AI summary by TwinMind
The Engineering Inside Wireless Earbuds || How do Wireless Earbuds and Audio Codecs Work?
Branch Education
17m 41s
Wireless Earbuds Overview
- Complete mystery how wireless earbuds work vs wired headphones
- 9 distinct technologies inside tiny plastic earbuds:
- Speaker, audio codecs, Bluetooth, system on chip (SoC), printed circuit board (PCB), accelerometers, lithium ion battery, microphones, noise cancellation
- Focus on audio codecs specifically
- Video structure: 4 parts covering teardown, conceptual overview, digital audio representation, detailed explanations
Apple AirPods 2 Teardown
- Opening extremely difficult/destructive due to glue
- Components from outside to inside:
- Outer plastic earpiece + mesh dust cover
- Rubber protective shell + optical sensor
- Speaker with 4 key components: diaphragm, suspension/spider, voice coil, magnet
- Complex circuitry folded into tiny package
- Circuit board structure:
- 3 separate PCBs folded together with flexible wires
- Top board: 2 speaker connection points, microchip for Bluetooth connectivity/audio decoding, accelerometers, programmable SoC
- Second board: low power stereo audio processing chip/audio codec
- Flat cable antenna adjacent to battery
- Microphone at back for noise reduction
- Third board connected to battery
- Additional small board with MEMS microphone (grain of rice size)
- Mesh dust cover + charging contacts at bottom
Conceptual Operation Overview
- Bluetooth communication channel established between paired headphones/smartphone
- Process flow:
- Smartphone grabs audio data from flash storage
- Decompresses audio, stores in working memory as long set of numbers
- Compresses/divides into packets per Bluetooth specs
- Converts packets to electromagnetic waves/photons
- Sends via Bluetooth to earbuds
- Earbuds receive, disassemble, decompress packets back to value sets
- Audio codec converts digital values to analog electrical waveform
- Waveform sent to voice coil attached to diaphragm
- Voice coil moves back/forth per waveform
- Diaphragm movement creates air pressure waves
- Pressure waves sensed by ear, interpreted as sound by brain
Audio Codec Fundamentals
- Codec = coding and decoding
- Software/hardware converting data between formats while compressing/decompressing
- Audio codec functions:
- Decoding: digital values/numbers → analog waveform (digital-to-analog conversion/DAC)
- Encoding: analog microphone signal → digital values (analog-to-digital conversion/ADC)
- Used in all technology for data type conversion
- Video codecs actively decompress video data during viewing
Digital Audio Representation Basics
- Analog audio waveform → digital list of numbers
- Process simpler than expected - no complex mathematics
- Method: place analog waveform on graph, capture values at set time intervals
- Digital audio = long list of values/points waveform passes through
- Example: 23 microseconds between each data point
- Audio codec process: plots points on graph, connects dots, smooths line, sends analog waveform to speaker
- Audio files in editing software show constituent points when zoomed in
- Process called digitization/analog-to-digital conversion
Sampling Rate Details
- X-axis = time, data points/samples spaced 23 microseconds apart
- Alternative: 1 microsecond spacing = 1 million samples/second (1 MHz) = 100+ MB for 60 seconds audio
- 23 microsecond spacing rationale:
- Based on average human ear sound perception ability
- Human hearing range: up to ~20 kHz (1 wave every 50 microseconds)
- Higher frequencies inaudible to humans
- Engineers ignore frequencies humans can't hear
- Capturing 20 kHz max frequency requires 2 data points per full wave
- Results in 1 data point every 23 microseconds
- Standard sampling rates:
- 44.1 kHz (44,100 samples/second) - most common for music/podcasts
- 48 kHz - second most common
- Why slightly more than double human hearing frequency: Nyquist theorem specifics in comments
Bit Depth and Quantization
- Question: how many bits (1s/0s) per sample for binary representation
- 1 bit per sample test:
- Each sample = 1 or 0
- Recreation looks nothing like original - insufficient
- 2 bits per sample test:
- 4 possible values per sample
- Still poor match to original waveform
- Optimal bit depth varies:
- Audio CD: 16 bits per sample
- 16 bits = 2^16 = 65,536 different Y-axis values
- Technical term: "bit depth of 16 bits"
- Higher quality: 24 or 32 bits per sample
- Process terminology:
- Analog signal → values = quantization
- Assigning bits to values = pulse code modulation
Audio File Formats
- Uncompressed digital audio (16-bit values at 44.1 kHz):
- ~10.5 MB per 60 seconds, 2-channel audio
- MP3 (320 kbps stereo):
- Reduces to ~2.4 MB per 60 seconds
- Processes every millisecond of audio
- Psychoacoustic algorithm removes elements humans can't hear well:
- Exceptionally low volume sounds
- Very high pitched sounds
- Sounds very close together
- Lossy compression format
- Lossless formats (ALAC, FLAC):
- Don't discard data
- Compress by finding patterns/redundant data
- Represent patterns more efficiently than uncompressed
- File compression = complex topic requiring dedicated videos
Future Content Plans
- 9 technologies in wireless earbuds require separate episodes
- Current video coverage status mentioned
- Working on additional topic videos
- Complete wireless headphone understanding requires full series
