L
Admin
[Technical article / FAQ, rfforum Editorial Team]
Length-match notes often stop at "within X mils." At multi-Gb/s and at GHz aggressors, skew converts differential energy to common mode, which then radiates or couples into RF victims that never appear in a pure SI eye diagram.
1. What skew actually buys you
- Intra-pair skew → common-mode; connectors, vias, and bend asymmetry often dominate over the traced length delta
- Common-mode current on a cable or chassis seam is an EMC problem even when the eye at the RX still passes
- RF coexistence: a "digital" SerDes pair can desense a nearby LNA or GPS patch without failing BER
2. Freeze list before you sign the pair
- Skew budget in time (ps) at the edge rate you ship, plus via/connector asymmetry notes — not only mils on one layer
- Whether common-mode filters / choke / stitching are in the traveler for cable-attached ports
- Nearby RF keep-out and whether you measured desense or only SI
- Stackup: reference-plane changes mid-pair that force mode conversion even with matched lengths
3. Habit
Archive skew, via symmetry, and a common-mode / desense note next to every high-speed pair that shares a board with RF. An eye that passes is necessary, not sufficient, on a mixed-signal RF product.
Discussion: on your last SI-vs-desense fight, did the fix land in skew/via symmetry, common-mode filtering, or RF keep-out — and do you require a common-mode check on pairs that exit the box?
— rfforum 编辑部 / Editorial Team
Length-match notes often stop at "within X mils." At multi-Gb/s and at GHz aggressors, skew converts differential energy to common mode, which then radiates or couples into RF victims that never appear in a pure SI eye diagram.
1. What skew actually buys you
- Intra-pair skew → common-mode; connectors, vias, and bend asymmetry often dominate over the traced length delta
- Common-mode current on a cable or chassis seam is an EMC problem even when the eye at the RX still passes
- RF coexistence: a "digital" SerDes pair can desense a nearby LNA or GPS patch without failing BER
2. Freeze list before you sign the pair
- Skew budget in time (ps) at the edge rate you ship, plus via/connector asymmetry notes — not only mils on one layer
- Whether common-mode filters / choke / stitching are in the traveler for cable-attached ports
- Nearby RF keep-out and whether you measured desense or only SI
- Stackup: reference-plane changes mid-pair that force mode conversion even with matched lengths
3. Habit
Archive skew, via symmetry, and a common-mode / desense note next to every high-speed pair that shares a board with RF. An eye that passes is necessary, not sufficient, on a mixed-signal RF product.
Discussion: on your last SI-vs-desense fight, did the fix land in skew/via symmetry, common-mode filtering, or RF keep-out — and do you require a common-mode check on pairs that exit the box?
— rfforum 编辑部 / Editorial Team
