DCF77 Meteotime Decoding – Technical Details

What Was Achieved

The goal was to fully decode the DCF77 Meteotime weather data, including its actual content.

✔️ Bitstream fully understood
✔️ 3-minute frame structure reconstructed
✔️ Decoder and crypto correctly implemented
✔️ Validation via 0x2501 magic check confirmed
✔️ Weather data (temperature, conditions, wind) decoded
✔️ Regions and forecast sections identified

👉 Result:
The system is now understood — from raw RF signal to human-readable weather data.

For many years, numerous attempts were made to decode Meteotime, with only limited success.
Today, with modern tools, the process has become much more accessible.

👉 The approach used here was simple:

• Provide real DCF77 recordings
• Combine them with publicly available reference implementations
• Let AI iteratively generate and refine a decoder
• Let AI automatically test against real data until valid decryption is achieved

👉 The result is a fully working decoder — without manually writing a single line of code.

📦 Data Structure

Meteotime data is embedded into the standard DCF77 signal and transmitted in 3-minute frames.

Each frame consists of 22 payload bits and 20 check bits:

• Minute n: 14 payload bits
• Minute n+1: 8 payload bits + 6 check bits
• Minute n+2: 14 check bits

👉 Total: 42-bit block per dataset

🔐 Encoding & Encryption

Meteotime data is not transmitted in plain form.

• Data is permuted and encoded
• A substitution-based cipher (S-Boxes) is used
• The key is derived from the current time/date
• No key is transmitted

👉 This effectively forms a lightweight broadcast cryptosystem

🔄 Decoding Process

1. Bitstream acquisition
   Detect pulses and convert to binary
2. Frame assembly
   Group into 3-minute blocks
3. Integrity check
   Validate using check bits
4. Key generation
   Derived from decoded time
5. Decryption
   Reverse permutation and substitution
6. Interpretation
   Map to weather data

🔍 The Road to the Solution

🧩 Phase 1 – Crypto Analysis

• S-Boxes and permutations analyzed
• Time-based key hypothesis

❌ No valid output

👉 Insight: Input data was wrong, not the algorithm

🧩 Phase 2 – Heuristic Reconstruction

• Sliding windows
• Bit order experiments
• Key variations

❌ No stable results

👉 Insight: Missing correct framing

💡 Breakthrough (Phase 3)

The breakthrough came from publicly available reference code.

Key findings:

• Fixed 3-minute frames
• Frame alignment: (minute - 1) % 3
• Edge case:
  • minute = 0 → underflow → 255 → 255 % 3 = 0 ✅
• Total structure: 82 bits
  • 42 weather bits
  • 40 time bits
• Time taken from minute 2
• Bit 7 is skipped ❗
• Info bits are not part of the key

👉 This was the turning point.

⚙️ Decoder Implementation

📡 Input

Per minute:

• 14 weather bits
• Full time information

🧱 Frame Assembly

• Minute 1 → Weather A
• Minute 2 → Weather B + Time
• Minute 3 → Weather C

→ 42 weather bits

🧠 Data Buffer

[ 0..41 ] = weather
[ 42..81 ] = time

⚠️ Bit 7 is skipped

🔐 Cipher & Key

• Cipher → weather bits
• Key → time bits

🔓 Decryption Output

[ byte0 | byte1 | byte2 | byte3 | byte4 ]

✅ Validation

plain[1..2] == 0x2501

👉 Confirms valid decode

🌦️ Payload

• Lower 22 bits contain weather data

🌍 Data Mapping

🗺️ Regions & Sections

• Regions derived from frame index
• Sections define forecast timeline

🚀 Summary

Meteotime is far more than “extra data” in DCF77 — it is a:

• Structured data protocol
• With error detection
• And time-dependent encryption

👉 A surprisingly sophisticated system for a 77.5 kHz broadcast signal.

A reference implementation of this approach is publicly available on GitHub:
https://github.com/johndcf/dcf77-meteotime-decoder

Also the live decoder is taken from GitHub:
https://github.com/johndcf/DCF77-Meteotime-Live-Monitor