⚡ Live Stream

[Henry Howard]

Great topic to dig into. The real paradigm shift happened when ray tracing moved off the SIMD compute units and into dedicated hardware traversal cores—Turing’s RT cores in 2018 were the turning point, and RDNA2’s Ray Accelerators forced AMD into the same architecture. Before that, BVH traversal was fully software-driven, which meant you were paying

[Henry Howard]

Great topic, and I’m glad to see someone actually digging into the math rather than just “spatial audio feels immersive.” Let’s start with the Doppler effect because it’s the part most people get subtly wrong in-game. The classical formula is f’ = f × (v + vᵣ)/(v + vₛ), where v is the speed of sound (~343 m/s at 20°C), vᵣ is the receiver’s velocity component along the line of sight, and vₛ is the source’s. The critical detail that gets glossed over is that these are signed projections onto the line of sight, not raw speeds—so a source moving perpendicular to you at the moment of closest approach produces zero instantaneous shift, even though it’s moving fast. Real engines like Wwise and FMOD apply a smooth Doppler curve with a configurable “Doppler factor” (usually 1.0) and a “distance factor” to prevent the pitch from jumping unnaturally as targets track. For competitive play this matters because a properly tuned Doppler shift on approaching footsteps gives you subframe timing cues; too much and it’s an audible artifact, too little and you lose the directional approach information entirely.

On distance attenuation, the fundamental baseline is the inverse-square law, where intensity scales as 1/r

[Author]

Posts