Skip to content

ROBOT36.md

Latest commit

 

History

History
251 lines (191 loc) · 7.31 KB

ROBOT36.md

File metadata and controls

251 lines (191 loc) · 7.31 KB

Robot36 Color Mode - Technical Specification

Overview

Robot36 is a fast SSTV mode designed for quick amateur radio transmissions. It uses interlaced YUV encoding with a transmission time of approximately 36 seconds.

Image Specifications

Parameter Value
Resolution 320×240 pixels
Aspect Ratio 4:3
Total Lines 240 (120 even + 120 odd)
Line Time ~150ms per scan line
Total Duration ~36 seconds
VIS Code 8

Frequency Mapping

Signal Frequency Purpose
Sync Pulse 1200 Hz Line boundary marker (9ms)
Black Level 1500 Hz Minimum brightness
Gray Level 1900 Hz Center frequency
White Level 2300 Hz Maximum brightness

Frequency Range: 800 Hz (1500-2300 Hz) represents 0-255 pixel values

Scan Line Structure

Total: 150ms per line

|--9ms--|--3ms--|--------88ms--------|--4.5ms--|1.5ms|-------44ms-------|
  Sync   S.Porch   Luminance (Y)      Separator Porch  Chrominance
 1200Hz  1500Hz    320 pixels         Even/Odd  1900Hz  320 pixels
                   1500-2300Hz        indicator         1500-2300Hz

Timing Breakdown

  1. Sync Pulse: 9ms @ 1200 Hz

    • Purpose: Line boundary detection
    • Used by decoder to identify start of new line

  2. Sync Porch: 3ms @ 1500 Hz

    • Purpose: Transition period after sync
    • Gives receiver time to stabilize

  3. Luminance (Y) Channel: 88ms

    • 320 pixels of brightness information
    • Frequency range: 1500-2300 Hz (black to white)
    • Transmitted every line (even and odd)

  4. Separator Pulse: 4.5ms

    • Even lines: Negative frequency (< 1900 Hz)
    • Odd lines: Positive frequency (> 1900 Hz)
    • Purpose: Tells decoder which chrominance component follows

  5. Porch: 1.5ms @ 1900 Hz

    • Transition period before chrominance

  6. Chrominance Channel: 44ms

    • 320 pixels of color information
    • Even lines: R-Y (red color difference)
    • Odd lines: B-Y (blue color difference)

Color Encoding: Interlaced YUV

Robot36 uses an interlaced color scheme to reduce transmission time while maintaining reasonable color quality.

Interlacing Strategy

Line 0 (even): Y₀ + (R-Y)₀    ← Store in buffer
Line 1 (odd):  Y₁ + (B-Y)₁    ← Combine with line 0, output 2 RGB rows

Line 2 (even): Y₂ + (R-Y)₂    ← Store in buffer
Line 3 (odd):  Y₃ + (B-Y)₃    ← Combine with line 2, output 2 RGB rows

...continues for 240 lines

Decoding Process

  1. Even Line (line 0, 2, 4, ...):

    • Store 320 Y (luminance) values
    • Store 320 R-Y (red difference) values
    • Wait for next line

  2. Odd Line (line 1, 3, 5, ...):

    • Receive 320 new Y values
    • Receive 320 B-Y (blue difference) values
    • Combine with stored even line data:
      • Even row: RGB = YUV_to_RGB(Y_even, R-Y_even, B-Y_odd)
      • Odd row: RGB = YUV_to_RGB(Y_odd, R-Y_even, B-Y_odd)

YUV to RGB Conversion (ITU-R BT.601)

// Input:
//   Y:  Luminance (0-255)
//   RY: R-Y color difference (0-255, centered at 128)
//   BY: B-Y color difference (0-255, centered at 128)

// Convert to difference signals
const ryDiff = RY - 128;  // -128 to +127
const byDiff = BY - 128;  // -128 to +127

// ITU-R BT.601 conversion
R = Y + 1.402 × ryDiff
G = Y - 0.344136 × byDiff - 0.714136 × ryDiff
B = Y + 1.772 × byDiff

// Clamp to valid range
R = clamp(R, 0, 255)
G = clamp(G, 0, 255)
B = clamp(B, 0, 255)

Signal Processing Pipeline

1. Sync Detection

  • Monitor for 1200 Hz tone
  • Measure pulse width (must be 9ms ±2ms)
  • Calculate frequency offset for calibration

2. Line Extraction

  • Extract audio samples between consecutive sync pulses
  • Should contain ~150ms of audio data
  • Pass to line decoder

3. Bidirectional Filtering

// Forward pass (left to right)
for (i = 0; i < lineLength; i++) {
  filtered[i] = EMA.average(samples[i]);
}

// Backward pass (right to left)
EMA.reset();
for (i = lineLength - 1; i >= 0; i--) {
  filtered[i] = EMA.average(filtered[i]);
}

4. Separator Detection

// Average separator pulse frequency
separatorFreq = average(samples[separatorStart:separatorEnd]);

// Determine line type
if (separatorFreq < 0) {
  lineType = EVEN;  // R-Y chrominance
} else {
  lineType = ODD;   // B-Y chrominance
}

5. Pixel Sampling

// Sample 320 pixels from 88ms Y channel
for (x = 0; x < 320; x++) {
  sampleIndex = yStart + (x * ySamples) / 320;
  Y[x] = frequencyToLevel(samples[sampleIndex]);
}

// Sample 320 pixels from 44ms chrominance channel
for (x = 0; x < 320; x++) {
  sampleIndex = chromaStart + (x * chromaSamples) / 320;
  Chroma[x] = frequencyToLevel(samples[sampleIndex]);
}

Chroma Noise Reduction

To improve color quality, a 5-pixel median filter is applied to chrominance channels:

// For each pixel position
const window = [chroma[x-2], chroma[x-1], chroma[x], chroma[x+1], chroma[x+2]];
window.sort();
filteredChroma[x] = window[2];  // Median value

Additionally, chroma saturation is reduced by 30% to suppress color noise:

const CHROMA_REDUCTION = 0.7;
chromaFiltered = 128 + (chroma - 128) * CHROMA_REDUCTION;

Implementation Notes

Sample Rate Handling

  • Supports 44.1 kHz and 48 kHz sample rates
  • All timing calculations scale proportionally
  • Frequency calibration compensates for drift

Buffer Management

  • 7-second circular buffer for audio samples
  • Separate buffer for FM demodulated values
  • Line extraction uses buffer wrap-around logic

Error Handling

  • Validate separator frequency (reject if out of range)
  • Check line length (should be ~150ms ±10%)
  • Skip lines with invalid sync pulse width
  • Fall back to alternating even/odd if separator ambiguous

Performance Characteristics

Metric Value
Processing Time Real-time (36 seconds)
Memory Usage ~30 MB (buffers + image)
CPU Usage 5-10% (single core)
Minimum SNR 15 dB for reliable sync

Common Issues & Solutions

Issue: Purple/Red Color Cast

Cause: Incorrect YUV to RGB conversion or wrong chroma channel assignment Solution: Ensure R-Y and B-Y channels are correctly identified via separator

Issue: Missing Lines

Cause: Sync pulses not detected or incorrect timing Solution: Verify 1200 Hz sync detection, check pulse width classification

Issue: Horizontal Banding

Cause: Insufficient filtering or frequency calibration drift Solution: Apply bidirectional EMA filtering, track frequency offset

Issue: Interlacing Artifacts

Cause: Even/odd line pairing mismatch Solution: Validate separator frequency, implement fallback logic

Testing Recommendations

  1. Sync Detection Test: Verify 9ms pulses at 1200 Hz are detected
  2. Timing Test: Confirm ~150ms between sync pulses
  3. Separator Test: Validate even/odd line classification
  4. Color Test: Use test patterns with known colors (red, green, blue, white)
  5. Interlacing Test: Verify consecutive even/odd lines produce correct RGB output

References

  • Original implementation: Robot36 Android App by xdsopl
  • SSTV specification: Amateur Radio magazines and ARRL documentation
  • Color space: ITU-R BT.601 standard
  • DSP techniques: Lyons, "Understanding Digital Signal Processing"