Post

It's Queensland Games Fest again this weekend! I'm not doing the Lightning Talks this year, but I thought I might share a bit of my talk from last year

Comb filtering is pattern of constructive and destructive phase interference between time-displaced versions of an identical audio signal, forming equally-spaced “teeth” in the frequency domain. Basically it is the resulting effect of a very short echo. I'll give an example next 😅

When I showed this at QGF, we ran into tech issues and the video was silent 😳 So I had to try to make these sounds with my mouth 😅

When you overlay two copies of the same simple sine wave, depending on the offset, the peaks will either line up (making it louder) or they will cancel out (making it quieter or silent)

Oh no, that gif didn't animate 🫠 Basically: same plus same = twice as loud, same plus opposite = silent. Any offset between these values gives an outcome somewhere in the middle

Because all sounds are made up of combinations of sine waves (thanks, Fourier!), for a more complex sound the one delay/offset will cause different phase interactions across its different constituent frequencies. So some parts get louder and some get quieter. This is what makes the comb effect

You can experience this for yourself if you place your ear near a hard wall in a noisy environment. As you get closer to the wall, you can hear the "comb" move upwards, and as you move away from it it moves downwards. You can get this effect a bit just holding your hand flat near your ear, too.

This is because you are hearing both the direct version of a sound and its reflection off another surface at the same time. The extra time it takes the reflection to travel through the air means the reflected sound arrives at your ear later than the direct sound.

If the distance is great enough, you hear this delay as an echo, but in most indoor environments the delay is so short that instead you hear it as comb filtering! Each metre of extra distance in the reflected sound's path adds roughly 3ms of delay

In games, we'll sometimes hear comb filtering when it isn't intended, and this is very jarring. Maybe you smash a bunch of crates at the same time, and the game tries to play back the same smashing sound sample multiple times at once. Even a delay of less than 1 frame is significant in audio.

But we can also do comb filtering on purpose! Your brain is really good at interpreting comb filtering as acoustic information about your environment, and calculating the correct delay for a comb filter can give players an intuitive understanding of where they are in a space.

Plus we can use it to dynamically simulate everyday sounds like distant aeroplanes or exploding fireballs. For example, I synthesised jetpacks and flamethrowers for Isopod with pretty much nothing but white noise generators and comb filters, and I think they turned out pretty good!