HDMI Over Cat6 Extender: How It Works and What to Buy

A rack-mounted source at one end of a conference room, a display at the other, 22 m of cable between them, and a screen that drops to black every time the room warms up. That is the usual entry point into HDMI extenders. Copper HDMI assemblies above roughly 5 m at 18 Gbps start to run out of margin. The fix is not a better cable. It is to stop carrying TMDS or FRL over a long copper HDMI run and hand the signal to a differential pair that was designed for distance.

Why copper HDMI runs out of margin past 5 m

An HDMI cable carries four TMDS pairs at 18 Gbps, or three FRL lanes at up to 48 Gbps, with no forward error correction and no retransmission. Every meter of copper adds attenuation and skew between pairs, and the receiver equalizer has a fixed budget to undo it. Passive assemblies generally hold 18 Gbps to about 5 m with 24 AWG conductors; thinner 28 and 30 AWG assemblies lose that margin closer to 2 m to 3 m. The failure is not graceful. You get sparkles, then intermittent black frames, then HDCP renegotiation that drops the link for two seconds at a time. A thicker cable pushes the cliff a little further out. It does not remove it. The equalizer sits inside the display, so cable and sink behave as one system, not two independent parts.

Single Cat6, dual Cat6, and the chipset you are buying

An extender pair is a transmitter that converts HDMI to a differential signal and a receiver that converts it back. The conversion cost is what separates the product classes. A commodity balun chipset handles 1080p and often 4K30, costs little, and needs a second Cat6 cable once bandwidth climbs. An HDBaseT chipset handles video, audio, Ethernet, control, and power on one cable, which is why it dominates commercial rooms. The cable itself matters less than people expect: unshielded Cat6 is fine for most HDBaseT links, and shielded Cat6A only helps when the route runs beside VFDs or dimmer racks. Buy the chipset, not the marketing tier. Distance ratings and resolution ceilings follow the chipset, not the copper, so compare datasheets at the resolution you plan to run.

HDBaseT versus proprietary chipsets

HDBaseT is a published standard with an interoperability test suite, so a transmitter from one vendor generally pairs with a receiver from another. Proprietary chipsets cost less and often add features the standard does not cover, such as 4K60 4:4:4 over a single cable using a light compression step. The trade is lock-in: a failed transmitter may only be replaceable with the same model. For a building with a ten-year service life, interoperability is worth the price difference.

What dual Cat6 designs do differently

Two cables give the transmitter twice the differential pairs. One cable carries three data pairs plus clock, the other carries the remaining channel and often the return path. That extra bandwidth lets a cheaper chipset carry 4K60 without compression, at the cost of a second pull and a second termination at each end. On a retrofit where one Cat6 already exists, the dual design can still win if the second run is short.

Power, EDID, and control have to travel too

Video is the easy part. The failures that generate support calls are almost always power, EDID, or control. Power over cable removes one wall wart but adds a rule about which end must be fed. EDID decides what resolution the source offers, and a wrong EDID table produces a source that outputs 1080p into a 4K display with no obvious error. IR and RS-232 pass-through keep the display controllable from the room, but only if the emitter is placed over the right window on the display bezel. Plan all three at design time; none of them can be added cleanly after the drywall closes. A receiver with a dead IR window looks exactly like a control processor fault during commissioning, which costs an extra hour.

Power over cable and which end is fed

Most HDBaseT transmitters accept the supply and push power down the same Cat6 to the receiver, which is convenient when the display end has no outlet. Some designs work in reverse, and a few accept power at either end. Check the specification before ordering, because feeding the wrong end on a pair that is not designed for it will simply leave the receiver dead. Budget the power supply at the rack end, where a UPS already exists.

EDID management and the copy switch

An extender needs to tell the source what the display can accept. Copy mode reads the EDID from the display at the receiver and passes it upstream, which is correct for a single display. Learn mode stores a fixed EDID in the transmitter, which is what you want when the display is powered off at source boot, or when a switcher sits in front. Set the switch before the source is connected, because many sources latch the EDID at boot.

Resolution ceilings, latency, and the fiber crossover

Distance and resolution trade against each other in every copper extender. The published numbers assume a clean cable, a single patch field, and a display that accepts the timing. Real routes add punch-downs, patch panels, and a service loop, each of which costs margin. Latency is rarely the constraint: the conversion and buffering add well under 1 ms, which is invisible to a presenter and irrelevant to a control system. The constraint is bandwidth per pair, and it is fixed by the chipset. Once the required resolution and distance both sit near the top of the table, the copper design stops being economical. The table below is a starting point; the vendor datasheet for the exact model is the number that matters at commissioning.

4K60 4:4:4 versus 4K30 over Cat6

Uncompressed 4K60 at 4:4:4 needs 18 Gbps of payload, and a single HDBaseT link carries that to roughly 40 m before the error rate climbs. Many products solve the problem by dropping to 4:2:0 chroma subsampling, which is visually acceptable on film and visibly wrong on thin colored text and CAD lines. If the room displays spreadsheets or drawings, insist on 4:4:4 and accept the shorter distance.

When fiber AOC is the cheaper answer

Hybrid active optical cables carry the same HDMI signal over four fiber strands with copper for power, and they hold 4K60 4:4:4 to 100 m without an extender pair. Above roughly 50 m in a congested pathway, the installed cost of an AOC often beats two devices, two power supplies, and two terminations. The cable is directional and fragile at the connector boot, so plan the pull with that in mind.

Extender classes and what each one actually delivers

Real-World Application Scenarios

Corporate conference rooms are the classic case: a floor box at the table, a display on the front wall, and 18 m of Cat6 already in conduit from a previous AV system. A single-cable HDBaseT pair turns that existing run into a 4K link and keeps the source in the rack. Lecture halls add distance and a second display, so a 100 m receiver at the podium plus a loop-through at the rear screen covers both without a matrix. Retail counters and quick-service restaurants use a short-reach balun to move the media player into a locked cabinet, which removes the easiest theft target in the store.

Related Guides on This Site

If the building already has coax in the walls, compare this approach with HDMI over coax before pulling new cable. Installers who need the pinout and shield details for the cable itself should read our HDMI cable with Ethernet reference, then check the Cat6 distance limit guide before committing to a run length. Rooms that also need peripherals at the far end usually pair the video link with USB over Cat6.

Frequently Asked Questions (FAQ)