Digital Audio Quality and MP3s
Digital Audio Standards
Since its market presence in 1981, the de facto standard for digital audio has become "CD Quality".
Yet, for some audio industry applications, this quality level is not always good enough.
However, since it is a quality level that is commonly known, it makes a good baseline for comparisons.
For the record, "CD Quality" audio is 44,100 samples per second of 16-bit audio data in 2 channels (stereo).
It doesn't matter how many times your CD player oversamples (reads, checks, and checks again - 8x - 10x - 20x) the audio data, or how many extra "error correction 20-bit" calculations are made.
The final output is going to be 16-bit audio.
On a Windows PC, without professional audio hardware, the output also is limited to 16-bit, 2 channel stereo.
When supported by your software, the internal calculations and audio manipulation can be done on 32-bit audio data, but the final output is still going to be 16-bit audio.
This limitation is due to a) the limits of the existing hardware and b) the limits of Microsoft's DirectSound API.
Stereo vs. 'other-than-stereo'
By manipulating the audio information, via software or hardware, it is possible to enhance output to more than 2-channel stereo.
Typically, this information comes from an audio source that has MORE information than 2 discrete channels. Dolby Digital is one example of this.
Another approach is commonly referred to as 'stereo spacializers' that separate the frequencies of the audio signal.
This modified audio information is split between the two channels to give an audio effect of 'space' or 'difference' between the left and right channel.
Nothing is really added to the audio signal, but the existing signal is modified.
Another approach is to mono-mix the output, so that the left and right discrete channels are merged into a single output channel.
Stereo separation and phase are lost, but only a single channel needs to be amplified.
In some environments, like dinner music in a banquet hall, or top down music in a large ballroom, this type of output sounds better than a two-channel stereo signal.
Also, when stored on a computer's hard drive, mono music takes exactly 50% as much space as stereo.
Variables to consider
You must include all of the possible variables when discussing audio quality and computer audio.
These variables include
- Original Source Music
- Compression (Storage)
Original Source Music
Newer music is digitally mastered with an intended format of CD audio.
This is not the case with older source material.
Many CDs, in fact, are simply digital copies of old analog tapes.
In fact, with a computer and a CD-R, you could make a CD audio track of an off-key shower singer who is singing along to the echos of a cheap radio in a tile bathroom.
The moral of this story is that final output can never be better quality than the original recorded source.
The most reliable way to get undistorted music from a CD onto a computer's hard drive is by digitally extracting the music.
Also, some sound cards allow you to digitally record data from a CD, MD, or DAT via a digital I/O connector.
In other cases, the data must come from an outside source and be converted into a digital format via the ADC (Analog Digital Converter) on your sound card.
The quality of this recording method varies based upon the input level, the recording level, and other mysterious factors.
Read the page about digital extraction for more information on recording.
At Audio Box, there is a very specific level of "quality" that is referred to as the first level of compression.
The first level of compression is defined as "the first level of audio compression that allows at least one percent of the listening audience to hear the difference between a CD player source audio and computer source audio."
Ideally, when switching between computer source audio, and other audio sources, the listener(s) should NOT be able to distinguish a difference.
If any difference is perceivable, the computer should sound better.
Audio Box has conducted numerous tests with a wide variety of audiences, rooms, music types, and equipment configurations.
With a given audience of twenty or more people, the following results have been observed:
- Statistically, at least one listener will be present who can distinguish the difference between CD source audio and computer source audio AT ANY COMPRESSION LEVEL HIGHER THAN 3:1.
- Statistically, two or more listeners will be present who can distinguish the difference between CD source audio and computer source audio AT ANY COMPRESSION LEVEL HIGHER THAN 6:1.
- A majority of listeners can distinguish the difference between an MP3 audio file, at any compression level, and an uncompressed audio file.
- A majority of listeners can hear quality loss from a TAPE RECORDING of an MP3 file at any compression level higher than 10:1.
In summary, the first level of compression is 3:1 or 6:1 compression, depending upon how sensitive your listening group is.
This definition presumes a "perfect" listening environment.
Obviously, a less-than-perfect listening environment introduces other factors that can lower audio quality.
While reviewing listener comments, groups of listeners had common things to say about different audio compression levels.
The written feedback about each compression format was usually consistent about what audio traits listeners heard or could not hear.
These comments include:
-The music sounds flat, like someone took the detail out of it.
MPEG Layer 2 (12:1)
-The high frequencies are muddy.
-The music has lost important details.
MPEG Layer 3 (12:1)
-The highs are very sharp...
-The mid-range is missing, but the highs are reproduced well.
MPEG Layer 3 (24:1)
-The music sounds like a transistor radio.
When you play a CD, the signal is transmitted through the left and right outputs.
Sometimes, in the case of a professional deck, you can manipulate pitch (frequency), speed (tempo), or add other effects to the audio.
On a computer, the output doesn't have to be ANYTHING like the source music stored on the hard drive.
With professional software, you can do all of the above manipulations and more: put all output in the left channel, switch the output channels, amplify the output, silence the output, add an echo, etc.
How does this affect quality?
When changing the qualties of the "stored" music, you are actually creating "new" audio data.
For example, if you amplify the music prior to output and increase the pitch by 5%, you introduce the potential for digital distortion due to overprocessing.
By doing all internal processing with 32-bit audio data, the potential for digital distortion is lowered.
Ultimately, on today's computers, the final audio output must be 16-bit.
Before the final output, however, all Audio Box products do the processing/mixing/manipulation with 32-bit audio.
Companies that rely on Microsoft's DirectSound architecture are limited to processing 16-bit audio data.
Therefore, audio manipulation introduces more audio distortion and artifacts.
From the hard drive, the music data is sent by the audio software to the software mixer.
The volume level(s) of the mixer determine how the data is mixed and sent to the sound card.
In the sound card, the digital audio data is converted into analog data.
The DAC (Digital Analog Converter) on your sound card varies a great deal from card to card and chip to chip.
Also, laptops tend to have smaller, cheaper chips of lower quality than full sized computers.
In 1998, a number of companies released sound cards with extremely high quality PCI audio cards that include high quality DAC chips.
Three of these companies are Creative Labs (SBLive!), Turtle Beach (Montego II), and Diamond Multimedia (MX-300).
The signal-to-noise ratio on these cards easily rivals that of amatuer MD and CD players.
Also, these cards tend to introduce less of the low frequency hum that laptops and ISA sound cards bring to your PA system.
From the sound card, the audio data is sent to your mixer and/or amplifier, and then finally to the speaker(s).
Obviously, different qualities of sound cards, mixers, amplifiers, and speakers affects the final quality level of the music.
You may wonder how the listener may affect the overall perceived audio quality of computer audio.
The reality statisticlly proven by listening tests is that listeners become accustomed to hearing a certain type of sound.
In fact, dance groups that have become accustomed to computer audio will regularly score the 'perceived audio quality' from computer HIGHER than CD audio!
Years ago, the debate was "vinyl sounds better than a CD - in my opinion."
A few years later, it was "I don't like MiniDiscs because I can hear the difference the compression makes."
Today and tomorrow, it might be "I don't like computer audio because I am MPEG sensitive."
What is MPEG sensitive?
Well, the MPEG compression technology relies on an accoustical model that makes assumptions about how people hear.
As compression levels go up, less and less audio data is given - based upon these assumptions.
When the compression has removed enough information from the signal, you HEAR the audio artifacts.
This may be the sensation that the music is stopping/starting all the time, it is missing elements of certain frequencies, or it contains one of a number of other 'audible artifacts'.
Since perceived audio quality is a personal interpretation, and individual listeners hear the same digital information differently, the level of MPEG compression that produces audio artifacts varies from person to person.
To make matters worse, listeners become accustomed to a certain type of sound, and any variation of that sound (even an improvement) usually is perceived as a loss of quality.
Size vs. Quality
The moral of the audio quality story is simple: bigger is better.
Compression may be useful for sending tunes over the Internet on a slow connection, but extreme compression has no business being hooked up to a PA system.
The solution is to get a bigger hard drive and to carry fewer songs.
Luckily, drives are getting bigger and less expensive, and music files are staying the same size.
If you are interested in conducting a less expensive self-test for compression and MPEG sensitivity, you can try these two tests:
Radio signal compression vs. CD
If you have a CD player in your car, tune in a station that is likely to play a song you like from a CD that you own.
Listen to the radio, and have your CD loaded and ready.
When they play the song you have loaded, listen to the radio for a few seconds, then press play on the CD.
The perceived "quality gain" that you should hear represents the difference of quality between CD audio and radio quality.
The FM signal quality is significantly lower than the quality allowed by modern computer compression technologies, so the change in quality should be very evident to the human ear.
For this test to be accurate, the volume level of the radio and the CD must be equal.
A difference in volume introduces other factors that change the perceived audio quality.
MP3 vs Cassette Quality
If you are tired of hearing how MP3 files are "CD Quality," you can test your ears to see if you can tell that MP3 is LESS THAN TAPE QUALITY.
Record a song onto your computer, but don't compress it.
Make a cassette recording of this song.
Now, compress the song to each available "CD Quality" level your encoder has.
Typically, these bitrates would include 128, 112, and 96 kbps.
Record the MP3 files onto another cassette.
Now, using a dual-deck system, switch back and forth between playing the recording of the uncompressed music and the MP3 encoded music.
Be your own judge, and see if you really can hear the difference.
Now, if you can hear the difference on a cassette tape, shouldn't you ask yourself how close to "CD Quality" the compression format you chose is?
Compression technology is moving ahead, and hard drives are getting bigger.
Someday, a hard drive (or other data storage device) will hold so much audio that music won't need to be compressed, because the storage media will have so much capacity.
In the meantime, we must listen to the rhetoric of different companies that are selling their compression technologies.
The thing to remember is that these technologies are usually focused on sending audio data across the Internet with maximum compression, not on preserving maximum sound quality.
Personal portable MP3 players are already available.
Also, car stereos based on the MP3 and TWINVQ (VQF format) compression technologies are becoming available to consumers.
Maybe the industry will train its listeners to become accustomed to the poor sound quality of extremely compressed audio.
Perhaps we all will become accustomed to a new compression technology that has yet to be introduced to the mass market.
Audio Box products are designed to support every mainstream audio compression format.
This allows you to choose from the best compression formats to fit your needs.
As compression technology advances, Audio Box will continue to develop the tools and applications that assure access to the best available formats.