3.3 ARINC 429 Transmission, Timing and Troubleshooting

3.3 ARINC 429 Transmission, Timing and Troubleshooting

A word that is correctly assembled in an LRU still fails on the flight line if the edges, gaps or loading are wrong. This section is the servicing half of topic 5.4 for ARINC 429: how bipolar return-to-zero actually looks on an oscilloscope, how bit rate and slew rate differ, what the twenty-receiver limit means electrically, and how a dedicated 429 analyser is used without becoming a twenty-first problem.

---

Bipolar return-to-zero (RZ): NULL, HI and LO

ARINC 429 does not use simple NRZ, where a 1 stays at +10 V for the whole bit time. It uses bipolar return-to-zero (BPRZ).

Each bit period is split in half:

• First half — the data state: HI (A − B ≈ +10 V) for a binary 1, or LO (A − B ≈ −10 V) for a binary 0.
• Second half — NULL (A − B ≈ 0 V), regardless of whether the bit was 1 or 0.

The waveform therefore has three levels, not two. A string of 1s is not a continuous +10 V plateau; it is a train of positive pulses separated by NULL. A string of 0s is a train of negative pulses separated by NULL. That return to NULL is what makes the code self-clocking: every bit contains an edge back to zero, so a sink can recover bit timing from the pair without a separate clock wire.

If an oscilloscope shows only two levels, you are either measuring single-ended, AC-coupled so that NULL and a weak HI collapse together, or looking at a shorted pair that cannot develop a differential voltage. Three levels (+10 V, 0 V, −10 V on A − B) are the first health check.

---

Bit times at the two speeds

Bit period is the reciprocal of bit rate:

• 100 kbit/s (high speed): $T_{bit} = 1 / 100,000 = 10\ \mu\text{s}$. Each half-bit (HI/LO pulse, or NULL rest) lasts 5 μs.
• 12.5 kbit/s (low speed): $T_{bit} = 1 / 12,500 = 80\ \mu\text{s}$. Each half-bit lasts 40 μs.

A 32-bit word therefore occupies:

• High speed: $32 \times 10\ \mu\text{s} = 320\ \mu\text{s}$ of signalling, plus the word gap.
• Low speed: $32 \times 80\ \mu\text{s} = 2.56\ \text{ms}$ of signalling, plus the word gap.

An analyser or oscilloscope that reports 10 μs bit widths is looking at a high-speed bus; 80 μs bit widths mean low speed. If you have selected the wrong speed on the analyser, the tool will not lock and you may wrongly condemn the source LRU.

---

Word gap

Words are not packed edge to edge. After bit 32 the pair must remain in NULL for at least four bit-periods before the first label bit of the next word. That idle is the word gap (also called the inter-word gap).

• High speed: minimum gap $4 \times 10\ \mu\text{s} = 40\ \mu\text{s}$, so the shortest legal word-to-word time is about $360\ \mu\text{s}$.
• Low speed: minimum gap $4 \times 80\ \mu\text{s} = 320\ \mu\text{s}$.

The gap may be longer. Many parameters refresh every 20 ms, 50 ms or 200 ms depending on the characteristic; the extra idle is simply more NULL. The four-bit-period figure is a minimum, not a fixed spacing. There is no start bit inside the word: the first HI or LO pulse after a legal gap is the label’s most-significant bit.

A gap that is too short merges two words. Sinks then slice the stream at the wrong bit, every subsequent label looks random, and parity fails in bursts. A source that never returns to NULL (stuck HI or stuck LO) is electrically a different fault — usually a driver or a short — not a gap-programming error.

---

Bit rate versus slew rate

Bit rate is how many bits per second the source emits (12.5 or 100 kbit/s). Slew rate (rise and fall time) is how fast A − B moves between NULL and HI or LO.

Industry teaching values from the electrical specification are typically:

• High speed: rise/fall of the order of 1.5 μs ± 0.5 μs (fast edges, short looms).
• Low speed: rise/fall of the order of 10 μs ± 5 μs (slow edges, longer looms).

Those times are industry electrical figures, not EASA-published constants, but the relationship is an examination idea: low-speed buses are slowed on purpose so reflections on a long, multi-stub loom die before the sampling instant, and so radiated emissions stay down. If you speed up the edges on a long low-speed run, you can pass a 12.5 kbit/s bit-rate check and still fail sinks because of ringing. Conversely, slow, rounded edges on a 100 kbit/s bus eat the 5 μs half-bit until receivers mis-sample.

Do not answer a speed question with a slew-rate number, and do not answer a ringing/EMI question with 12.5 versus 100 kbit/s alone.

---

Bus loading and the 20-receiver limit

The source driver imposes the differential waveform through a 75 Ω ± 5 Ω output impedance, divided between Lines A and B. The cable characteristic impedance is commonly about 60–80 Ω, while each receiving sink is deliberately high impedance. ARINC 429 does not require a separate 78 Ω resistor across the far end of the pair.

A teaching load estimate: twenty receivers at the 8 kΩ minimum effective input impedance appear as 400 Ω in parallel. That is the standard’s minimum aggregate receiver impedance. A twenty-first receiver is outside the stated topology, and extra or poorly repaired stubs add capacitance and reflections that can reduce amplitude and distort edge timing. Sinks may then show intermittent parity errors, missing labels or flickering SSM.

The 20-receiver limit is both a specification cap and a practical maintenance limit. Count every display, computer, recorder interface and analyser. If the bus already has 20 sinks, use only the monitoring method and test point authorised by the aircraft procedure.

---

Open, short and shield faults

An open in one conductor is not always a dead silent bus: some sinks still see a messy, small signal via capacitance and report sporadic labels. A hard A-to-B short is the cleanest electrical picture — the source cannot develop a differential voltage, so the trace sits at NULL. Shield faults are the intermittent ones that appear only in the air or when a particular transmitter is keyed.

After a loom repair, confirm cable type, shield continuity, branch length, splice quality and receiver connections against the wiring manual. Do not add a far-end shunt resistor: ARINC 429 relies on the specified transmitter output impedance and high-impedance sinks.

---

Using an ARINC 429 analyser

A dedicated 429 analyser (sometimes called a bus analyser or DITS tester) is the right tool once you have confirmed three-level differential activity.

Receive (monitor) mode. Connect as a high-impedance sink across A and B, set the bit rate to 12.5 or 100 kbit/s, and observe:

• labels in octal;
• SDI, SSM (Normal Operation / Functional Test / NCD / Failure Warning on BNR words) and parity;
• data in raw binary/BCD and, if a characteristic file is loaded, in engineering units;
• word rate (refresh) and gap;
• a parity-error counter.

If label 203₈ appears with SSM Normal Operation and odd parity at the expected refresh rate, the source and the loom as far as your tap are doing their job; a blank display then points downstream (sink LRU, power, or that unit’s input circuit). If 203₈ is absent on the bus but the air-data computer is powered, the fault is the source driver, its connector, or the pair before your tap.

Generate (inject) mode. On a bench or a bus that the procedure isolates, the analyser becomes a source and injects a known word (for example label 206₈, SDI 00, a chosen airspeed, SSM Normal Operation, odd parity). If the display then shows that airspeed, the sink path works and the aircraft source is suspect. Never generate on a live multi-sink bus in parallel with the real source: that is a second transmitter.

An oscilloscope remains useful for amplitude, gap and slew; the analyser remains useful for label identity, SSM and parity. Use both on an intermittent: the scope catches a shield or loading problem that still produces “almost valid” words.

---

Common Module 5 traps

Memorise this short list; it is where three-option items harvest marks:

1. Parity is odd, not even. Bit 32 makes the total number of 1s in the 32-bit word odd.
2. Twenty receivers, not 32, 16 or unlimited. The twenty-first tap is out of specification.
3. Speeds are 12.5 kbit/s and 100 kbit/s, not 10, 125 or 1 Mbit/s, and not mixed on one pair.
4. Bit rate is not slew rate. Low speed uses slower edges; high speed uses faster edges.
5. One pair, one transmitter. Bidirectional traffic needs two buses.
6. The label’s most-significant bit is sent first. Host-interface documentation may show a bit-reversed stored label octet; do not confuse that representation with the serial word convention.
7. SSM 11 on a BNR word is Normal Operation, not Failure Warning. Failure Warning is 00 on the usual BNR map.
8. A roughly 60–80 Ω cable figure is typical industry material, while the transmitter output is 75 Ω ± 5 Ω; neither implies a separate far-end shunt terminator or an EASA-published exam constant.

Worked timing check: a high-speed source that emits one 32-bit word and then idles for 20 ms is legal. The word takes 320 μs, the minimum gap is 40 μs, and the remaining ~19.6 ms is extra NULL. An engineer who expects words back-to-back at 2778 words/s on every label has confused the theoretical minimum gap with the parameter refresh rate in the characteristic.

When a bus is silent, the efficient order is: confirm power to the source LRU, confirm you are on the correct pair, measure A − B for three levels, set the analyser to the correct speed, then decide whether you have an open/short/shield or branch problem, a dead transmitter, or a sink that is simply not using the labels that are present.