Transcoding lossy formats

Q: When I convert my AAC songs on iTunes (from my original CDs) into MP3, is the result the same as if I had imported the original CD into MP3 format in the first place?

– George

A: The process of converting digital audio files between formats or bit-rates is referred to as transcoding, and this term applies regardless of whether you are converting between two different formats (ie, MP3 and AAC), or simply converting to a different bit-rate in the same format (ie, 192kbps MP3 to 128kbps MP3).

In this case, it is also important to note that we are talking about “lossy” formats. This means that the encoder actually removes audio information from the track in order to produce the resulting compressed MP3 or AAC audio file. This is normally information representing audio frequencies and harmonics that most people cannot hear anyway, although the loss of audible quality will increase dramatically at lower bit-rates.

Transcoding between lossy formats is generally considered a bad idea in terms of quality, since the results will not be the same as if you had re-ripped the track from the original CD.

This may run contrary to what some people expect, since with digital audio the encoding should be a consistent process being run by a computer. If you rip the same audio CD to the same bit-rate with the same encoder, you will always get the same results. The encoder doesn’t make subjective or random decisions when encoding audio—it simply follows a pre-defined algorithm.

There are two problems with transcoding that will result in reduced quality, however:

The first issue has to do with the fact that different encoders will make different decisions about what audio data gets “thrown out” when performing the compression (remember that lossy encoders actually discard audio data). Since the differing encoders use different algorithms, when you transcode between two lossy formats, you are actually getting the worst of both encoders.

In other words, you may find that the AAC encoder has discarded audio information that the MP3 encoder would not have. Unfortunately, once that information has been discarded by the AAC encoder, there’s no way for the MP3 encoder to get it back without going back to the original CD. In practical application, these differences should be relatively subtle, but they do make a difference, particularly at lower bit-rates.

The second issue has to do with something called “artifacting.”  No lossy encoder produces a perfect sound, and in the compression process, audio “artifacts” are generally introduced. These are imperfections in sound quality—effectively sounds and frequencies that were not present in the original recording. In essence, the compression process is actually adding (or at least distorting) sound.

Even if this compressed file is returned to an original WAV file (ie, burned to a CD), these artifacts will remain part of the audio. A reanalysis of the audio, whether from a burned CD or through transcoding, is going to pick up these artifacts. This will skew the reanalysis of the file in question, thereby producing a different result.

While the first issue will not affect transcoding between bit-rates with the same encoder, this second issue affects all types of transcoding between lossy formats.

In reality, the quality loss is not directly cumulative, however, nor is it as dramatic as some would suggest. There is a lot of misinformation regarding transcoding, and some will try to suggest that if a 192kbps AAC file contains 50% of the audio quality of the original CD, then a transcode to 192kbps MP3 would produce a file that is only 25% of the original quality (50% of 50%). This is not accurate, however, as there is a high degree of overlap between the information that the various lossy encoders will discard. In fact, it is not really possible to put a specific measurement on the quality loss, but it is not as high as the percentages that some often quote.

Therefore, while there is a theoretical loss in quality, it may not be perceptible, depending upon the bit-rates that you are encoding from and to, the encoders themselves, and your own ears and listening equipment. Audio quality is highly subjective, and we recommend that users conduct their own listening tests to determine what formats and methods work best for their own preferences. The best way to do this would be to take a track that you already have in your iTunes library in AAC format and convert it to MP3. Then, rip that same track from the CD directly to MP3 format. Give them both the same name and other track information, and then listen to them in whatever your typical listening conditions would be, without knowing which one is which, and see if you can hear a difference. This may not be as sophisticated a method as the “blind ABX tests” that you may have heard of, but is usually sufficient for most people.