Hz signal aliases to 100 Hz. Add an analog low-pass filter before the ADC.

Question 57

A chip's analog-to-digital converter (ADC) reads a sensor through a 100 kilohm divider with the shortest sample time. Readings are low and seem to depend on the channel read before. What is the best fix?

Accepted Answer

Lengthen the sample time, or buffer the input with an op-amp.

Question 58

A general-purpose input/output (GPIO) pin wired to a switch reads random values when the switch is open. What is the most likely cause?

Accepted Answer

The input is floating, with no pull-up or pull-down to set its idle level.

Question 59

On a chip with a data cache, DMA (direct memory access) fills a buffer from the network. The CPU then reads old data from that buffer. What is the fix?

Accepted Answer

Invalidate the cache for that buffer before the CPU reads it.

Question 60

A garage door keypad counts one press of a button as three or four presses. What is the most likely cause?

Accepted Answer

Contact bounce makes several fast edges for each press.

Question 61

Which software method is a sound way to debounce a push button?

Accepted Answer

Sample every few ms and accept a change once it holds for several samples.

Question 62

A toy robot has a bump switch wired to an edge interrupt. Each bump triggers dozens of interrupts and stalls other work. What is a good approach?

Accepted Answer

On the first edge, mask that interrupt and confirm the state with a timer.

Question 63

A water meter's firmware refreshes the watchdog from a 10 ms timer interrupt. Later, the main loop hangs, but the device never resets. Why?

Accepted Answer

The timer interrupt keeps refreshing the watchdog while main is stuck.

Question 64

A warehouse robot runs five real-time operating system (RTOS) tasks. How should the watchdog be refreshed so a hang in any task causes a reset?

Accepted Answer

A monitor refreshes it only after every task has recently checked in.