Following are the Interview Questions for the Post of DFT Engineer. The questions are collected from my friends, who attended the interviews and Some of them are from the interviews that I've attended.
ROUND 1:
- EDTbypass is passing. EDT chain patterns are failing. What could be the reasons?
- (lockup latches not present between edtclock, shiftclock.)
- Two subchips are there. Shifting power is huge. HOw to reduce power dissipation?
- What are the options you have, without affecting coverage and test time? (HW or SW options)
- Setup vioations are there in your design. Which patterns would fail? (s@ or @speed?)
- Hold vioations are there. Which patterns would fail?
- -ve, +ve flops are connected, +ve, -ve flops are connected without lockup latch.
- Which one is better in terms of setup and hold timing?
- Why do we need a lockup latch at all? what happens if we dont use it?
- Two flops will have the same values during shift. Whats the effect of it?
- JTAG : how do you connect tdi, tdo, tms, trst, from board level TAP to all chips on the board.
- explain an algorithm for MBIST.
- Where do we use parallel testbench? Whats the use of parallel simulations?
- TEstpoint? why do we need them?
- If you have simulations failures, what are the things you would look at?
- (setup, libs, clocks, what else?)
- In synthesis, if area is not meeting the target, what would you do?
- If some flops in two clock domains are failing, (S@ patterns), and those failures are expected.
- HOw would you make them pass them on silicon? (no design changes are allowed.)
- Atspeed coverage is always less than STUCK-at coverage. Why?
- P1500? are you aware of it?
- AND gate - optimal input pattern set to detect all S@ faults.
- how to swap two numbers in C (using a single line of code)
- 2**300, 3**200 - which one is bigger?
- explain EDT logic.
- Simulation mismatches you have seen. how did you debug?
- LOC - LOS adv and disadv.
- What is the significance of DFT?
- why atspeed coverage is low than S@ coverage?
- why is target coverage 90% at intel? and 99% at marvell?
- What is that hindering you from achieving more coverage?
- what are the issues you've faced in pattern simulations?
- What is the need to do Timing simulations of ATPG patterns, even after the timing is analysed completely?
- What is needed between two power domains, in the DFT perspective? (level shifters?)
- Two different clock domains are there. How do you target the at-speed faults?
- How do you improve fault coverage?
- What are the issues you've seen in simulations? (design bugs)
- Why do we need OCC?
- Why do we need EDT?
- How do you test multicycle paths at-speed?
- What happens if you don't specify any NCPs?
- What are EDT aborted faults?