Audio Fundamentals, Compression & Real-World Applications

TAKEAWAYS

• Audio compression reduces file size by discarding redundant or imperceptible data (lossy) or using mathematical patterns (lossless).
• Sampling rate (e.g., 44.1 kHz) and bit depth (e.g., 16-bit) determine audio quality—higher values mean better fidelity but larger files.
• MP3 (lossy) and FLAC (lossless) are industry standards for music, while AAC dominates streaming (Apple Music, Spotify).
• MIDI synthesizes sound via instructions (small files) but lacks the realism of recorded audio.
• Perceptual coding exploits human hearing limitations (e.g., masking) to improve compression efficiency.
• Real-world trade-offs: WhatsApp uses Opus (low-latency, adaptive bitrate) for calls, while Ncell’s music app prioritizes AAC for balance between quality and data savings.

---

1. Audio Fundamentals: From Analog to Digital

Audio is a continuous analog signal (sound waves) that must be converted to digital form for computers to process. This involves three key steps:

1.1 Analog-to-Digital Conversion (ADC)

• Sampling: Capturing the amplitude of the analog signal at discrete intervals (measured in Hz).
  • Example: A 44.1 kHz sampling rate means 44,100 samples per second.
  • Nyquist Theorem: Sampling rate must be ≥ 2 × highest frequency in the signal to avoid aliasing.

• Quantization: Assigning numerical values to each sample (e.g., 16-bit = 65,536 possible levels).
• Encoding: Storing samples as binary data (e.g., WAV, AIFF).

1.2 Key Audio Parameters

Worked Example: CD Audio vs. MP3

• A CD-quality audio file:
  • Sampling rate: 44.1 kHz
  • Bit depth: 16-bit
  • Channels: 2 (Stereo)
  • Uncompressed size: ~10 MB per minute (700 KB/sec).
• An MP3 (compressed) version of the same audio might be ~1 MB/minute (100 KB/sec).

---

2. Audio Compression Techniques

Compression reduces file size by removing redundancy or irrelevant data. Two main types:

2.1 Lossless Compression

• No data loss; original audio can be perfectly reconstructed.
• Methods:
  • Run-Length Encoding (RLE): Replaces repeated values with a count. Example: "aaaaabbbbcccdde" → (5a,4b,3c,2d,e) → Compression ratio = 12/10 = 1.2 (20% reduction).
  • Dictionary Methods: Stores repeated patterns once (e.g., FLAC, ALAC).
  • Entropy Coding: Assigns shorter codes to frequent values (e.g., Huffman coding).
• Use Cases: Archival audio, professional editing (FLAC, WMA Lossless).

2.2 Lossy Compression

• Removes imperceptible or redundant data; irreversible but achieves high compression ratios.
• Key Techniques:
  • Perceptual Coding: Exploits how humans hear:
    • Masking: Loud sounds hide quiet ones (e.g., a drum drowns out a hi-hat).
    • Frequency Irrelevance: Humans don’t hear ultra-high/low frequencies well.
  • ** Psychoacoustic Models**: Analyzes audio to discard inaudible components.
  • Hybrid Approaches: Combines time-domain and frequency-domain analysis (e.g., MP3, AAC).

Worked Example: MP3 Compression in Ncell Music

• Original WAV: 10 MB/minute (44.1 kHz, 16-bit, stereo).
• MP3 (128 kbps): ~1.6 MB/minute (80% smaller).
• Trade-off: Slight loss in high-frequency details (inaudible to most listeners).

---

3. Specialized Audio Formats

3.1 MIDI (Musical Instrument Digital Interface)

• Not audio data; stores instructions (notes, timing, instruments) for synthesizers.
• Advantages:
  • Tiny file sizes (e.g., a 5-minute song = ~50 KB).
  • Editable (change instruments, tempo without re-recording).
• Disadvantages:
  • No "real" audio; relies on a sound font/synthesizer.
  • Poor for complex sounds (e.g., vocals, live recordings).
• Use Cases: Video game soundtracks, ringtone editors, karaoke apps.

Comparison: MIDI vs. Digital Audio for Video Games

Real-World Example: Pathao’s In-App Music

• Uses AAC (lossy) for background music in the driver app to:
  • Save mobile data (critical for low-bandwidth users).
  • Balance quality and file size (~1 MB per song).

---

4. Real-World Applications of Audio Compression

4.1 WhatsApp Voice Messages (Opus Codec)

• Why Opus?
  • Adaptive bitrate: Adjusts quality based on network speed.
  • Low latency: Optimized for real-time calls (~200 ms delay).
  • Compression: Reduces a 1-minute voice note from ~10 MB (WAV) to ~500 KB.
• How It Works:
  1. Audio sampled at 48 kHz.
  2. Opus applies perceptual coding to discard inaudible frequencies.
  3. Encrypted and sent over WhatsApp’s servers.

4.2 Ncell Music Streaming (AAC)

• Challenge: Nepal’s average internet speed (~10 Mbps) struggles with high-bitrate audio.
• Solution:
  • Uses AAC at 128–192 kbps (balance of quality and data).
  • Adaptive streaming: Switches between 64 kbps (slow network) and 320 kbps (fast network).
• Result: A 3-minute song downloads in ~1–2 MB vs. ~30 MB uncompressed.

4.3 eSewa’s IVR System (MIDI + Text-to-Speech)

• Problem: Storing pre-recorded voice prompts for transactions (e.g., "Enter your PIN") wastes space.
• Solution:
  • Uses MIDI-like synthesized speech for static prompts.
  • Dynamic prompts (e.g., balance updates) use low-bitrate MP3 snippets.
• Savings: Reduces storage by 90% compared to WAV files.

---

5. Why Compression Matters for Video and Animation

5.1 Data Volume in Multimedia

• 1 minute of uncompressed video:
  • 4K @ 60 FPS: ~1.5 GB (requires 150 Mbps bandwidth).
  • Audio (stereo, 48 kHz): ~10 MB/minute.
• Compressed (H.264 + AAC):
  • ~50 MB for the same video (97% smaller).

5.2 Compression Techniques in Video

Worked Example: Daraz’s Product Videos

• Problem: Sellers upload 1-minute videos (1080p, 30 FPS) → ~500 MB uncompressed.
• Solution: Daraz’s platform compresses to ~20 MB using:
  • H.264 (interframe compression).
  • AAC audio (128 kbps).
• Result: Faster uploads, lower bandwidth for viewers.

---

Exam Tip

1. Lossy vs. Lossless:

   • Lossy (MP3, AAC) = high compression, irreversible, used for music/streaming.
   • Lossless (FLAC, WAV) = no quality loss, used for archiving.
   • Exam trick: Always mention perceptual coding for lossy formats.

2. Sampling Rate/Bit Depth:

   • CD quality = 44.1 kHz, 16-bit.
   • Broadcast radio = 48 kHz, 24-bit.
   • Higher values = better quality but larger files.

3. Real-World Scenarios:

   • WhatsApp: Opus codec for calls.
   • Ncell Music: AAC for streaming.
   • eSewa IVR: MIDI/text-to-speech for prompts.
   • Tip: Relate numerical examples to these apps (e.g., "AAC at 128 kbps reduces a song to 1 MB").

4. Trade-offs:

   • MIDI vs. Digital Audio: MIDI saves space but lacks realism.
   • JPEG vs. MPEG: MPEG uses interframe compression for video; JPEG is for images.

5. Compression Ratio Calculation:

   • Given: Original size = 100 KB, Compressed = 10 KB.
   • Ratio = Original/Compressed = 10:1.

6. Diagrams:

   • Always draw sampling process, psychoacoustic model, or MPEG frame types if asked for steps.
   • Label every component (e.g., "I-frame," "P-frame," "Quantization step").

---

Final Note: Audio compression is about balancing quality, file size, and real-world constraints. Master the trade-offs (e.g., MIDI vs. WAV, MP3 vs. FLAC) and real-world examples (WhatsApp, Ncell, eSewa) to score full marks!