Video assemblies for displays, docking and equipment links. HDMI Type A, Type C and Type D plugs, 20-pin DisplayPort and USB-C video, built to your drawing with coaxial high speed pairs, controlled impedance and the rated bandwidth held over the length you actually order.
The connector decides the tooling and the pair count. The bandwidth column is what the assembly is built to hold over its finished length, not what a source at the other end might send.
| Connector | Pins | Typical product | Bandwidth we build to |
|---|---|---|---|
| HDMI Type A | 19 | TV, monitor, console, set-top box | 10.2 / 18 / 48 Gbps |
| HDMI Type C, mini | 19 | Cameras, compact devices | 10.2 Gbps |
| HDMI Type D, micro | 19 | Cameras, embedded links | 10.2 Gbps |
| DisplayPort, 20 pin | 20 | Monitor, dock, daisy chain | 21.6 / 32.4 Gbps |
| USB-C, video alt mode | 24 | Dock, portable monitor | up to 40 Gbps |
The figure is aggregate across the high speed lanes, because that is how the lanes are used. A cable that reaches its top rate only on a short bench sample is a cable sold on a number rather than built to one, which is why we quote the rate against a length.
At video rates the limiting factor is rarely how much copper is in the conductor. It is whether the pair holds its differential impedance along the whole run, and whether the shield keeps the outside world out of a signal that is measured in millivolts.
A twisted pair is good enough at 10 Gbps. Past that, each high speed lane wants its own coaxial construction: a solid dielectric around the conductor, a bonded foil wrap, and a drain path that terminates at both plugs rather than floating at one end.
Two things kill an otherwise correct build. A shield that stops short of the shell, leaving an unshielded gap right where the cable flexes, and an impedance mismatch at the transition from cable to plug. Both are tooling and process decisions, not inspection findings.
What we specify at quotation. The pair construction, the differential impedance window, shield coverage and the drain termination, against the finished length you order. If the drawing names a rate we cannot hold at that length in passive copper, we say so before tooling rather than after the first shipment.
Every rate has a length past which passive copper stops holding. Asking for both is a normal request, and the answer is sometimes a different construction rather than a longer cable.
| Rate | Practical passive length | What changes to go longer |
|---|---|---|
| 10.2 Gbps | up to 10 m | Larger conductor, tighter impedance window |
| 18 Gbps | 3 to 5 m | Heavier shield, more pair isolation |
| 32.4 Gbps | 1.5 to 3 m | Coaxial pairs, tighter bend radius |
| 48 Gbps | 1 to 2 m | An active element or an optical run |
These are working numbers for passive copper, not the ceiling of what the interface can define. Lengths past them are built with an active element, and we will quote that as a different assembly rather than promising the passive one will perform.
| Element | Options | Effect |
|---|---|---|
| High speed pairs | Coaxial, shielded twisted pair, 28 to 34 AWG | Rate ceiling and how well it holds over length |
| Differential impedance | 100 Ω ±10 %, tighter on request | Reflections at the plug transition |
| Shield | Bonded foil, braid, foil plus braid, 85 % braid coverage and up | EMI ingress and crosstalk between lanes |
| Drain termination | 360° crimp to shell, or terminated at one end by design | Whether the shield actually works |
| Jacket | PVC, TPE, TPU, braided nylon over TPE | Flex life, hand feel, retail look |
| Shell | Overmoulded PVC, die-cast zinc, aluminium | Retention, heat spread, perceived quality |
| Contacts | Gold flash, 15µin, 30µin over nickel | Insertion cycles before resistance climbs |
| Bend radius | 4× OD static, 8× OD in flexing use | Whether the shield survives the install |
Connector shells, contact plating and pair construction are the three lines that decide whether a video cable still works in month eighteen. Length and colour are the easy ones.
A video cable usually dies at the two ends, not in the middle. These are the four checks that predict it, and each one is a number you can put on a drawing.
| Test | What it measures | Typical target |
|---|---|---|
| Flex and swing | Conductor fatigue and jacket splitting at the strain relief | 5,000 to 10,000 cycles at ±60° |
| Mating cycles | Contact plating wear and shell retention | 10,000 insertions |
| Pull force | How much load the moulded end takes before it separates | 50 N held for one minute |
| Temperature rise | Heating at the contacts under the rated current | under 30 °C above ambient |
Flex and mating tests are destructive, so they are quoted per request and the report names the sample size. Quoting a cycle count without a test behind it is exactly the kind of claim this page is trying not to make.
The link settles at the highest rate every element in the chain supports, and the cable is only one element. A display that tops out lower, a source with a weaker transmitter, or a repeater in the middle will pull the whole negotiation down. Two cables can measure identically on a bench and behave differently on a long run, because the margin is in the eye pattern, not the spec sheet.
Passive copper is practical at one to two metres. Three metres is the edge of what holds without help, and it depends on the pair construction and the plug. Past that an active element or an optical run is the honest answer. Anyone quoting a five metre passive 48 Gbps cable is either measuring something else or has not tested it on a real sink.
It has to cover the whole signal path and reach the shell at both ends. A shield that stops ten millimetres short of the plug leaves an unshielded window exactly where the cable flexes, which is where the failures come from. Braid coverage above 85 % with a bonded foil underneath is the usual construction for the higher rates.
Yes, if your drawing leaves those conductors out. It reduces the pair count and the cost, and it is a common request for equipment links where the protocol never uses the return path. It also means the assembly will not work with a consumer source that expects that channel, so the decision belongs on the drawing rather than in a conversation after the samples arrive.
We build passive copper in the lengths on this page. For active or optical runs we quote against a module supplier's reference design rather than treating it as a passive cable with a chip added. Tell us the rate and the length and we will say which of the two it has to be.
Send a sample. We measure conductor gauge and pair construction, section the overmould to see what the tooling has to reproduce, and check where the shield terminates. We will tell you where the existing part is under-specified, because copying a design that fails at eighteen months is not a service.
Tell us the connector pair, the rate you need to hold, the finished length and the device it plugs into. We will come back with unit price, the pair construction we would use, and whether passive copper gets there.