Electronics Guide

Sampling Process

The sampling process captures instantaneous amplitude values of the audio waveform at precisely timed intervals. According to the Nyquist-Shannon sampling theorem, the sampling frequency must be at least twice the highest frequency component in the audio signal to enable perfect reconstruction. This minimum rate, called the Nyquist rate, establishes the fundamental relationship between sampling frequency and audio bandwidth.

Before sampling, an anti-aliasing filter removes frequency components above half the sampling rate. Without this filtering, high frequencies would fold back into the audible spectrum as aliasing artifacts, creating distortion unrelated to the original signal. The filter must transition sharply from passband to stopband, requiring careful analog design or oversampling techniques that relax filter requirements.

The sample-and-hold circuit captures each instantaneous value and maintains it constant during the time required for analog-to-digital conversion. Aperture jitter, variations in the precise sampling instant, introduces noise proportional to the signal's slew rate at the sampling moment. High-performance audio converters minimize jitter to preserve signal quality, particularly at high frequencies where slew rates are greatest.

Quantization

Quantization maps each sampled amplitude to the nearest value from a finite set of discrete levels. The number of quantization levels, determined by the bit depth, establishes the resolution and dynamic range of the digital representation. Each additional bit doubles the number of levels and adds approximately 6 dB to the theoretical dynamic range.

The quantization error, the difference between the true sample value and its quantized representation, behaves as additive noise when the signal is sufficiently complex. For signals much larger than the quantization step size, this quantization noise is approximately uniform and uncorrelated with the signal, resulting in a noise floor that sets the system's dynamic range limit.

Linear PCM uses uniformly spaced quantization levels, providing consistent resolution across the entire amplitude range. This approach suits high-quality audio where maximum fidelity is paramount. Non-uniform quantization schemes, such as mu-law and A-law companding used in telephony, allocate more resolution to small amplitudes where human hearing is most sensitive, achieving acceptable voice quality with fewer bits.

Dithering

Dithering adds low-level noise to the signal before quantization, converting the deterministic quantization error into random noise. This seemingly counterintuitive technique actually improves perceived audio quality by eliminating the correlation between quantization error and signal, which otherwise manifests as audible distortion on low-level signals.

Rectangular probability density function (RPDF) dither uses uniformly distributed random noise with amplitude equal to one quantization step. This basic dither decorrelates the quantization error from the signal but does not eliminate the noise modulation that occurs as signal level varies.

Triangular probability density function (TPDF) dither, created by summing two RPDF sources, provides complete decorrelation and eliminates noise modulation. The noise floor becomes constant regardless of signal level, providing psychoacoustically superior results at the cost of a slightly higher noise floor compared to undithered signals at some levels.

Noise-shaped dithering applies filtering to the dither signal, pushing quantization noise to frequencies where human hearing is less sensitive. By concentrating noise above 10 kHz, noise shaping can achieve perceived dynamic range improvements of several decibels beyond what the bit depth alone would suggest, particularly valuable when reducing bit depth for distribution formats.

PCM Data Organization

PCM samples must be organized into a defined structure for storage and transmission. Key organizational choices include byte order (endianness), sample alignment within bytes, and the arrangement of samples in multichannel audio.

Little-endian format stores the least significant byte first, as used by Intel processors and WAV files. Big-endian format stores the most significant byte first, as used by Motorola processors and AIFF files. Audio interfaces and file formats must clearly specify endianness to ensure correct interpretation of sample data.

Sample alignment affects how samples with bit depths not equal to multiples of eight are packed into bytes. Common approaches include padding to the next byte boundary, packing consecutive samples to minimize padding, or storing samples in larger word sizes with explicit padding. The choice affects storage efficiency and processing complexity.

Multichannel audio may be organized as interleaved samples, where all channels for one time instant are grouped together, or as separate channel streams. Interleaved organization simplifies streaming and maintains channel synchronization naturally. Separate channels can be more efficient for processing that operates on individual channels independently.

Standard Sample Rates

The audio industry has settled on several standard sample rates, each optimized for particular applications:

• 44.1 kHz: The CD standard, providing frequency response to 20 kHz with minimal margin. Chosen to match the bandwidth of analog video recording systems used for early digital audio mastering. Remains the most common rate for music distribution.
• 48 kHz: The professional video standard, providing slightly extended bandwidth and simpler relationships with video frame rates. Standard for broadcast, film, and video production.
• 88.2 kHz and 96 kHz: Double-rate formats that ease anti-aliasing filter design and provide ultrasonic headroom for processing. Common in professional recording where quality margins are paramount.
• 176.4 kHz and 192 kHz: Quad-rate formats used in high-resolution audio distribution and archival applications. The audible benefit beyond 96 kHz is debated, but these rates provide maximum flexibility for post-processing.

Sample rate conversion between these standards requires careful interpolation to avoid aliasing and preserve audio quality. Integer-ratio conversions like 44.1 to 88.2 kHz are simpler than non-integer ratios like 44.1 to 48 kHz, which require rational approximation techniques.

Common Bit Depths

Bit depth directly determines the dynamic range and noise floor of the digital audio signal:

• 8-bit: Provides only 48 dB dynamic range, audibly noisy for music but acceptable for voice in early computer systems and games. Largely obsolete for quality applications.
• 16-bit: The CD standard, offering 96 dB theoretical dynamic range. With proper dithering, provides noise floors low enough for critical listening in quiet environments. Remains the most common distribution format.
• 24-bit: Professional recording standard with 144 dB theoretical dynamic range, far exceeding any analog component. The practical benefit is headroom during recording and processing rather than audible noise floor improvement.
• 32-bit float: Uses floating-point representation with effectively unlimited headroom and approximately 24-bit mantissa precision. Ideal for processing where intermediate calculations may exceed normal levels.

Actual dynamic range is always less than theoretical due to converter noise, jitter, and analog circuit limitations. Modern high-quality 24-bit converters typically achieve 115-125 dB actual dynamic range, still far exceeding 16-bit limits.

High-Resolution Audio

High-resolution audio generally refers to formats exceeding CD quality, typically 24-bit depth at sample rates of 88.2 kHz or higher. The audibility of improvements beyond CD quality remains controversial, with controlled tests often failing to demonstrate reliable perception of differences.

The engineering arguments for high-resolution audio center on processing headroom and filter behavior rather than audible frequency extension. Higher sample rates allow gentler anti-aliasing filters with less phase distortion in the audible band. Additional bit depth provides headroom during mixing and processing without risk of clipping.

Distribution of high-resolution audio requires significantly more bandwidth and storage than CD-quality formats. Lossless compression reduces file sizes substantially but cannot approach the efficiency of lossy codecs. The premium pricing and specialized equipment required for high-resolution playback limit its adoption to audiophile markets.

Sample Rate Conversion

Converting between sample rates requires interpolation to create sample values at new time positions. Simple approaches like linear interpolation introduce artifacts; high-quality conversion uses polyphase filter structures with many taps to approximate ideal sinc interpolation.

Integer-ratio conversion, such as doubling or halving the sample rate, can be performed efficiently with interpolation and decimation filters. The conversion 44.1 to 48 kHz, commonly needed when moving between music and video domains, involves the ratio 160:147 and requires more complex rational resampling algorithms.

Asynchronous sample rate converters dynamically track the actual sample rates of input and output, accommodating not only nominal rate differences but also the drift and jitter inherent in separate clock domains. These converters use digital phase-locked loops and adaptive filter coefficient selection to maintain quality under varying conditions.

Lossless Compression

Lossless audio compression exploits the structure and redundancy in audio signals to reduce file size while enabling bit-perfect reconstruction of the original PCM data. Typical compression ratios range from 40% to 60% of original size, varying with content characteristics.

The Free Lossless Audio Codec (FLAC) has emerged as the dominant open-source lossless format. FLAC divides audio into blocks, applies linear prediction to model sample-to-sample correlation, and entropy codes the prediction residuals. The format supports metadata, seeking, and error detection, making it suitable for both archival and streaming applications.

Apple Lossless Audio Codec (ALAC) provides similar compression performance with native support in Apple devices. The format uses adaptive prediction and Golomb-Rice coding, achieving results comparable to FLAC while integrating with Apple's ecosystem.

Windows Media Audio Lossless (WMA Lossless) and other proprietary formats offer similar capabilities within their respective platforms. The lack of a single universal lossless standard requires consideration of playback device compatibility when choosing formats.

Perceptual Lossy Compression

Perceptual audio coding exploits the characteristics of human hearing to discard information that listeners cannot perceive. By modeling auditory masking, where loud sounds render nearby quieter sounds inaudible, these codecs achieve dramatic compression while maintaining acceptable quality.

MP3 (MPEG-1 Audio Layer III) pioneered widespread lossy audio compression and remains ubiquitous despite its age. The codec transforms audio into frequency subbands, applies psychoacoustic modeling to determine masking thresholds, and quantizes each band to just above the masking threshold. Bit rates from 128 to 320 kbps provide quality ranging from acceptable to near-transparent.

Advanced Audio Coding (AAC) improves upon MP3 with more efficient transform coding, better handling of transients, and more sophisticated psychoacoustic models. AAC achieves quality comparable to MP3 at roughly 80% of the bit rate. The format is standard for Apple devices, YouTube, and many streaming services.

Opus, developed for internet communications, offers excellent quality across a wide bit rate range from very low rates suitable for speech up to high rates rivaling AAC for music. Its low latency and bit rate adaptability make it ideal for real-time communications and streaming. Opus is royalty-free and increasingly adopted as a universal audio codec.

Codec Comparison

Selecting an audio codec involves balancing multiple factors:

• Quality at target bit rate: Modern codecs like Opus and AAC generally outperform MP3 at equivalent bit rates, particularly below 128 kbps
• Latency: Opus offers modes with latency as low as 5 ms, while MP3 and AAC require hundreds of milliseconds, making codec choice critical for interactive applications
• Computational complexity: Encoding complexity varies widely, affecting power consumption and real-time capability on limited hardware
• Compatibility: MP3's universal support makes it a safe choice despite lower efficiency; newer codecs may require software updates or specific hardware support
• Licensing: Patent licensing affects commercial deployment; royalty-free codecs like Opus avoid these costs

Spatial Audio Codecs

Emerging codecs address the needs of immersive and spatial audio formats. These codecs must efficiently compress multichannel or object-based audio while preserving the spatial cues essential for 3D sound reproduction.

MPEG-H Audio encodes both channel-based and object-based content, allowing renderers to adapt playback to the specific speaker configuration. The codec includes metadata describing object positions and movements, enabling personalized rendering on devices from headphones to large speaker arrays.

Dolby Atmos, while primarily a rendering system, includes coded audio formats that carry bed channels, objects, and metadata. The associated codecs like Dolby Digital Plus with Atmos extensions or Dolby AC-4 provide efficient compression of immersive content for streaming and broadcast.

Channel Configurations

Standard channel configurations are identified by notation indicating main channels, low-frequency effects channel, and height channels:

• 2.0 (Stereo): Left and right channels, the baseline for most consumer audio
• 5.1: Left, center, right, left surround, right surround, and low-frequency effects (LFE). The dominant home theater format.
• 7.1: Adds rear surround left and right to 5.1, common in larger home theaters and commercial cinemas
• 7.1.4: Adds four height channels to 7.1 for Dolby Atmos and similar immersive formats
• 22.2: NHK's Ultra HD audio format with 24 channels including upper, middle, and lower layers

Channel ordering within audio streams varies by format. SMPTE, ITU, and manufacturer-specific orderings differ, requiring careful mapping when interfacing between systems. Metadata describing channel assignments helps receivers correctly route channels to speakers.

Bass Management

The LFE channel and bass management are often confused. The LFE channel carries dedicated low-frequency content mixed specifically for that channel. Bass management is the receiver or processor function that redirects low frequencies from main channels to capable subwoofers based on speaker capability settings.

Proper bass management requires crossover filtering to split each channel into low and high frequency components. The low-frequency portions route to the subwoofer along with the LFE channel, while high-frequency portions route to the respective speakers. Crossover frequency, typically 80-120 Hz, depends on the main speakers' low-frequency capability.

The LFE channel itself has limited bandwidth, typically 20-120 Hz, and is recorded at a level 10 dB below the main channels (on professional formats) to provide headroom for high-impact effects. Consumer decoders apply corresponding gain when reproducing the LFE channel.

Object-Based Audio

Object-based audio represents sound sources as discrete objects with position metadata rather than fixed channel assignments. This approach enables content to adapt to any playback system, from headphones to large immersive speaker arrays, by rendering objects to the available speakers at playback time.

Each audio object carries position data (azimuth, elevation, distance) that may be static or animated over time. Additional metadata may specify object size, diffuseness, and exclusion zones. The renderer interprets this metadata to determine the appropriate contribution of each object to each speaker.

Dolby Atmos combines a bed of channel-based audio with up to 128 audio objects. The bed provides ambient and background content efficiently as traditional channels, while objects provide precise positioning for discrete sound sources. This hybrid approach balances efficiency with flexibility.

MPEG-H Audio provides a fully open standard for object-based audio, supporting both prerendered channels and positioned objects. The format enables accessibility features through separate dialogue objects that can be boosted or substituted for different languages.

Ambisonics

Ambisonics represents the entire sound field using spherical harmonic components rather than discrete channels or objects. First-order Ambisonics uses four channels (W, X, Y, Z) representing omnidirectional and three figure-eight patterns. Higher orders add more channels, improving spatial resolution.

The format is speaker-agnostic: the same Ambisonic recording can be decoded for any speaker arrangement or for binaural headphone reproduction. This flexibility makes Ambisonics attractive for virtual reality and applications where the playback configuration is unknown at production time.

Higher-Order Ambisonics (HOA) at third order requires 16 channels; seventh order requires 64 channels. The improved spatial resolution comes at significant bandwidth cost, making HOA most practical for studio production with conversion to other formats for distribution.

Content Metadata

Descriptive metadata identifies and categorizes audio content:

• Title, artist, album: Basic identification used for display and library organization
• Genre, year, track number: Classification enabling sorting and filtering
• Cover art: Visual imagery associated with the content, often embedded in the file
• Lyrics: Synchronized or unsynchronized text of vocal content
• Credits: Contributors including performers, composers, engineers, and producers

ID3 tags remain the most common metadata format for MP3 files, with ID3v2.4 supporting extensive fields and embedded images. Vorbis Comments provide similar functionality for Ogg and FLAC files. Each container format has its own metadata provisions that parsers must specifically support.

Technical Metadata

Technical metadata describes properties required for correct playback and processing:

• Sample rate and bit depth: Fundamental parameters defining the audio format
• Channel configuration: Number of channels and their assignments to speaker positions
• Codec parameters: Settings required to initialize decoders for compressed formats
• Duration: Total playback time enabling progress display and seeking
• Bitrate: Data rate information for streaming and buffering decisions

Container formats like MP4, Matroska, and WAV include header structures that carry technical metadata. Proper parsing of these headers is essential before audio decoding can begin.

Loudness Metadata

Loudness metadata enables consistent playback levels across content from different sources. Without loudness normalization, listeners must constantly adjust volume, and quiet content may be lost in noisy environments while loud content causes discomfort.

The ITU-R BS.1770 loudness measurement algorithm provides the standard method for determining integrated loudness in LUFS (Loudness Units relative to Full Scale). Broadcast standards like EBU R128 specify target loudness levels and require metadata embedding to enable receiver-side normalization.

ReplayGain was an early loudness normalization system for consumer audio, storing track and album gain values. Modern systems use more sophisticated algorithms but serve the same purpose: enabling playback at consistent perceived loudness without modifying the audio data itself.

Dynamic Range Metadata

Dynamic range metadata enables receivers to compress the dynamic range of content to suit the listening environment. A home theater system might reproduce the full theatrical mix, while portable playback might apply compression so quiet passages remain audible over ambient noise.

Dolby Digital and its successors include dialnorm (dialogue normalization) and dynamic range control metadata. The encoder specifies the average dialogue level and provides profiles defining how the receiver should compress dynamics. Different profiles suit different listening scenarios.

MPEG-H Audio provides sophisticated dynamic range control with object-aware processing. Individual audio objects can have different compression applied, for example compressing music while preserving dialogue dynamics for intelligibility.

SMPTE Time Code

SMPTE time code expresses time as hours, minutes, seconds, and frames in the format HH:MM:SS:FF. Frame rates match standard video rates: 24 fps for film, 25 fps for PAL video, and 29.97 or 30 fps for NTSC. The frame reference derives from the film and video origin of the standard.

Drop-frame time code, used with 29.97 fps video, periodically skips frame numbers to maintain correspondence between time code and clock time. Without this adjustment, time code would drift ahead of clock time by approximately 3.6 seconds per hour. The drop pattern omits frames 0 and 1 of each minute except every tenth minute.

Linear Time Code (LTC) encodes SMPTE time code as an audio-frequency signal recorded on tape or transmitted over audio channels. The biphase modulated signal can be read at any speed, enabling synchronization during fast-forward and rewind as well as normal playback.

Vertical Interval Time Code (VITC) embeds time code in the vertical blanking interval of video signals. Unlike LTC, VITC is readable even when the tape is stationary, but requires video circuitry to extract.

MIDI Time Code

MIDI Time Code (MTC) conveys SMPTE time code over MIDI connections, enabling synchronization of MIDI sequencers and other devices with video and audio systems. MTC uses quarter-frame messages that distribute time code information across eight MIDI messages per frame.

Full-frame messages can establish or re-establish time position, useful when starting playback from arbitrary positions. Quarter-frame messages then maintain continuous synchronization during normal transport.

The eight quarter-frame messages carry two nibbles of time code data each: hours, minutes, seconds, and frames with rate information. This fragmented transmission trades latency for MIDI bandwidth efficiency and smooth position updates.

Word Clock and Digital Synchronization

Word clock synchronizes the sample rate between digital audio devices. When multiple devices must share audio, they must agree not only on nominal sample rate but on the precise timing of each sample. Word clock provides this synchronization reference.

A word clock signal is a square wave at the sample rate frequency, typically distributed via BNC connectors at 75-ohm impedance. One device serves as the clock master, generating the reference; all other devices slave to this reference, adjusting their internal oscillators to match.

Professional interfaces like AES/EBU and S/PDIF embed clock information in the data stream, allowing receiving devices to extract clock from the audio signal. This simplifies cabling but means the clock source cannot be separated from the audio source.

Super clock and word clock variants at multiples of the sample rate (256 times or 512 times) provide higher-resolution timing references for systems that benefit from oversampled clock recovery.

Timestamp Protocols

Network audio systems use timestamp protocols to maintain synchronization over packet networks where transit times vary. IEEE 1588 Precision Time Protocol (PTP) enables sub-microsecond synchronization between network devices.

Timestamps accompany audio data, indicating the precise time at which samples should be rendered. Receiving devices buffer incoming data and render at the indicated timestamps, compensating for network delay variations. The buffer size represents a tradeoff between latency and tolerance for network jitter.

Audio Video Bridging (AVB) and its evolution Audio Video Transport Protocol (AVTP) build on PTP to provide guaranteed-latency audio transport over Ethernet. The network reserves bandwidth and enforces timing, ensuring audio packets arrive within bounded delay.

Audio Frame Concepts

An audio frame typically represents a fixed number of samples processed as a unit. Frame size trades off between latency (smaller frames mean lower latency) and processing efficiency (larger frames amortize per-frame overhead). Common frame sizes range from 64 samples for low-latency applications to 4096 samples for efficient file-based processing.

Codec frames in compressed audio may differ from processing frames in PCM. MP3 frames contain 1152 samples; AAC frames contain 1024 or 960 samples. The encoder and decoder must process in these fixed increments regardless of the application's preferred buffer size.

Overlapping frames support processing that requires context beyond the current frame. The Modified Discrete Cosine Transform used in AAC and many other codecs uses 50% overlap, where each sample is part of two consecutive frames. This overlap enables smooth transitions without blocking artifacts.

Packetization for Networks

Network transmission divides audio into packets sized for efficient transport. Real-time Transport Protocol (RTP) is the standard framework for audio packet transport, providing sequence numbering, timestamps, and payload type identification.

RTP packet size balances header overhead against network efficiency. Very small packets carry excessive header overhead; very large packets suffer greater impact from single-packet losses. Typical choices range from 1 to 20 milliseconds of audio per packet depending on the application's latency requirements.

The RTP header includes a sequence number enabling detection of lost, duplicate, or reordered packets. Receivers can request retransmission of lost packets if time permits, or apply concealment if real-time constraints preclude recovery.

Error Resilience

Audio packet systems must handle errors that corrupt or lose data. Techniques for error resilience span prevention through redundancy to concealment when errors occur:

• Forward Error Correction: Sends redundant data allowing reconstruction of lost packets without retransmission
• Redundant coding: Sends the same audio in multiple packets so loss of one packet does not lose content
• Interleaving: Distributes consecutive samples across multiple packets so single-packet loss creates distributed rather than burst errors
• Error concealment: Interpolates or repeats samples to mask errors that cannot be corrected

The appropriate techniques depend on the error characteristics of the channel and the latency budget. Broadcast systems may use extensive FEC and interleaving; low-latency communication systems may rely primarily on concealment.

Container Formats

Container formats wrap audio data with metadata and organizational structures for storage and streaming. Common containers include:

• WAV: RIFF-based container for PCM audio, widely supported but limited metadata
• AIFF: Apple's equivalent to WAV, also storing PCM audio with metadata
• MP4/M4A: ISO base media file format supporting AAC, ALAC, and other codecs with rich metadata
• Ogg: Open container supporting Vorbis, Opus, and FLAC audio
• Matroska (MKA): Flexible open container supporting virtually any audio codec
• FLAC: Native container for FLAC audio with integrated metadata

Container choice affects metadata capabilities, seeking efficiency, streaming compatibility, and decoder requirements. Platform compatibility often drives container selection more than technical merits.

HTTP-Based Streaming

HTTP-based streaming delivers audio as a series of small files or chunks fetched via standard web protocols. This approach leverages existing HTTP infrastructure including content delivery networks, proxies, and caching.

HTTP Live Streaming (HLS), developed by Apple, divides content into segments typically 6-10 seconds long, listed in a playlist file. The player downloads segments sequentially, buffering ahead to absorb network variations. Adaptive bitrate variants of the same content enable quality switching based on available bandwidth.

MPEG-DASH (Dynamic Adaptive Streaming over HTTP) provides a standardized alternative to HLS with similar segmented delivery and adaptive bitrate support. The manifest describes available representations and segments, enabling sophisticated client adaptation strategies.

HTTP streaming inherently has high latency due to segment duration and buffering. Low-Latency HLS and Low-Latency CMAF reduce segment sizes and optimize delivery chains to achieve latencies approaching 2-3 seconds, still higher than true real-time protocols but acceptable for near-live applications.

Real-Time Streaming

Applications requiring sub-second latency use protocols designed for real-time delivery. These protocols typically use UDP transport to avoid TCP retransmission delays and accept some packet loss in exchange for timing consistency.

Real-time Transport Protocol (RTP) provides the standard framework for real-time audio delivery. RTP handles packetization, sequencing, and timing; the companion RTCP protocol provides feedback on reception quality. WebRTC builds on RTP for browser-based real-time communication.

Secure Reliable Transport (SRT) combines low latency with error recovery, using selective retransmission when time permits and Forward Error Correction for guaranteed recovery. Originally developed for broadcast contribution, SRT has become popular for low-latency streaming to large audiences.

RIST (Reliable Internet Stream Transport) provides another approach to reliable low-latency streaming, developed through the Video Services Forum. RIST supports multiple profiles from simple to advanced, enabling deployment matching each application's requirements.

Network Audio Protocols

Professional audio-over-IP protocols move audio between devices within facilities, replacing traditional point-to-point audio cabling with network infrastructure:

• Dante: Proprietary protocol using standard Ethernet with sub-millisecond latency and automatic discovery. Widely adopted in professional audio installations.
• AES67: Open standard for high-performance audio streaming, providing interoperability between different manufacturer systems. Based on RTP with PTP synchronization.
• AVB/Milan: IEEE standards providing guaranteed bandwidth and timing over Ethernet. Milan builds on AVB with specific profiles for professional audio.
• Ravenna: Open standard compatible with AES67, offering additional features for broadcast applications.

These protocols enable flexible audio routing through network switches rather than physical patch bays. A single network cable can carry hundreds of audio channels, dramatically simplifying installation and reconfiguration.

Bluetooth Audio

Bluetooth provides wireless audio delivery to consumer devices including headphones, speakers, and automotive systems. Various codecs and profiles address the challenge of delivering quality audio over the bandwidth-limited Bluetooth link.

The standard SBC codec provides baseline compatibility but limited quality. Optional codecs including aptX, AAC, and LDAC offer improved quality at the cost of requiring support in both transmitting and receiving devices.

Bluetooth Low Energy Audio (LE Audio) with the LC3 codec provides improved efficiency and quality along with new features including broadcast audio for sharing and multi-stream support for true wireless earbuds. Auracast enables broadcasting audio to unlimited receivers in public spaces.