5GBASE-T camera interface: bandwidth, cables and distance

If your camera must move more pixels per second than 1 GigE can carry, but it mounts 40 m from the control cabinet on conduit you would rather not re-pull, 5GBASE-T was written for that job. Ratified in 2016 as IEEE 802.3bz, it carries about 600 MB/s of image data over the full Ethernet channel on ordinary CAT 6, powers the camera on the same cable, and needs only a standard Ethernet port in the PC. This guide sets it against 2.5GBASE-T, 10GBASE-T, USB 3, Camera Link and CoaXPress on throughput, distance, power, latency and system cost, works one data-rate example with the arithmetic shown, and commits to a choice for each case.

The decision first

Pick 5GBASE-T when the camera's sustained data rate sits above 1 GigE's roughly 120 MB/s and at or below 540 MB/s — 90% of the usable 600 MB/s, the headroom a production line should keep — and the run is longer than USB 3's 5 m. You get the full channel on CAT 6, Power over Ethernet and no frame grabber.

Pick 2.5GBASE-T, about 300 MB/s, when the installed cable is CAT 5e and re-pulling is off the table. Pick 10GBASE-T when the sensor outruns 600 MB/s: it reaches about 1.2 GB/s, but 55 m on CAT 6 and full length only on CAT 6A. Pick CoaXPress when a line-scan web or a closed loop needs microsecond determinism or more than 1 GB/s, and budget for a frame grabber. Pick USB 3 when the camera sits within 5 m of the host: the cheapest chain, the shortest reach. For the 1 GigE baseline itself, start with GigE Vision camera selection: bandwidth, cables and software.

The numbers behind 5GBASE-T

IEEE 802.3bz arrived in 2016 to deliver 10GBASE-T's benefits at lower cost and power: the 10GBASE-T line coding, scaled down to 2.5 and 5 Gbit/s. You plan with payload, not line rate. Figure about 120 MB/s of usable image throughput per 1 Gbit/s of line rate, so 5GBASE-T hands you roughly 600 MB/s.

The cable rule falls straight out of the Nyquist frequency. 10GBASE-T puts most of its energy below 400 MHz, which is CAT 6A territory. 5GBASE-T halves that to 200 MHz, inside CAT 6's specified 250 MHz, so a full 100 m channel is in spec. 2.5GBASE-T halves it again to 100 MHz, exactly CAT 5e's specified bandwidth. That is the whole cable story: CAT 5e for 2.5G, CAT 6 for 5G, CAT 6A for 10G at length.

Interface comparison at a glance

Two patterns matter more than any single cell. USB 3 starts from the same 5 Gbit/s line rate as 5GBASE-T but delivers about 360 MB/s, because its protocol overhead is far heavier, and it stops at 5 m. And the two grabber interfaces are the only deterministic ones: a CoaXPress card carries at most eight cameras, a Camera Link card two, and both tie the camera to a PC slot.

One stock camera per tier of that ladder, straight from the area scan cameras finder:

All 3 share: Humidity 20% RH to 95% RH (no condensation); Certification CE, RoHS, KC. Values as printed in each product's data sheet; open a part for its drawing and full specifications.

Each step up buys frame rate at full resolution; the interface follows the sensor, not the other way round.

Worked example: 31.4 MP at 17 fps over a full CAT 6 channel

Inputs: a 31.4 MP sensor at 6464 × 4852, 8-bit output at 1 byte per pixel (color cameras stream Bayer at the same byte per pixel, so the arithmetic holds for both), a target of 17 fps, and one existing CAT 6 run at about the full channel length from camera to switch.

The arithmetic, written out:

• 6464 × 4852 = 31,363,328 pixels, about 31.4 MP
• × 1 byte per pixel = 31.4 MB per frame
• × 17 fps = 533 MB/s sustained
• 0.9 × 600 MB/s = 540 MB/s budget — 533 fits, with 7 MB/s of headroom

On slower links the same camera collapses: 120 MB/s caps it at 3.8 fps on 1 GigE, 300 MB/s at 9.6 fps on 2.5GBASE-T, and USB 3's 360 MB/s at 11.5 fps, inside its 5 m limit. Push the sensor to 20 fps and the rate becomes 628 MB/s, over budget; that camera belongs on 10GBASE-T or CoaXPress.

Decision: one 5GBASE-T camera per CAT 6 run — no repeater, no grabber, no separate power drop. How much frame rate a link really delivers at full resolution has its own walk-through in GigE Vision for a high speed camera: how much frame rate fits?.

Power on the same cable

5GBASE-T carries Power over Ethernet, so the data run is also the supply. Budget it in three steps: a standard PoE port delivers about 13 W to the camera after cable losses; PoE+ (802.3at, Type 2) raises that to about 25 W, which covers most area scan cameras; four-pair 802.3bt lifts the ceiling to roughly 51 W for heated sensors and onboard illumination. Above that, run a separate supply and let the copper carry data only.

USB 3 carries 5 V bus power over its short reach. Camera Link and CoaXPress power the camera through PoCL and PoCXP on the grabber cable. Among the high-rate interfaces, 5GBASE-T is the one that feeds a full-size sensor and still leaves a single cable to the cabinet.

Latency, jitter and determinism

The PHY adds about 3 µs of link latency on 5GBASE-T — the same order as 10GBASE-T, against roughly 30 µs for USB 3 and about 4 µs for the point-to-point grabber links. At frame level those numbers vanish beside an exposure of hundreds of microseconds and processing in the milliseconds, which is why Ethernet owns general inspection.

What Ethernet does not give you is determinism: the medium is shared and best-effort, so packet arrival jitters in a way a grabber link's does not. For closed-loop work, budget the entire loop — trigger, exposure, transfer, processing, output — as laid out in Image based control latency: budget the full feedback loop. When cameras must share a clock, enable PTP (IEEE 1588) in the cameras and the switch; the failure modes are covered in Synchronize two cameras at high frame rate.

The commitment: 3 µs link latency suits any inspection driven by frame rate. Specify CoaXPress or Camera Link when the loop needs microsecond-jitter image arrival, as line-scan webs do.

Cable plant rules for 5GBASE-T

• Pull CAT 6 for every 5GBASE-T run and treat the full channel length as the budget: camera patch, horizontal run and panel patch all count.
• CAT 5e is a 2.5GBASE-T cable. Do not plan 5G over it.
• Solid-core cable for the fixed horizontal run; stranded cable only for the short patches at each end.
• Use shielded CAT 6 with shielded jacks where the route passes drives, contactors or welding equipment, and bond the shield through the connector at the cabinet.
• In drag chains, use high-flex cable with locking connectors: the LS-CBL-0199 is a 3 m GigE camera cable built for drag chains, with screw locks, and the LS-CBL-0117 adds a locking-stud RJ45 in a 15 m length for fixed runs.
• On the camera side in vibration or washdown areas, M12 X-coded connectors carry 5G and 10G and lock against vibration.

Long copper runs past machinery are also where the picture dies first, and the diagnosis steps for video signal loss over long Cat5 runs carry over to vision links. Data cables for the camera end of the link, from the camera cables category:

NIC and switch budgeting for several cameras

Budget each full-rate 5GBASE-T camera at 4.8 Gbit/s of capacity. A 10 GigE uplink delivers about 9.4 Gbit/s usable, which holds one full-rate camera with room to grow, or two cameras trimmed to 580 MB/s each. Three 2.5GBASE-T cameras at 300 MB/s fit the same uplink with margin at 7.2 Gbit/s total; a fourth fits only if you cap each at 270 MB/s.

On the PC side, 5GBASE-T cameras enumerate as GigE Vision devices on standard hardware — no grabber, no vendor driver stack. The MV-GE4004 is a quad 2.5G Ethernet adaptor with 4 ports for a 2.5GBASE-T line-up, and the MV-GT1102P is a dual 10GigE frame grabber with 2 ports when the uplinks land in one PC. Enable jumbo frames on NIC, switch and camera before commissioning: small frames cost interrupt load and throughput on every GigE Vision link. The port-by-port arithmetic for mixed camera counts is walked through in GigE camera network planning: switch ports and bandwidth, and the frame grabbers category covers the CXP and Camera Link side of the fence.

When 5GBASE-T is the wrong answer

Above 540 MB/s sustained. A 25 MP sensor at 40 fps streams 1 GB/s, which is 10GBASE-T territory: 55 m on CAT 6, the full run on CAT 6A. The MV-CH250-25TC-F-NF is a 25 MP 10GigE camera on an F-mount, with a fan, built for that tier.

Line-scan webs and hard real-time loops. These need the point-to-point, low-jitter grabber link; four CXP-12 links reach 50 Gbit/s on one card. The trade-offs are laid out in Camera Link frame grabber compatibility for line scan and Frame grabber image acquisition for Camera Link and CoaXPress.

Inside a 5 m bench or cell. USB 3 at about 360 MB/s is the lowest-cost chain, and its high-flex cables survive motion: the LS-CBL-0219 is a 3 m USB3.0 camera cable rated for drag chains. Sensor trade-offs at that reach are compared in USB3 machine vision camera selection: compare sensor trade-offs.

A legacy grabber plant. Camera Link Deca still moves 850 MB/s over 7 m, and when the grabber, slot and cabling already exist, the interface decision was made years ago.

Checks before you order

1. Compute the sustained rate with the formula above and keep 10% headroom: the working budget is 540 MB/s on 5GBASE-T.
2. Match cable to rate: CAT 6 for the full channel on 5G, CAT 5e only for 2.5G, CAT 6A for 10G past 55 m.
3. Place the camera's draw in a PoE class: the 13 W, 25 W or 51 W ceilings.
4. Count ports: one 5G camera per port, pairs onto 10 GigE uplinks at 580 MB/s each at most.
5. Enable jumbo frames and PTP across NIC, switch and cameras before commissioning day.
6. Fix the pixel count the task needs before you specify cable, because pixels set the rate and the rate sets the copper:

With rate, cable and power on one page, build the rest of the chain in the system configurator, browse the camera finder by interface, or bring the sensor and run length to our contact page and we will size the link.