Industry update — On 9 July 2026, EASA and EUROCONTROL gathered more than 100 aviation experts — operators, manufacturers, air navigation service providers, pilots and controllers — for a workshop on Europe’s GNSS interference Action Plan, drawn up after requests from 13 EU Member States and the European Commission to address the growing operational impact of satellite-navigation radio-frequency interference (EASA, 9 July 2026). Whatever a navigation receiver finally computes still has to reach the rest of the aircraft — and on most transport aircraft flying today, that journey happens one 32-bit ARINC 429 word at a time.
In short: the ARINC 429 protocol carries avionics data as 32-bit words — an 8-bit octal label, a 2-bit SDI, a data field, a 2-bit SSM and an odd parity bit — sent over a one-way shielded twisted pair at 12.5 or 100 kbit/s, from exactly one transmitter to as many as 20 receivers.
ARINC 429 is the quietest piece of infrastructure in commercial aviation. No packets, no addresses, no arbitration, no retries — and it has carried air data, navigation and system status between line-replaceable units for nearly five decades. If you are designing avionics equipment, a test bench or a flight-test recorder, you will meet it, and you need to know exactly what a word looks like on the wire before you can decode one. This guide walks the protocol field by field — word format, octal labels, data rates, wiring, electrical levels — and shows how an FPGA IP core turns it into logic you can instantiate.
The short answer
ARINC 429, formally the Mark 33 Digital Information Transfer System (DITS), sends 32-bit words down a one-way shielded twisted pair. Bits 1–8 are the label, an octal code identifying what the word contains; bits 9–10 are the SDI; bits 11–29 are the data field (in BNR words, bit 29 is the sign); bits 30–31 are the SSM, which says whether the value is valid; bit 32 is odd parity. Words go out at either 12.5 or 100 kbit/s in bipolar return-to-zero, separated by a gap of at least four bit times of null — which is why no clock line is needed. One transmitter drives the pair; up to 20 receivers listen. Two-way exchange means two buses.
What ARINC 429 actually is
ARINC 429 is published and maintained by ARINC Industry Activities, now part of SAE ITC, under the title Mark 33 Digital Information Transfer System (DITS), Part 1: Functional Description, Electrical Interface, Label Assignments and Word Formats — the best one-line summary of the standard you will find. It defines a wire, a word, and a table of what the words mean.
It was first released as ARINC 429-1 in April 1978 and has been revised steadily since, according to AIM’s protocol tutorial. It is described in reference literature as the predominant avionics data bus on higher-end commercial and transport aircraft — a position it holds not because it is fast, but because it is trivially deterministic. There is no bus master, no collision domain, no negotiation: a transmitter repeats its words at a fixed rate, forever, and anything that wants the data listens. Three properties define it, and they explain almost every design decision that follows:
- Unidirectional (simplex). A pair carries data one way only. Two boxes that must talk to each other need two buses.
- Broadcast, not addressed. A transmitter does not send to anyone: it puts labelled words on the wire, and each receiver decides whether that label interests it.
- Self-clocking. No clock wire. The bipolar return-to-zero encoding and the mandatory gap between words let a receiver recover timing from the data itself.
The 32-bit word, field by field
Everything on an ARINC 429 bus is a 32-bit word — no variable-length frames, no headers, no negotiation. Bit numbering runs 1 to 32 in transmission order, although diagrams are conventionally drawn reversed, 32 down to 1: a detail that has confused more than one engineer reading a datasheet next to a logic-analyser capture.
| Bits | Field | Width | What it carries |
|---|---|---|---|
| 1–8 | Label | 8 bits | An octal code identifying the type of data in the word (for example, label 203 is barometric altitude from an air data computer). Transmitted most significant bit first. |
| 9–10 | SDI — Source/Destination Identifier | 2 bits | Identifies the intended receiver or, more frequently, the transmitting subsystem — the way one of several identical installed units says which one it is. |
| 11–29 | Data | 19 bits | The payload. In BNR (binary) words bit 29 carries the sign, so the magnitude occupies bits 11–28. BCD and discrete words use the same span differently. Data bits are transmitted least significant bit first. |
| 30–31 | SSM — Sign/Status Matrix | 2 bits | Says whether the value can be trusted, and for BCD words also carries sign or direction. Interpretation depends on the data type. |
| 32 | Parity | 1 bit | Set so that the whole 32-bit word contains an odd number of ones. ARINC 429 channels typically use odd parity. |
The SSM is the field people skip — and shouldn’t
A receiver that decodes the data field and ignores the Sign/Status Matrix is reading numbers it has no right to believe. The two SSM bits mean different things for BNR and BCD words. For BNR, the four states are Failure Warning, No Computed Data, Functional Test and Normal Operation — only the last means the value is live. For BCD, they are Plus/North/East/Right/To/Above, No Computed Data, Functional Test and Minus/South/West/Left/From/Below, so the same two bits double as sign and direction.
That is the whole integrity model of the protocol: parity catches a corrupted word, the SSM tells you whether the source itself believes what it just sent. Both are cheap to check in hardware, and both belong upstream of any use of the value.
Building equipment that has to talk ARINC 429 from an FPGA? Our IP core handles the word format, label filtering and parity, so your logic only ever sees valid data.
Labels and octal numbering
The label is the first thing on the wire and the only routing mechanism the protocol has. Eight bits give 256 possible labels, and they are always written as three octal digits — 000 to 377 — because eight bits split cleanly into three octal groups, which makes a label readable at a glance in a specification table. Every ARINC 429 label list you will ever be handed uses that convention.
Two practical points follow. First, the label is transmitted most significant bit first, while the data field is transmitted least significant bit first. That deliberate asymmetry is a classic source of bring-up bugs: a decoder written with a single bit order produces plausible-looking but wrong labels. Second, a label number is not universally meaningful on its own. Assignments are made per equipment type — which is exactly why Part 1 of the standard is titled “Label Assignments and Word Formats.” Label 203 is barometric altitude from an air data computer; the same octal value on another equipment type may mean something else. Decode against the interface control document of the box you are actually wired to, never against a generic table found online.
Rates, encoding and electrical levels
ARINC 429 defines exactly two bit rates, and they are not mixed:
- Low speed — 12.5 kbit/s, with an allowable range of roughly 12 to 14.5 kHz. Used for slow-changing parameters where bandwidth is irrelevant and tolerance is generous.
- High speed — 100 kbit/s ±1%, a much tighter tolerance. Used where update rates matter.
According to AIM’s tutorial the two rates cannot share a transmission bus: a bus is either low speed or high speed, decided at design time. And note what 100 kbit/s is: a raw line rate. Every word costs 32 bits plus at least a four-bit-time gap, so a high-speed bus tops out around a couple of thousand words per second. Budget an interface in words per second, not in bits.
Bipolar return-to-zero on the wire
The line code is bipolar return-to-zero (BPRZ) with three states: HIGH, NULL and LOW. A HIGH bit drives the differential pair to about +10 V for the first half of the bit cycle then returns to zero; a LOW bit drives it to about −10 V and likewise returns to zero. That 10 V differential is made from roughly +5 V on line A against −5 V on line B. Because every bit returns to null mid-cycle, the receiver gets an unambiguous edge in every bit period; the mandatory gap of at least four bit times of null between words gives it the word boundary. Together, that is what removes the clock wire.
Transmitter output impedance is specified as 75 Ω ±5 Ω, divided equally between lines A and B; a receiver presents a minimum input impedance of 8 kΩ — deliberately high, so extra listeners barely load the bus.
Wiring: one talker, many listeners
The physical medium is a shielded 78 Ω twisted pair. On it sits exactly one transmitter — the source — feeding between 1 and 20 receivers, in a star or bus-drop topology. Twenty is a hard architectural ceiling set by the loading arithmetic above, not a guideline.
The consequence drives port counts and therefore pin budgets. A box receiving from four sources and transmitting to three destinations needs seven physical ARINC 429 channels: four Rx and three Tx, each on its own pair, each with its own line driver or receiver. There is no shared medium to economise on. That is why avionics units advertise their ARINC 429 capability as a channel count — and why an FPGA implementation is attractive, since channels become an instantiation parameter rather than another chip on the board.
How an FPGA IP core implements ARINC 429
Mapping the protocol onto fabric splits cleanly in two, and knowing where the split falls saves a board respin.
The FPGA does the digital protocol. Transmit side: a 32-bit word arrives from the host interface into a FIFO, a serialiser shifts it out label-first, parity is computed and inserted in bit 32, the encoder generates the return-to-zero waveform and the required inter-word gap. Receive side: the incoming stream is oversampled, framed on the null gap, de-serialised into 32 bits, parity-checked, filtered by label so the host only sees words it cares about, and pushed into a receive FIFO that raises an interrupt.
The FPGA does not do the electrical layer. No FPGA I/O bank drives ±10 V differential. An ARINC 429 line driver sits between the fabric and the connector on transmit, and a receiver or level-shifting front end sits on the input side. Plan for those parts — and for their power rails — at schematic time, not after the RTL works in simulation.
What our own ARINC 429 core provides
Our current generic ARINC 429 core, M429GEN, was released in November 2024. Its published characteristics map directly onto the protocol described above:
- AXI-Lite host interface — a standard register port, so the core drops into an AMD or Altera design without custom bus glue.
- RZ line coding, serialiser specified up to 2.5 Mbit/s — comfortably above both standard rates, leaving headroom for the higher-rate variants some programmes ask for.
- Independent Rx and Tx channels, each with its own FIFO, with the TX and RX line counts set by VHDL generics — the channel-count problem solved at elaboration time.
- Multi-label filtering in hardware, so the host is never woken for traffic it will discard.
- Configurable parity — odd, even or none — because test equipment sometimes has to generate deliberately non-conforming words.
- Independent interrupts per channel.
The earlier M429T1R1 macro remains in the catalogue for legacy designs, with a 16-bit host bus and a clock rate up to 70 MHz. IP is delivered as source code — VHDL or Verilog — and DO-254 documentation is available on request. We do not publish LUT, flip-flop or BRAM utilisation for the ARINC core: those depend on your channel count and target family, and we would rather give you a real number for your configuration than a marketing one.
Where ARINC 429 stops being the right answer
The protocol is excellent at what it does, and clearly wrong for several things it is sometimes asked to do:
- Bandwidth-hungry payloads. At 100 kbit/s maximum, ARINC 429 cannot carry video, radar returns or bulk file transfer. Those belong on ARINC 664/AFDX or a dedicated high-speed link. Our comparison of ARINC 429, ARINC 629 and MIL-STD-1553 sets out where each one belongs.
- Command/response traffic. There is no request mechanism and no acknowledgement. If your architecture needs a controller to poll subordinate units and receive a reply, that is MIL-STD-1553’s model, not this one.
- Dense meshes. Every direction needs its own pair, so wiring count — and weight — grows fast. Past a certain number of interconnected units, a switched network wins.
- Long-term capture at scale. Logging many channels for hours is a storage problem, not a bus problem — see designing avionics data recorders on FPGA.
- Certification claims. We provide DO-254 documentation on request; we do not publish a design assurance level for the core. Any DAL claim has to be established for your programme, with your evidence — never inherited from a vendor web page.
In summary
Three things decide whether an ARINC 429 interface works on the bench: decoding the label with the right bit order (MSB first, unlike the data field), honouring the SSM before any value is used, and keeping the ±10 V electrical layer outside the FPGA where it belongs. Everything else — the 32-bit word, the octal labels, the two rates, the four-bit-time gap — is a serialiser, a FIFO and a label filter.
Frequently asked questions
How many bits is an ARINC 429 word, and what are its fields?
Every ARINC 429 word is exactly 32 bits, with no variable-length framing. Bits 1 to 8 are the label, an octal code naming the data type; bits 9 and 10 are the SDI, identifying the source or intended destination; bits 11 to 29 are the data field, with bit 29 acting as the sign bit in BNR words; bits 30 and 31 are the Sign/Status Matrix, which reports validity; and bit 32 is the parity bit. Words are separated on the wire by a gap of at least four bit times of null.
Why are ARINC 429 labels written in octal?
Because the label is an 8-bit field, and eight bits divide neatly into three octal digits, giving the familiar range 000 to 377 for the 256 possible labels. It is a readability convention carried through the standard’s label assignment tables and every equipment interface document. Label assignments are made per equipment type: label 203 is barometric altitude from an air data computer, but always decode against the interface control document of the specific unit you are connected to.
What is the difference between 12.5 kbit/s and 100 kbit/s ARINC 429?
They are the only two bit rates the standard defines, and they cannot share a bus — a pair is either low speed or high speed, fixed at design time. Low speed is nominally 12.5 kbit/s with a wide allowable range of about 12 to 14.5 kHz; high speed is 100 kbit/s with a tight ±1% tolerance. Because each word costs 32 bits plus a minimum four-bit-time gap, budget an interface in words per second rather than raw bits per second.
How many receivers can share one ARINC 429 bus?
One transmitter drives a shielded 78 Ω twisted pair, and between 1 and 20 receivers may listen on it, wired in either a star or a bus-drop topology. Twenty is the specified architectural limit, made possible by receivers presenting a minimum input impedance of 8 kΩ so that each additional listener barely loads the line. Because the bus is unidirectional, two units that need to exchange data in both directions require two separate buses.
Does ARINC 429 use odd or even parity?
ARINC 429 channels typically use odd parity: bit 32 is set so that the total number of ones across the 32-bit word is odd. It is a single-bit integrity check on the word, and it is not the only validity signal in the protocol — the Sign/Status Matrix in bits 30 and 31 tells you whether the transmitting equipment itself considers the value usable, reporting states such as Normal Operation, No Computed Data, Functional Test or Failure Warning. A correct receiver checks both before using the data.
Can an FPGA drive an ARINC 429 bus directly?
Not the electrical layer. An FPGA handles the digital side very well — serialising and de-serialising the 32-bit word, generating and checking parity, producing the return-to-zero waveform and the inter-word gap, filtering by label, buffering through FIFOs — but no FPGA I/O bank produces the ±10 V differential bipolar levels ARINC 429 requires. You need an external line driver on transmit and a receiver or level-shifting front end on the input, budgeted into the schematic and the power rails from the start.
How we write these guides: protocol behaviour comes from the ARINC 429 specification and the published references cited below; claims about our own IP come from our product documentation. Resource utilisation, latency and design assurance level are configuration-specific and quoted for your target, never as a generic figure. Last reviewed 25 August 2026.
Need ARINC 429 channels inside an AMD or Altera FPGA? Send us your channel count, target family and host bus — we will point you at the right core, delivered as VHDL or Verilog source with DO-254 documentation available on request.
- ARINC Industry Activities (SAE ITC) — ARINC 429P1-19, Mark 33 Digital Information Transfer System (DITS), Part 1: the standard itself, covering functional description, electrical interface, label assignments and word formats.
- ARINC 429 — reference article: word structure, SDI and SSM state tables, BNR/BCD formats, odd parity, label bit order and the 78 Ω twisted-pair medium.
- AIM — ARINC 429 Tutorial: bit rates and tolerances, bipolar return-to-zero levels, output and input impedances, the four-bit-time inter-word gap, star and bus-drop topologies and the 1-to-20 receiver limit.
- EASA — Protecting the backbone of aviation: EASA and EUROCONTROL host workshop on Europe’s GNSS interference Action Plan (9 July 2026): the industry update opening this article.
