PCB Propagation Delay Calculator
How long does a signal take to cross your trace? Convert trace length to flight time (ps / ns), turn a delay budget into a maximum length, check skew between two traces against a budget (DDR-style length matching), and estimate via delay — for stripline and microstrip on any laminate. 100% client-side, nothing leaves your browser.
Board & geometry
h and w use the same unit (mm recommended) — only their ratio matters for microstrip. These feed every tab below.
Length to time
Time to length
Skew matcher
Both traces are assumed to share the same geometry and laminate above (same delay per unit length). For differential pairs, match within the pair first, then pair-to-pair.
Via delay estimate
Worked examples
Every number below is computed by the same engine that powers the calculator — recompute any of them in the tabs above.
Example 1: 100 mm trace on FR4 stripline
Geometry: symmetric stripline, er = 4.4. Step 1: delay per cm = 33.3564 ps/cm x sqrt(4.4) = 69.9690 ps/cm. Step 2: length = 100 mm = 10 cm. Step 3: total delay = 69.9690 x 10 = 699.69 ps = 0.6997 ns.
Example 2: 100 mm microstrip on 1.6 mm FR4 (w = 3.0 mm)
Geometry: microstrip, h = 1.6 mm, w = 3.0 mm, er = 4.4. Step 1: er_eff = (4.4+1)/2 + (4.4-1)/2 x (1+12x1.6/3.0)^(-1/2) = 3.324932. Step 2: delay per cm = 33.3564 x sqrt(3.324932) = 60.8234 ps/cm. Step 3: 10 cm gives 608.23 ps — about 13 percent faster than the stripline case above.
Example 3: Time to length: 250 ps budget on FR4 stripline
Budget: 250 ps. Delay per cm: 69.9690 ps/cm. Max length = 250 / 69.9690 = 3.5730 cm = 35.73 mm. Any stripline trace longer than 35.73 mm on this stackup eats more than 250 ps of timing budget.
Example 4: Skew matcher: 100 mm vs 105 mm (DDR-style check)
Two FR4 stripline traces, 100 mm and 105 mm: skew = |105-100| mm x 69.9690 ps/cm = 0.5 cm x 69.9690 = 34.98 ps. Against a 50 ps budget: 34.98 ps is inside the budget, PASS with 15.02 ps of margin.
Example 5: Via through a 1.6 mm board
Board thickness 1.6 mm = 0.16 cm, FR4 stripline. Via delay estimate = 69.9690 ps/cm x 0.16 cm x 1.25 = 13.99 ps. Small compared with a 700 ps trace delay, but the stub resonance — not this number — is what usually limits via performance at GHz.
Method & validity
Formulas
Stripline (trace centred between two planes, field fully in the dielectric):
tpd = (1/c) · √εr per unit length.
Microstrip (trace on the outer layer, field partly in air):
εreff = (εr+1)/2 + (εr−1)/2 · (1 + 12h/w)−1/2 (Hammerstad–Jensen),
tpd = (1/c) · √εreff per unit length.
The constant 1/c per centimetre is 33.3564 ps/cm, from the exact speed of light c = 299,792,458 m/s. On FR4 (εr = 4.4) stripline this gives 69.97 ps/cm.
Validity limits
- Hammerstad–Jensen is a quasi-static approximation: good for typical digital geometries (w/h roughly 0.1 to 10). At tens of GHz, dispersion raises the effective permittivity slightly.
- εr itself disperses with frequency — FR4 quoted at 4.4 (1 MHz) often measures 4.0–4.2 at GHz. Use the laminate datasheet value at your operating frequency for critical work.
- Stripline here means symmetric stripline. Asymmetric or broadside-coupled structures need a field solver.
- Loss (tan δ), surface roughness, and fibre-weave skew are not modelled — this is pure time-of-flight.
- The via estimate is a rule of thumb (see its tab), not a substitute for 3D EM simulation.
Boundary with the sibling tool
PCB Impedance Calculator is impedance-first: it finds trace width for a target Z0 and shows propagation delay only as a secondary output. This tool is delay-first: length↔time conversion, skew budgets, and via estimates. Use the impedance tool to size your traces, this one to budget their timing.
Frequently asked questions
What is PCB signal propagation delay?
Propagation delay is the time a signal edge needs to travel along a copper trace — the time of flight from driver to receiver. On FR4 it is roughly 60 to 70 picoseconds per centimetre, so a 100 mm stripline trace delays the signal by about 699.69 ps (0.700 ns). It matters because at high clock rates this delay is a large fraction of the bit period, and unequal delays between traces (skew) can break setup and hold timing.
How is propagation delay calculated?
For stripline, where the trace is buried in dielectric and the field lives entirely in the laminate, t_pd = 33.3564 ps/cm x sqrt(er). The 33.3564 ps/cm is the time light needs for one centimetre in vacuum (1/c, with c = 299,792,458 m/s). For FR4 with er = 4.4 this gives 33.3564 x sqrt(4.4) = 69.97 ps/cm. For microstrip, where part of the field is in air, use the effective dielectric constant from the Hammerstad-Jensen formula, er_eff = (er+1)/2 + (er-1)/2 x (1+12h/w)^(-1/2), then t_pd = 33.3564 ps/cm x sqrt(er_eff).
Why is microstrip faster than stripline on the same board?
A microstrip trace sits on the outer layer with air above it, so its field is only partly inside the laminate. Its effective dielectric constant is lower: for h = 1.6 mm, w = 3.0 mm, er = 4.4, er_eff = 3.3249 instead of 4.4. That gives 60.82 ps/cm versus 69.97 ps/cm for stripline, so the signal travels about 15 percent faster on the surface layer. Same physics, less dielectric loading.
What is a typical propagation delay on FR4?
Symmetric stripline on FR4 (er = 4.4): about 69.97 ps/cm, i.e. 177.72 ps per inch — roughly 5.6 inches per nanosecond. Microstrip on 1.6 mm FR4 with a 3.0 mm trace: about 60.82 ps/cm (154.49 ps/inch). Real boards vary with the laminate's actual er (FR4 is often 4.0 to 4.4 at GHz frequencies), so treat these as estimates, not guarantees.
How do I match trace lengths for a skew budget, e.g. for DDR?
Divide the skew budget by the delay per unit length. Example: a 50 ps budget on FR4 stripline (69.97 ps/cm) allows 50 / 69.97 = 7.15 mm of length mismatch. The Skew Matcher tab does this directly: enter both trace lengths and it reports the skew in picoseconds with PASS/FAIL against your budget. For example, 100 mm versus 105 mm on FR4 stripline gives 34.98 ps of skew, which passes a 50 ps budget.
How long can a trace be for a given delay budget?
Length = budget / delay-per-unit-length. A 1 ns (1000 ps) budget on FR4 stripline allows 1000 / 69.97 = 14.29 cm, i.e. 142.9 mm. A 250 ps budget allows 3.57 cm (35.7 mm). The Time to Length tab computes this for any budget, geometry, and dielectric.
Do vias add propagation delay?
Yes — roughly the delay of a stripline trace as long as the board is thick, plus extra for the pads and any stub. The rule of thumb used here is via delay = board thickness x stripline delay x 1.25. For a 1.6 mm board on FR4: 69.97 ps/cm x 0.16 cm x 1.25 = 14.0 ps. Treat this as an estimate only: at multi-GHz speeds the via's stub resonance and impedance discontinuity usually matter far more than this pure transit time.
Does the dielectric constant change with frequency?
Yes. Laminate permittivity disperses with frequency and also varies between manufacturers and resin content. FR4 is often quoted as er = 4.4 at 1 MHz but can measure 4.0 to 4.2 at GHz frequencies. This calculator uses the single er value you enter, so for critical high-speed work use the laminate datasheet value at your operating frequency. Note the square-root dependence: a 5 percent error in er changes the delay by only about 2.5 percent.
What is the difference between propagation delay and rise time?
Propagation delay is the transit time of the edge along the trace (time of flight); rise time is how fast the edge itself transitions. They are independent: a signal with a 100 ps rise time on a 700 ps trace still arrives 700 ps later. Delay matters when it becomes comparable to the rise time — the common rule is that a trace whose delay exceeds about one sixth of the rise time behaves as a transmission line and needs controlled impedance and termination.
When does propagation delay actually matter in a design?
Whenever the flight time is a meaningful fraction of your timing budget: high-speed serial links, DDR memory (data-to-strobe skew), clock distribution, and length-matched buses. A handy check: if the one-way delay exceeds roughly one sixth of the signal's rise time, treat the trace as a transmission line. For slow control signals with nanosecond rise times and short traces, the delay is negligible and length matching is unnecessary.
Free tool by NRTechStudio