Resistors

A resistor limits current flow. It dissipates electrical energy as heat. Resistors are used to: set current through LEDs, create voltage dividers, pull signals to known voltage levels, limit inrush current, and set gain in amplifiers.

Colour code (4-band): first two bands = digits, third band = multiplier, fourth = tolerance. Mnemonic: Black Brown Red Orange Yellow Green Blue Violet Grey White = 0 1 2 3 4 5 6 7 8 9. So Brown-Black-Red = 1, 0, ×100 = 1000Ω = 1kΩ.

Standard values follow the E12 series: 1.0, 1.2, 1.5, 1.8, 2.2, 2.7, 3.3, 3.9, 4.7, 5.6, 6.8, 8.2 — then repeat at ×10. This is why you use 4.7kΩ pull-ups, not 5kΩ.

Series and Parallel Resistors

Series resistors divide voltage. Parallel resistors divide current. In series, more resistance → more voltage drop across that resistor. Two equal resistors in parallel give half the resistance of one.

Voltage divider: two resistors in series from VCC to GND. Output voltage tapped between them = VCC × R2/(R1+R2). Used to scale a 5V signal down to 3.3V for a microcontroller input, or to set a reference voltage.

Capacitors

A capacitor stores energy in an electric field between two conducting plates separated by an insulator (dielectric). It opposes changes in voltage. Key behaviour: it charges up to the supply voltage, then stops conducting. So in DC steady state, a capacitor is an open circuit. In AC circuits, it passes current — higher frequency = lower opposition (impedance).

Types: Ceramic (non-polarised, small values, for decoupling — 100nF near IC power pins), Electrolytic (polarised, large values, for bulk filtering — 10–1000µF), Tantalum (compact electrolytic, more stable), Film (precision, audio).

CRITICAL: Electrolytic capacitors have a positive and negative lead. Connect backwards and they will overheat, swell, and potentially explode. The negative lead is marked with a stripe. Always check polarity.

Series caps: total capacitance decreases (same formula as parallel resistors). Parallel caps: total capacitance adds up. When you add a 10µF and 100nF in parallel near an IC, you get 10.0001µF — the 10µF handles low-frequency bulk demand, the 100nF handles high-frequency transients (because it has lower inductance and responds faster).

Decoupling capacitors: when an IC switches internally, it demands a sudden burst of current. Without a cap nearby, this current must travel from the power supply through long traces — those traces have inductance which causes voltage spikes. A 100nF cap placed right next to the IC's VCC pin supplies that burst current instantly, keeping the power rail clean.

Inductors

An inductor stores energy in a magnetic field when current flows through a coil of wire. It opposes changes in current (the dual of a capacitor). In DC steady state, an inductor is a short circuit (just wire resistance). In AC circuits, it opposes high-frequency current — higher frequency = higher impedance.

Inductors are used in: switching power supplies (buck/boost converters), EMI filters, RF tuning circuits, and current sensing. You'll mostly encounter them in power supply designs at Foxconn.

Diodes

A diode allows current to flow in one direction only — from anode (+) to cathode (−). The cathode is marked with a stripe on the component body. Silicon diodes have a forward voltage drop of approximately 0.6–0.7V — meaning when conducting, the voltage across the diode is ~0.7V regardless of current (within rated limits).

Applications: Rectification (convert AC to DC), reverse polarity protection (protect circuits from accidentally reversed power connections), freewheeling (across motor/relay coils to absorb inductive kickback), ESD protection clamps on IC pins.

Zener diode: special diode that conducts in reverse at a precise voltage (the Zener voltage). Used as a voltage reference or clamp. If you need to clamp a 5V signal to 3.3V, a 3.3V Zener in reverse across the line will clamp it.

LED (Light Emitting Diode): a diode that emits light when forward-biased. Forward voltage is colour-dependent: Red ~2V, Green ~2.1V, Blue ~3.2V. Always use a series resistor to limit current (typically 10–20mA). Without a resistor, an LED will draw unlimited current and burn out instantly.

BJT — Bipolar Junction Transistor

A BJT has three terminals: Base, Collector, and Emitter. It is a current-controlled device — a small current into the Base controls a larger current from Collector to Emitter. The ratio is called hFE or beta (β), typically 100–300. So 1mA into the base can control 100–300mA through the collector.

NPN transistor: Current flows from Collector to Emitter when the Base is driven HIGH (positive current into Base). The transistor turns ON when VBE ≥ 0.7V. Used to switch loads connected to the positive supply — the load goes between VCC and Collector, emitter connects to GND.

PNP transistor: Current flows from Emitter to Collector when the Base is pulled LOW (current flows out of Base). Used for high-side switching — load connects between Emitter and the load, Collector goes to GND.

Two operating regions: Saturation (fully ON — like a closed switch, VCE ≈ 0.2V) and Cut-off (fully OFF — like an open switch). For switching applications you want the transistor firmly in one region or the other — never in between (that wastes power as heat).

MOSFET — Metal Oxide Semiconductor Field Effect Transistor

A MOSFET has three terminals: Gate, Drain, and Source. It is a voltage-controlled device — voltage on the Gate controls current from Drain to Source. Almost no gate current flows (extremely high input impedance), making it very efficient to drive.

N-channel MOSFET: Turns ON when Gate voltage is sufficiently higher than Source (typically VGS > 2–4V for logic-level MOSFETs). Current flows from Drain to Source. Used for low-side switching — load between VCC and Drain, Source to GND. This is the most common type in embedded power circuits.

P-channel MOSFET: Turns ON when Gate voltage is sufficiently lower than Source (VGS is negative). Used for high-side switching. Less common because performance is lower than N-channel.

RDS(on): the ON-resistance of a MOSFET when fully turned on. Lower is better — it determines how much voltage is dropped and how much power is wasted. A MOSFET with RDS(on) = 10mΩ at 5A wastes only P = I²R = 25 × 0.01 = 0.25W — far better than a BJT.

Why MOSFETs over BJTs in power circuits: MOSFETs switch faster, waste less power (no base current needed), and have lower on-resistance. Used in all modern power supplies, motor drivers, and battery management systems.

LDO — Linear Dropout Regulator

An LDO takes a higher input voltage and regulates it down to a fixed output voltage by burning off the excess as heat. It works like a variable resistor in series with the supply — always dissipating power equal to (Vin − Vout) × Iout.

Example: AMS1117-3.3 takes 5V input and gives 3.3V output. If your circuit draws 500mA, the LDO wastes (5−3.3) × 0.5 = 0.85W as heat. At 1A it wastes 1.7W — it will get very hot and needs a heatsink or it will go into thermal shutdown.

Advantages: Simple, cheap, no switching noise, no external inductor needed, very low output ripple — good for noise-sensitive analog circuits.

Disadvantages: Inefficient, especially when Vin is much higher than Vout. Efficiency = Vout/Vin. A 5V-to-3.3V LDO is only 66% efficient. All extra power becomes heat.

"Dropout voltage": the minimum difference between Vin and Vout for the LDO to regulate properly. A standard LDO needs ~1.5V dropout. An LDO needs Vin ≥ Vout + dropout. If Vin falls too close to Vout, the output droops. LDOs are used extensively in consumer electronics for post-regulation.

Buck Converter (Step-Down Switching Regulator)

A buck converter uses a switch (MOSFET), an inductor, and a capacitor to step voltage DOWN efficiently. The MOSFET switches on and off at high frequency (typically 100kHz–3MHz). When ON, current builds in the inductor. When OFF, the inductor releases that energy to the output. By controlling the duty cycle (ratio of ON to OFF time), you set the output voltage: Vout = Vin × duty cycle.

Example: A 12V to 5V buck converter with 75% efficiency at 1A wastes only 12×1/0.75 − 5×1 = 1W compared to an LDO wasting (12−5)×1 = 7W. Massive difference at scale.

Advantages: High efficiency (75–95%), handles large voltage differences, used for all main power rails in modern electronics.

Disadvantages: More complex, requires external inductor and capacitors, generates switching noise that can interfere with analog circuits.

Boost Converter (Step-Up Switching Regulator)

A boost converter steps voltage UP. Same principle — switch, inductor, capacitor — but configured differently. The inductor charges when the switch is ON, then releases energy to the output in series with the input when the switch is OFF, giving a higher output voltage.

Example: boosting a 3.7V Li-ion battery to 5V for USB output, or 3.3V to 12V for a display backlight. Used wherever you need a higher voltage than your source provides.

Common Power Failure Symptoms

An oscilloscope shows you voltage over time — it draws a graph. The X axis is time, the Y axis is voltage. This is what a DMM cannot show: how a signal changes moment to moment. It is your most powerful diagnostic tool after the DMM.

Key Controls

Timebase (horizontal, time/div): Sets how much time is represented by each grid division on the X axis. If set to 1ms/div and there are 10 divisions, you're seeing 10ms of signal. To see a 1kHz signal (1ms period), set timebase to 0.2ms/div so one full cycle spans about 5 divisions — filling the screen nicely.

Voltage scale (vertical, V/div): Sets how much voltage each Y division represents. A 3.3V signal on a 1V/div setting would show about 3.3 divisions of height — good visibility. If the signal is clipped at the top and bottom, increase the V/div. If the signal is a tiny line, decrease it.

Trigger: The trigger system decides when to start drawing a waveform. Without it, the waveform scrolls continuously and you can't see it clearly. A rising-edge trigger at a threshold of ~1.65V (half of 3.3V) tells the scope: "every time the signal crosses 1.65V going upward, start a new sweep." This makes a repetitive waveform appear stable on screen. If the trigger isn't set correctly, the waveform appears to drift or show multiple overlapping copies.

Probe ground clip: Always connect the probe's ground clip to the circuit's ground first. The probe tip then measures voltage relative to that ground. If you connect the ground clip to a non-ground point, you'll short it to ground and potentially damage the circuit.

Probe compensation: Before using a new probe, test it on the scope's built-in calibration signal (usually a 1kHz square wave at the front panel). Adjust the small trim capacitor on the probe until the square wave's corners are sharp and flat. An uncompensated probe gives inaccurate readings especially at high frequencies.

×1 vs ×10 probe: A ×10 probe has a 9MΩ series resistor inside, making it draw 10× less current from the circuit — much less loading. But it attenuates the signal by 10× so you need to set the scope to compensate. For most electronics debugging, use ×10 mode. Only use ×1 for very small signals that you can't afford to attenuate.

What to look for on a scope in FA work

• Power rail noise: a rail that reads 3.3V on a DMM might show 50mV of switching ripple at 500kHz on a scope. This ripple can cause ADC errors, communication failures, and erratic MCU behaviour. Look for: clean flat rail vs noisy rail.
• Clock signal quality: check frequency (matches expected?), duty cycle (should be 50% for most clocks), rise/fall times (should be fast — slow edges indicate signal integrity issues), and any ringing or overshoot on the edges.
• Communication signals: UART data — is it present? Is the idle line HIGH (correct for UART)? SPI — does CS go LOW before data starts? I2C — are there ACK bits? Is SDA being held low (bus lockup)?
• Glitches and spikes: short voltage spikes that a DMM completely misses. These can cause MCU resets, data corruption, and intermittent failures. Use single-shot trigger mode to capture one-time events.
• PWM signals: measure duty cycle, frequency, and whether it's changing correctly in response to commands.

Scope vs DMM — when to use which

Power Analyser

A power analyser measures electrical power with high accuracy. Unlike a DMM which gives you one reading at a time, a power analyser simultaneously and continuously measures: voltage, current, real power (W), apparent power (VA), reactive power (VAR), power factor, frequency, and efficiency. It logs these over time and can generate reports.

Power factor: the ratio of real power (doing useful work) to apparent power (total power drawn). For a purely resistive load, PF = 1.0. For motors, capacitors, and inductors, PF < 1.0 — meaning you're drawing more current from the supply than the load actually uses for work. Power factor matters for mains-powered products.

In an NPI FA context: used to characterise power consumption of a new product across all operating modes (idle, active, peak), verify the product meets its rated power specs, measure efficiency of power supplies, and identify products that draw more current than expected (which might indicate a failure or design issue).

Difference from DMM: A DMM measures one thing at a time and gives you a snapshot. A power analyser gives you real-time simultaneous multi-parameter measurement with logging. You connect it inline (current path through the analyser, voltage tapped in parallel).

Spectrum Analyser

A spectrum analyser displays signal amplitude versus frequency — the "frequency domain" view. While an oscilloscope shows you what a signal looks like over time (time domain), a spectrum analyser shows you what frequencies are present and at what levels.

How to read it: X axis = frequency (e.g., 1MHz to 1GHz), Y axis = amplitude in dBm (decibels relative to 1mW). Each peak on the display represents a frequency component. A fundamental clock at 100MHz will show a peak at 100MHz. Harmonics will appear at 200MHz, 300MHz, etc. (typically at lower amplitudes).

In Foxconn NPI FA context: consumer electronics (phones, tablets, laptops) must pass EMC (Electromagnetic Compatibility) regulations before sale. The spectrum analyser checks that the device does not emit electromagnetic radiation above permitted levels at any frequency. It's also used to validate RF performance (Wi-Fi, Bluetooth, cellular) — checking that the transmitter is at the right power level, right frequency, and not spreading energy into adjacent channels.

If you haven't used one: "I haven't operated a spectrum analyser directly, but I understand its purpose and how to interpret a spectrum plot — identifying fundamental frequencies, harmonics, and spurious emissions. I'm ready to learn the specific instrument in use here."

Logic Analyser

Although not listed in the JD, you have this on your resume and you have used it. A logic analyser captures multiple digital signals simultaneously and decodes them into readable data. Unlike a scope which shows waveform shape, a logic analyser focuses on timing relationships between many signals and protocol decoding.

You connect it to UART TX, SPI (MOSI, MISO, SCK, CS), or I2C (SDA, SCL) lines and it will decode the actual bytes being transmitted. This is how you verify that an MCU is sending the right commands to a sensor, or that a sensor is responding correctly. Without a logic analyser, you'd have to count bits on an oscilloscope manually — tedious and error-prone.

UART — Universal Asynchronous Receiver Transmitter

UART uses two wires: TX (transmit) and RX (receive). It is asynchronous — there is no shared clock. Instead, both devices must be configured to the same baud rate (bits per second) before communication. Common baud rates: 9600, 115200, 921600. At 115200 baud, each bit lasts 1/115200 = 8.68 microseconds.

Frame structure: Start bit (always LOW) → 8 data bits (LSB first) → optional parity bit → Stop bit (always HIGH). The idle state is HIGH. When a transmission begins, the line goes LOW for one bit period (start bit) — this wakes up the receiver. The receiver then samples each bit in the middle of its period to determine if it's HIGH or LOW.

Common failures: Baud rate mismatch (garbled output — looks like random characters), wrong voltage levels (5V TX → 3.3V RX without level shifting — works sometimes but unreliable), TX and RX swapped between devices (connect TX of one to RX of the other, not TX to TX), parity mismatch.

Debugging: Scope on TX line — you should see a signal going LOW briefly then HIGH for the start bit, then data bits, then returning HIGH. If the idle state is LOW, there's a wiring problem (inverted signal or connection to wrong pin). Check that the line isn't floating (no pull-up, no connection).

I2C — Inter-Integrated Circuit

I2C uses exactly 2 wires: SDA (Serial Data) and SCL (Serial Clock). It supports multiple devices on the same two wires — each device has a 7-bit address (sometimes 10-bit). The master generates the clock and initiates all transactions. Slaves only respond when addressed.

Open-drain: Both SDA and SCL use open-drain (or open-collector) signalling — devices can only pull the line LOW, never drive it HIGH. The line is pulled HIGH by a pull-up resistor connected to VCC. This is why pull-up resistors are mandatory — without them, the line can never go HIGH and nothing works.

Pull-up values: Typically 4.7kΩ for standard mode (100kHz), 2.2kΩ for fast mode (400kHz). Too high = slow rise times, signal doesn't reach HIGH before next clock edge. Too low = too much current, may exceed sink capability of devices. The right value depends on bus capacitance and speed.

Transaction structure: START condition (SDA goes LOW while SCL is HIGH) → 7-bit device address → Read/Write bit → ACK bit from slave (slave pulls SDA LOW for one clock) → data bytes, each followed by ACK → STOP condition (SDA goes HIGH while SCL is HIGH).

Common failures: Missing pull-up resistors (SDA/SCL stuck LOW or floating), wrong device address (ACK not received), bus lockup (slave stuck mid-transaction holding SDA LOW — fix by clocking SCL 9 times then sending STOP), address conflict (two devices with same address), voltage level mismatch (5V device on 3.3V bus without level shifting).

You have real experience here: SHT30, BMP180, SHT40, ALS sensors — all I2C. Hydroponics project, STM32 project. This is genuine knowledge you can talk about.

SPI — Serial Peripheral Interface

SPI uses 4 signals: MOSI (Master Out Slave In), MISO (Master In Slave Out), SCLK (clock), and CS/SS (Chip Select, active LOW). It is synchronous (shared clock) and full-duplex (send and receive simultaneously). Much faster than I2C — can run at tens of MHz.

CS pin: Each slave device has its own CS pin. The master pulls CS LOW to select a device, transfers data, then releases CS HIGH to deselect. Only one CS can be active at a time. This is how you have multiple SPI devices on the same MOSI/MISO/SCLK lines.

SPI modes (CPOL/CPHA): Four modes define clock polarity and phase — when exactly the data is sampled and when it changes. The datasheet for every SPI device specifies which mode it needs. Getting this wrong means the device receives shifted or inverted data. Mode 0 (CPOL=0, CPHA=0) is most common — clock idles LOW, data sampled on rising edge.

Common failures: CS not going LOW before clock starts (device never selected), CPOL/CPHA mismatch (data shifted, garbage received), clock frequency too high for device (data not sampled correctly), MOSI and MISO swapped, CS left permanently LOW (prevents other SPI devices from working).

USB — Universal Serial Bus

USB is a differential pair (D+ and D−) plus power (VBUS = 5V) and GND. Differential signalling means the receiver looks at the difference between D+ and D− — this makes it highly immune to common-mode noise. USB uses a complex protocol with descriptors, endpoints, and enumeration.

USB CDC (Communication Device Class): Makes the USB device appear as a serial (COM) port to the computer. This is what Arduino and ESP32 boards use for programming and debug output. The MCU runs a USB stack that implements the CDC class, and the computer loads a CDC driver (or uses the built-in one) and creates a virtual COM port.

Enumeration: When you plug in a USB device, the host (computer) requests its descriptor — a block of data that identifies the device type, manufacturer, VID/PID, and capabilities. If enumeration fails, the device shows as "Unknown Device" in Device Manager. Common causes: wrong descriptor data, D+ and D− swapped, missing pull-up resistor on D+ (needed for Full Speed USB).

X-ray Inspection

X-ray inspection is non-destructive — you can inspect a component without removing or cutting it. X-rays pass through the PCB and components at different rates depending on density and thickness. The resulting image shows the internal structure of solder joints that are hidden under components.

Why it's needed: BGA (Ball Grid Array) packages have their solder balls completely hidden under the IC body. You cannot inspect them visually or with a scope. X-ray is the only way to see if the balls are properly formed, or if there are voids (air pockets) or bridges between adjacent balls.

What X-ray shows: Solder voids (dark circular areas inside a solder joint — reduce reliability), solder bridges between adjacent balls/pads, missing solder balls, misaligned components, component tombstoning (one end lifted off the pad).

Voids in solder joints: Small voids (<25% of pad area) are usually acceptable. Large voids reduce the mechanical strength and thermal conductivity of the joint. IPC standards define acceptable void percentages for different applications.

Cross-sectional Analysis

Cross-sectioning is destructive — you physically cut through the sample, then polish the cross-section to a mirror finish, and examine it under a microscope. This reveals the internal structure of solder joints, through-holes, and component packages.

What it reveals: Crack propagation through a solder joint (from thermal cycling or mechanical stress), intermetallic layer thickness between solder and pad (too thick = brittle joint from over-heating), voiding inside the joint, pad delamination from the PCB substrate, plating thickness in through-holes.

Process: The sample is encapsulated in epoxy resin, then ground with progressively finer abrasives, then polished to optical flatness. The cross-section is then viewed under an optical microscope (up to 1000×) or prepared for SEM imaging.

SEM — Scanning Electron Microscope

SEM gives extremely high-resolution images (up to 100,000× magnification — compared to 1000× for optical microscopes) by scanning a focused beam of electrons over the sample surface. Electrons interact with the surface and produce signals that are detected to form an image.

What SEM reveals that optical microscopy cannot: Fine cracks that are invisible optically, surface texture at the nanometer scale, fracture surface morphology (how a joint fractured — ductile fracture looks different from brittle fracture), minute corrosion products, fine solder structure.

Sample preparation: Samples must be conductive (or coated with a thin conductive layer like gold) and placed in a vacuum chamber. This makes SEM a specialised, expensive tool — usually available in larger labs or sent to external analysis labs.

EDS — Energy Dispersive X-ray Spectroscopy

EDS is always used in conjunction with SEM. When the electron beam hits the sample, it causes each element in the sample to emit characteristic X-rays at specific energies. By measuring these X-ray energies, EDS identifies exactly which chemical elements are present — and in what quantities.

What EDS tells you: Is the solder the correct alloy (SAC305 = Sn 96.5%, Ag 3%, Cu 0.5%)? Is there contamination on the surface (sodium, chlorine, sulphur = corrosion agents)? Is the pad plating the right material (HASL vs ENIG vs OSP)? What are those white deposits (tin whiskers? flux residue? corrosion products?).

Practical use in FA: If you suspect a component is counterfeit (not genuine), EDS can check whether the die composition matches what the manufacturer specifies. If corrosion is found, EDS identifies the corrosive species, pointing to the contamination source.

Honest answer framework for tools you haven't used

FMEA (Failure Mode and Effects Analysis) is a proactive tool — you do it before failures happen, not after. The goal is to predict what could go wrong, assess how bad it would be, and take action to prevent it before the product ships. It's the difference between engineering quality in versus inspecting quality in.

FMEA Table Structure

An FMEA is a structured table. Each row represents one potential failure mode for one component or process step. The columns are:

• Component/Function — What component or process step are we analysing? (e.g., Input capacitor C1, Solder reflow process)
• Failure Mode — How could this component/step fail? (e.g., C1 open circuit, C1 shorted, C1 reversed polarity)
• Effect of failure — What happens to the product if this failure occurs? (e.g., Power supply fails, device completely dead)
• Severity (S) 1–10 — How bad is the effect? 10 = safety hazard or regulatory failure. 7–9 = loss of function. 4–6 = degraded performance. 1–3 = minor annoyance.
• Cause of failure — What could cause this failure mode? (e.g., Electrolytic capacitor placed with reversed polarity during assembly)
• Occurrence (O) 1–10 — How likely is this cause to occur? 10 = almost certain. 7–9 = high. 4–6 = moderate. 1–3 = rare.
• Current controls — What is already in place to prevent or detect this failure? (e.g., Polarity marking on silkscreen, incoming inspection, AOI inspection after reflow)
• Detection (D) 1–10 — How likely is the current control to detect the failure before it reaches the customer? 1 = almost certain to detect. 10 = almost impossible to detect. Note: LOW Detection score = GOOD detection (counterintuitive — lower is better for Detection).
• RPN — Risk Priority Number = S × O × D. Range: 1 to 1000. Higher RPN = higher priority for corrective action.
• Recommended action — What should be done to reduce the RPN? Focus on reducing Severity (design change), Occurrence (better assembly process), or Detection (add inspection step).

Worked FMEA example — capacitor failure

Component: C1 (100µF electrolytic input capacitor). Failure mode: Reversed polarity during assembly. Effect: Capacitor overheats, vents gas, power supply fails — device completely dead. Severity = 8 (loss of function). Cause: Assembly operator places cap in wrong orientation. Occurrence = 4 (moderate — happens occasionally with unmarked polarised caps). Current control: Visual inspection by operator. Detection = 6 (visual inspection misses some errors — moderate detection). RPN = 8 × 4 × 6 = 192. This is high priority. Recommended action: Add clear polarity marking to silkscreen AND add Automated Optical Inspection (AOI) after reflow. New Detection = 2 (AOI catches almost all), new RPN = 8 × 4 × 2 = 64. Acceptable.

Pareto Analysis — prioritising which failures to fix first

The Pareto principle (80/20 rule) says approximately 80% of your failures come from 20% of the failure modes. A Pareto chart ranks failure modes from most frequent to least, with a cumulative percentage line. Focus corrective action on the top 2–3 failure modes first — they give the most yield improvement for the least effort.

DOE is a statistical method for efficiently discovering which process variables (factors) affect an output (response) and by how much. Instead of changing one variable at a time (OFAT — slow, misses interactions), DOE changes multiple variables simultaneously in a structured pattern and uses statistics to separate their effects.

Key terminology

• Factor: An input variable that you can control and vary. Examples: solder reflow peak temperature, paste volume, component placement pressure, conveyor speed.
• Level: The specific values a factor takes during the experiment. Two levels (low/high) is common — e.g., temperature at 240°C (low) and 260°C (high).
• Response: The output you're measuring to see how the factors affect it. Examples: solder joint strength (grams-force), void percentage (%), product yield (%), electrical resistance (mΩ).
• Interaction: When the effect of one factor depends on the level of another. Example: high temperature might only cause joint defects when paste volume is also high — neither factor alone causes the issue. OFAT testing misses interactions; DOE catches them.
• Full factorial: Test every combination of every factor at every level. For 3 factors at 2 levels: 2³ = 8 experiments. For 5 factors at 3 levels: 3⁵ = 243 experiments. Gets unwieldy quickly.
• Fractional factorial / Taguchi orthogonal arrays: Run only a carefully chosen fraction of all combinations. Taguchi's L9 array runs only 9 experiments instead of 27 for 4 factors at 3 levels. You lose the ability to detect some interactions but gain huge efficiency.

Real example — solder reflow optimisation

Problem: 15% of BGA components showing voids on X-ray. You suspect three factors: (A) peak reflow temperature, (B) time above liquidus, (C) solder paste type. Each at two levels (low/high). Full factorial = 2³ = 8 runs. For each run, reflow 20 boards and measure average void percentage. Statistical analysis (ANOVA or main effects plots) reveals: factor A (temperature) has the largest effect, factor C (paste type) has moderate effect, factor B (time) has minimal effect. You set temperature and change the paste type. Void percentage drops from 15% to 3%. This is DOE in practice.

5-Why Analysis

The 5-Why technique asks "Why?" repeatedly until you reach a root cause that, if fixed, prevents the problem from recurring. The magic number is approximately 5 — most root causes are found within 5 iterations. But it's not a rigid rule — sometimes 3 whys are enough, sometimes 7 are needed.

The key insight: Each answer to "Why?" becomes the next problem statement. You're not looking for blame — you're looking for the system or process failure that allowed the problem to occur.

Detailed example — board fails power-on test

• Problem: Board fails power-on test — 3.3V rail measures 0V
• Why is the 3.3V rail 0V? → The LDO regulator is not outputting voltage
• Why is the LDO not outputting? → Its input capacitor (C1) is shorted
• Why is C1 shorted? → It was installed with reversed polarity
• Why was it installed backwards? → The assembly drawing did not clearly indicate polarity, and the PCB silkscreen had no "+" marking
• Why did the drawing lack polarity information? → The PCB design checklist did not include a requirement to verify polarity markings for polarised components
• Root cause: Missing checklist item in PCB design review process → Corrective action: Update design review checklist to mandate polarity verification for all polarised components, update silkscreen guidelines

Notice the root cause is not "operator error" — it's a process gap. The operator had no way to know which way the cap went. Fixing the process (updating the checklist and silkscreen) prevents every future operator from making the same error.

Fishbone (Ishikawa) Diagram

The Fishbone diagram is a visual brainstorming tool for 5-Why — it helps you organise all possible causes before investigating. The "head" of the fish is the problem statement. The "bones" are categories of causes. Standard categories for manufacturing are the 6Ms: Man (training, fatigue, skill), Machine (equipment calibration, wear), Method (process procedures, sequence), Material (component quality, wrong spec), Measurement (test accuracy, calibration), Environment (temperature, humidity, ESD).

You brainstorm potential causes in each category, then investigate to confirm or rule out each one. This prevents tunnel vision — you don't jump to the most obvious cause without considering others.

A proximity sensor detects whether an object is near (typically within 0–20cm) without physical contact. In smartphones this is used to detect when the phone is held against your ear during a call — the screen turns off to prevent accidental touches and to save battery.

How a proximity sensor works

The sensor contains two components: an IR LED (infrared light-emitting diode) and an IR photodetector. The LED emits pulses of infrared light (typically 940nm — invisible to humans). When an object is near, some of this IR light reflects back and hits the photodetector. The IC measures the intensity of reflected IR. More reflected IR = object is closer. The IC compares this measurement to programmable thresholds and generates an interrupt signal when the object crosses the near/far threshold.

The IR LED is pulsed (not continuous) to reduce power consumption and to allow synchronous detection — the IC only looks for reflected light during the LED pulse window, which reduces interference from ambient IR sources (sunlight, incandescent bulbs).

Crosstalk — the main FA challenge

Crosstalk occurs when IR from the LED reaches the photodetector via a path that doesn't involve an external object — typically by reflecting off the device's own housing, glass cover, or any nearby surface. This causes the sensor to report "close object" even in open air. Crosstalk is a common NPI failure mode because it depends heavily on the mechanical assembly — the gasket around the sensor, the glass transmittance, the sensor placement relative to the housing.

Diagnosing crosstalk: measure the sensor output in open air at maximum distance from any reflective surface. If the reading is above the near threshold, crosstalk is present. Solution: better optical isolation in the sensor packaging, revised mechanical design, software offset calibration.

Common FA issues

• IR LED not driving: If the current-limiting resistor for the IR LED is wrong value, open-circuit, or the LED itself is damaged, no IR is emitted and the sensor reads 0 (far) always. Test: measure voltage across the LED — should show ~1.2–1.4V forward voltage drop during the LED drive pulse. Check on scope during drive cycle.
• Threshold registers wrong: The IC uses upper and lower threshold registers to determine near/far state. If these aren't programmed correctly, the interrupt fires at wrong distances. Verify via I2C: read the current threshold register values and compare to the designed values.
• Black tape or opaque adhesive over window: In assembly, adhesive from a screen protector or label can cover the sensor window. Infrared light cannot pass through opaque materials — the sensor will always report "far." Inspect the sensor window carefully under IR-pass filter (or smartphone camera, which can see IR).
• Ambient IR saturation: In very bright sunlight, the IR photodetector may saturate from ambient solar IR even without the LED pulsing. Better IC designs have ambient light cancellation — they measure ambient IR first, then subtract it from the LED-on measurement. If this cancellation isn't configured, bright sunlight causes false near readings.

Logic Gates

Logic gates are the building blocks of all digital circuits. Each gate performs a Boolean operation on its inputs to produce an output. In electronics, these are implemented as transistor circuits inside ICs.

Flip-flop (D-type): A memory element that stores one bit of information. The output Q takes the value of input D on each clock edge. It holds that value until the next clock edge. Flip-flops are the building blocks of registers, counters, and all sequential digital logic.

Op-Amp — Operational Amplifier

An op-amp is a differential amplifier — it amplifies the difference between its two inputs (non-inverting input + and inverting input −). Ideal op-amp has infinite gain, infinite input impedance, zero output impedance. Real op-amps have very high but finite gain (~100,000).

Key rule for op-amp with negative feedback: The op-amp adjusts its output to make the two input voltages equal (virtual short circuit). This is the foundation of all op-amp circuit analysis.

Voltage follower (buffer): Output connected directly to (−) input. Output = Input. Gain = 1. Purpose: isolates the source from the load — the op-amp provides current without loading the source. Used when you need to drive a low-impedance load from a high-impedance source.

Non-inverting amplifier: Gain = 1 + R2/R1 (where R1 and R2 are the feedback resistors). Output is in-phase with input. Used to amplify sensor signals before ADC.

Comparator mode: Op-amp without negative feedback. Output swings to maximum positive when (+) > (−), maximum negative when (−) > (+). Used as a threshold detector — "is this voltage above my reference?" Used in battery under-voltage detection, thermostat circuits, overvoltage alarms.

Signal Integrity basics — for PCB debugging

Ringing: On a scope, you see oscillations at the edges of a square wave — the signal overshoots and undershoots before settling. Caused by transmission line effects — impedance mismatch between the driver, trace, and receiver. Solution: series termination resistor (typically 22–100Ω) at the driver output.

Ground bounce: When many outputs switch simultaneously, the sudden current spike through the ground inductance causes the ground to momentarily rise. This shifts all voltage references and can cause false logic triggers. Solution: good decoupling caps, multiple ground vias, star ground topology.

Cross-talk: A signal on one PCB trace induces a small voltage on an adjacent trace through capacitive or inductive coupling. Particularly affects high-speed signals and sensitive analog inputs. Solution: spacing traces apart, using ground guard traces between sensitive signals, keeping digital traces away from analog.

This is your primary story. Use it for any question about: hardware debugging, failure analysis, problem-solving, root cause analysis, working under pressure. Fill in the actual details from your memory below — even rough details become powerful when told with the right structure.

Asking good questions shows you've thought seriously about the role and are evaluating them, not just hoping to be selected. It changes the dynamic from "please hire me" to "let's see if this is a good fit." Always ask at least 2.