For the ADS-L registry

skyBlip applies for AMT entry 58, and for the exception OGN-Tracker has.

skyBlip builds open-source electronic conspicuity equipment for general aviation. The devices transmit and receive ADS-L 4 SRD-860 air-to-air, receive the ADS-L traffic uplink from ground stations, and receive FLARM-generation traffic. Under ADS-L.4.SRD860.F.2.2 we apply for entry 58 of the Address Mapping Table, manufacturers page 49, and we ask that it carry the whole 24-bit address with no manufacturer prefix, exactly as entry 7 carries OGN-Tracker addresses. A skyBlip would then mint its address inside a space that is ours, whether it was shipped by us or built by its owner.

This page is what accompanies the application to registry@ads-l.aero: who is applying, what the equipment does on the air, why the manufacturer prefix of F.2.2 does not fit this fleet, and one question we would like the registry’s view on.

Where it stands

Manufacturer page 49, AMT entry 58, prefix 0, assigned on 5 October 2026.

The registry gave skyBlip manufacturer page 49, which is AMT entry 58, and manufacturer prefix 0 inside it. The assignment carries a prefix, so it is the structured scheme of F.2.2 rather than the flat 24-bit space we applied for: the 16 bits below the prefix are ours to fill densely, one per unit. The application below is as we sent it on 20 September 2026.

Who we dealt with

Log

Who is applying

An open-source project, built against the published specification.

The firmware is public, so every claim below can be read rather than trusted: the address logic is one file, core/settings/address.cpp, and the wire format another, core/protocol/adsl.cpp. The host test suite generates a behavior index, docs/BEHAVIOR.md, which lists what the firmware is checked to do and is regenerated from the tests themselves. Conformance has a suite of its own, below.

The equipment

Two airborne devices, two ground stations, one radio design.

FLARM-generation traffic is decoded and shown to the pilot. It is never re-transmitted: an airborne skyBlip transmits its own position and nothing else, and the uplink is a ground station function.

On the air

Conformance

The specification is a test suite here, read clause by clause.

Issue 2 has 80 numbered clauses. firmware/test/adsl/ holds at least one case for each of them, named after the clause it reads, and scripts/check_adsl_clauses.py fails the build if a clause has no case or if a case names a clause the document does not contain. Where the specification prints an algorithm or a worked example, the case carries the document’s own version of it rather than restating ours: E.2’s XXTEA, E.3.1’s CRC-24, G.1.6’s exponential encoder and the example tables of G.1.7 to G.1.9.

104 cases, of which 64 run and pass and 40 are skipped. A skipped case is written as the clause requires it and skipped over a one-line note saying why, so what this device does not do is a list rather than a silence. Fourteen of the skips are clauses with nothing executable in them, such as the block diagram and the OSI table, and two are duties of the manufacturer rather than of the device. The rest is the honest part:

The whole list, with the note on every skip, is the suite’s own output: build/skyblip_tests --no-skip --test-case="ADS-L*".

The address

What we do today, and what an entry would change.

Today a skyBlip has no range to draw from, so it mints its own: the 24-bit address is derived from the MCU serial number and transmitted under an AMT entry in the self-minted group, 0 to 4. Because that space is shared with every other tracker that derives its identity the same way, the device moves off the prefixes known to be crowded — 0xD0, 0xDD to 0xDF and 0x11 — and refuses the two addresses that mean “none” and “dead SPI read”, 0x000000 and 0xFFFFFF.

F.2.2 offers a manufacturer a page of the AMT and an 8-bit prefix inside it, and asks that the remaining 16 bits be filled densely. That shape assumes a production line: someone issues address n+1 to the next unit and writes it down. This firmware is GPL-3.0 and builds from source, so a share of the fleet is assembled by its owner, flashed at a kitchen table, and passes through no counter we could run. Such a device mints its address the only way an offline device can, from the MCU serial number, and 16 bits are too few to absorb that: derived addresses collide at even odds by about 300 devices.

So the application is for one AMT entry with the structuring of F.2.2 waived: 6 bits of AMT and 24 bits of address, no 8-bit prefix in between. We are asking for a deviation and say so plainly, but it is one the table already grants three times over. Entries 6, 7 and 8 are flat 24-bit spaces, and OGN-Tracker at entry 7 is a fleet of exactly this shape: devices built rather than bought, each deriving its identity locally, with a public database resolving who is who. Assigned that way, the entry costs the table one page out of 55 and ends the guesswork for every skyBlip ever built, not only for the ones we ship.

The entry we ask for is 58, manufacturers page 49, for one reason only: 58 reads as SB. It is the single piece of vanity in this application and the registry should overrule it without hesitation, since a page issued out of order is a favour and the exception above is the part that matters. Any entry in 9 to 63 works for us.

A 24-bit space does not make collisions impossible either. It makes them rare enough to settle one at a time: about 4,800 derived addresses before even odds of a single colliding pair. What settles them is a register, which is the first of the two answers in the question below.

References

• EASA — ADS-L 4 SRD-860, ED Decision 2022/024/R, Issue 2. F.2.2 sender address and the AMT, G.1 Traffic, G.2 Status.
• skyBlip firmware — github.com/fcatuhe/skyblip. The implementation, its host test suite, and the generated behavior index.

skyBlip  ·  Food for thought