The Operator's Cookbook Vol. 2, Part 1: The Airband Rig — AM Voice and ACARS on One Dongle

Part 1 of The Operator’s Cookbook, Volume 2, a 14-part second run of complete, copy-paste GopherTrunk builds — one working rig per part. Volume 1 took one box from a $40 P25 starter to a kitchen-sink config. Volume 2 is built on what landed in the project afterwards: an airband rig, paging, NXDN, fleet signalling, alerts, transcription, decryption with your own keys, a phone in your pocket, a band survey, a meter walk and a Raspberry Pi sized honestly. The skeleton is unchanged, with one addition: every recipe says in a plain sentence what has been verified on air and what has not. We open with the build Volume 1 could not do at all: VHF airband, which is AM, and the ACARS data link that shares the band.

TL;DR: Airband is a scan list of mode: am channels under scanner.conventional, driven by one role: voice SDR. An AM channel ignores squelch_dbfs: it opens on the carrier’s carrier-to-noise ratio, squelch_cn_db (default 12 dB; noise alone reads about 3–7 dB), so a gain change does not move it. The voice chain finds the carrier within ±4.5 kHz and recentres on it, so a few ppm of tuning error costs nothing. Add decoders: [acars] to an AM channel on 131.550 MHz and decoded blocks land in the acars_log table, GET /api/v1/acars/messages and the ACARS panel — which needs storage.path. Status: AM voice is on-air verified (#1219); ACARS is verified against acarsdec in both directions, with the reporter’s live run still pending (#1231).

Key takeaways

  • Airband squelch is a ratio, not a level. squelch_cn_db compares the carrier with the channel’s own noise floor; both move together with gain, so the threshold means the same thing after you touch the gain.
  • frequency_hz is the carrier, not the channel name. An 8.33 kHz channel “118.005” is the 118.000 MHz carrier; the ±4.5 kHz channel filter covers 25 kHz and 8.33 kHz spacing alike.
  • ACARS is a tap on a channel you already scan. decoders: [acars] runs on the AM channel’s own IQ, so no dongle is dedicated to it — at the price that a block which starts while the scanner is elsewhere is missed.
  • The two halves have different evidence. The AM chain was confirmed live on aircraft radio; the ACARS decoder is pinned to an independent decoder and real-air blocks, not yet to a live scanner run.

Cheat sheet

Concern What it does Where it lives
The scan list fixed channels, scanned in order on one SDR scanner.conventional[], role: voice
AM channel envelope detection, ±4.5 kHz channel filter mode: am
AM squelch carrier-to-noise in dB, gain-independent squelch_cn_db (default 12)
Off-frequency carrier found within ±4.5 kHz and mixed to centre log composer: AM carrier found off the tuned frequency
Hum under the voice 300 Hz audio high-pass, on by default recordings.fm_audio_highpass_hz
ACARS decode 2400 bit/s MSK off an AM channel decoders: [acars]
Where ACARS lands SQLite log, REST, web panel acars_log, GET /api/v1/acars/messages, /acars
Did it open, and on what level against threshold at the moment of opening log conv: squelch opened

In this post

  • How Volume 2 relates to Volume 1 — two conventions for all fourteen parts.
  • What you’re building — a scan list of AM channels with a data tap.
  • The config — the complete file, every key verified.
  • First run — what healthy looks like — a carrier opening, a block landing.
  • When it doesn’t work — symptom → cause → fix, then the variations.

How Volume 2 relates to Volume 1

Volume 1 is still the place to start: its Part 1 gets a dongle recognised and a first system locked. Volume 2 assumes you can already run gophertrunk sdr list and gophertrunk run -config config.yaml.

Two conventions hold for all fourteen parts. Log lines are printed as Volume 1 printed them — level abbreviated to INF / WRN, timestamp dropped — but message text and key=value fields are exactly what the daemon writes, so they grep. And each recipe uses the project’s four status words literally: synthetic-verified (decodes its own test signals), reference-pinned (matches an independent decoder), capture-verified (decodes a recording off the air) and on-air verified (an operator ran it live and confirmed it). A recipe never rounds one up to the next.

What you’re building

One dongle stepping through a list of airband channels — a tower, a ground frequency — the way a handheld scanner does. When a pilot or controller keys up, the scanner dwells, demodulates AM, and files the transmission as a call in History with a WAV on disk, exactly like the analog FM dwells of Volume 1, Part 6, which this recipe extends. One more entry in the same list sits on an ACARS frequency and decodes the aircraft data link into text: registration, flight number, label, message.

Two things separate airband from Volume 1’s FM channels. Civil aviation voice is amplitude-modulated, so an FM chain on a tower frequency records silence; mode: am swaps the discriminator for an envelope detector that divides by the tracked carrier, which makes the carrier itself the gain reference. And the squelch is not a level. A dBFS threshold is a number about your gain knob, and under automatic gain it screens nothing; an AM channel measures carrier-to-noise from its own spectrum instead. Squelch, Part 7 is the deep dive on that meter and the captures that shaped it.

antenna RTL-SDR role: voice scanner.conventional scan loop: tune next channel Tower → Ground → ACARS → … mode: am squelch squelch_cn_db: carrier ÷ noise, dB dwell until carrier drops + hangtime_ms AM voice chain find carrier ±4.5 kHz → envelope → 3 kHz WAV + History row (am-conv) decoders: [acars] 2400 bit/s MSK → block check acars_log → REST → ACARS panel the data tap runs only on the channel that lists it
One dongle, one scan list: every AM channel opens on carrier-to-noise and records voice, and the channel that lists the ACARS decoder also turns its bursts into log rows.

The shopping list

Item Price (rough) Notes
RTL-SDR Blog V3/V4 ~$35 airband AM is an easy decode, and the squelch does not care about gain
Airband antenna $0–$40 118–137 MHz, vertical; the kit whip at about 60 cm works, a dedicated airband antenna works better
Computer $0 anything from the last decade

Frequencies are per airport — the aviation scanner guide covers where to find them — and the ACARS operator page lists the common data frequencies: 131.550, 131.525, 130.025, 129.125 and 131.125 MHz among them. Airband reception is lawful in many countries and restricted in some; check yours first.

The config

log:
  level: info

storage:
  path: "../data/calls.db"      # the ACARS log is a table in this database

recordings:
  dir: "../recordings"

sdr:
  sample_rate: 2_400_000
  devices:
    - serial: "00000001"        # from `gophertrunk sdr list`
      role: voice               # the scan list drives a voice SDR
      gain: "auto"

scanner:
  conventional:
    - label: "Tower"
      frequency_hz: 118_700_000 # the carrier, not an 8.33 kHz channel name
      mode: am
      squelch_cn_db: 12         # carrier-to-noise in dB; 12 is the default
      hangtime_ms: 1500
      priority: 4
      talkgroup_id: 2147500101  # pin it (Volume 1, Part 6)
    - label: "Ground"
      frequency_hz: 121_900_000
      mode: am
      talkgroup_id: 2147500102
    - label: "ACARS primary"
      frequency_hz: 131_550_000
      mode: am                  # required: acars is refused on an fm channel
      decoders: [acars]
      hangtime_ms: 250          # no voice here: keep the burst "calls" short
      talkgroup_id: 2147500103

audio:
  enabled: true                 # hear dwells on the host speakers

The frequencies are placeholders. The decisions behind the rest:

role: voice. The conventional scanner takes the last voice device in the pool; on an airband-only rig that is your only dongle.

mode: am and squelch_cn_db. An AM channel ignores squelch_dbfs. The scanner averages a short power spectrum of the channel and divides the strongest bin within ±4.5 kHz of the centre by the 20th-percentile bin across ±12 kHz. Receiver noise reads roughly 3–7 dB at any gain; 12 dB opens on a weak but intelligible carrier. Nothing in that is absolute, so gain: "auto" is safe.

frequency_hz. Enter the carrier. An 8.33 kHz channel name is a label: “118.005” denotes the 118.000 MHz carrier. The ±4.5 kHz channel filter passes both spacings.

decoders: [acars]. Accepted only on a mode: am channel. A block lasts up to about 0.9 s, and once the decoder has locked a block’s sync the scanner holds the channel for up to one second so a hop cannot cut it off (Squelch, Part 11 has the detector). The short hangtime_ms is deliberate: a data burst raises a carrier, so it opens an AM call like any transmission, and with no voice to wait for there is no reason to hold the channel after each one.

storage.path. Decoded blocks reach you only through the SQLite log; without it the REST route answers 503 and the panel stays empty.

First run — what healthy looks like

gophertrunk run -config config.yaml

A first run on a new machine stops at the Terms of Service: read them with gophertrunk terms, acknowledge with gophertrunk terms accept. Then the scanner announces how it will tune:

INF conv: scanner LO offset tuning serial=00000001 lo_offset_hz=518000 mode=auto sample_rate_hz=2400000
INF api: listening addr=127.0.0.1:8080 tls=false

At 2.4 MS/s the scanner parks the tuner 518 kHz below each channel and mixes the channel back, so the tuner’s DC spur never sits on your carrier (Squelch, Part 4).

An idle list is silent in the log. These are the lines the daemon printed when the project’s committed real-air AM slice — a 126.400 MHz departure channel, interpolated to 2.4 MS/s and mounted as a baseband.replay tuner — ran through an AM scan-list channel:

INF conv: squelch opened freq_hz=126400000 label=Departure mode=am level_cn_db=60.6 squelch_cn_db=12
INF synthetic call started device=replay-am grant="scanner/am-conv tg=2147500101 src=0 freq=126400000 "
INF recorder: call started device=replay-am wav=…/recordings/scanner/2147500101/20261121_193112_2147500101.wav tg=2147500101 provoice=false vocoder=""

level_cn_db against squelch_cn_db is the whole squelch decision: 60.6 dB of carrier over a 12 dB threshold. The protocol tag is am-conv, the system is scanner, and on your rig device= is your dongle’s serial. When the carrier drops and the hangtime expires, synthetic call ended and recorder: call ended close the call with its duration.

ACARS prints no log line per block; the proof is the log table. A synthetic AM burst carrying one of the real-air blocks pinned in internal/radio/acars/acars_test.go, replayed through the ACARS primary entry, produced this from curl 'http://127.0.0.1:8080/api/v1/acars/messages?limit=1':

[
  {
    "id": 5,
    "received_at": "2026-11-21T19:35:54.141660126Z",
    "mode": "E",
    "address": "PH-BXR",
    "ack": "!",
    "label": "5V",
    "block_id": "4",
    "downlink": true,
    "msg_no": "S53A",
    "flight_id": "KL1681",
    "crc_ok": true,
    "raw_hex": "45aed0c8adc2585215b5d63402d3b5b3c1cb4c31b638318314fc",
    "serial": "replay-acars",
    "frequency_hz": 131550000
  }
]

The same row shows in the web console’s ACARS panel (/acars). A block that needed bits repaired before its check validated carries a corrected count. That run also showed the known limit plainly: the 0.66 s burst opened an ordinary AM call. With the default 1.5 s hangtime it left a 2.17 s recording; with hangtime_ms: 250, 0.92 s — and the block decoded either way.

Status, in one sentence each. AM voice on the scan list is on-air verified: on 29 Sep a reporter on a build with the carrier-tracking fix heard “very good and clear” audio, on frequency, from real aircraft radio (#1219). ACARS is reference-pinned, and capture-verified on one recording — GopherTrunk decodes all seven real-air messages in the acarsdec project’s own test audio to the fields acarsdec prints, and acarsdec decodes GopherTrunk’s synthesised audio — but the reporter’s own capture and live scanner run are still pending (#1231).

When it doesn’t work

Symptom Likely cause Fix
Scan list never dwells; a start-up WARN mentions Voice SDRs no voice device in the pool The line is daemon: scanner.conventional / manual_tune_enabled configured but no Voice SDRs in the pool; skipping. Give the scanner a role: voice SDR; on a combined trunked + airband rig add a second, because the scanner takes the last
Daemon refuses the config, naming the ACARS entry the channel is FM acars needs mode: am
A channel never opens though traffic is audible elsewhere carrier under the threshold, or the wrong frequency Fix the antenna first, then lower squelch_cn_db a little — noise alone reaches about 7 dB, so stay well above that
A channel opens on noise or a neighbour interference inside the channel Raise squelch_cn_db; the advanced guide’s ATIS example uses 15
An INFO line says the AM carrier was found off the tuned frequency the entry or the dongle is 1 kHz or more off Nothing is lost — the chain recentres within ±4.5 kHz — but correct frequency_hz, or set the device ppm
Low hum under every transmission mains or other interference below 300 Hz Leave recordings.fm_audio_highpass_hz at its default 300; it applies to AM too
ACARS panel empty, REST route answers 503 no database Set storage.path
Short AM recordings beside every ACARS row by design today: a burst raises a carrier Keep them short with a small hangtime_ms on the ACARS entry
Scanner parks on one channel for hours a continuous carrier such as ATIS is on the list Take it off, or use the wideband variation below, where nothing scans

Two rows have a history. A reporter’s first on-air captures “tuned best at 123.453, not 123.450”: the transmitter was on frequency, but the carrier sat 3.3 kHz from where the channel filter was centred, and the filter was cutting a sideband. The chain now finds the carrier (the strongest peak within ±4.5 kHz, at least 15 dB over the window’s median) and mixes it to the centre first; the line to grep is composer: AM carrier found off the tuned frequency; tracked and recentred. And in those same recordings most of the decoded energy sat below 300 Hz — hangar interference — which is the hum row.

One caveat: GET /api/v1/scanner reports each channel’s level_cn_db, but on an idle AM channel the figure is seeded while the meter is still filling and can read far below zero. Trust the value in conv: squelch opened.

How this recipe shapes operator practice

  • Stop tuning squelch against gain. On AM the threshold is a ratio; a misbehaving channel means antenna, frequency or interference.
  • Enter carriers, not channel names. The recentring line in the log is how you catch a wrong 8.33 kHz conversion.
  • Treat a scan-list data tap as sampling, not logging. A complete ACARS log needs a receiver that never leaves the data channels.

Variations

  • Continuous coverage of one airport. A role: wideband dongle takes a conventional: list of the same entries, each on its own always-on tap. Nothing scans, so a long transmission on tower never hides ground and a continuous ATIS is harmless. Every channel must fit the dongle’s window around center_freq_hz:

    sdr:
      sample_rate: 2_400_000
      devices:
        - serial: "00000001"
          role: wideband
          gain: "auto"
          center_freq_hz: 119_000_000
          conventional:
            - { label: "Tower", frequency_hz: 118_300_000, mode: am }
            - { label: "Ground", frequency_hz: 119_900_000, mode: am }
            - { label: "ATIS", frequency_hz: 119_450_000, mode: am, squelch_cn_db: 15 }
    

    The same shape centred near 131.3 MHz, with decoders: [acars] on each entry, watches several ACARS channels at once. The advanced guide documents it.

  • A dongle per band. device: "<serial>" on a channel moves it to that voice SDR; each SDR scans its own list in parallel. Per-channel gain and audio_gain_db balance a mixed list.
  • Aircraft positions too. ADS-B at 1090 MHz is a separate receiver on its own dongle — Beyond Voice, Part 8.

Where this goes next

Airband needed a new demodulator and a new squelch. Part 2 needs neither voice nor a scan list: it pins one dongle to a group of paging channels and decodes POCSAG and FLEX off all of them at once — including the fix that made a paging-only box work without a browser tab open.

FAQ

Can an RTL-SDR receive aircraft radio? Yes. Civil airband voice on 118–137 MHz is AM. In GopherTrunk it is a scanner.conventional channel with mode: am on a role: voice dongle; each transmission is recorded and logged as a call.

Why does GopherTrunk use squelch_cn_db instead of squelch_dbfs for AM? A dBFS threshold moves with gain and means nothing under automatic gain control. squelch_cn_db is the carrier’s level over the channel’s own noise floor, so it is the same number at any gain: noise alone reads about 3–7 dB and the default threshold is 12.

How do I decode ACARS with GopherTrunk? Add an AM scan-list channel on an ACARS frequency such as 131.550 MHz with decoders: [acars], and set storage.path. Decoded blocks appear in GET /api/v1/acars/messages and the ACARS web panel, tagged with their channel.

Is GopherTrunk’s ACARS decoder verified on air? Not yet by a live scanner run. It decodes all seven real-air messages in the acarsdec project’s test recording, and acarsdec decodes GopherTrunk’s synthesised audio; the live confirmation on issue #1231 is outstanding.

Why do I get short recordings on my ACARS channel? An ACARS burst is a keyed AM carrier, so it opens the channel’s squelch like a voice transmission and is recorded as a short call. A small hangtime_ms on the ACARS entry keeps each one to the burst plus that hangtime.

Series navigation

Part 1 of 14 · Next → Part 2: The Paging Rig — POCSAG and FLEX on One Dongle