Tips on How to Recognize Lo-Fi Audio Files
Revised 17.NOV.06

The objective of this page is to provide information on how to successfully discover whether a file is from a good source (ie: original Audio CD or high bitrate MP3) or from a bad one (ie: an MP3 with a bitrate lower than 192kbps). I like to use 192kbps as a reference because I believe that such bitrate provides quality that is good enough for most (normal) people. Audiophiles, however, can only be satisfied with recordings coming out of a vinyl disc and output through a big pair of $2,000+ speakers. For anyone else, bitrates in the 192-256 kbps range should be more than enough.

It's important to point out that, the human ear can only perceive sounds within the frequency range of 20Hz to 20,000Hz (aka 20kHz). Sounds with frequencies outside this range might wake up your dog, but they will not be heard by you or any other (normal) human. For those who have already crossed the 30-years-of-age barrier, the situation gets even worse! At such age, most people cannot hear sounds at frequencies above 19kHz! That's because pristine hearing is a characteristic only commom to children. The older people get, the narrower their frequency scope sensibility becomes. In fact, many adults can't even hear sounds above 18kHz.

So to sum this up: each individual has a different sensibility to sound, but many adults will not be able to hear sounds at frequencies above 18,000Hz....and very few will be able to hear frequencies above 19,000Hz.

Armed with this information, it's easier to decide what to keep and what to re-encode. However, don't rely so much on the science of Acoustics. Some people can spend a whole day listening to music through a "walkie-talkie" and have the time of their lives! Remember: ignorance is a bliss! :-)

What you need to know about MP3 is: the lower the bitrate you choose for your encodings, the narrower will be the frequency range used in your file. MP3s encoded at 128kbps (popularly and *wrongly* referred to as "CD Quality") are in the frequency range of 20Hz-16kHz (20Hz to 16,000Hz). Lots of people have those kind of MP3s and cannot tell how bad they sound -- once again, ignorance is a bliss! MP3s encoded at 192kbps usually have a broader spectrum: 20Hz-18kHz (20Hz to 18,000Hz). That's why they sound better than their 128kbps counterparts.

Below you can see the "Spectrum Analysis" of a song encoded at 3 different bitrates. Compare their graphs to the original WAV rip and observe the different shapes their graphs can take. To run these Analysis I used "SoundForge", but there are other options out there. For example, you could use "CoolEdit 2000" (now Adobe Audition) or "SigView". On the freeware camp, you can try Audacity".

On the 1st screenshot you can see what the spectrum of the original WAV file looks like. Note that there are no frequencies missing and the graph is displayed on its fullness, from 20Hz to 20kHz. Actually, on Audio CDs the spectrum goes as high as 22kHz, which is overkill since not even the ears of a child would be able to perceive sounds at such high frequency. That extra 2kHz on the Audio CD range was purposedly added by Sony and Philips engineers when they developed the Compact Disc specification. The reason for that "extra mile" was to make sure audiophiles would have less to complain about when the discs hit the market. The move proved pointless as the audiophiles still bitched that vinyls were much better. Well, technically speaking, vinyls are better, but you gotta be Lee Majors with freaking bionic ears to be able to notice any difference between the 2 formats! :-)




On the screenshot below you see the Spectrum Analysis of the WAV rip encoded to MP3 at 128kbps CBR. Note how the graph of the 128kbps song takes a sharp dive at about 16kHz. That, again, is a characteristic commom to files encoded at this bitrate. 128kbps should be *utterly* avoided as it deprives your ears from the sounds within the 16kHz-18kHz range...Sounds that you (most likely) can still hear if you are not a grandpa. Those 2kHz that were filtered from the range during the 128kbps compression might look negligible, but they are not. With such file, you'd be missing a lot from the audible sound spectrum. Do not subject your ears to such deprivation! Stop using 128kbps for your MP3s. That is not "CD Quality"!




The screenshot below shows the Spectrum Analysis of the original WAV file encoded at 192kbps CBR. Note how the frequency range is a little wider and the graph is a bit "fuller" towards the end. That encoding sounds good and very few people would be able to tell the difference between such MP3 and the original Audio CD or WAV file. It's important to note that different songs compress at different amounts at a given bitrate. In any given album, "fast & harsh" songs encoded at 192kbps will yeld "fuller" and wider graphs than the "slow & quiet" tracks. In any case, a graph showing a sharp dive at 16kHz (see previous screenshot) usually indicates a poorly compressed song.




Despite the good quality that a 192kbps CBR encoding can yeld, I strongly suggest everyone to use an even better encoding parameter.... It's called the "Alt-Preset Standard". AP Standard is an encoding profile available on the Lame MP3 Codec.

Lame is the best MP3 codec out there and it happens to be free! MP3s created with Lame have superior sound quality (given that you use decent encoding parameters). When using Lame to backup your music, make sure you employ one of the several Alt-Preset profiles to encode your tracks. "AP Standard" offers a good compromise between quality and file size, but feel free to experiment with "AP Extreme" as well as "AP Insane", which is a major overkill. To rip CDs I use the excellent freeware CDex. Besides ripping discs, it can also convert WAVs to MP3s and vice versa. Another good Ripper/Converter is EAC, also freeware. Sorry but Windows Media Player and iTunes are for amateurs.

Now the last screenshot.... Here the song was encoded using the "AP Standard" of the aforementioned Lame MP3 Codec. Note that the spectrum ranges from 20Hz to about 19,000Hz, at which point the graph dives substantially and flattens at the inaudible level of -130dB. Also notice that, although there was a drop at 16kHz, that drop was subtle. From all the Graphs showed so far, this one is the "fullest", which means that most of the sound present in the the original WAV rip was preserved!




What about those high bitrate MP3s (192kbps and up) that sound like garbage and whose Spectrum Analysys graphs resemble those of mediocre 128kbps files? Well, there's an explanation for such aberrations. Those files were most likely ripped from a CD-R, not an original Audio CD! And it is also very likely that the CD-R was made of a bunch of low quality 128kbps MP3s. Once quality is lost, it can never be re-gained. Converting a 64kbps song to 256kbps will not yeld any quality gain whatsoever, mate!

This thought might cross your mind: "If you can get the widest spectrum by keeping the songs as WAV files, what's really the point of MP3s?" The answer is: 1) WAV files are about 7 times larger than a decent MP3; 2) You don't need those extra frequencies on the 19kHz-20kHz range because it's very unlikely your ears can notice their absence, anyway. The same thinking applies to lossless formats like FLAC and APE. They are useful for people who trade shows online, but they are a major overkill to anyone else who just need to backup their Audio CDs or transfer them to a hard drive MP3 player like Creative's Zen Vision:M, Cowon's A2 or Apple's iPoo (faddy and highly-overrated).

In sum, if a Spectrum Analysis graph shows a steep drop at 16kHz, you are "probably" looking at a low quality audio file. I say probably because again, some songs are very quiet and slow and therefore their graphs can be deceiving. So before you go about re-encoding all your audio files, grab a decent pair of headphones and listen to them first!