CW Transcription

Morse code decoding, conversation tracking, and auto-fill

CW Transcription decodes Morse code audio from your radio, helping you copy callsigns and exchange information. The decoder uses real-time audio processing with adaptive frequency and speed tracking.

Enabling Transcription

  1. Open the Logger tab
  2. Tap the CW button
  3. Grant microphone permission when prompted

The transcription panel appears below the logger.

How It Works

  1. Audio capture - Microphone picks up your radio’s audio
  2. Tone detection - Goertzel algorithm identifies CW tones at the target frequency
  3. Frequency tracking - Automatically locks onto the dominant tone and follows drift
  4. Speed estimation - Adaptive WPM tracking adjusts to the sender’s speed in real time
  5. Decoding - Morse patterns converted to text using timing analysis
  6. Display - Decoded text shown in real-time with conversation separation

Audio Processing

Goertzel Filter

The decoder uses the Goertzel algorithm for efficient single-frequency tone detection:

  • Targeted detection - Computes signal energy at a specific frequency rather than performing a full FFT
  • Low CPU overhead - Runs continuously in real time without significant battery impact
  • Configurable bandwidth - Detection window can be narrowed to reject adjacent signals
  • Sample-by-sample processing - Operates on the audio stream with minimal latency

The Goertzel filter is more efficient than FFT when monitoring a single tone, making it ideal for CW decoding on mobile devices.

Adaptive Thresholds

Background noise is analyzed to set:

  • Tone detection threshold (auto-adjusts to ambient noise floor)
  • Noise gate level
  • Signal-present confidence score

Audio Input Sources

  • Device microphone - Hold phone near speaker
  • Wired audio - Connect via audio cable for cleaner signal
  • WebSDR audio - When connected to a WebSDR receiver, CW transcription can decode the receiver’s audio stream directly

Frequency Tracking

Automatic

The decoder automatically finds and tracks the CW tone:

  • Detects dominant tone frequency in the audio passband
  • Adaptive tracking follows the tone as it drifts (common with analog receivers or temperature changes)
  • Works across typical CW pitch ranges (400-900 Hz)
  • Lock indicator shows when the decoder has acquired a stable tone

Manual

If automatic tracking struggles:

  1. Tap the frequency indicator
  2. Adjust the target frequency manually
  3. Tap to lock

Manual lock is useful in high-QRM conditions where multiple tones compete for the decoder’s attention.

WPM Estimation

Adaptive Speed Tracking

The decoder continuously estimates and adapts to the sender’s speed:

  • Initial detection - Analyzes the first few characters to estimate base speed
  • Running average - Maintains a weighted moving average of dot and dash lengths
  • Speed display - Current WPM shown in the transcription panel
  • Range - Tracks speeds from approximately 5 WPM to 40 WPM
  • Speed changes - Adapts within 2-3 characters when the sender’s speed changes

Farnsworth Timing

The decoder handles Farnsworth-spaced CW (where character speed is faster than word speed) by independently tracking:

  • Intra-character timing (dots and dashes within a character)
  • Inter-character spacing (gaps between characters)
  • Inter-word spacing (gaps between words)

Transcription Display

Live Transcript

Characters appear as they’re decoded. The display shows:

  • Decoded text in a scrolling view
  • Current character in progress (partial decode indicator)
  • Timing indicators for element detection

CW Conversation Tracker

The CW Conversation Tracker uses a state machine to separate and identify the two sides of a QSO:

State Machine:

  1. Listening - Waiting for initial transmission
  2. Station A transmitting - Receiving from the first station
  3. Pause detected - Gap between transmissions
  4. Station B transmitting - Receiving from the second station
  5. QSO in progress - Alternating between stations

Identification heuristics:

  • Callsign detection - When a callsign is decoded, the tracker associates it with the current transmission
  • Timing patterns - Long pauses indicate station transitions
  • Frequency shift - Slight frequency differences between stations (RIT offset) help distinguish them

Chat-Style Conversation View

For extended QSOs, the conversation view presents decoded text in a chat-bubble layout:

  • Left bubbles - Station A transmissions
  • Right bubbles - Station B transmissions (or your own, if identified)
  • Timestamps between exchanges
  • Callsign labels when identified
  • Color coding distinguishes the two stations

This view makes it easy to follow the flow of a QSO, especially at higher speeds where real-time copying is challenging.

Callsign Detection

The decoder identifies likely callsigns in the transcript:

  • Highlighted in the transcript with a distinct background color
  • Common patterns recognized (W/K/N/A prefix for US, VE for Canada, international prefixes)
  • Confidence indicator shows how likely the detected string is a valid callsign
  • Tap to auto-fill the Logger callsign field

Auto-Fill Integration

When a callsign is detected:

  1. Callsign appears highlighted in the transcript
  2. Tap to fill the Logger callsign field
  3. QRZ lookup triggers automatically
  4. Continue copying the QSO

WebSDR Integration

When connected to a WebSDR receiver, CW transcription integrates with the tuning system:

  • Auto-retune - If the CW tone drifts significantly, the decoder can request the WebSDR to retune for optimal reception
  • Frequency sync - The decoder’s target frequency stays in sync with the WebSDR’s passband center
  • Transcript sync with playback - During recording playback, the transcript scrolls in sync with the audio position

Tips for Best Results

Audio Setup

  • Hold phone near speaker, or use audio cable
  • Reduce background noise
  • Moderate volume (not too loud, not too quiet)
  • For best results, use the WebSDR audio path which bypasses the microphone entirely

Tuning

  • Center the signal in your passband
  • Use narrow filter if available
  • Reduce QRM by tuning away from interference

Limitations

CW transcription works best with:

  • Clean, well-spaced sending
  • Single signal (no QRM)
  • Standard timing

Challenging for:

  • Very fast sending (35+ WPM)
  • Poor sending (irregular timing)
  • Heavy QRM/QRN
  • Multiple overlapping signals

Accuracy

Machine CW decoding isn’t perfect. Use transcription as an aid, not a replacement for developing your own ear. When in doubt:

  • Ask for repeats
  • Verify callsigns via QRZ lookup
  • Use your own ears for critical info

See Also