“What distinguishes a senior RTL engineer?” The question is really about answers: what a senior answer contains that a mid-level one omits.
A strong answer names the full lifecycle: architecture, implementation, verification, physical awareness, and bring-up observability. This lesson makes that concrete.
A senior answer has three parts in order. Mechanism first: the one-sentence technical truth. Then trade-offs: the options and the pick. Then evidence: what would prove it.
Mid-level answer | Senior answer adds |
|---|---|
Names the mechanism | Names when the mechanism fails and what replaces it |
Describes the block | Compares options and recommends one under a stated priority |
It simulates cleanly | Names the evidence: sim coverage, STA corners, CDC review, silicon correlation |
The question: “A register-to-register path fails setup by 0.30 ns. What do you do?” The report is the one from Reading a critical path report: arrival 2.10 ns, required 1.80 ns, split into tcq 0.25, logic 0.85 and net 1.00.
The mid-level answer, said in one breath: “It fails setup, so I would upsize the cells on the path, and then it should pass.”
Part | Mid-level | Senior |
|---|---|---|
Mechanism | “It fails setup.” Restates the report. | Slack is required minus arrival: 1.80 − 2.10 = −0.30 ns. Of the arrival, the net (1.00 ns) is the largest piece, more than the logic (0.85 ns). |
Trade-off | “Upsize the cells.” One fix, no cost named. | Placement or buffering attacks the net with no latency change, at some power and hold-margin cost. A pipeline cut attacks the logic but adds a cycle the handshake must absorb. With the net largest, try placement and buffering first. |
Evidence | “Then it should pass.” No check named. | Rerun STA in every required view, check hold as well as setup, report WNS and TNS, and rerun the regression if latency changed. |
Read each row across: the left cell is what the mid-level answer said in that slot, the right cell is what the senior answer put there.
The senior mechanism uses the numbers from the report. That is what makes the trade-off row possible.
The senior trade-off names two options and the cost of each, then picks one and says why.
The senior evidence names checks someone else could run. The mid-level answer gives nothing to check.
Evidence is the sharpest separator. Staff-level answers name trade-offs and evidence together, never one without the other.
Take this mid-level answer and rewrite it with the three parts, in order, in at most four sentences. Question: “Pass a one-cycle pulse from a 200 MHz clock to a 50 MHz clock.” Mid-level answer: “Put it through a two-flop synchroniser.”
Mechanism: the pulse is 5 ns wide and the destination samples every 20 ns, so a two-flop synchroniser can miss it entirely. Trade-off: a toggle synchroniser is small but needs pulses spaced by about three destination edges; a request/acknowledge handshake tolerates any clock ratio but is slower; pick the toggle if the source guarantees the spacing. Evidence: an assertion on the spacing rule, a bench that counts pulses sent against pulses received, and the structural CDC check.
“How do you answer something you do not know?” State the boundary plainly. Then reason from mechanism as far as it goes, and propose the check that would settle the rest.
“Mid-level code works. Why rewrite it?” Working code is not the bar. The senior question is what it costs: timing margin, debuggability, and the next engineer who must change it.
“Give a one-line review verdict.” Either the invariant holds under every legal sequence, or here is the sequence that breaks it. Verdicts without triggers carry no weight.
A weaker answer describes personal diligence: long hours, careful reading. Effort is not a technical argument, and interviewers cannot verify it.
Another weak answer stays inside one block. Senior scope crosses boundaries: what the block assumes, what it guarantees, and what breaks upstream when it is wrong.
The second. Mechanism plus evidence is the senior shape; a bare works-now claim is not checkable.
Naming the evidence that would prove the claim. That habit forces the mechanism to be precise enough to test.