VLSI Interview Questions With Answers and Worked Numbers
VLSI interviews for physical design, design and verification roles test whether you can reason about timing, power and transistors with numbers, not just definitions. Here are fifteen common questions with the answer an interviewer expects, and worked numbers for the ones that are usually asked as calculations. Every number below is recomputed in the site's tests.
Timing
- Q1. What are setup and hold time? Answer: setup time is how long before the clock edge the data must be stable; hold time is how long after the edge it must stay stable. Violating either can make the flip-flop go metastable or capture the wrong value.
- Q2. Which one can you fix by slowing the clock? Answer: only setup. Hold depends on the shortest path from one edge to the same edge at the next flip-flop, so the clock period does not appear in it; hold is fixed by adding delay to fast paths.
- Q3. What is clock skew, and is it good or bad? Answer: the difference in clock arrival time between two flip-flops. If the capture clock arrives later, setup gets that much more time but hold gets that much less; positive skew helps one and hurts the other.
Q4. Work out setup and hold slack
A path runs at 500 MHz (a 2 ns period). Clock-to-Q is 0.15 ns, the logic delay is 1.4 ns, setup time is 0.1 ns, and the capture clock arrives 0.05 ns late.
| Check | Formula | Result |
|---|---|---|
| Setup slack | T + skew − (t_cq + t_logic + t_su) = 2 + 0.05 − 1.65 | 0.40 ns, met |
| Maximum clock, no skew | 1 / (t_cq + t_logic + t_su) = 1 / 1.65 ns | 606 MHz |
| Hold slack (t_cq,min 0.1, t_logic,min 0.05, t_hold 0.08) | t_cq,min + t_logic,min − t_hold − skew = 0.15 − 0.13 | 0.02 ns, met but tight |
The same 0.05 ns of skew added to the setup slack and subtracted from the hold slack, which is the point of Q3.
Power
- Q5. Where does power go in CMOS? Answer: dynamic switching power, short-circuit power while both networks conduct during a transition, and static leakage. Dynamic power is P = αCV²f, with α the activity factor, C the switched capacitance, V the supply and f the clock.
- Q6. Why is lowering the supply the strongest lever? Answer: power goes with V squared. Example: α = 0.1, C = 1 nF, V = 0.9 V and f = 1 GHz give 0.1 × 1 nF × 0.81 × 1 GHz = 81 mW; at 0.81 V (10% lower) it is 65.6 mW, 19% less, before slowing the clock at all.
- Q7. What is clock gating? Answer: stopping the clock to registers that don't need to change, which cuts α for them and the power of the clock tree beneath the gate. It uses an integrated clock-gating cell with a latch, so the enable can't glitch the clock.
Devices and circuits
- Q8. Why is the PMOS in an inverter wider than the NMOS? Answer: holes have lower mobility than electrons, so a PMOS of the same size conducts less; widening it (often 2 to 3 times) balances rise and fall times and centers the switching threshold.
- Q9. What is the body effect? Answer: a source-to-body voltage raises the threshold voltage, so a transistor whose source is above its body (such as the upper NMOS in a stack) switches more weakly.
- Q10. Latch or flip-flop? Answer: a latch is transparent while its clock is active; a flip-flop samples only at the edge. Latches are smaller and allow time borrowing, but make timing analysis harder; most designs are flip-flop based.
Q11. Estimate a wire's delay with Elmore
An RC ladder has three sections of 100 Ω, with 10 fF, 10 fF and 20 fF at the three nodes. The Elmore delay to the far end adds each resistance times all the capacitance downstream of it: 100 Ω × 40 fF + 100 Ω × 30 fF + 100 Ω × 20 fF = 4 + 3 + 2 = 9 ps. The upstream resistors count most, which is why drivers are sized up and long wires are broken with repeaters.
Metastability and the flow
- Q12. What is metastability? Answer: when data changes inside the setup and hold window, a flip-flop can sit between 0 and 1 for an unbounded time. The chance it is still unresolved falls exponentially with the time allowed, which is why a synchronizer adds a second flip-flop.
- Q13. What happens between RTL and GDSII? Answer: synthesis to a gate netlist, then floorplanning, placement, clock tree synthesis, routing, and sign-off checks: static timing analysis, power, physical verification (DRC and LVS) and equivalence checking.
- Q14. What does static timing analysis check that simulation doesn't? Answer: every path's timing against constraints, without vectors, across process, voltage and temperature corners. Simulation checks function on the vectors you gave it.
- Q15. What is the difference between DRC and LVS? Answer: DRC checks the layout against the foundry's geometric rules (widths, spacings); LVS checks that the layout's extracted circuit matches the schematic or netlist.
How to answer well
Use numbers when you can: an interviewer who hears 'positive skew of 50 ps adds 50 ps of setup margin and removes 50 ps of hold margin' knows you understand it. Draw the timing diagram for any setup or hold question, and for power questions say which term of αCV²f a technique attacks.
Related
Sources
Weste and Harris, CMOS VLSI Design, 4th edition (power in chapter 5, timing in chapter 10, interconnect and the Elmore delay in chapter 6); Rabaey, Chandrakasan and Nikolić, Digital Integrated Circuits, 2nd edition, chapters 5 and 7.
Last verified
The slack, power and Elmore figures are computed in the site's tests (lib/memoryNumbers.test.js) on 16 November 2026.