SAME — Specific Area Message Encoding

Implementation: app_utils/eas_fsk.py (encoder primitives) · app_utils/eas.py (EASAudioGenerator, header construction, attention tone, EOM) · app_utils/eas_decode.py (offline decoder) · app_utils/eas_demod.py (live AFSK demodulator)

Standards: FCC 47 CFR §11.31 (regulatory definition) · NRSC-4-B §4 / Annex D (industry standard, summarised in ../NRSC4B_SAME_STANDARD.md)

SAME (Specific Area Message Encoding) is the audio-layer signalling protocol used by the U.S. Emergency Alert System (EAS) and NOAA Weather Radio (NWR) to convey machine-readable alert headers. Every EAS activation begins with a SAME header burst, is followed by an audio attention tone and a voice (or synthesised) message, and ends with a SAME End-of-Message (EOM) burst.

EAS Station™ produces SAME audio for every outgoing broadcast — auto- forwarded CAP/IPAWS, OTA EAS relay, and operator-authored manual events — and consumes SAME audio off-air via the SDR / line-in receivers, decoding it back into structured alerts.

1. Physical layer

The 520 5⁄6 baud rate produces exactly 4 mark-cycles or 3 space-cycles per bit period (1.92 ms/bit), giving the AFSK waveform integer-cycle symbols that align cleanly at every bit boundary — this is what makes SAME decodable on minimal receivers like NWR-only weather radios.

EAS Station™'s renderer (generate_fsk_samples) is phase-continuous: phase is carried across every symbol boundary so there are no glitches at mark-to-space transitions. The fractional-bit timing (samples_per_bit = sr / 520.83…) is handled by a per-bit carry accumulator so the integer sample count rounds out correctly across many bits — at 16 kHz, the mean is ~30.72 samples/bit but individual bits round to 30 or 31 to maintain the exact 25000/48 baud rate over the burst.

2. Burst structure

A complete SAME header transmission consists of three identical bursts, each repeated for redundancy. A receiver decodes each burst independently and accepts the message if any one (or, for stricter implementations, two of three) decodes cleanly with a valid checksum-equivalent comparison.

Each burst is a continuous bit stream made of three concatenated parts:

2.1 Preamble

16 bytes of 0xAB (binary 10101011), transmitted LSB-first with no framing. As bits this is 1101 0101 repeated, which alternates roughly every bit and gives the receiver a strong clock-recovery signal at the half-baud rate. After the preamble the receiver is locked and ready to sample data bits at the correct phase.

SAME_PREAMBLE_BYTE = 0xAB, SAME_PREAMBLE_REPETITIONS = 16 in app_utils/eas_fsk.py.

2.2 ASCII header

Exactly the format mandated by FCC 47 CFR §11.31(c). Fields are hyphen-delimited; the entire string is terminated with a single ASCII carriage return (0x0D).

ZCZC-ORG-EEE-PSSCCC+TTTT-JJJHHMM-LLLLLLLL-<CR>

Each character is encoded as:

• 7 ASCII data bits, LSB-first, followed by
• One null bit (zero) — the eighth bit of the byte, which the FCC rule specifies as "either 0 or 1" but is conventionally always 0.

This is the responsibility of encode_same_bits in app_utils/eas_fsk.py. Crucially there are no start or stop framing bits between characters — the bit stream is continuous within the burst.

2.3 ENDEC fingerprint (optional, EAS Station™ extension)

Immediately after the carriage return, EAS Station™ optionally transmits 3 copies of the byte 0xA9 (binary 10101001, 24 mostly-alternating mark/space pulses, ~46 ms). This is not part of the SAME standard — it is a common ENDEC-vendor convention for station fingerprinting, used to identify the originating ENDEC model from the audio waveform. It is controlled by the endec_fingerprint setting in EASSettings and implemented via encode_terminator_bits(0xA9, 3).

Because it falls after the SAME <CR>, every standards-compliant decoder treats it as inter-burst silence/noise and ignores it.

3. Attention tone

Between burst 3 of the header and the start of the voice message, EAS broadcasters emit an 8-second to 25-second dual-tone "attention signal":

EAS Station™ emits the standard 853 + 960 Hz mix for 8 seconds by default (configurable via attention_tone_seconds). NOAA Weather Radio uses a single 1050 Hz "warning alarm tone" instead; the codebase supports both forms via the _tone_freqs selector in app_utils/eas.py.

The attention tone is not part of the SAME signalling — its purpose is to alert humans (it survives FM broadcast EQ and squelch tails). SAME decoders typically use the gap before the tone as their detection window for the EOM-or-tone classifier.

4. End-of-Message (EOM)

Three identical EOM bursts terminate the broadcast:

Each EOM burst has the same 16-byte 0xAB preamble as the header bursts, followed by the ASCII string NNNN. The EOM has no carriage return terminator — FCC §11.31 specifies only the four ASCII characters. EAS Station handles this distinction via the include_cr parameter to encode_same_bits (default True; set False for EOM).

After EOM 3, ~1 second of trailing silence is appended before the post-alert signal (if any) is rendered.

5. Composite audio assembly

Putting it all together, every EAS Station™ broadcast follows this exact audio timeline, assembled in EASAudioGenerator.build_files:

The "voice message" segment can include per-broadcast pre_alert_samples and post_alert_samples operator uploads bracketing the narration — distinct from the system-level pre/post-alert signals described in ALERT_SIGNALS.md. All audio is finally serialised to a 16 kHz (default) signed-int16 mono WAV by samples_to_wav_bytes.

6. Origination → broadcast pipeline

EAS Station™ builds SAME audio from one of four input paths, all converging into EASAudioGenerator.build_files:

All four paths funnel into the same SAME header construction:

ZCZC-ORIG-EEE-PSSCCC+...+PSSCCC-TTTT-JJJHHMM-LLLLLLLL-<CR>

…with ORIG, EEE, the location code list, validity, issue time, and station ID populated from EASSettings plus the alert-specific fields.

7. Reception path

Live SAME demodulation lives in app_utils/eas_demod.py. The pipeline is:

Field parsing and validation share code with app_utils/eas_decode.py — the offline decoder used for bench testing and forensic replay of recorded captures. app_utils/eas_codes.py and app_utils/event_codes.py provide the originator/event code lookup tables used both during decode (for human-readable display) and during encode (for validation).

8. Tests

9. References

• FCC 47 CFR §11.31 — Emergency Alert System Codes; SAME definition (the regulatory source of truth). https://www.ecfr.gov/current/title-47/chapter-I/subchapter-A/part-11
• NRSC-4-B §4 + Annex D — United States RBDS Standard, SAME industry specification for FM broadcasters. See companion ../NRSC4B_SAME_STANDARD.md.
• NWS Instruction 10-1712 — NOAA Weather Radio SAME usage and event code list.
• EAS Handbook — FCC operational guidance for broadcasters.

10. Implementation notes

• All SAME timing constants are exact rationals (Fraction) at the module top of app_utils/eas_fsk.py to avoid floating-point drift.
• The encoder is deliberately decoupled from the audio output device: it produces an int16-range Python list[int] that samples_to_wav_bytes finalises to a WAV. This keeps the modem trivially testable and lets the same code drive aplay, an Icecast stream, a sound card, or a buffer fed to the airchain mixer.
• Phase-continuous FFSK rendering is shared with the MDC1200 encoder (see MDC1200.md) — both protocols call into generate_fsk_samples with their own baud / mark / space parameters.