 |
Introduction
The issue of "what encoder to use" was answered
with the accepted norm as well as some semi-scientific "proof". The
Fraunhoffer MP3 Producer 2.1 codec is regarded as one of the best compression codecs (if
not the best) for converting WAV files to MP3 format. For [virtually] CD-quality
compressed audio, you are offered to use bitrates of: 112kbit/s, 128kbit/s, 160kbit/s,
192kbit/s, and 256kbit/s.
Methodology
Soundclip: INOJ: Love You Down (chorus
segment)
Duration: 7.2s
Reasoning: I needed a track w/ a good amount of treble. It was previously
known that Producer was accurate for the audio spectrum below 16 KHz. After that is where
the differences between bitrates lie. Why 7.2 seconds? To generate the spectrum analysis
charts below using the entire song would mean having my P2-300 processing away for
literally hours. It already took nearly a minute just to make the charts for the
7.2 seconds of audio.
Process:
- Encode the clip to WAV-MP3.
- Load WAV-MP3 to Cool Edit Pro
- Generate graph.
- Overlay MP3's graph over original WAV.
- Repeat for each bitrate tested.
Results
In the graphs,
sky blue
denotes the particular MP3 tested, and
red denotes the original CD footprint.
112kbit/s 8% of original size
1:12 ratio |
And this is why many in the MP3 community frown upon 112kbit
encoding. The dropoff occurs around 15 KHz. The ever-so-hated Xing
(and company) codec produces similar results dropping off at 16 KHz.
Bottom line? Don't even bother with 112kbit/s encoding--the
negligible compression increase is not worth the loss in quality. |
128kbit/s 9.1% of original size
1:11 ratio |
Well, this is better. No dramatic drop-offs here. There is faithful
reproduction through and beyond the 16 KHz point. No wonder this is regarded as the
recommended bitrate for personal MP3 archival. For an extra 16kbit/s over 112kbit/s you
get a very close reproduction of the original with losses mostly in the higher end treble. |
160kbit/s 11.4% of original size
1:9 ratio |
The use of 160kbit/s has escalated greatly in recent months. This
can be attributed to its improved resolution in the 17KHz-plus range. |
192kbit/s 13.7% of original size
1:7 ratio |
 |
Going to 192, you start to see the trend of better high-end reproduction.
This is best shown above left; the purple line is the 160kbit/s version while the blue is the
192kbit/s. Notice how much more audio data is gained in the 17-20 KHz region. Of course it
does come at a price. These files weigh in at 120% the size of
160kbit/s files and 150% the size of their 128kbit/s counterparts. |
256kbit/s 18.2% of original size
1:5 ratio |
As they say, this is "as good as it gets." This one is
nearly identical to the original right through the 17 KHz range. And even up to that
point, it is very faithful to the original--not dropping off until the 21 KHz range, which
you'd have to be a damn drop dead audiophile to squabble about! (No offense to drop dead
audiophiles, of course :) |
The Bottom Line
To wrap things up, I made an overlay of all the graphs and color
coded them in a nice rainbow fashion. The warmer it is, the closer to the original. With
this, I leave the "what should I..." question up to you. At all costs, avoid
112kbit/s. Spend the extra kilobytes and improve your treble appreciably by encoding
with at least
128kbit/s. To add a noticeable kick to the treble, I would try 192kbit/s. For optimal
archival quality, definitely go with 256kbit/s.
~Fin~
|