Ethernet stops at 100 metres. NextAccess keeps going — over coax, telephone pair and Cat cabling — because G.hn splits the wire into thousands of adaptive lanes and schedules every subscriber on it. Pick how your building is wired — the page shows what applies.
The curve bends with distance instead of falling off a cliff — long after standard Ethernet has given up.
G.hn works between 2 and 200 MHz — the low end of the coax spectrum, where cable loses the least signal per metre. That is why it reaches hundreds of metres on coax, and shares the cable with TV.
On copper every apartment already has its own pair back to the comms room. Each one is jumpered across at the MDF to its own headend port, so nothing on the wire is shared with the neighbours — every unit gets its pair's full line rate.
| Distance | MIMO · 2 pairs | SISO · 1 pair |
|---|---|---|
| 100 m | ~1,400 Mb/s | ~1,390 Mb/s |
| 300 m | ~900 Mb/s | ~500 Mb/s |
| 500 m | ~420 Mb/s | ~200 Mb/s |
Where two pairs reach the unit, MIMO roughly doubles throughput past ~200 m. MIMO uses 2–100 MHz; SISO uses 2–200 MHz on one pair. Lab figures, same data as the chart above.
Instead of one fast signal, G.hn divides the band into up to 4,096 narrow subcarriers that all run at once. Before carrying traffic, both ends measure every lane and load it with as many bits as it can hold — up to 12.
Pairs in the same riser cable leak signal into each other — the reason DSL slows down as more neighbours connect. Because every unit's pair lands on the same NextAccess headend, VectorBoost manages them as a group.
On coax, a single headend port feeds up to 16 NTUs through ordinary passive splitters. The headend is the traffic controller: every 40–160 ms it broadcasts a schedule telling each NTU exactly when it may transmit, and for how long.
A preamble lets the receiver lock on and re-measure the line. Guard intervals swallow echoes from splitters and connectors. Forward error correction repairs flipped bits without retransmission. Net result: roughly 85–90% of the raw line rate is delivered as data.
One coax port loaded with one to four NTUs, all running the same bidirectional test at the same moment. Raw line rate ≈ 1.85 Gb/s; delivered throughput holds near 1.7 Gb/s as clients are added.
| Active NTUs | Down / NTU | Up / NTU | Port total |
|---|---|---|---|
| 1 | 847 Mb/s | 845 Mb/s | 1,692 Mb/s |
| 2 | ~417 Mb/s | ~424 Mb/s | 1,680 Mb/s |
| 3 | ~269 Mb/s | ~290 Mb/s | 1,676 Mb/s |
| 4 | ~204 Mb/s | ~214 Mb/s | 1,670 Mb/s |
Each colour is one NTU's download + upload; the track is the raw line rate. Lab test: 512-byte IPv4, bidirectional, 1:16 passive splitter, 1–2 m drops, <20 dB attenuation.
An even split only happens when every NTU asks for the maximum at once. Real buildings don't behave like that. Because the schedule is rebuilt many times a second, whoever needs capacity gets the slots nobody else is using — any single subscriber can burst to the full port rate the moment the others go quiet.
DSL was built for long telephone loops from a street cabinet. Inside a building the runs are short, the cable is often coax, and several units may share one riser. G.hn was designed for exactly that.
| VDSL2 (17a, vectored) | G.fast (106/212 MHz) | NextAccess G.hn | |
|---|---|---|---|
| Runs over | Twisted pair only | Twisted pair (coax variants rare) | Coax and twisted pair |
| Spectrum | Up to 17.7 MHz | Up to 106 or 212 MHz | 2–200 MHz |
| Short-run rate | ~100–150 Mb/s down | ~1 Gb/s aggregate | ~1.7–1.8 Gb/s aggregate |
| At ~300 m of pair | ~60–100 Mb/s down | ~300–500 Mb/s aggregate | ~900 Mb/s MIMO · ~500 Mb/s SISO |
| Up / down split | Fixed by the band plan, download-heavy | Set per line by the operator | Re-allocated every cycle to whoever needs it |
| Units per cable | One pair per subscriber | One pair per subscriber | Up to 16 per coax port · one per pair on copper |
| TV on the same cable | — | — | Yes, on coax |
| Customer handoff | DSL modem / router | G.fast modem / router | NTU with Ethernet, any router |
VDSL2 and G.fast figures are typical published ranges (ITU-T G.993.2 and G.9701); results depend on cable gauge, crosstalk and profile. NextAccess figures are from the twisted-pair and coax lab results above.
Send us the cable type and riser lengths — we'll model the per-unit throughput before you commit.