loading demo track
Avionics Test Engineering Intern · Fundamentals interviewer
A finished interview, so you can see what comes out of one. In a real track it is live: the questions come from your own bank, chosen by the same scheduler that runs your practice, and the interviewer follows up when an answer is worth pushing on — twice here, in both cases where the candidate was partly there. The score below is computed, not written into the fixture — the same difficulty-weighted arithmetic that scores a real interview.
Strong on signal integrity and instrumentation — the termination answer and the rise-time correction were both what you would want to hear. The bring-up thread was the weaker one: the right instincts, but the sequencing hypothesis arrived late and without a measurement that would have distinguished it.
8.3
weighted / 10
Follow-ups improved the signal-integrity answer and weakened the bring-up one. Pushed on termination, the answer got more specific and named the tradeoff unprompted, which is the pattern of real understanding. Pushed on the short, the answer moved to a new hypothesis rather than to a way of testing the existing one — under time pressure that becomes guessing, and interviewers read it that way.
Consistently led with the approach and reached the conclusion second. That works in conversation but costs you in a debrief: an interviewer taking notes wants the finding first — 'it is a reflection, and here is how I would confirm' — with the reasoning behind it.
Interviewer
Thanks for making the time. I'd like to spend this session on how you reason through real problems rather than on definitions, so think out loud as much as you can. You're probing a 3.3 V SPI bus at 20 MHz and you see significant ringing on the clock edges — the link works most of the time but drops a transaction every few thousand. Walk me through what could be causing it and how you'd confirm.
You · scored 8.5
The ringing is almost certainly a reflection rather than picked-up noise. The driver's output impedance doesn't match the trace, so part of the edge reflects off the far end and comes back. What matters is the edge rate rather than the 20 MHz — if the rise time is around a nanosecond there's content past 300 MHz, so a few inches of trace is already electrically long. I'd measure the ringing period and compare it against twice the propagation delay of the trace; if they line up it's a reflection. I'd also want to rule out my own probe before believing any of it, because a standard ground lead rings on its own.
Follow-up
Good — and let's say the period does match the round trip. You decide to terminate. Which termination, where does it go, and what does that choice cost you?
You · scored 9.0
Series termination at the driver, sized so the resistor plus the driver's output impedance equals the trace impedance. On a point-to-point SPI clock that's the cheapest fix and it draws no static current. The cost is that the launched edge is half amplitude until the reflection returns, so it only works with the receiver at the far end — if there were devices partway along the line I'd need parallel termination at the end instead, and pay for it in static current.
Interviewer
That's right. Different situation: you power a new board for the first time and current draw hits the supply limit immediately. What do you do?
You · scored 7.5
First thing is to not keep it powered — I'd set the supply's current limit low, low enough that nothing can be damaged, and bring it up in that mode so I can probe while it's faulted rather than after something has already burned. Then I'd try to localise it: check whether the short is on the input rail before any regulator or downstream of one, by measuring resistance to ground on each rail with the board off. If a rail is near zero ohms I'd look for the obvious causes — a backwards part, solder bridge, a cap fitted in the wrong footprint — and I'd use a thermal camera or my finger to find what's getting warm at low current.
Follow-up
Say every rail measures a few ohms to ground rather than a dead short, and nothing gets warm. What now?
You · scored 6.0
A few ohms across everything sounds more like a part that's on and drawing current than a fabrication short. I'd check the sequencing — if a rail comes up out of order some parts will conduct through their protection diodes and look like a low impedance. So I'd bring the rails up one at a time in the intended order and watch where the current appears.
Interviewer
Last one, and it's arithmetic. You measure a 7 nanosecond rise time on a 100 MHz scope. Taking the scope's own rise time as 0.35 over its bandwidth and combining in root-sum-square, what's the true edge, and would you trust it?
You · scored 9.5
The scope's rise time is 0.35 divided by 100 megahertz, so 3.5 nanoseconds. Root-sum-square means the true one is the square root of 7 squared minus 3.5 squared — that's the square root of 36.75, so about 6.06 nanoseconds. I'd treat it as roughly 6 and not more precise than that: the instrument is contributing a fifth of the measurement in quadrature and the 0.35 constant is a rule of thumb, so if the decision hinged on 6 versus 7 I'd go find a faster scope rather than correcting arithmetically.