What is the MP3 format?

MP3 is a lossy audio compression format. MP3 is based on the fact that the human ear can only hear/perceive certain combinations of audio frequencies. It removes the parts of the audio that you wouldn't normally miss. This means that once a song is encoded, you can't easily "go back" to the original WAV or AIFF.

MP3 encoders can produce files which are 10-40 times smaller than a WAV or AIFF file. This is significant because downloading a 40 MB WAV file from the Internet will take 4 hours (on dialup), whereas a 1 MB MP3 file will take 5 minutes.

What is an MP3 encoder?

An MP3 encoder is a program that "listens" to your audio (for instance, your WAV file or your AIFF file) and compresses it into the MP3 format.

An MP3 encoder uses a psychoacoustic model (how your brain interprets sound) to guess which parts of the audio can be removed or transformed, without significantly altering what you hear.

Are all MP3 encoders the same?

The quick answer is no.

Some encoders are optimized for quality, and some are optimized for speed. Some are not optimized at all. Some are over-optimized. Most have "default" settings that make music sound like crap. The most common sound problems with MP3 files are distortion, artifacting, and too-narrow band passes.

What is distortion in an MP3 file?

Distortion happens during the encoding process when "bits" of the original signal are lost. Human ears can only perceive a certain level of sound, so it's generally acceptable to take out some of the "bits" of the signal. However, if you take out too many "bits", you can make certain songs or instruments sound scratchy or grainy.

What is artifacting in an MP3 file?

When the encoder is trying to decide which "bits" are necessary and which "bits" are not, it might pick the wrong ones, or too many. You might hear this as a "watery" effect or a "warbly" effect. If you listen to your MP3 in a noisy environment, this might not matter much. However, later you will listen to it on headphones and think to yourself that the song sounds watery or warbled.

What is too-narrow band pass?

The human ear is generally considered to perceive audio in the 20Hz - 20000Hz range. Some people can hear lower frequencies (down to about 16KHz) and some people can hear higher ones (up to about 22KHz).

An encoder will sometimes assume that a low threshold will be "acceptable" to most humans. It will then cut off your signal at 16Khz, 18KHz, etc. Why? Because it wants to squeeze just a tiny bit more juice out of your audio, which can result in a song that doesn't sound as clear or as bright as it was on CD. You might think the song sounds "flat."

On the flipside, the encoder might assume you're not going to play your music on a high-end system with big sub-woofers. It might cut off the lower bass frequencies, below 100Hz, 80Hz, etc. When you listen on headphones you probably won't notice. However, in your car stereo, you will definitely notice that the bass no long has any kick to it.

How do I make MP3 files that sound as good as the original audio?

MP3 compression has been around for about 10 years. During that time, a lot of people have made objective and subjective tests of encoding techniques. A few years back, there was a very helpful website called r3mix.net. Unfortunately, they folded. However, their legacy remains in places like afterdawn.com and mp3tech.

Essentially, it comes down to this: an opensource project known as LAME is the best MP3 encoder on the planet, but only if you use it correctly.

How do I download LAME?

Unfortunately, due to legal defintions, LAME is not "technically" an encoder, so they don't provide the executable on their website. You can get the executable from this site in Russia.

Is LAME some kind of really difficult to use DOS program?

Yes, and no. It's actually built for the command line (either UNIX or DOS). It's not really hard to use, if you're comfy with the command line. However, if you prefer graphics and checkboxes and buttons and things, you should get one of the frontend programs for LAME.

What is the best front-end for LAME?

There are two, and both are free. RazorLame is a user friendly front-end with several "presets" built in. It's easy to use. Get a version of RazorLame that has LAME bundled with it from mp3-tech.org.

Also, dbPowerAmp Music Converter is much more powerful, but it's a lot harder to use.

Why do my MP3 files still sound dull, or warbly?

You probably used the default settings that came with your encoder. The default settings are probably "fast" settings, which leave a lot to be desired in terms of sound quality.

Why do you want to spend the next 10 years of your life listening to crappy MP3 files just because it took 1/2 the time to encode them? That doesn't make sense. Use the high quality settings, and spend the extra 5 minutes now so that you'll enjoy your MP3 files for years to come.

What is the best size/quality setting for RazorLame?

This question will form endless debate. However, there is a preset built into RazorLame that will do exactly this.

Open RazorLame => Edit => LAME Options => Load Options

Then choose either Archive Quality CBR (r3mix.net).rlo or Archive Quality VBR (r3mix.net).rlo.

It will take a little bit longer to encode these files, but the sound quality is worth it.

*** WARNING ***

Make sure you uncheck the box that says Delete source file after processing on the Advanced tab. Unless you want to permanently lose your WAV file, of course.

What is the best size/quality setting for dbPowerAmp?

Go into Compression Settings (locations vary by version -- but make sure you have selected MP3/Lame as your output format). Go into Advanced Options. Under preset, select R3Mix for VBR files. For CBR files, select the bitrate to 192Kbps, then go into Advanced Options and select Constant Bitrate and slide quality to the highest setting.

What is the difference between CBR (constant bitrate) and VBR (variable bitrate)?

Constant bitrate requires that, no matter what the audio signal sounds like, the MP3 will always use the same compression ratio over the entire audio passage.

Variable bitrate means that the encoder will pick a higher compression ratio if it thinks the passage "sounds" fine, or a lower one if the passage is distorted. The compression ratio can change from moment to moment, to allow higher and lower bitrates in the same file.

Should I use CBR or VBR?

Let's use two examples.

Pretend there is a period in your song which is silent for 5 seconds. A constant bitrate (say 256Kbps) would make that 5 seconds the same bandwidth (256Kbps) as the entire song, which could end up wasting a lot of valuable bandwidth or hard drive space. Variable bitrate would look at that same 5 seconds of silence and realize that it can fit into a smaller bitrate (say, 32Kbps). Over that 5 second period, the VBR passage used one eighth (1/8) of the bandwidth as the CBR passage.

Now, pretend there is a 10 second intense passage in your song. A constant bitrate (this time, let's use 128Kbps) would force that passage to 128Kbps, and that section might then sound warbled and dull. Variable bitrate would look at that passage and determine that it would "sound" better at 256Kbps. In the case of this 10 second passage, your VBR passage uses up twice (2x) the bandwidth as the CBR passage.

In the end, the size of the files end up being approximately the same. The difference is that the variable bitrate sounds better because it intelligently allocated the bandwidth instead of mindlessly wasting it, like constant bitrate.

In almost all situations you should use VBR. The only two exceptions are 1) when the player requires CBR (for instance, some hardware MP3 players don't like VBR files) or 2) when a company requires CBR (for example, mp3.com used to require 128Kbps files and would reject all other formats).