Back to Blog

    Hardware Engineering Whiteboard Interview: Best Practices to Solve Problems with Confidence

    Chinmay Kulkarni19 min read
    Smiling engineer soldering a small circuit board at a lab bench with two oscilloscopes in a bright lab

    A hardware engineering whiteboard interview is a live problem-solving session: you design or analyze a circuit, logic block, or small system on a board while an interviewer watches and asks follow-ups. Because they hear your reasoning as you go, the process is part of your answer, and one method works for analog, digital, and system questions: clarify, state assumptions, sketch and decompose, solve, sanity-check, and summarize. Two fully worked examples below show the method in action.

    What a Hardware Whiteboard Interview Tests

    A whiteboard round tests whether you can solve an engineering problem out loud, with another engineer in the room. Northeastern's career office notes that on-site technical interviews can include solving "design problems on a whiteboard" (Northeastern University Career Development). For hardware roles, that might mean a filter, a transistor bias point, a state machine, or a block diagram.

    University career offices agree on the main point: your reasoning has to be visible. UMBC's career center says "the most important thing to remember when answering technical questions is to verbalize your thought process" (UMBC Career Center). Illinois Grainger's career services tells students to "think out loud" and to "provide a rationale why your thought could work well" (Illinois Grainger Engineering Career Services).

    An Ohio State engineering career services article makes the same point from industry. Bhanusri Sridharan, an OSU ECE alum working as a 3D NAND sort engineer at Intel, gives this advice: "the key tip is to always talk through to the interviewer on what's going on in your head to solve the problem!" (Ohio State Engineering Career Services).

    A 2016 Microsoft Life article, written for software candidates, makes a similar point: hiring managers "generally want to see how you think" (Microsoft Life). The format differs by field, but the idea carries over to hardware.

    There's a broader reason this matters, too. In NACE's Job Outlook 2025 survey of 237 employers, nearly 90% said they look for evidence of problem-solving ability on college students' resumes, and nearly 80% look for strong teamwork skills (NACE). That's resume data, not interview data, but a whiteboard round is one of the few places you can show both skills live.

    What Interviewers Are Listening For

    The career-office advice above boils down to a handful of signals you can make audible:

    SignalWhat it sounds like at the board
    You understood the problem"Before I start: what are the supply rails, and what's driving this input?"
    Your assumptions are explicit"I'll treat the op-amp as ideal for now and come back to bandwidth."
    You have a plan"Gain stage first, then the filter, then I'll check the output swing."
    Your work is checkableLabeled nodes, units on every number, one equation per line
    You test your own answer"If C goes to zero, this should reduce to a flat gain of -2. It does."
    You take hints well"Good point. Source impedance adds to R1, so let me redo the gain."

    For more on how technical depth and communication fit together, see why communication matters in hardware engineering interviews.

    The 6-Step Method for Any Whiteboard Problem

    Use the same six steps on every problem, whether it's an op-amp, a state machine, or a system block diagram, because a fixed routine gives you something to do in the first minute, when nerves are worst.

    1. Clarify the problem. Restate it in your own words and ask about missing specs. Northeastern says clarifying questions show "careful thinking" and prevent "solving the wrong problem." UMBC's advice: "Make sure you understand the question and have all the information you need to solve the problem."
    2. State your assumptions. Write them in a corner of the board, such as "ideal op-amp" or "input synchronous to the clock." Microsoft recruiter Abby Arvanitidis, quoted in the 2016 article, advises candidates to "validate your assumptions."
    3. Sketch and decompose. Draw a block diagram, schematic, or state diagram before any math. Then split the problem into pieces you can solve one at a time.
    4. Solve. Start with the simplest model that could work, then refine it. Narrate each step and say why you chose it.
    5. Sanity-check. Check units, limiting cases, orders of magnitude, and real constraints like supply rails or clock period. Arvanitidis also says to "always validate or test your solutions."
    6. Summarize. Restate the answer, the key numbers, and what you'd check next with more time.

    Add short checkpoints inside step 4: after each piece, say what you've finished and what's left, as in "Gain is done. Next is the filter corner." That gives the interviewer a natural moment to redirect you.

    Hardware Clarifying-Questions Checklist

    You don't need to ask every question, only the ones whose answers would change your design.

    Problem typeQuestions worth asking
    Analog and boardSupply rails? Input range and frequency? Source impedance? Load? Accuracy or tolerance needed? Power budget?
    Digital and RTLClock frequency? Reset type (synchronous or asynchronous, active high or low)? Mealy or Moore? Latency allowed? Overlapping matches?
    SystemWho uses it, and how? Throughput or sample rate? Power source and battery life? Interfaces? Cost or size limits?

    Worked Example 1: Inverting Amplifier With a 10 kHz Roll-Off

    Prompt: Design a circuit that inverts a sensor signal with a gain of -2 and rolls off noise above about 10 kHz.

    The component values below are illustrative choices for this walkthrough, not the only right answer. For more analog prep, see these analog hardware engineering interview tips.

    Clarify the Amplifier Spec (Step 1)

    What to say: "A few questions first. What supply rails do I have, and what's the input range? What's driving the input, what's the load, and how accurate does the gain need to be?"

    Suppose the interviewer gives you the gain, supply, and swing specs from TI's inverting amplifier reference design: a gain of -2 V/V, a ±15 V dual supply, a ±7 V input, and a ±14 V output (Texas Instruments). TI's own design targets a 3 kHz bandwidth, while this prompt asks for about 10 kHz. Assume the source has low impedance and the load is high impedance.

    State the Op-Amp Assumptions (Step 2)

    What to write: "Ideal op-amp with infinite gain and no input current; non-inverting input at ground, so the inverting node is a virtual ground; a single-pole roll-off is acceptable."

    Say out loud that you'll revisit the ideal op-amp assumption at the end, which tells the interviewer you know where the model breaks.

    Sketch and Decompose (Step 3)

    Split the problem into two pieces: a gain stage and a low-pass filter. Draw the standard inverting amplifier, where R1 runs from the input to the inverting node, R2 runs from that node to the output, and the non-inverting input goes to ground.

    Then propose the shortcut: put a capacitor C in parallel with R2, so one op-amp handles both the gain and the filter.

    Solve for Gain and Corner Frequency (Step 4)

    Gain. The inverting node sits at 0 V, so the current through R1 is Vin/R1. No current enters the op-amp, so the same current flows through R2. That gives the standard result:

    Vout / Vin = -R2 / R1

    Pick R1 = 10 kΩ and R2 = 20 kΩ for a gain of -2, then say what R1 means for the source. TI's design notes that "the input impedance of this circuit is determined by the input resistor, R1," so the input impedance here is 10 kΩ.

    Filter. With C across R2, the feedback impedance becomes R2 in parallel with 1/(sC). The transfer function is:

    H(s) = -(R2 / R1) × 1 / (1 + s·R2·C)

    That's a single pole at f_c = 1/(2π·R2·C), the same RC corner formula Analog Devices gives, f_c = 1/(2πRC) (Analog Devices Wiki). Solve for C with R2 = 20 kΩ and f_c = 10 kHz:

    C = 1 / (2π × 20 kΩ × 10 kHz) ≈ 796 pF

    The nearest standard E12 value is 820 pF, which puts the corner at 1/(2π × 20 kΩ × 820 pF) ≈ 9.7 kHz, close enough to "about 10 kHz." If you'd rather land just above 10 kHz, a 750 pF part (E24) gives about 10.6 kHz.

    Sanity-Check Units and Limits (Step 5)

    Run these checks out loud before you call it done:

    • Units: ohms times farads is seconds, so 1/(2πRC) comes out in hertz. Here R2·C = 20 kΩ × 820 pF = 16.4 µs.
    • Limiting cases: if C goes to zero, H reduces to -R2/R1, a flat gain of -2. At DC the capacitor is open, so the gain is also -2. At high frequency the capacitor shorts R2 and the gain falls toward zero.
    • The corner: at f_c, the gain magnitude drops to 1/√2 of its passband value, about 1.41 V/V. Analog Devices calls this the half-power frequency.
    • Roll-off rate: one pole means -20 dB per decade. At 100 kHz the gain is about 2/√(1 + (100/9.7)²) ≈ 0.19 V/V.
    • Output swing: 2 × 7 V = 14 V, which fits inside the ±15 V rails and matches TI's ±14 V output spec. Check the op-amp's datasheet, though: not every part can swing within 1 V of its rails.

    Summarize the Amplifier Design (Step 6)

    What to say: "Inverting amplifier with R1 = 10 kΩ and R2 = 20 kΩ, for a gain of -2 and a 10 kΩ input impedance. An 820 pF capacitor across R2 sets a single-pole corner near 9.7 kHz. A ±7 V input gives ±14 V out, inside the rails. Next I'd check the op-amp's bandwidth and offset."

    Follow-Ups to Expect

    A good interviewer will push past the ideal model, so be ready for these:

    • Op-amp bandwidth. The noise gain is 1 + R2/R1 = 3, so the closed-loop bandwidth is roughly the gain-bandwidth product divided by 3. That needs to sit well above 10 kHz so your RC, not the op-amp, sets the corner. Slew rate matters too: a full 14 V peak at 10 kHz would need about 2π × 10 kHz × 14 V ≈ 0.88 V/µs. The filter has already cut the amplitude by then, so treat that as a conservative bound.
    • Offset. The op-amp's input offset voltage shows up at the output multiplied by the noise gain of 3. Input bias current through R2 adds its own error. A resistor equal to R1 in parallel with R2 (about 6.7 kΩ here) from the non-inverting input to ground cancels most of the bias-current error, as long as the two input bias currents are well matched.
    • Source impedance. Any source resistance adds in series with R1. With an illustrative 1 kΩ source, the gain becomes -20/11 ≈ -1.82, about 9% low. Fixes include buffering the source or subtracting the source resistance from R1.
    • Steeper roll-off. If -20 dB per decade isn't enough, a second-order stage such as Sallen-Key or multiple-feedback is the next step.

    Worked Example 2: A 101 Sequence Detector FSM

    Prompt: Design a circuit that outputs 1 when it sees the bit sequence 101 on a serial input, overlaps allowed.

    Sequence detectors are a staple of digital design rounds. For related RTL questions, see these common ASIC design interview questions.

    Clarify the Detector Behavior (Step 1)

    What to say: "Overlaps are allowed, so 10101 should give two detections, right? Do you want a Mealy machine, where the output can go high in the same cycle as the last bit, or a Moore machine, where it goes high a cycle later? Is reset synchronous, does one bit arrive per rising clock edge, and should the output be a one-cycle pulse?"

    Suppose the answers are: either machine type is fine, reset is synchronous and active high, one bit arrives per rising edge, and the output is a one-cycle pulse.

    State the Clocking Assumptions (Step 2)

    What to write: "Mealy machine first, since it needs fewer states; input synchronous to the clock; reset puts the machine in S0."

    If the input came from another clock domain, you'd mention a synchronizer before the state machine; saying so takes five seconds and shows you're thinking about real hardware.

    Sketch the State Diagram (Step 3)

    Name each state by how much of the pattern you've seen so far:

    • S0: no useful bits yet (reset state)
    • S1: the last bit was 1
    • S2: the last two bits were 10

    The key to overlap is where you go after a match. When S2 sees a 1, the pattern is complete and the output is 1, but that final 1 could also start the next 101, so the machine goes to S1, not S0. Likewise, when S1 sees another 1, it stays in S1, since the newest 1 is still a valid start.

    Write the state table next to the diagram:

    Present stateMeaningx = 0: next state / zx = 1: next state / z
    S0No match yetS0 / 0S1 / 0
    S1Saw 1S2 / 0S1 / 0
    S2Saw 10S0 / 0S1 / 1

    Encode and Derive the Logic (Step 4)

    Three states need two flip-flops. Use the encoding S0 = 00, S1 = 01, S2 = 10 (written as Q1Q0), and treat the unused code 11 as a don't-care. Reading the table row by row gives three short equations, where x' means NOT x:

    D0 = x, D1 = Q0 · x', and z = Q1 · x

    D0 and D1 are the next values of Q0 and Q1. Every x = 1 arc goes to S1 (01), so D0 is just x. The only arc into S2 (10) is S1 on a 0, so D1 = Q0·x'. The only arc with z = 1 is S2 on a 1, so z = Q1·x.

    The Moore version needs a fourth state, because the output can depend only on the state. Add S3, meaning "just detected 101," with output 1. S2 on a 1 now goes to S3. From S3, a 0 goes to S2 (the last two bits are 10) and a 1 goes to S1. Four states still fit in two flip-flops.

    Present stateOutput zx = 0: next statex = 1: next state
    S00S0S1
    S10S2S1
    S20S0S3
    S31S2S1

    Say the trade-off out loud: the Moore output depends only on the state, so input glitches can't reach it and it's stable for the whole cycle. It still comes from decoding the state bits, so for a truly glitch-free output, register z or pick an encoding that gives z its own flop. The cost is that it shows up one clock later than the Mealy output.

    Sanity-Check With a Trace (Step 5)

    Run an illustrative input, 1 0 1 0 1, through the Mealy machine starting from reset:

    Clock edgePresent stateInput xNext stateOutput z
    1S01S10
    2S10S20
    3S21S11
    4S10S20
    5S21S11

    The output sequence is 0 0 1 0 1: two detections, where the second one reuses the middle 1, so overlap works. The equations agree: at edge 3, Q1Q0 = 10 and x = 1, so z = Q1·x = 1.

    Then check the structure of the machine itself.

    • Reset: the machine starts in S0, which matches the reset assumption you wrote down.
    • Completeness: every state has an arc for both x = 0 and x = 1, so no input leaves the machine undefined.
    • Unused state: if a fault ever puts the flops in 11, the equations send it to S2 (on a 0) or S1 (on a 1) at the next edge, so it can't lock up. It can produce one false z = 1 on the way out, so say that too.

    Summarize the State Machine (Step 6)

    What to say: "Three-state Mealy machine with two flip-flops: D0 = x, D1 = Q0·x', z = Q1·x. Overlap works because a match returns to S1. A Moore version needs four states, still two flops, with the output one cycle later. I can write the RTL next if you'd like."

    Practice Whiteboard Interview Questions by Type

    These electrical engineering whiteboard interview questions are practice prompts written for this guide, not any company's actual questions. Start each one with the move in the "First thing to write" column, then follow the six steps.

    Question typeExample promptFirst thing to write on the boardCommon trap
    Passive circuitsA 0 to 5 V step drives a 1 kΩ resistor in series with a 1 µF capacitor. Sketch the capacitor voltage.τ = RC = 1 ms, and v(t) = 5(1 - e^(-t/τ)) VSkipping the initial condition, or mixing up ms and µs
    Op-ampsWhat's the gain of a non-inverting amplifier, and why is its input impedance high?Virtual short (V+ = V-), gain = 1 + Rf/RgUsing the virtual short in a circuit with no negative feedback
    Transistor biasingAn NMOS has VGS = 1.2 V, Vth = 0.5 V, and VDS = 1 V. Which region is it in?Region conditions: cutoff if VGS < Vth; triode if VDS < VGS - Vth; saturation if VDS ≥ VGS - Vth (here 0.7 V ≤ 1 V, so saturation)Mixing up MOSFET and BJT "saturation," which mean different things
    Digital logicBuild a 3-input majority function with a K-map, then with a 4:1 mux.The truth table, then the K-map (AB + BC + AC)Jumping to gates without a truth table and missing a group
    State machinesDetect 1101 on a serial input, overlaps allowed.The clarifying questions: overlap, Mealy or Moore, resetReturning to S0 after a partial mismatch instead of the longest matching prefix
    Timing (setup and hold)Given clock-to-Q, logic delay, and setup time, find the maximum clock frequency. Is hold met?T ≥ t(clk-to-Q) + t(logic, max) + t(setup)Thinking a slower clock fixes a hold violation (a same-edge hold check doesn't depend on the period)
    Timing (clock divider)Divide a clock by 3 with a 50% duty cycle.Input and output waveforms, cycle by cycleUsing one clock edge only, which gives a 1/3 or 2/3 duty cycle
    System designDraw a block diagram for a battery-powered temperature data logger, including its power tree.Requirements: sample rate, accuracy, battery life, interfaceNaming part numbers before the requirements are pinned down
    DebuggingA board's ADC readings are noisy. How do you find the cause?What to measure: input signal, ADC reference, supply rails, groundSwapping parts before measuring, or probing with a long ground lead that picks up noise itself

    If a row exposes a gap in fundamentals, review these circuit concepts for hardware engineering interviews. For block-diagram rounds, work through these hardware system design interview questions.

    How to Use the Board (and a Virtual One)

    A tidy board makes your reasoning easy to follow and your mistakes easy to catch. Before you write anything, split the board into zones:

    • Given and assumptions (left edge): the specs, your clarifying answers, and assumptions like "ideal op-amp."
    • Diagram (center, largest zone): the schematic, block diagram, or state diagram.
    • Math (right side): one equation per line, top to bottom.
    • Summary (bottom right): the boxed final answer and key numbers.

    A few habits keep each zone readable:

    • Label every node, signal, and component, and use those labels in your equations.
    • Put units on every number you write, including intermediate results.
    • Write larger than feels natural, so the interviewer can read it from across the room.
    • When you change direction, say why, and strike through the old line instead of erasing it right away, so you can go back.

    Virtual Whiteboards

    Remote rounds might use a shared online whiteboard, a drawing tablet, or a shared document. Practice with the same kind of tool beforehand, because drawing with a mouse is slow and awkward until you've done it a few times.

    Narrate more than you would in person, and say what you're drawing as you draw it: "This box is the LDO, and this arrow is the 3.3 V rail into the MCU." Check early that the interviewer can see your screen clearly. If drawing a schematic is too slow, ask whether a typed netlist or a short component list works instead.

    What to Do When You Get Stuck

    Getting stuck is normal, and what matters is what you do next. Northeastern's advice is direct: "Don't spend 10 minutes stuck on one thing," and if you don't know something, "Say so honestly" (Northeastern University Career Development). Asking for help is fine too: in the OSU article, Google program manager Yashvardhan Gusani says "it is totally OK to ask for a hint."

    When you stall, keep talking and try one of these:

    • Go back to fundamentals. KCL, KVL, Ohm's law, or a truth table can restart your progress.
    • Simplify. Make the parts ideal, drop a source, or try a single input case.
    • Try a limiting case and ask what happens at DC, at very high frequency, or when a resistor goes to zero.
    • Say what you'd measure. "On the bench, I'd probe this node to confirm the bias point" shows how you'd solve it for real, the same habit that carries hardware debugging interview questions.

    If a schematic request is what freezes you, this guide on what to do when an EE interview asks you to draw a circuit walks through the panic and the recovery.

    Common Whiteboard Interview Mistakes

    Most weak whiteboard rounds come down to a few habits you can fix with practice:

    • Drawing a circuit before asking a single clarifying question
    • Reaching for a complex design before a simple baseline works
    • Solving in silence, then explaining at the end
    • Using numbers without units, or values that break the supply rails
    • Spending half the time on a derivation the problem doesn't need
    • Erasing large sections without saying why
    • Going quiet when stuck instead of saying what you're trying next

    Handling the Pressure of Being Watched

    Being watched can hurt performance, so practice under observation, not only alone. In one controlled study at NC State, 48 computer science students solved a technical problem either at a whiteboard with an interviewer watching while they explained aloud, or alone in a private room with no requirement to talk (NC State News; ESEC/FSE 2020 paper).

    The difference was large: study co-author Chris Parnin said people in the traditional interview "performed half as well as people that were able to interview in private." The sample was small, and the researchers say "a larger sample size would be needed to draw firm conclusions."

    The participants were CS students, not EE candidates. Still, the result is a useful warning: solving a problem quietly at a desk isn't the same skill as solving it with someone watching.

    So rehearse the conditions you'll face:

    • Practice out loud, at a real board. Gusani's recommendation in the OSU article is "using a whiteboard and talking out loud."
    • Get watched. Ask a classmate, TA, or friend to sit in and ask follow-ups.
    • Use a timer and practice finishing all six steps in the time you'd likely get.
    • Record yourself, because watching the playback shows long silences, unlabeled diagrams, and missing units that you won't notice in the moment.

    A Practice Plan and Day-Of Checklist

    Short, regular sessions are easier to sustain than one long cram before the interview, so try a simple weekly loop like this one.

    1. Pick one prompt per session from the table above, or use a daily design challenge.
    2. Set a timer, stand at a whiteboard, and work the six steps out loud.
    3. Record the session and review it the same day.
    4. Rotate between analog, digital, and system problems so you're ready for any of them.
    5. Once a week, have someone watch and push you with follow-ups.
    6. Keep a list of follow-ups you couldn't answer, and study those first.

    To find more prompts by topic, browse the Voltage Learning hardware interview question bank.

    Before you walk in, or log on, run through this list:

    • I'll restate the problem and ask clarifying questions before drawing.
    • I'll write my assumptions in their own zone and sketch a diagram before any math.
    • I'll narrate each step and summarize after each piece.
    • I'll check units, limiting cases, and supply or timing limits.
    • If I'm stuck, I'll say so, go back to fundamentals, and ask for a hint if needed.
    • I'll end with a short summary and what I'd check next.

    Sources

    1. Tech Sector Job Interviews Assess Anxiety, Not Software Skills, NC State University News, July 14, 2020 (reporting Behroozi et al., ESEC/FSE 2020)
    2. Does Stress Impact Technical Interview Performance?, Behroozi, Shirolkar, Barik and Parnin, ESEC/FSE 2020
    3. What Are Employers Looking for When Reviewing College Students' Resumes?, National Association of Colleges and Employers (NACE), December 9, 2024
    4. Tackling the Technical Interview, The Ohio State University Engineering Career Services, February 15, 2021
    5. Interview Type: Technical, Northeastern University Career Development, accessed October 2026
    6. Technical Interviews, UMBC Career Center, accessed October 2026
    7. Interview Fundamentals, University of Illinois Grainger Engineering Career Services, accessed October 2026
    8. How to ace a technical interview at Microsoft, Microsoft Life, January 21, 2016
    9. Activity: Low Pass and High Pass Filters, For ADALM2000, Analog Devices Wiki, accessed October 2026
    10. Inverting amplifier circuit (CIRCUIT060020), Texas Instruments, accessed October 2026

    FAQ

    Should I write Verilog on the whiteboard in a hardware interview?

    Only if the interviewer asks for it. A block diagram or state diagram is usually clearer on a board and shows your design before the syntax. Once the diagram and state table are agreed on, offer to write the RTL, and keep it short and readable.

    What if I don't know the answer to a whiteboard question?

    Say so honestly, as Northeastern University's career office advises, then reason from fundamentals out loud. Start from KCL, KVL, Ohm's law, or a truth table, try a simplified or limiting case, and ask for a hint if you're still stuck. Northeastern also warns against spending 10 minutes stuck on one thing.

    How long should I spend asking clarifying questions?

    Usually a minute or two, enough to pin down the spec. Focus on the questions whose answers would change your design, such as supply rails, input range, load, and accuracy for analog problems, or clock, reset, Mealy versus Moore, and overlap for digital ones.

    How do I practice whiteboard interviews alone?

    Stand at a real whiteboard, set a timer, and solve one prompt out loud using the same six steps every time. Record yourself and review the playback for long silences, missing units, and unlabeled diagrams. When you can, add a friend or TA as an observer, since solving while watched feels different from solving alone.

    Are virtual whiteboard interviews different?

    The method is the same, but the tool is slower. Practice with a shared whiteboard, tablet, or document beforehand, describe what you're drawing as you draw it, and confirm early that the interviewer can see your screen. If drawing a schematic is too slow, ask whether a typed component list works instead.

    Chinmay Kulkarni, founder of Voltage Learning

    Chinmay Kulkarni

    Founder of Voltage Learning. BS and MS in Electrical Engineering from Georgia Tech; former hardware engineer at Qualcomm, where he contributed to hardware that shipped in the iPhone 15, and later interviewed and evaluated hardware engineering candidates in Qualcomm's university recruiting.

    Related Articles