Loading Polly and the canyon...

Parrot Echo: how sound bounces back

Press Call! Polly shouts, the sound spreads as waves, hits the cliff and comes back as an echo. The wait is 2 x distance / speed of sound. Drag to orbit, wheel or pinch to zoom.
Echoes at Polly (list from the image-source formula)
Microphone at Polly: 1-D wave grid (cyan), predicted echo times (orange dashes)
Pressure along the line from the back to the front cliff
The science

Speed of sound. In dry air sound travels at c = 331 + 0.6 T m/s (T in degrees C), the textbook linear fit of c = sqrt(gamma R T_K). It is 343 m/s at 20 C, slower in cold air, faster in hot air, and it does not depend on pitch or loudness.

Echo delay. The sound goes to the cliff and back: t = 2 d / c. A wall 100 m away at 20 C gives 0.583 s. The echo is a separate sound only if it arrives after the call has ended (plus roughly 0.05 s for the ear, a typical value): about 17 m for a 0.1 s call, 5 m gives a blurred "reverb".

Two cliffs. Between parallel cliffs at +d1 and -d2 the echoes can be written with image sources at x = 2 d1, -2 d2, 2 d1 - 2(d1+d2)... Each arrival time is (distance of the image from Polly) / c and each wall multiplies the pressure by R.

Wave grid. The pages integrate the linear acoustic equations dq/dt = -c du/dx, du/dt = -c dq/dx (q = p/rho c) on a staggered grid (leapfrog, Courant number 0.9) between the cliffs. A wall of reflection coefficient R is an impedance boundary u = q (1-R)/(1+R). The call is a sin2-enveloped 4-cycle wave of length Tc, a slowed and stretched picture: a real call is kHz and its wave is thousands of times shorter than what could be drawn.

What the 3D picture shows. The rings on the ground and in the air slice are the spherical wave of the call and of each image source, p = Rk w(t - r/c) / sqrt(1 + r/r0), summed (the spreading is softened so far rings stay visible). It is the exact method-of-images solution for two flat parallel walls; the grid and the pictures agree on arrival times. The 1-D grid has no spreading loss; the list of levels uses 1/distance spreading (level relative to the call heard at 1 m) and ignores sound absorption by air, ground reflections, wind and the fact that real cliffs scatter sound.

Checks that do not reuse the renderer. The page re-runs the wave grid with a short call and compares: the echo time with 2d/c, the echo strength with R, the arrival time at a different pitch (it must not change), the second canyon echo with the image-source value, the speed measured on the 2-D map with c, and the fit 331 + 0.6 T with the thermodynamic formula.

Simplifications. Polly is drawn about 30 times larger than her real size relative to the distances (distances are on the markers). Sound is shown in slow motion (slider); the real delay is in the readouts. No wind, no temperature layers, no ground echo. The audio is synthesised; echo loudness in the audio is a listening curve, not part of the physics. Models: parrot, canyon, stump, marker built in Blender.

Made with 🟠 Blender · Three.js