What's all this 'fidelity' stuff about anyway?

Nearly everybody has heard the term "high fidelity".  Most understand that fidelity is a desirable thing to have, and many believe that the word 'fidelity' is synonymous with 'quality'. In any casual (non-technical) setting, they are the same.  But what does an audio engineer mean when he speaks about fidelity?

Audio quality is a subjective measurement that is different for each person.  We all know quality when we hear it, but when I listen to a 128k bitrate mp3, I may hear beautiful music, while somebody that is used to 320k files will think it's terrible.  Yet another person might be fine with 96k.  A bit of sound (a signal) can have hissing, scratching, ringing, echoes, and any number of other effects (collectively known as artifacts) that can affect quality.  But how do you quantize that?  Obviously a sound with a lot of unpleasant artifacts is of lower quality than one that doesn't have them, but can you turn that into numbers?  

Okay, yeah, that was a rhetorical question.  The answer, of course, is yes.  Over the years, audio engineers (people much smarter than me) have invented a number of characteristics and properties of a signal (metrics, in scientific parlance) that can be calculated mathematically, and that indicate how good or bad the signal is going to sound.  These metrics, taken collectively, determine the fidelity of the signal.  Over the next several articles, I'm going to go over several of the more important fidelity metrics and try to explain in fuzzy detail what each metric means, what a signal sounds like when you get it right (or wrong!), and why it is important for high fidelity.

Audio index

Comments

• Anonymous
  June 22, 2006
  Welcome to the blogosphere.  I look forward to reading what you have to say.
• Anonymous
  June 24, 2006
  when I listen to a 128k bitrate mp3, I may hear beautiful music, while somebody that is used to 320k files will think it's terrible
  
  You bring up a good point.  The concept of fidelity applies equally strongly to listening devices.
  
  Consider a VOIP session - each user has a speaker and a microphone.
  The quality of the sound you hear is limited not only by
  * the fidelity of your speaker
  * the lossiness of the internet transport
  but also by
  * the fidelity of the other person's microphone
• Anonymous
  June 27, 2006
  Maurits: Exactly true.  A lot of end-to-end scenarios are like that.
  
  The "easiest" part is the middle software, because software is so malleable, because bandwidth is getting so cheap, and because compression is getting so good at picking which parts can be safely lost and which are important.
  
  Speakers are likewise getting better all the time.  A speaker that has been optimized for full-bandwidth music is going to output a voice chat pretty well.
  
  Now, if only we could start raising the quality on microphones...