Artifacts Of Compression

The conversation of data compression often comes up in regards to audio. Do I compress? How Much? What file type?

There are probably as many opinions out there as bit rates. So, how’s one to tell what’s best? Well, as data compression goes, it’s always a trade-off between size vs. quality. The general rule states that the higher the bit rate, the closer to the source quality the file will be. As bit rates increase, so does file size.

For the last 25 or so years the Compact Disc has set the standard for consumer audio. Though studios often record using higher standards, 16 bit audio at a sample rate of 44,100 kHz has been the way listeners have enjoyed music. With the adoption of the internet and the surge in file sharing during the Napster days, the MP3 format became dominant. The MP3 brought down the size of audio files, enabled sharing over the internet and at the same time, promised quality with little compromise.

As time moved on, disk space grew, hard drive prices dropped and audio snobs began complaining about quality. Arguments slipped into the general public about data rates, compression artifacts and those would be snobs began spreading the word about FLAC. Most of the conversation revolved around music collections, mostly from ripped CD’s. In fact, most didn’t complain about web audio. But when it came to their personal CD collection, all civility flew out the window.

So, as someone with average ears, a consumer playback system and most likely no idea what a bit rate is, how were you supposed to have an opinion? And why? If the audio sounded good, it probably was. Even better if you could email it… right?

Well, finally there’s a proven way to test the theory. It’s called ABX testing, and you use your own ears. The general idea is that what sounds good often is, and in a blind test, if you can’t tell the difference – why bother.

ABX audio testing is a way for a user to load two identical song files, one as control (often CD quality) and one compressed (often using MP3). The listener plays track A, then track B. He attempts to discover the difference in quality between A and B and then plays track X. If X is like A he votes that way, likewise for similarity to B. Often the test is run 10 times with results hidden until the end. Once the listener finishes, a score is given using a percentage of 100. The closer to the full percentage (100%) he gets, the greater ability he has at hearing the differences between A and B, and as such, between the compressed and uncompressed audio. If the user ranges near 90%, it’s likely that he has great ears, -or- the difference between A and B is great… say a CD quality track and one that is compressed at 64 k/bits – or both.

Below is a tutorial (For Mac).

Download ABXer ( ) and install.

Grab a CD from your shelf, this is important. You might not be able to tell what bit rate your existent iTunes tracks are, so starting fresh is always best.

CD’s notable in audio excellence:

  • Everything Must Go – Steely Dan
  • So – Peter Gabriel
  • Joshua Judges Truth -Lyle Lovett
  • Who Is Jill Scott – Jill Scott
  • Aenima – Tool
  • Spirit Of Eden – Talk Talk
  • The Dark Side Of The Moon – Pink Floyd

Beyond the above titles, you can pick really anything rich in sound, but it helps immensely if you’re very familiar with the track.

Put the CD in the computer and before you open iTunes, click on the CD icon on the desktop. A new window will appear with the track names and numbers. Highlight the track your interested in and drag it off to the desktop. It should copy there.

Once the copy from the disc is complete, open iTunes. After a short bit the CD will likely show in the track list window and an icon of a CD will appear in the left side bar with the album name. If this is the case, then you’re set. If not, please highlight the CD icon in the sidebar, showing the contents of the CD.

CD Icon In Side Bar

Now comes the compression settings part. What bit rate you choose depends on several factors. If the audio destined for the web, for personal listening or archival. As far as bit rates go, whether it’s AAC or MP3, sizes generally go 128, 256 and 320. The greater the bit rate the higher quality the audio file and the less compression used. But data size also increases. Because hard drives are cheap, internet speeds are relatively fast and quality is important, I suggest a rough starting point of 128 kbits for web directed stuff and a bit rate of either 256 or 320 for your music collection. Archivists will likely choose ‘Lossless’, though for our purposes this won’t enter into the conversation.

With iTunes open, head up to the menu bar and select ‘iTunes’ then ‘Preferences’.

Select Preferences

Then ‘Import Settings’.

Import Settings Dialog

Then in the two drop-down menus choose ‘MP3’ and ‘Custom’.

Custom MP3 Option

A new dialog box should pop up. Let’s choose ‘128 kbps’ as the bit rate, leave VBR unchecked, sample rate of ‘44.100 kHz’, channels set to ‘Auto’, stereo mode to ‘Normal’ and check the lower two boxes ‘Smart Encoding Adjustments’ and ‘Filter Frequencies Below 10 Hz’.

MP3 Settings Window

Click OK. Then click OK again to accept the previous dialogue box, the OK once more to exit the preferences window.

Back in iTunes with the CD icon selected and the CD track list showing, highlight the same track you previously copied off the disc to your desktop. Right-Click that file (control + click on Mac) and select ‘Create MP3 Version’.

Create MP3 Dialog

This should take a bit, the progress bar will show up in the main iTunes window above. Once completed ( a green check next to the track will indicate so), click on your main Music library icon in the left sidebar, up top left.

Music Tab In Sidebar

In the search bar up on the top left, type in the name of the track and let iTunes search for it. It should show now in the main window. If in fact you’re already ripped the CD a while ago and you find another copy there, head up the the Name/Artist bar above the track listing.

Artist Bar

Right-click on that bar (control + click on Mac) and select ‘Bit Rate’ as an option. Once selected you’ll see that option appear next to Artist, Album, Name, Genre and so forth. Because you’ve previously selected to import the track at a bot rate of 128 and used MP3 you’ll now see that next to the track in iTunes. Highlight that track and drag it off to your desktop.

Now the fun begins. At this point you should have two tracks, same name but different file extensions (.wav or .aiff and .mp3). Don’t worry if the track names aren’t identical, as long as it’s the same track you’re OK. There might be a number pre-pending the name, maybe not. But the name of the track itself should be the same.

Go ahead and open ABXer now. Once the main window is up you’ll need to select your files. The two browse boxes up top are where you include those two tracks on your desktop.

Two Track Selection

Select the two tracks, one for each dialog box. Which order doesn’t matter. the program will randomize the tracks for you. Importantly, you’ll need to make sure that the file extensions are correct and don’t match. This is crucial! One should be a .wav or .aiff and the other an .mp3.

File Extensions

Once selected click the main ‘Start Run’ button in the middle of the window. This leaves you with the main ABXer test window.

Main ABX Window

The test number shows up top. The two tracks you inserted are represented by ‘Play A’ and ‘Play B’ and the ‘Play X’ is the randomized middle. You should now play A and B quite a few times. Listen closely for audio differences, bass response, pristine highs, digital artifacts, hollowness, muddiness and crackle. The cleaner sounding track will be the CD copy and the artifacted one the MP3 version. Once you’ve played A and B a few times, now play X. Is X more like A or B? If like A then select that box below, same for B. Repeat this for each take, playing A and B to discern the difference then playing X and finding what it resembles most.

Final Results

The final results page pops up and shows you your score. Like I mentioned above, a higher percentage means either you have good ears and can tell the difference or that the quality of the compressed version was so different from the original that it was easy to tell. This of course get’s much harder as you increase bit rates.


In the end, what sounds good IS good. However, many of us haven’t trained our ears accordingly and though a listening test will help, some basic guidance will also aid in making decisions. In general, the more complex the audio, the greater the bit rate should be. Music is often best compressed at a higher bit rate, whereas voice over, lectures and audiobooks can get away with greater compression.

CD bit rates run at 1401 kbps, while MP3’s are variable, our test running at 128 kbps. Below is a file size comparison.

File Sizes

As you can see, the file size difference is great. Often a bit rate of 128 kbps will yield a 11:1 ratio of data reduction!

Many will offer advice, some will swear by their techniques, however, most discussion is just hyperbole. Most people can’t tell the difference! Now, while this is true in most cases, lower bit rates DO show when playback systems rise in quality and size. As the size of the speakers increase, the deficiencies that compression brings about start to rear their head. Most listeners will enjoy music using earbuds, home HI-Fi’s and in the car. These environments are often littered with competing sounds, compromised dimensions and cheap or small woofers and amplifiers, as such, artifacts rarely show. But, that doesn’t mean they aren’t there.

No one can tell you what you should do, but the playback environment and hard drive space DO come into question. If the audio is destined to be consumer on the web, a lower bit rate will ensure faster streaming. If the listening environment is a home 5.1 system, then a higher bit rate will ensure a more robust sound.

Compress as you see fit, and when someone wants to give you their 2cents, tell them that you’ve undergone listening tests and that you’re confident that you now know what you’re talking about!