RobustPCBWhere Robust Design Starts
CalculatorsالعربيةUpgradeSign inSign up

What's new

Full changelogEngine 1.26.4

Crosstalk & Spacing

PCB & Copper

How much does this trace couple into its neighbour, and what spacing meets my noise budget?

Inputs

Board profile

Fills Dielectric constant from your board profile.

Microstrip: trace to plane. Stripline: plane to plane (b), trace centred.
The pitch. Edge gap = s − w.
Laminate datasheet, at your signal frequency.
How far the two traces run side by side.
Driver datasheet or IBIS model.
s
1 ns
Your allowance at the victim, as % of the swing.
%
Swing and copper thickness
Driver output swing, from its datasheet.
V
From the fab stack-up.

Results

Near-end crosstalk 

42.9 mV on a 3.3 V swing
1.3%
Pass

Checked: 2 × NEXT ≤ your 5 % budget, the 2 × being a design margin for model error: 2.60 % here, with 35 µm copper
Show working
  1. Effective permittivityε_eff at w/h = 0.15 mm / 0.1 mm = 1.5, ε_r = 4.2, t = 35 µm (Hammerstad and Jensen) = 2.933
  2. Saturation lengthL_sat = t_r × c / (2 × √ε_eff) = 1 ns × 299.8 mm/ns / (2 × 1.7125) = 87.53 mm
  3. Backward couplingK_b = ½ × 0.48 × e^(−0.96 × g/h) = 0.24 × e^(−0.96 × 0.15 / 0.1) = 5.69 %
  4. Near-end crosstalkNEXT = K_b × l / L_sat = 5.69 % × 20 / 87.53 = 1.3 % = 42.9 mV
Spacing for the budget s

centre to centre, edge gap 0.08183 mm
0.2318mm
Saturation length Lsat

not saturated: the 20 mm run is 23 % of it
87.53mm
3-W spacing 

centre to centre, Montrose's rule of thumb
0.45mm
All results (5)
Backward coupling coefficient Kb

at this edge gap
5.69%
Johnson and Graham estimate 

75.4 mV on a 3.3 V swing, a second estimate
2.28%
Johnson and Graham spacing 

centre to centre, for the same budget, a second estimate
0.2853mm
Edge gap g

s − w
0.15mm
Effective εr εeff
2.933
NEXTJohnson and Graham

Accuracy

Source
  • Bogatin
  • Simonovich
  • Johnson and Graham
  • Hammerstad and Jensen
  • Montrose
  • IPC-2141

Bogatin, Signal and Power Integrity, Simplified the equation

Simonovich, I-Connect007 column on backward crosstalk the equation

Johnson and Graham, High-Speed Digital Design: A Handbook of Black Magic the equation

Hammerstad and Jensen, Accurate Models for Microstrip Computer-Aided Design the equation

Montrose, Printed Circuit Board Design Techniques for EMC Compliance the equation

IPC-2141, Design Guide for High-Speed Controlled Impedance Circuit Boards fitted to measurements

Precision
Three independent 2-D field solves at w 0.15, h 0.1, s 0.3 mm, ε_r 3.0 give K_b of 6.2 to 6.4 % with 35 µm copper and 4.5 to 4.7 % at zero thickness, so this estimate is about 11 % low to 25 % high there.

That was checked at one geometry, g/h = 1.5. It ignores copper thickness, so its error can go either way, and the 2 × in the verdict absorbs it. On microstrip past g/h = 2 the fit understates the coupling, and the page shows Unverified there.

Verified against

3 independent cases. See the working.

This is a design aid. The engineer remains responsible for the design and for checking the standard itself.

Parameter sweep

Vary one input over a range and see the answer and verdict at each step, as a table and a curve.Pro

Worst-case corners

Put a tolerance on each input and get the worst-case band around the answer.Pro

Cross-section

The principle

An edge on the aggressor trace induces noise on the victim beside it. At the victim's near end the noise grows with the coupled length until that length reaches the saturation length, half the distance the edge travels in its rise time, and then stays at Kb times the swing. This page estimates Kb from the same edge-coupled pair formula Controlled Impedance uses, judges the near-end noise against your budget, and gives the spacing that meets it.

Backward coupling
Kb=14(CmC+LmL)K_b = \frac{1}{4}\left(\frac{C_m}{C} + \frac{L_m}{L}\right)
Bogatin. The near-end noise of a saturated coupled run, as a share of the aggressor swing.
Weak coupling
Kb≈12 Ze−ZoZe+ZoK_b \approx \frac{1}{2}\,\frac{Z_e - Z_o}{Z_e + Z_o}
Simonovich, first order in the coupling. Without the ½ the ratio is twice the near-end noise.
From the pair impedance
Kb≈12⋅0.48 e−0.96 g/hK_b \approx \frac{1}{2}\cdot 0.48\,e^{-0.96\,g/h}
IPC-2141 writes the edge-coupled pair as Zdiff = 2Z₀(1 − k), so Zo = Z₀(1 − k). This page takes Ze ≈ Z₀(1 + k), which gives Kb = k/2. For stripline, Kb ≈ ½·0.347·e^(−2.9 g/b).
Saturation length
Lsat=tr v2,v=cεeffL_{sat} = \frac{t_r\, v}{2}, \qquad v = \frac{c}{\sqrt{\varepsilon_{eff}}}
Bogatin. εeff is the Hammerstad and Jensen value Controlled Impedance uses. A stripline's is εr.
Near-end crosstalk
NEXT=Kb⋅min⁡(1,lLsat)\t{NEXT} = K_b \cdot \min\left(1, \frac{l}{L_{sat}}\right)
Bogatin. The verdict asks for 2 × NEXT within your budget: the 2 × is a design margin for model error.
Johnson and Graham estimate
NEXT≈K1+(D/H)2\t{NEXT} \approx \frac{K}{1 + (D/H)^{2}}
Johnson and Graham, for traces over a solid plane. Their K is below 1 and depends on the rise time and the coupled length. This page sets K = min(1, l/Lsat). Microstrip only, shown as a second estimate.
  • KbK_bbackward (near-end) coupling coefficient
  • Cm,LmC_m, L_mmutual capacitance and inductance per length
  • C,LC, Lself capacitance and inductance per length
  • Ze,ZoZ_e, Z_oeven-mode and odd-mode impedance
  • ggedge gap, s − w
  • ssspacing, centre to centre (the pitch)
  • hhheight to the plane (microstrip)
  • bbplane-to-plane spacing (stripline), entered as the dielectric height
  • trt_r10–90 % rise time of the driver
  • vvspeed of the edge on the line
  • εeff\varepsilon_{eff}effective dielectric constant
  • llcoupled length
  • D,HD, Hspacing and height, in Johnson and Graham
More detail

What the verdict checks

Pass when twice the near-end crosstalk fits your budget, Check when only the crosstalk itself fits, and Fail when it does not. The 2 × covers the model's error, about 11 % low to 25 % high against field solves at ε_r 3.0. That was checked at one geometry, g/h = 1.5. On microstrip past g/h = 2 the fit understates the coupling, so the page shows Unverified there.

Far-end crosstalk

Not computed here. On microstrip it grows with the coupled length and with faster edges. In stripline, in one uniform dielectric, it is ideally zero.

The 3-W rule

Montrose's rule of thumb puts trace centres three widths apart. It is a starting point, not a noise figure: the spacing above comes from your own budget.

This is one aggressor and one victim. Several aggressors add. The estimates here are for a victim terminated at both ends, the case Bogatin's model describes. With a driver at its near end and an open receiver at its far end, the noise reflects, and the receiver sees a different amount. Simulate that case. For a critical bus, confirm the coupling with a 2-D field solver and your fabricator's stack-up.
References: E. Bogatin, Signal and Power Integrity, Simplified (Kb, the saturation length). B. Simonovich, I-Connect007 (Kb from the even- and odd-mode impedances). IPC-2141, the edge-coupled pair impedance, as Controlled Impedance uses it. H. Johnson and M. Graham, High-Speed Digital Design. E. Hammerstad and Ø. Jensen (εeff). M. Montrose, Printed Circuit Board Design Techniques for EMC Compliance (the 3-W rule). Related: Controlled Impedance & Delay, Trace Inductance & di/dt.

Engine version ⁨1.26.4⁩