SIGNALS

You call them. They put you on hold. That's bad enough. Then they play the cheesy on-hold music. That makes it worse. But even that's not as bad as the actual sound quality of the music. Oh. My. Lord.

The sound quality of on-hold music makes the sound of a dentist's drill seem as soothing as the sweet whisperings of your lover cradled in your arms. I am not exaggerating. On-hold music is absolutely terrible. Underwater, garbled, muffled, swirly, distorted. We can land people on the moon. We can put cheese in an aerosol can. Why can't we get decent music on a phone?

The answer is both simple and complex. But let me preface the explanation by saying that realistically, at least for now, there is no solution. So don't get your hopes up.

The simple answer is this: phones are made for talking, not for playing music. That's why while voice quality is marginal, music quality is abysmal.

The more complex answers are numerous and interestingly, it seems like many people misunderstand the biggest culprit. You often hear that bandwidth limitations are the problem. Yes, high frequencies are removed, and low frequencies are filtered as well. Yes, this affects music fidelity, but it's the least of your worries.

Another oft-cited problem is that the music is not properly recorded or conditioned for phone transmission. Yes, music with a high dynamic range poses greater problems, but again, that's not really the central issue.

The biggest problem refers back to our simple answer: phone systems are designed to convey speech. In fact, they are ruthlessly optimized for that task. Specifically, the codecs that encode speech and then reproduce it use models of the human vocal tract for that job. They are only designed to "hear" speech sounds, and then "speak" those sounds. If they encounter other kinds of sounds (like music sounds) they have no good way to encode or reproduce those sounds. They try to convey music as if it was speech.

Here is a thought experiment: You, as a human, hear music. But you can only interpret music sounds as if they are speech sounds. Now you must reproduce those sounds, using only your voice-optimized larynx. The sound of the music you convey is terrible because you were never designed to do that. Now repeat the experiment with several talkers in a row. This accounts for the fact that the audio path will travel through different systems, and thus in and out of tandem codecs, adding to the misery. In addition, these codecs were designed decades ago, and were necessarily crude because of hardware and software limitations of the time. And yesterday's crude designs have standardized today's poor performance.

Still other factors conspire to make things even worse. For example, phone systems relentless try to distinguish between voices and background noise. Assuming they are noise, they attempt to remove any noises that are not identified as voice sounds. And a lot of music content, at least to those algorithms, sounds like noise. Uh oh! There is no end to the degradation. For example, some systems are not only optimized for speech, they are optimized for particular languages, and thus not for others. That can make music sound even more weird.

Well, you get the picture. In the quest to optimize the system for speech, music is screwed. A system that could successfully convey both would require way too much transmission bandwidth, engineering finesse, and industry cooperation (ie. cost). And that cost could never be justified - aside from on-hold music, how often do the systems convey music?

I suppose on-hold music could be composed in a way to better survive phone transmission, but it would have to be very voice-like. Or I suppose on-hold music could always be a cappella songs. Hmm, instead of trying (unsuccessfully) to convey on-hold music, how about the system reads you Shakespeare sonnets. Or a grocery shopping list. Or even Signals blogs. Anything but music.