RTL Design / Module 5: Sequential Design from Scratch
Lesson 5.5

Sync vs async reset — industry choice & why

Reset is not an afterthought — it's a team-wide convention that affects timing closure, DFT scan insertion, CDC, and whether your chip boots reliably. This lesson compares asynchronous and synchronous reset in depth so you can read any company's style guide and know why they chose what they chose.

Core idea

Async reset can clear flops immediately, independent of clock — but deassertion must be synchronized. Sync reset clears only on a clock edge — simpler timing, but reset must be held until the next clock.

1

Side-by-side comparison

AspectAsynchronous resetSynchronous reset
Assert timingImmediate — no clock neededNext posedge clk after rst asserted
Deassert timingMust sync — recovery/removal riskClean — deasserts on clock edge
Sensitivity listposedge clk or negedge rst_nposedge clk only
STA constraintsRecovery & removal checks on rstReset is just another sync input
DFT / scanScan chains need async reset handlingEasier scan insertion — reset is sync
Power-onChip enters known state before first clockFlops may be X until first clocked reset
Industry trendAssert async, deassert sync (hybrid)Pure sync in many FPGA flows
Async vs sync reset timing

Async assert is safe; async deassert near a clock edge causes recovery/removal metastability.

Analogy

Async reset is an emergency stop button — works even if the conveyor (clock) is off. Sync reset is a scheduled shutdown — only takes effect when the conveyor reaches the marked stop point. You want the emergency stop to work instantly, but you want the "all clear" signal to arrive on a schedule.

2

RTL templates — same polarity, different lists

async_reset_ff.v
// Async assert, async deassert (needs synchronizer at chip level!)
always @(posedge clk or negedge rst_n) begin
  if (!rst_n)
    state <= IDLE;
  else
    state <= next_state;
end
sync_reset_ff.v
// Pure synchronous reset — rst_n is just another input
always @(posedge clk) begin
  if (!rst_n)
    state <= IDLE;
  else
    state <= next_state;
end
Trap

Mixing async and sync reset styles in the same design without documentation causes STA failures and simulation/synthesis mismatches. Follow the project style guide — one convention per chip.

3

DFT implications

Scan test chains shift test data through every flip-flop. Reset behavior during scan mode matters:

  • Async reset on scan FFs: Must be controllable in test mode — often gated or overridden by scan_reset
  • Sync reset: Scan insertion tools handle it cleanly — reset is applied synchronously during capture
  • Reset tree: Large designs buffer reset through a tree; async trees have skew that sync deassert must account for
DFT hook

Leave a test_mode or scan_en signal that can override reset polarity during ATPG. DFT engineers will ask for this in review — design it in from the start.

4

Timing: recovery & removal

STA checks two reset-specific margins:

CheckWhat it meansFails when
RecoveryReset deasserts far enough before clock edgerst_n rises too close to posedge clk
RemovalReset deasserts far enough after clock edgerst_n rises too soon after posedge clk

Why sync deassert wins: A reset synchronizer (Lesson 5.6) guarantees deassertion happens away from the clock edge, eliminating recovery/removal violations chip-wide.

5

Team conventions — what you'll see in industry

ConventionWho uses itRationale
Async assert / sync deassertMost ASIC teamsFast boot + clean timing
Pure sync resetMany FPGA projectsSimpler tools, BRAM-friendly
Active-high rstSome legacy IPMatch polarity in wrapper — don't fight it
Power-on reset (POR) + soft resetMobile SoCsPOR handles power-up; software triggers sync soft reset

Interview tip: When asked "async or sync reset?" — answer "async assert, sync deassert with a reset synchronizer" and explain recovery/removal. That shows production awareness.

6

Reset strategy checklist

  • Know your project's reset polarity (rst_n vs rst) and stick to it
  • Never async-deassert reset into logic without a synchronizer
  • Document which blocks use async vs sync reset in the micro-arch spec
  • Verify reset values of every register — don't leave FSMs in illegal states
  • Include reset in your self-checking testbench (Lesson 5.7)
  • Check DFT requirements before taping out scan chains

🎯 What to remember

  • Async reset asserts immediately; deassertion is the danger zone for timing.
  • Sync reset simplifies STA and DFT but needs a clock to take effect.
  • Industry default: async assert + sync deassert via reset synchronizer.
  • One reset convention per chip — document it and lint against violations.
Advertisement