Now THIS is some interesting audio from the era of the MP3. (I’d never heard this before!)

Earlier this month, episode 1047 of The Ongoing History of New Music explored the rise and fall of the MP3. Part of the story involved an explanation into how the Fraunhofer Institute struggled to rid their compression algorithm of errors, glitches, and all manner of distorted audio. They found that compression full songs down by a factor of 10 or 12 was far easier that shrinking the sound of a single human voice.

Why? The principles of psychoacoustics.

All audio comes at us in layers and layers of sound. We tend to only hear the sound “on top” (i.e. the louder sounds), which completely mask the quieter sounds in the environment. The MP3 algorithm compresses digital files by removing audio that the ear and brain can’t perceive. But what if there’s only audio “on top?” What if there’s nothing being masked? Standard psychoacoutic theories are to get harder to use.

Famously, the Franuhofer Institute calibrated their algorithm by listening to Suzanne Vega’s acapella version of her song, “Tom’s Diner,” released in January 1984. It’s a very simple recording of just Vega’s singing, which means if the algorithm was going to get all glitchy, this was the song that would reveal flaws in the math.

The Fraunhofer people listened to this version of “Tom’s Diner” thousands of times as they worked things out. When they got it right, they knew that the MP3 had a chance to be the world standard for audio compression.

Cool, right? But how did they do this? If psychoacoustic theory broke down, what did the Fraunhofer people strip out to make “Tom’s Diner?” I never even thought about this until I received an email Karl, who is working on a computer science degree. He’s studying MP3s and took a very deep dive into the work of the Fraunhofer boffins.

He sent me this cool piece of audio: It’s what the Fraunhofer people stripped out of “Tom’s Diner” to turn it from a full-sized .wav file into an MP3.

Thanks, Karl!