Skip navigation

I’ve recently gotten back into high end audio equipment with the purchase of a pair of Sennheiser HD 800 headphones. I’ve had a quality set of home shelf speakers for many years now but they aren’t something I can take into work or anywhere for that matter. For anyone who’s still with me on this, you know the difference a pair of high end speakers or headphones can make in the sound and clarity of music compared to even expensive ear buds or car stereo speakers. There really is no comparison.

The downside to this is that any of your music that has been compressed into MP3, M4A, etc. sounds thin, hollow and lifeless if recorded well to begin with. This has inspired me to begin to rebuild much of my music library that isn’t already in FLAC and along the way I thought I’d run a spectrum analysis of a song to document various levels of compression and the inherent data loss they incite.

Audio Sprectrum

The top graph represents the FLAC 44.1kHz sample, the middle represents 256kbs Mp3, and the bottom represents ~174kbps VBR Mp3. The frequency cutoff, banding and definition loss are obvious just looking at the graph and if you know the song well even more so upon listening. It is true you won’t be able to notice much of a difference withoput high quality audio components but given today’s high availability of cheap storage there is no good reason not to store your catalog in lossless or at least low-loss formats (>256kbs). Need to fit alot of music onto a smaller mobile listening device? Encode the music from the lossless sources when you need to, keeping the original as a backup (example).