Level 2 · Construction & Debugging

Intermediate MMMC Interview Questions

Learn to construct, activate, validate, and debug real MMMC environments across Cadence and Synopsys flows.

0 of 10 marked complete

What to practise at this level

Focus on object reuse, activation, coverage, and view-first debugging. Be ready to distinguish universal concepts from vendor-specific commands.

  1. 01 Why can the exact same physical path pass timing in one view and fail in another?Intermediate · Because "the same path" is only the same topology — it isn't the same electrical or timing condition. Change the mode, the corner, or both, and you change the active clocks, the cell delays, the wire delays, and possibly the check itself.
  2. 02 Can one constraint mode be reused with multiple delay corners?Intermediate · Yes, and it usually should be. A constraint mode captures timing intent for one behavior (say, functional mode), and that same intent commonly needs to be checked against several different physical conditions — so it gets paired with several different delay corners to build several different views.
  3. 03 Can one delay corner be reused across several different modes?Intermediate · Yes. A delay corner only describes a cell-plus-wire physical condition — it has no idea what behavior the chip is performing. So the same delay corner can legitimately pair with a functional constraint mode, a scan constraint mode, or a low-power constraint mode to build separate views.
  4. 04 Why is "SS means setup, FF means hold" an unsafe rule to lean on?Intermediate · Because the actual worst-case check depends on the full launch-and-capture relationship, the interconnect condition, temperature behavior (including inversion effects at some nodes), the clock network, and project-specific methodology — not on a two-letter process label.
  5. 05 What's the difference between creating an analysis view and activating it?Intermediate · Creating a view just registers it as an object that exists in the database — it doesn't participate in any check yet. Activating it (declaring it as part of the setup list, the hold list, or both) is what actually makes STA use it during analysis.
  6. 06 A view exists in the database but isn't active for setup or hold. What does that actually mean for signoff?Intermediate · It means that view is contributing nothing to the current timing results, no matter how correctly it was built. If your signoff plan requires that condition to be checked, and the view sits inactive, that requirement is currently unmet — even though nothing looks broken in any report.
  7. 07 Walk through how you'd debug one unexpected timing violation.Intermediate · Start from the environment, not the cell. Confirm which view or scenario is failing, then work outward: is the mode correct, is the SDC correct, is the library/PVT correct, is the RC/parasitic data correct, is variation handled consistently — and only after all of that checks out do you look at the physical path itself.
  8. 08 What's the conceptual difference between a Cadence analysis view and a Synopsys scenario?Intermediate · They solve the same underlying problem — combining behavior and physical condition into one complete timing world — but they are not the same database object, and their commands are not interchangeable.
  9. 09 Why can a path become newly critical only after routing, when it looked fine before?Intermediate · Because pre-route timing uses estimated interconnect, and actual routing can introduce longer detours, extra vias, different layer assignments, more coupling exposure, and different real loading — all of which can push net delay and slew well past what was estimated.
  10. 10 What is scenario or view explosion, and why not just analyze every mathematical mode-times-corner combination during optimization?Intermediate · View explosion is the uncontrolled growth of the Cartesian product of modes, cell conditions, and RC conditions into far more combinations than are actually meaningful. Analyzing all of them burns runtime and memory, produces conflicting optimization goals, and buries real signal in noise — without necessarily improving coverage, since many combinations are invalid or redundant anyway.