The power line channel is the one that fails for reasons nobody wrote down. A surge arrester goes in on the same board. A lighting retrofit lands on the same circuit. Both are perfectly good electrical work, and both can take the network down.
ISO/IEC 14908-3 is unusually direct about this. It declines to specify the medium at all, on the grounds that the power line already exists and nobody writing the standard has any control over how it was installed or what else got plugged into it. What it specifies instead is what a device has to survive.
That makes it the most useful of the four parts if you’re troubleshooting rather than designing.
Table of Contents
The signal
| Band | 125 kHz to 140 kHz, per EN 50065-1 |
| Modulation | BPSK, NRZ data on carrier |
| Carrier frequency | 131,579 kHz ± 0,02 % |
| Symbol rate | 5 482,45 symbols/s ± 0,02 % |
| Timing reference | Not the mains frequency |
That last row is a design rule, not a note. Devices must not use the line frequency for timing or synchronisation. AC power may run the device; it must not clock it.
The 0,02 % tolerances on both carrier and symbol rate are tight. That is a narrowband channel by design, and it’s why the band placement matters so much.
What throughput you actually get
Every byte from the MAC layer goes out as an 11-bit word:
| Bits | Content |
|---|---|
| 1–8 | The data byte, NRZ, uncoded |
| 9 | Even parity over the first 9 bits |
| 10–11 | Always 01 |
Three bits of overhead in every eleven, before framing. Around that sits a bit sync pattern of 24 alternating 10 bits, a word sync word (11001111011), and an end-of-frame made of two EndOfPacket words (11100110011).
Bit sync gives the receiver clock timing. Word sync gives it bit polarity and word boundaries.
Work the arithmetic: 5 482,45 symbols/s at one bit per symbol, times 8/11 for the word encoding, leaves roughly 4 kbit/s of payload before you subtract framing. That is the number to plan around. It is about twenty times slower than the twisted-pair channel.
Timing is a different world
If you have twisted-pair habits, this is the table to internalise:
| Twisted pair | Power line | |
|---|---|---|
| Beta1 | 868 µs transmit | 3,4 ms ± 0,1 ms |
| Beta2 randomising slot | 168 µs | 2,0 ms ± 0,1 ms |
| Priority slots | 4 | 8 recommended |
| Carrier detect | — | 1,7 ms max |
| Transmit start delay | — | 100 µs max |
Beta2 is twelve times wider. With the MAC’s base randomising window of 16 slots, a node waits an average of tens of milliseconds before it transmits — before you add carrier detect on the receiving end.
The practical consequence: do not carry transport timer settings across from a twisted-pair segment. The Xmit_Timer ≥ 3 × packet cycle rule still applies, but the packet cycle is an order of magnitude longer. Retry counts tuned for twisted pair will time out prematurely here.
Coupling rules
One rule dominates, and it’s a safety rule as much as a signalling one:
Signalling goes between phase and neutral only. No functional connection to the protective conductor. Ever. You may connect earth for protective or other functional purposes, but never for signalling.
For multi-phase devices, any or all phases may be coupled with respect to neutral. Cross-phase coupling uses phase couplers per EN 50065-4-1.
If a connector is used rather than a direct connection, it must impose less than 0,1 dB of loss in either direction — from the network to the device, and from the device to the network.
Compliance requires EN 50065-1, plus either EN 50065-2-1 (residential, commercial, light industrial) or EN 50065-2-2 (industrial), depending where the device is going. Check which one a device claims before putting it in a plant.
The five safety hazards the standard names
Worth quoting to anyone who thinks of this as a comms job rather than an electrical one. Power line network cables face at least five direct electrical safety hazards:
- High-energy transients coupled in from external environmental sources
- Differences between the safety grounds that network components connect to
- High voltages on neutral or ground wiring
- Open safety grounds
- High short-circuit current available at the interface
The standard adds that these have to be dealt with not only in the operational system, but specifically during changes to an existing network — the moment when someone opens an enclosure and assumes the earth reference is what it says on the drawing.
What kills the channel
Surge protection
The standard flags this explicitly. Surge protection and frequency-selective protection devices installed on the power network may attenuate the channel waveform enough to prevent operation in part or across the entire network.
The precaution is one sentence: don’t choose a device that substantially attenuates 125 kHz to 140 kHz.
That sentence has to reach the person specifying the protection, who is usually not the person specifying the network. This is the single most common preventable failure on power line segments, and it is invisible until commissioning.
Dimmers
The standard’s own conformance suite tells you what it expects to fight. One of four required receiver tests is immunity to impulsive noise explicitly described as the kind produced by triac-controlled dimmers.
The test waveform is specified precisely: a damped sinusoid with A = 75 V, f = 120 kHz, decay constant b = 2,4 × 10⁵, occurring once every half cycle of the mains at an arbitrary phase offset — generated in practice with a commercial triac dimmer driving a 100 W bulb. A compliant device must hold packet error rate under 2 % with that running.
If a power line segment goes intermittent after a lighting retrofit, this is where to look first. The people who wrote the standard expected dimmers to be the problem.
Notches in the channel
Third receiver test: a series RLC notch of R = 8,5 Ω, L = 150 µH, C = 0,01 µF, producing a 10 dB notch with a Q of 5 centred near 130 kHz — right on the carrier. Devices must hold PER under 2 % through it at a received level of −60 dBV.
Any resonance on the installation that lands near 131,6 kHz does this to you for free.
Broadcast interference
Fourth test, and it explains a class of geographic failure. The standard identifies power lines as antennas for radio broadcasts, and notes European and North American AM broadcast is present at high levels. Devices are tested against interfering tones from 10 kHz to 1 MHz — every 5 kHz up to 150 kHz, then every 50 kHz to 1 MHz — with PER required below 2 %.
Tolerated tone levels tighten sharply as you approach the channel. At 10 kHz a device must cope with +5 dBV. Across 100–145 kHz, immediately around the carrier, the specified level drops to −56 dBV. In the broadcast regions above 150 kHz it relaxes back to −12 dBV.
That −56 dBV figure is the point. In its own band, the channel tolerates almost nothing.
Conformance thresholds worth knowing
The metric throughout is packet error rate over more than 1 000 packets.
| Test | Condition | PER limit |
|---|---|---|
| Quiet line | −60 dBV to ≥ 9 dBV received | < 0,1 % |
| Interfering tone | 10 kHz–1 MHz per level table | < 2 % |
| Distorted channel | 10 dB notch, Q = 5, at −60 dBV | < 2 % |
| Impulsive noise | Triac dimmer waveform, at −60 dBV | < 2 % |
Transmit amplitude, measured through a CISPR 16 V-network with a tuned receiver at 131,5 kHz: above 0 dBV (2,828 V peak-to-peak) and below 11 dBV (10,0 V peak-to-peak) with the test switch open, above −12 dBV (0,7 V peak-to-peak) with it closed.
Receive-mode effective input impedance must be at least 200 Ω.
The dynamic range implied by the quiet-line test is large — from 2,828 mV peak-to-peak up to 8 V. That range is why the channel works at all across an uncontrolled medium.
What the standard deliberately leaves out
No cable specification. No topology rules. No distance limits. No node count.
The reasoning is stated: the medium already exists in any environment using power line communication, and the specification has no control over its installation, physical properties, topology, or what else is connected to it.
So there is no equivalent of the twisted-pair distance table. What you get instead is a defined signal, a defined band, and a set of conformance thresholds. Whether a given installation works is an empirical question, and the honest answer to “how far will it go” is that you measure it.
FAQ
What frequency does LonWorks power line use? The 125 kHz to 140 kHz band from EN 50065-1, with a BPSK carrier at 131,579 kHz.
How fast is it? 5 482,45 symbols per second, which after the 11-bit word encoding leaves roughly 4 kbit/s of payload. Roughly twenty times slower than the twisted-pair channel.
Why did my power line network stop working after electrical work? Most likely a surge protection or frequency-selective protection device attenuating 125–140 kHz. The standard warns this can disable part or all of a network. Second candidate is a dimmer install.
Can I connect the signal to earth? No. Signalling is phase to neutral only, with no functional connection to the protective conductor. Earth may be connected for protective purposes.
How far will a power line segment reach? The standard sets no distance limit, because it specifies no medium. Reach depends on the installation, and you determine it by measurement.
Which EN 50065 part applies to my device? EN 50065-1 always, plus EN 50065-2-1 for residential, commercial and light industrial use, or EN 50065-2-2 for industrial environments.
Does the network care about mains frequency? No, and devices are forbidden from using it for timing or synchronisation. The channel works the same on 50 Hz and 60 Hz.