Ken Pohlmann, who started the first master's degree in music engineering in the U.S., believes understanding music is critical for EEs who design audio products.
That's because EEs can design a circuit or write software that perfectly meets the spec sheet, but an audio engineer might turn up his nose and say it's unlistenable. "We try to attune engineers to these issues," says the professor at the University of Miami in Coral Gables.
He's looking at low bit rate coding, used for MP3s and many other audio and video files. "With a CD, you're usually using 16-bit quantization, but with low bit rate coding systems, you're coding with 2 bits on the average, so you've only got a 12 dB signal to noise ratio. That's the price you pay for reducing file size by 8:1," he says.
By bringing in disciplines outside of electronics and software, like psychoacoustics, EEs learn to use techniques like masking, where a loud sound like the desired note masks the softer noise sounds. It's the job of the algorithm to pick the techniques that provide the best results.
Pohlmann sees perceptual coding as the great encoding technique of our time. The iPOD and DirecTV use LBRC while HDTV uses similar techniques in compressed video.
The Director of the Music Engineering Technology program thinks it's ironic this research comes as storage capacities and bandwidth are soaring. So why is data compression necessary?
"The reality is LBRC will always be needed, no matter how fast broadband becomes. Even as storage capacities get larger and processing power goes up, people want more in less space."
However, he thinks researchers have probably hit a wall with known techniques. His group here has filed patents for LBRC that use a more structured approach as opposed to today's frame-by-frame approach. "They take in, say, 8 msec chunks and analyze them. We take in longer frames and analyze them," says Pohlmann.