Quick answer: A passive copper HDMI cord is reliable to about 5 m at 1080p60 and about 3 m at 4K60, because HDMI 2.0 pushes 6 Gbit/s down each of three lanes and copper attenuates that band by several dB per metre. Past those lengths the link either drops to a lower format, shows sparkles, or refuses to handshake. An active optical cord replaces the three high-speed lanes with fiber, keeps copper only for 5 V, EDID and hot plug detect, and reaches 30 m or more at the same bandwidth. Length, not price, is what decides the architecture.
At 4K60 the pixel clock is 594 MHz and the TMDS character rate is 6 Gbit/s per lane, which puts the fundamental energy around 3 GHz. A 28 AWG copper pair loses roughly 0.3 to 0.5 dB per metre in that band, and the receiver equaliser has a finite budget. Once insertion loss and inter-pair skew eat the eye opening, the sink reports a format it can handle instead of the one you asked for.
Where a copper cord stops working
Attenuation rises with frequency, not with resolution
Cable loss is specified in dB per metre and climbs with frequency, so the same cord that passes 1080p60 for 10 m can fail at 4K60 for 4 m. The 18 Gbit/s of HDMI 2.0 needs all three lanes at full rate, and the receiver has to recover a clock from the data itself, which makes jitter tolerance tighter than a simple voltage threshold. Thin 36 AWG cords sold for portability are the first to fail.
Skew and the connector both count
The four pairs inside the cord must arrive within a fraction of a bit period of each other. A poor crimp, a bent pin or a coupler in the middle adds reflection and skew at the same time, which is why two short cords joined by a female-to-female adapter usually test worse than one long cord of the same total length. If a run must be extended, extend it actively, not with a barrel.
What an active optical cord changes
Fiber carries the four high-speed lanes
An AOC HDMI cable converts the electrical TMDS signal to light at the source end, runs four fibers, and converts back at the sink end. Fiber loss at 850 nm is around 3 dB per kilometre, so length stops being the limiting factor; the electronics in the two heads become the constraint, and they draw 200 to 500 mW from the source 5 V rail.
Copper still runs the control lines
EDID, hot plug detect, CEC and the 5 V rail stay on copper because they are low speed and need to work before the link trains. That is why an optical cord is directional: the source head contains the laser driver and the sink head contains the photodiode, and swapping the ends leaves the display without a signal. The AOC cable direction guide covers the marking conventions.
Length map by format
| Format | Lane rate | Practical passive limit | Above that |
|---|
| 1080p60 | 1.485 Gbit/s | 5 to 8 m at 28 AWG | Active copper or AOC |
| 4K30 | 2.97 Gbit/s | 5 m at 28 AWG | AOC from 5 m |
| 4K60 4:4:4 | 6 Gbit/s | 3 m at 28 AWG | AOC from 3 to 5 m |
| 8K60 4:2:0 | 12 Gbit/s | 2 m, certified only | Ultra High Speed AOC |
| Any format, 15 m | Up to 6 Gbit/s | Not achievable | AOC to 30 m |
Direction, power and installation
Check three things before pulling an optical cord through a wall. First, the heads are marked Source and Display; the wrong way round gives a black screen with no EDID. Second, the source port must supply enough current, and some splitters and switchers do not, which is why a powered splitter is safer than a bus-powered one. Third, the bend radius at the head is much tighter than on copper, because the photodiode sits directly behind the connector shell. A HDMI extension cable joined to a long cord is the usual shortcut, and it is also the usual cause of intermittent sparkle.
Practical scenarios
- Projector 12 m from the rack: run a 15 m AOC in a conduit with a pull string, power the source end from the receiver, and leave 30 cm of slack at each head so the fiber is never under tension.
- Conference table to wall plate: keep the in-table run under 2 m of copper and put the AOC behind the wall plate, so a damaged cord is replaced without reopening the table.
- Digital signage in a shop window: use AOC for the 20 m run, but check the player supplies 5 V at 500 mA continuously; signage players that sleep will drop the link and the display will re-handshake at 1080p.
- Gaming PC to a 4K120 TV across a room: 4 m is already marginal at 12 Gbit/s per lane, so specify a certified Ultra High Speed AOC rather than a thin passive cord with a marketing label.
FAQ
Can I join two HDMI cords with a coupler for a long run?
It will often work at 1080p and often fail at 4K60. The coupler adds a discontinuity that reflects energy back down the line and introduces skew between pairs, so the receiver sees a worse eye than the same total length in one piece. If the run must be split, split it at an active device such as a splitter or extender.
Do long cords need external power?
Most AOC designs take 5 V from the source and need no external supply. Long active copper cords and some 8K optical models include a USB power lead because the sink-side retimer draws more than the 50 mA a source port guarantees. If the cord has a USB tail, connect it; the link may train without it and then drop under load.
Why does a long cord show sparkles instead of no picture?
Sparkles are bit errors that the link layer does not retransmit, because HDMI TMDS has no forward error correction. A marginal eye produces occasional wrong characters, which the sink paints as white pixels. The format still reports as locked, so the fix is a shorter or active cord, not a different source setting.
