What Differential Impedance Actually Is
Differential impedance is the characteristic impedance that a pair of differential lines presents to a high-speed signal, expressed in ohms. It is not the DC resistance you read off a multimeter — it is the impedance the signal feels on the transmission line, set by the cable's geometry. In practice there are only two common targets: 90Ω and 100Ω.
Which one you need isn't a matter of taste. The protocol decides for you.
Where 90 Belongs and Where 100 Belongs
| Interface / Protocol | Target Differential Impedance |
|---|---|
| USB 2.0 / 3.x | 90Ω |
| LVDS | 100Ω |
| eDP / DisplayPort | 100Ω |
| HDMI | 100Ω |
| MIPI D-PHY | 100Ω |
| Gigabit Ethernet (per pair) | 100Ω |
The pattern is obvious at a glance: the vast majority of high-speed differential protocols are 100Ω, and USB is the one conspicuous 90Ω exception. So there's a working shortcut on the floor — see a USB pair, assume 90Ω; for any other high-speed pair, start at 100Ω and then confirm against the datasheet.
Why 10Ω Is Worth Fighting Over
As the signal travels from the chip to the connector and on through the cable, every abrupt change in impedance reflects part of the energy back. The reflection rides on top of the original signal, and the receiver's eye gets chewed up — worse the higher the rate and the longer the link.
A short link is no free pass either. Push the rate up and a 10Ω mismatch can eat what little margin you had, and the result is intermittent bit errors, a display that won't light, or instability when it gets hot — exactly the failures that are hardest to pin down.
How a Cable Hits Its Target Impedance
Differential impedance is set jointly by a handful of structural parameters: conductor diameter and the spacing between the two lines, the dielectric constant of the insulation, the twist pitch, and the distance to the shield. A custom assembly works the problem backwards — fix the target impedance first, then design those features so the finished cable lands near the target, and finally use something like a TDR to verify that impedance stays flat along the length.
In other words, the target impedance is a design input, not a number you measure afterward and hope for.
The One Rule for Selection
- The impedance target comes from the protocol and from the datasheets of the chips on both ends — check the spec first, then design the structure.
- When one assembly carries multiple protocols (a USB pair and an LVDS pair together, say), design each differential pair to its own target impedance; don't flatten them all to one value.
- When in doubt, hand the protocol, the chip part numbers and the data rate to engineering to confirm together — steadier than reaching for a remembered number.
