Skip to content
DenseChip
Learn
PracticeCode labInterview

DenseChip

Navigate

Menu

LearnPracticeCode labInterview

Tools

All toolsSV editorCode repoResume
DenseChip

Practice, courses, and chip-design tools in one focused workspace.

LearnPracticeCode labInterviewToolsLearning paths

© 2026 DenseChip

AboutContactPrivacyTerms
Skip to lesson

Course contents

  • Logic Synthesis Glossary
  • Logic Synthesis Cheatsheet
  • Synthesis Foundations14
  • Logic Optimisation and Mapping20
  • Constraints and Timing15
  • Sequential, FSM and Memory19
  • Datapath Synthesis14
  • Area, Power and Test14
  • Verification, ECO and Handoff15
  • Flows and Debugging15
  • Interview and Judgment11
    • Synthesis Interview Questions
      • Flow and Inference Questions
      • Optimization and Mapping Questions
      • Timing and Constraint Questions
      • Sequential and Datapath Questions
      • Power and Test Questions
      • Verification and ECO Questions
    • Synthesis Design Scenarios
Loading lesson

‹Synthesis Interview Questions
  1. Topics
  2. Logic Synthesis
  3. Interview and Judgment
  4. Synthesis Interview Questions
  5. Power and Test Questions

Part 9 · Interview and Judgment · Intermediate

Power and Test Questions

Power and test questions: dynamic power arithmetic, clock gating, and scan insertion.

On this page

    Dynamic power arithmetic

    Dynamic power is activity times capacitance times voltage squared times frequency. All values below are stated EXAMPLE parameters.

    Diagram
    EXAMPLE parameters: activity 0.2, capacitance 50 fF, voltage 0.8 V, frequency 500 MHz
    P = alpha x C x V squared x f
      = 0.2 x 50e-15 F x 0.64 x 5e8 Hz
      = 0.2 x 50 x 0.64 x 5 x 1e-7 W
      = 3.2e-6 W = 3.2 uW
    second case: 0.1 x 100e-15 F x 1.0 x 1e8 Hz = 1.0e-6 W = 1.0 uW
    EXAMPLE split of the 3.2 uW total: registers 50 percent, combinational 30 percent, clock network 20 percent
    50 + 30 + 20 = 100

    Gating and activity questions

    Gating questions test when the saving is real. The program below proves the two quoted power numbers.

    example.py
    1
    2
    3
    4
    5
    6
    7
    8
    9
    10
    11
    12
    13
    14
    15
    16
    17
    18
    19
    20
    import sys
    
    def power(alpha, c_ff, v, f_hz):
        return alpha * (c_ff * 1e-15) * v * v * f_hz
    
    def close(got, want):
        return abs(got - want) < 1e-12
    
    cases = [
        (power(0.2, 50.0, 0.8, 5e8), 3.2e-6),
        (power(0.1, 100.0, 1.0, 1e8), 1.0e-6),
    ]
    
    for i in range(len(cases)):
        got, want = cases[i]
        if not close(got, want):
            print("FAIL: case " + str(i + 1) + " got " + str(got) + " want " + str(want))
            sys.exit(1)
    
    print("PASS: power cases agree at 3.2 uW and 1.0 uW")

    Question

    Precise answer

    Follow-up probe

    Common wrong answer

    Write the dynamic power formula and evaluate the EXAMPLE case.

    P is activity times capacitance times voltage squared times frequency, giving 3.2 uW for the stated values.

    Which term moves most if voltage drops 10 percent?

    Power scales linearly with voltage.

    What must hold for clock gating to save power?

    The gated registers must hold value while disabled, the enable must be glitch-free, and the tree must stay balanced; savings vary with activity.

    What breaks with a plain AND gate on the clock?

    Gating the clock always saves power.

    Scan and test questions

    Scan questions test what test logic costs the function. Every scan flop carries a mux the functional path must pass.

    Question

    Precise answer

    Follow-up probe

    Common wrong answer

    What does scan insertion change in the netlist?

    Each flop gains a scan mux on its data input, chained serially for shift, adding area and a test mode the function must ignore.

    What does it cost in area terms?

    Scan is free because test logic is removed later.

    Why does shift timing get its own check?

    Shift runs the whole chain in one mode, so its path needs a dedicated constraint apart from the functional clock.

    What fails if shift is left unconstrained?

    Functional timing covers shift automatically.

    Common pitfall

    Gating a clock with ordinary logic. A glitch on a gated clock is a false clock edge, so use only the gating style the flow supports.

    Interview

    Likely probes: the power square law and gating safety.

    • Voltage enters squared, so a 10 percent drop cuts power about 19 percent.

    • The two EXAMPLE cases give 3.2 uW and 1.0 uW.

    • Scan adds one mux per flop plus routing, paid in area and shift timing.

    Check yourself Voltage drops 10 percent with all else fixed. What happens to dynamic power?

    Voltage squared gives 0.81, so power falls about 19 percent.

    Check yourself What does the EXAMPLE split of the 3.2 uW total add to?

    Registers 50 percent plus combinational 30 percent plus clock network 20 percent, which is 100 percent.

    Related topics

    • Verification and ECO

    • Clock gating and low power

    • DFT and scan insertion

    Previous

    133 / 139

    Next: Verification and ECO Questions

    Up next

    Up nextVerification and ECO Questions

    Sign in to save progress

    Practice this lessonReport an error

    See also

    • Flow and Inference Questions
    • Optimization and Mapping Questions
    • Timing and Constraint Questions