Custom Double-Precision Bessel

Ainulindale V2 replaces android's truncated polynomial approximations shortcuts with a custom, double-precision Bessel that iterates until absolute machine epsilon convergence. By eliminating the numerical rounding drift and approximation decay common at high β parameters, the engine fully unlocks the pure mathematical potential of the Kaiser windowing architecture. Stopband attenuation is no longer an estimated compromise, instead it satisfies unyielding, predictable audiophile specifications. The real-world acoustic result is a windowing profile capable of delivering razor-sharp transition bands and absolute deep-floor rejection of high-frequency imaging artifacts. This pristine isolation ensures that delicate spatial imaging cues, phase micro-dynamics, and ultra-low-level harmonic nuances remain perfectly intact rather than bleeding or smearing across the frequency spectrum.

High-Fidelity Reverse Format DC-Blocking Filter

To clean up incoming music streams, Ainulindale V2 introduces a real-time, single-pole DC-blocking IIR filter directly injected at the reverse format conversion layer. DC offset — a silent, 0 Hz subsonic displacement caused by poor masterings or clock drift, quietly robs the digital space of crucial headroom and introduces asymmetric waveform distortion. V2 strips away this underlying voltage flaw across 16-bit, 24-bit, and high-fidelity 32-bit audio at absolute hardware velocity. Using a fixed psychoacoustic pole constant it provides absolute immunity against thread-local race conditions across multiplexed playback threads, maintaining a pristine subsonic cutoff across all sample rates without risking timing errors or audio stuttering. This ensures zero precision loss across all audio formats, along with deep, musical sub-bass passing through completely untouched and phase-transparent.

Triangular Probability Density Function Dither

Additionally, Ainulindale V2 introduces a mathematically pure Triangular Probability Density Function (TPDF) dither engine. When high-resolution floating-point audio is forced into a lower bit-depth container without dither, the system abruptly truncates the signal. This creates quantization error that correlates directly with the audio waveform, manifesting as harsh, gritty harmonic distortion. TPDF dither completely breaks this correlation by summing two independent uniform random noise sources, perfectly linearizing the quantization process. To maximize perceptual dynamic range, this architecture integrates a 2nd-order closed-loop error feedback network. Rather than leaving a flat noise floor, the loop continuously isolates the quantization error and psychoacoustically shapes its spectrum — shifting the dither energy away from the ear's most sensitive mid-range frequencies and pushing it into the near-ultrasonic region above 15 kHz. This ensures that whenever android enforces a bit-depth reduction, ultra-low-level acoustic tails (like a decaying cymbal or reverb) dissolve seamlessly into absolute blackness, allowing the hardware to cleanly resolve micro-details sitting deep below the theoretical floor of the target bit depth.

Deferred Gain Staging

Furthermore, Ainulindale V2 implements an advanced Upstream Unity-Gain Interception architecture to combat the severe resolution loss caused by sub-optimal system volume scaling. Natively, when system volume is lowered, android multiplies the audio data by a tiny scalar coefficient early in the pipeline, squashing the digital signal into the lowest bits of its word length before mixing occurs and destroying critical dynamic range. V2 eliminates this bottleneck by caching the intended volume level and forcing unity gain into mixer core. This ensures the entire software stack processes the audio with maximum bit-depth precision. The actual volume reduction is deferred to the absolute final rendering threshold, paired with a first-order high-pass noise-shaping filter. By shifting the inevitable rounding errors away from human hearing sensitivity and into the ultrasonic spectrum, the soundstage maintains its holographic depth, keeping transient clarity pure and fatigue-free even at whisper-quiet listening levels. Because modern hardware abstraction layers natively enforce a transparent, 0 dB hardware-level unity gain baseline, this architecture is completely immune to destructive double attenuation or gain cascading distortion.

Studio-Grade Soft-Saturation Limiter

Finally, Ainulindale V2 completely bypasses and overrides android's native peak limiter — a rigid, piece-wise brickwall implementation that violently flattens transient peaks, collapsing the soundstage and inducing listening fatigue. In its place sits a True-Peak Aware Soft-Saturation Limiter driven by a continuous, transcendental wave-shaping function and backed by a 2x oversampled inter-sample tracking. Standard digital limiters are blind to what happens between samples. When the digital signal is reconstructed into analog audio by a DAC or compressed into a lossy codec, the smoothed waveform can overshoot the digital ceiling, causing harsh inter-sample distortion. V2 solves this by using a high-precision 4-tap half-band polyphase interpolation filter to calculate the hidden peak values between samples. Any signal below a transparent -1.0 dBFS threshold passes through 100% bit-perfect and untouched. The moment a native or inter-sample transient overshoot is detected, the engine tracks the minimum required gain reduction across both domains, engaging a smooth, analog tape-style transfer curve that elegantly rounds off the peaks into a safe 0 dBFS ceiling. By mapping the saturation knee with mathematical slope continuity, it completely eliminates the harsh, jagged edges of digital hard-clipping. Instead of generating a brittle spray of high-frequency aliasing, it transforms raw signal spikes into warm, low-order odd harmonics that mimic the natural saturation characteristics of premium vintage studio gear.