Encoders 128 kbit/s

Last listening test in this group was finished 14 July 2006. According to the results all contenders could be divided into four groups:

• MPC, LC-AAC(Winamp), OGG and LC-AAC(iTunes) showed similar results: 5.05-5.45 points
• CT HE-AAC(Winamp) and new Nero Free AAC encoders are slightly better: 5.92-6.10 points
• Good old mp3 by the latest recommended Lame is just slightly behind the main group: 4.13 points
• ATRAC3 (LP2) is the last with 3.77 points

Taking into account that Lame used less bits (112 kbit/s) for coding SoundExpert test samples, its performance gap doesn’t look so dramatic. Thanks to Lame developers MP3 quality is still comparable at the most popular bitrate – 128 kbit/s.

Despite the fact that Coding Technologies (CT) HE-AAC encoder from Winamp is on the top of 128 kbit/s group, the use of SBR at this bitrate is still questionable. It needs more complicated and thus energy consuming decoder which is still rarely found in portable players where this bitrate is commonly used. The same time new Nero Free AAC encoder achieved almost the same results using standard LC-AAC profile (and slightly less bitrate to be precise).

For preserving as much sound quality as possible @128 kbit/s SoundExpert recommends compressing music with these coders:

• AAC family encoders both LC and HE profiles from Nero and Winamp (except iTunes aac encoder)
• MPC

Disclaimer:

1. The above recommendations relate exclusively to perceived audio quality of encoders but not their popularity, compatibility with software/hardware players or any other features which can also be considered in practice (tag support for example).
2. The above recommendations refer exclusively to the encoders which have been already tested at SoundExpert and which ratings are listed below. If you think the list is incomplete, please drop us a line with your suggestion. After the new rating become reliable SoundExpert recommendations will be revised accordingly.

How to read the ratings

The ratings are updated live as new participants add their grades to the system. A rating bar consists of the following elements:

- Device or technology being tested.

- Value of actual perceived audio quality (rating) which is also indicated by a bar length . Anchor points could be interpreted as follows:

In most cases using this device/technology:

1.0 – you will hear heavily distorted sound
2.0 – you will hear unpleasant sound artifacts
3.0 – you will hear distinctly audible but tolerable sound artifacts
4.0 – you will hear faintly discernible sound artifacts
5.0 – you will not hear any sound artifacts
above 5.0 – all sound artifacts will be beyond threshold of human perception with corresponding perception margin

and - the high and the low of a rating. As each device is tested under nine different sound items, there are nine different local ratings for a device. In fact, the actual rating is an average of those nine local ratings. The highest and the lowest ones are indicated. Big gap between them means that sound quality of device/technology is not consistent enough. It will vary with type of sound material: music of different genres and complexity, voice with or without music, noisy/clear recordings etc. The lowest local rating is more important in this sense as it indicates worst case behavior of tested device.

- Reliability of rating. It is also indicated by the color of bar - more reliable ratings have less percentage values and darker bars. Reliability depends on number of grades returned by participants. In most cases 5% or less is OK.

- Ruler for convenient measuring of highs and lows.

For devices with small impairments (not audible in ordinary listening tests) SoundExpert amplifies their sound artifacts to some predefined extent. Ratings of such devices are calculated analytically taking into account both the grades received and the amplification applied. They are above 5th grade on the scale showing certain quality headroom of such devices.