Patch cord length planning: sizing cords for a rack build

The cord is part of the channel, not an accessory

Patch cords are usually the last items on a rack build purchase order, and they are usually chosen from whatever is in stock. That ordering hides an arithmetic that decides the acceptance test: the 100 m channel is not 100 m of horizontal cable. It is 90 m of horizontal cable plus a cord at each end, and each of those cords has a limit of its own.

On a 10GBASE-T design the margin is thinner than most people expect. The channel limit for 10GBASE-T over CAT6 is about 55 m, so a length decision made at the end of a project can consume budget the designer assumed was still available. Cord length is a design input, not a packing detail.

Every cord in our range is cut, terminated and tested before packing, and carries the same 100 ± 15 ohm impedance and ≤ 9.38 ohm per 100 m conductor resistance limits as the bulk cable it is made from — see the factory-terminated patch cord range. The channel arithmetic only holds if the cord is measured on the same terms as the cable behind it.

Measure the route, not the distance between two ports

A cord does not travel in a straight line. It leaves the panel, enters a horizontal manager, turns, runs to the switch and turns again. Measuring the straight-line gap between two ports consistently produces cords that are too short: the turns consume length, and a service loop at each end consumes more.

The method we use: run a pull string along the actual path, add the service loop, and round up to the next length on your ladder. Thirty to fifty centimetres per end is enough for re-termination and a later device swap. Each 90° turn adds roughly 1.57 times the bend radius, which is a couple of centimetres per turn — small on one cord, decisive across 48 ports in a cabinet.

Do not round up twice. A cord ordered one size longer than the calculation, just in case, is the most common reason a rack looks tidy on day one and unreadable by year two.

A length ladder that works in most racks

Most panel-to-switch patching falls into five separations. A ladder of 0.3, 0.5, 1.0, 1.5, 2.0 and 3.0 m covers the large majority of cabinets. The table below is a starting point to check against the real routing, not a substitute for measuring.

The pattern that keeps a rack readable is panel, then horizontal manager, then switch, with the cord leaving the manager horizontally instead of crossing the front of the switch. Once the ladder is fixed, the patch-panel and switch schedule can carry a length per row, and the cord order becomes a copy of that schedule.

What an over-length cord costs after handover

Slack is not neutral. A loop of spare cord inside a vertical manager covers port labels, and the label you cannot read is the one somebody needs at two in the morning during an outage. Slack also invites the cord to be dressed into a tighter turn than the minimum bend radius allows, and in a dense rack it blocks the airflow path across the switch face.

There is an electrical cost as well on powered rows. A cord three sizes too long adds conductor length to every link, and a coil of excess cord inside a manager is a thermal problem in a way that a single run is not.

What an under-length cord costs later

A cord that is too short does not fail at handover; it fails after the first device swap. Tension on the plug pulls on the jack contacts and on the panel termination, and the link starts to report intermittent errors that a tester will not reproduce on a quiet rack.

That failure is usually booked against the cable or the termination, because nobody photographs the cord. A service loop of 30 to 50 cm, secured inside the manager rather than draped across the airflow path, costs less than the fault call it prevents.

Bend radius sets the minimum length

For four-pair balanced twisted-pair cable the minimum bend radius is four times the cable diameter, installed, under ANSI/TIA-568.0-D. ISO/IEC 11801 is stricter during installation — eight times the diameter while pulling — and also four times the diameter installed for horizontal runs. A cord that has to turn inside a 1U manager needs enough length to make that turn at that radius.

Shielded cords change the geometry without changing the rack. A CAT6A or CAT7 cord has a larger outside diameter than an unshielded CAT6 cord, so the same manager needs a slightly longer cord, and it is precisely the rack where the customer asked for headroom above 10GBASE-T that can least afford a bend violation.

PoE makes cord length an electrical question too

Conductor resistance is published as a maximum per 100 m, and in this range that limit applies to the finished cord as well as to the reel: no more than 9.38 ohm per 100 m at 20 °C. Thin 28 AWG cords, popular for high-density patching, have higher resistance per metre than the 22 to 24 AWG used in horizontal cable, so they spend the same channel budget faster and drop more volts over the same distance.

Where a row carries PoE, use the shortest rung of the ladder that still routes cleanly, and do not coil spare length inside the manager. Bundle derating for powered cables depends on category, conductor size and the number of cables in the bundle, so confirm the figure for the part number you are buying instead of transferring a number from another supplier's datasheet.

Standardise a short ladder, or order to the as-built list

Two approaches work. The first is a fixed ladder: five or six lengths, enforced, spares held per length, and a rack standard that says which length belongs to which separation. It suits mixed estates and maintenance teams that re-patch often.

The second is an as-built list: for repeat builds of ten or more cabinets with the same internal layout, measure one rack properly and order exactly that length for every position. The saving is not the copper. It is that nobody chooses a length on site, so every cabinet comes out identical and the as-built documentation stays true.

Our cords are made to order: length, AWG and jacket material are fixed at that point, and each cord is cut, terminated and tested to the length list supplied with the order. That makes the as-built approach practical at modest volumes and keeps the cord list inside the quotation rather than inside a stock code. If you are building cabinets rather than patching a floor, the data centre and rack build page covers the rest of the bill of materials.

What to put in the cord order

A cord order that cannot be misread has six fields per line. Anything missing becomes a substitution decision taken by somebody else.

• Category and shield per row — CAT5E, CAT6, CAT6A, CAT7 or CAT8, and UTP, FTP or shielded construction, matching the horizontal cable in the same channel.
• Jacket material per room — PVC inside a closed equipment room, LSZH where the fire strategy or the local code names it.
• Conductor size (AWG) — state it wherever the customer's specification states it, and on any row carrying PoE.
• Colour — only where the customer's documentation or an ANSI/TIA-606 labelling scheme depends on it.
• Length and quantity per position, taken from the schedule rather than from memory.
• Test evidence — which test the acceptance will follow, so the cord report matches the channel report.

One more rule that costs nothing: never put a CAT6 cord into a CAT6A channel. The channel is only as good as its weakest component, and the resulting failure is recorded against the horizontal run, which is the expensive part to replace.

A worked example

A 48-port panel sits above a 48-port switch with three rack units between them, patched through one 1U horizontal manager. The routed path measures 0.45 m. Two turns at roughly 4 cm each add 0.08 m. Service loops of 30 cm at each end add 0.60 m. The total is 1.13 m, so the cord comes from the 1.5 m rung — not 0.5 m, because the loops are part of the route, and not 3 m, because that slack has to be stored somewhere.

Run the calculation once per rack type rather than once per port, and the length list writes itself. On a repeat build it also gives the customer a number they can hold a contractor to.