NOTICE: The user assumes the entire risk as to the accuracy and the use of this information and the links provided.
This document may be freely copied and distributed subject to the condition that it may not be distributed for profit.

Podcasting and MPEG-1 and MPEG-2 Layer III Audio

Synopsis: Frank's thoughts on the use of MPEG-1 and MPEG-2 Layer III audio files (.mp3) in podcasting, with a suggestion that bookmarkable MPEG-4 AAC-LC (.m4b) files are a more appropriate choice for the application.

Note: The original URL of the Web page containing the table shown on this Web page was
<http://www.opticom.de/producer/prod_prof_spec.htm>.

The original Web page is no longer available, but an archived copy can be found on the Internet Archive WayBack Machine Web site at the following URL.
http://web.archive.org/web/20030205035600/http://www.opticom.de/producer/prod_prof_spec.htm

OPTICOM GmbH, which still exists, was the company that was responsible for selling the software, notably MP3 encoders, produced by the Audio & MultiMedia (AMM) Group at the Fraunhofer IIS (Institut Integrierte Schaltungen or Institute for Integrated Circuits) in Germany where MPEG Layer III audio coding was originally developed.

Contents:

MPEG-1 and MPEG-2 Layer III Audio.

Detailed Specifications of MPEG-1/MPEG-2 .MP3 Modes (Table 1)

Why Podcasts (often) sound so bad.

Winamp screenshot (Figure 1)
WinPlay3 screenshot (Figure 2)

Making Podcasts sound better.

Standard MPEG Test File and Sample Files.

Summary of Common MPEG-4 File Types.

MPEG-1 and MPEG-2 Layer III Audio.

From a purely file format point-of-view, there is no such thing as an "MP3 file". We have MPEG-1 files, MPEG-2 files, and MPEG-4 files, but no MPEG-3 files as there is no such thing as MPEG-3. The "MPEG-3" designation was skipped (project abandoned). Next is MPEG-7, followed by MPEG-21. For the current status of MPEG projects, please see the following Web page.

ISO/IEC JTC 1/SC 29 Programme of Work
http://www.itscj.ipsj.or.jp/sc29/29w42911.htm

What we commonly call an "MP3 file" (that is, a file with a ".mp3" extension) is, from a strictly file format point-of-view, actually an MPEG-1 file or an MPEG-2 file that just happens to contain MPEG Layer III audio data (as opposed to MPEG Layer I audio data or MPEG Layer II audio data). For those who are familiar with the internal format of MPEG (Moving Picture Experts Group) files, the specific file type, either MPEG-1 or MPEG-2, can be easily determined by examining the header information, located at the beginning of the MPEG file, in a hex editor program. For Windows users, I heartily recommend Ian David Mead's excellent UltraEdit-32.

Also, many MP3 player programs provide this information. Nullsoft's Winamp 2.91, for example, can do this. While playing back an .mp3 file, simply perform an Alt-3 keyboard sequence to bring up Winamp's MPEG file info box + ID3 tag editor window to display the file information. Fraunhofer's old MP3 player program, WinPlay3, displays this information in its main UI (user interface).

It should be noted that there are some number of files available on the public Internet which, although they have extensions of .mp3, aren't in fact MPEG-1 or MPEG-2 Layer III files at all. Typically, they are instead MPEG-1 Layer II files and therefore really ought to have a file extension of .mp2. The majority of users will be unaware of this discrepancy due to the fact that most programs capable of decoding Layer III audio also support Layer II audio and will play such misnamed files without issuing a warning message.

Which type of file, MPEG-1 or MPEG-2, a given (properly named) .mp3 file is, is dependent upon the encoding parameters chosen at the time the file was created. The parameters which determine whether the MP3 encoding software will produce an MPEG-1 file or an MPEG-2 file are the sampling rate (measured in kHz or kilohertz - thousands of cycles per second), the datarate (or bitrate, measured in kbps or kilo bits per second - thousands of bits per second), and the number of audio channels (one channel for monaural or two channels for stereophonic).

The table below shows the relationship between the MPEG-1 and MPEG-2 datarates (bitrates), sampling rates, and number of channels as used for Layer III audio (.mp3 file) encoding. Ignore the "MPEG-2.5" portion of the table, as this, like mp3PRO encoding, never became a part of any of the official, published, MPEG ISO/IEC standards. For this reason, and owing to potential playback problems, files conforming to the MPEG-2.5 or mp3PRO specifications should not be posted on the public Internet for general consumption by an undefined class of users.

Table 1: Detailed Specifications of MPEG-1/MPEG-2 .MP3 Modes

As can be seen in the table above, the original MPEG-1 specification (ISO/IEC 11172-3), in its support of stereo encoding, only defined high bitrates (96 kbps through 256 kbps) and high sampling rates (32, 44.1, and 48 kHz). The purpose of including additional Layer III definitions in the MPEG-2 standard (ISO/IEC 13818-3) was to extend the permissible bitrates and sample rates downward. Thus, the MPEG-2 Layer III standard, which was published a couple of years after the original MPEG-1 standard, supports stereo bitrates as low as 48 kbps and stereo sampling rates of 16, 22.05, and 24 kHz.

Those who are interested in learning more about the technical characteristics of MPEG audio files are directed to the following article by Konrad Windszus posted on The Code Project Web site. Available for free download (registration required) is a Windows application program (source included) called MPEG Audio Info that can parse most MPEG audio files, including MPEG-1, MPEG-2, and MPEG-2.5 Layer III files, and display their technical parameters. MPEG-4 (.mp4, .m4a, .m4b) files are not supported, however.

CodeProject: MPEG Audio Frame Header. Free source code and programming help
http://www.codeproject.com/KB/audio-video/mpegaudioinfo.aspx

I highly recommend this program to anyone who regularly works with MPEG Layer I, Layer II, or Layer III audio files.

Why Podcasts (often) sound so bad.

There are two primary reasons why podcasts are so often lacking in audio quality. One is that they were recorded and edited by audio amateurs in ordinary domestic (home) environments using consumer-grade equipment instead of by audio professionals in properly designed, constructed, and maintained sound studios using professional-grade equipment, although this situation is showing improvement as higher quality equipment becomes available at more affordable prices and as more podcasters strive to improve the production values of their offerings. The other is that so many of them are encoded as (low bitrate, low sampling rate) MPEG-2 Layer III files.

Using an old Adam Curry DSC (Daily Source Code) podcast* as an example, upon examination we find that the file is an MPEG-2 Layer III file with a 22.05 kHz audio sampling rate, a datarate (bitrate) of 64 kbps, and is 2-channel stereo. The technical characteristics of this file are not atypical of what is found in the podcasting world. (Adam's more recent podcasts are MPEG-1 Layer III encoded with an audio sampling rate of 44.1 kHz at a datarate of 128 kbps and therefore sound much better than the example cited here.)

* Source: http://homepage.mac.com/dailysourcecode/DSC/DSC-2005-01-14.mp3 (No longer available.)

Shown below is a screenshot taken from Nullsoft's Winamp 2.91 program while playing the file DSC-2005-01-14.mp3 mentioned above.

The notation MPEG 2.0 layer 3 indicates that the file format is MPEG-2 and the audio encoding is Layer 3. The notation 64kbit, 80852 frames indicates that the file has a 64 kbps datarate (bitrate). The notation 22050Hz Joint Stereo indicates that the file has a 22.050 kHz audio sampling rate and is 2-channel stereo.

Figure 1: Winamp screenshot

Shown below is a screenshot taken from Fraunhofer's WinPlay3 V2.3b5 program while playing the file DSC-2005-01-14.mp3 mentioned above.

The notations II and III indicate, respectively, that the file format is MPEG-2 and the audio encoding is Layer III. The notation 64kBit/s indicates that the file has a 64 kbps datarate (bitrate). The notation 22 kHz indicates that the file has a 22.050 kHz audio sampling rate while the dual red loudspeaker symbols indicate that the file is 2-channel stereo.

Figure 2: WinPlay3 screenshot

Making Podcasts sound better.

Compared to MPEG-2 Layer III, a better choice, in terms of sound quality, would be to encode as (higher bitrate, higher sampling rate) MPEG-1 Layer III files, although this would have the very undesirable side effect of increasing the file size and besides, MPEG Layer III audio is a 14-year-old technologically obsolete technology that was long ago surpassed by the likes of RealAudio and Windows Media Audio (just to name two) anyway. In my opinion, it's high time that MPEG Layer III was retired.

If we really wanted to improve the sound quality of podcasts, then an even better choice than an MPEG-1 Layer III file, and one that would seem to befit the leading-edge nature of podcasts themselves, would be to produce AAC-LC (Advanced Audio Coding - Low Complexity) MPEG-4 files. In fact, an even more ideal situation would be to use a file extension of .m4b on these AAC-LC-encoded MPEG-4 files so that iPod players can provide bookmarking capability* for the files, as well as offer the ability to play them faster or slower than normal speed.

* Reference: http://www.hutteman.com/weblog/2004/09/12-212.html

It should be noted that a Windows user can play an .m4b file on their computer using Apple's free QuickTime Player program, although at this time only a few portable players, aside from the iPod itself, can play MPEG-4 audio files. Over time, however, this situation is expected to change since MPEG-4 AAC (ISO/IEC 14496-3) is the natural successor to MPEG-1 Layer III and MPEG-2 Layer III. An example of this is the Sony PSP (PlayStation Portable), which can play not only MPEG-4 AAC-LC-encoded .mp4 audio files, but also MPEG-4 Part 2 SP-encoded .mp4 video files (ISO/IEC 14496-2) and MPEG-4 Part 10 AVC-encoded .mp4 video files (ISO/IEC 14496-10 / ITU-T Rec. H.264) as well.

Windows users can also use Microsoft's Windows Media Player (WMP) program to play .m4b files provided that the appropriate DirectShow filters are installed on the system. I use this (1,303,407 byte .zip archive).

.m4b files can also be played back in the free VideoLAN VLC media player program, which is available for a wide variety of platforms including Mac, Windows, and Unix/Linux.

For additional information, please see the following document.

What is MPEG-4 AAC?

Adobe Acrobat Portable Document Format (PDF) file.

Dated April 4, 2003 - English - 2 pages - 44,072 bytes

Standard MPEG Test File and Sample Files.

The .zip file listed below contains a copy of the standard MPEG audio test file in the form of a two-channel stereo, 16-bit, 44.1 kHz, LPCM (Linear Pulse Code Modulation), RIFF (Resource Interchange File Format) .wav file. The wave file has zero DC Offset and a running time of 90 seconds. The file consists of a series of short musical selections that are considered to be difficult for lossy compressed audio codecs to encode without introducing audible artifacts and other distortions noticeable to the average listener. The file includes a short segment from Suzanne Vega's "Tom's Diner", discussed in a video presentation with Dr. Jürgen Herre over on the Fraunhofer site.

MPEG Standard Audio Test File 
.zip file = 14,178,454 bytes — .wav file = 15,876,084 bytes

The eight sample MPEG audio files listed below were encoded using the above .wav file as the source. They all have a 44.1 kHz audio sampling rate with the exception of the 64 kbps MPEG-2 Layer III .mp3 file, which has a 22.05 kHz sampling rate. I've indicated the file type (file extension) and file size in bytes for each of the files.

• MPEG-1 Layer I at 448 kbps — .mp1 file = 5,040,032 bytes
  
  
• MPEG-1 Layer I at 384 kbps — .mp1 file = 4,320,028 bytes
  
  
• MPEG-1 Layer II at 384 kbps — .mp2 file = 4,320,862 bytes
  
  
• MPEG-1 Layer II at 224 kbps — .mp2 file = 2,520,503 bytes
  
  
• MPEG-1 Layer III at 256 kbps — .mp3 file = 2,878,245 bytes
  
  
• MPEG-1 Layer III at 128 kbps — .mp3 file = 1,437,399 bytes
  
  
• MPEG-2 Layer III at 64 kbps — .mp3 file = 717,184 bytes
  
  
• MPEG-4 AAC-LC at 96 kbps — .mp4 file = 1,103,818 bytes

For the sake of completeness, it should be mentioned that the AAC codec was originally introduced as part of the MPEG-2 standard (ISO/IEC 13818-7), although MPEG-2 AAC files are rarely encountered these days. At the time of its introduction, the AAC algorithm was defined as being NBC, or Non Backward Compatible, because existing MPEG-1 audio decoders couldn't handle the format. Also, in the early days of the AAC codec, it was common to work with files that had an .acc file extension. Sometimes these .aac files contained an ADIF (Audio Data Interchange Format) or ADTS (Audio Data Transport Stream) header, while at other times they were headerless and contained only a raw ACC-encoded audio bitstream. Working with AAC data is much easier when the data is contained within an MPEG-4 container (wrapper).

Summary of Common MPEG-4 File Types.

• .mp4 — Strictly speaking, the .mp4 file extension simply indicates the use of an MPEG-4 container (ISO/IEC 14496-14), although such files usually contain MPEG-4 encoded audio/video data, either audio-only, video-only, or audio+video. Some vendors have taken advantage of the MPEG-4 Part 14 container specification and use the .mp4 file extension on files that don't contain MPEG-4 audio/video bitstreams. An example of this is the .mp4 files produced by the Sony PMW-EX1/PMW-EX1R and PMW-EX3 XDCAM EX format camcorders. These three camcorders produce .mp4 files that contain lossy compressed MPEG-2 video data and non-compressed LPCM audio data.
  
  
• .m4a — An audio-only MPEG-4 file.
  
  
• .m4b — An Apple-proprietary invention, the .m4b extension indicates a bookmarkable MPEG-4 audio file. Such files are given special treatment, such as the ability to set bookmarks and resume playback at the last break point, set playback speed, and are ignored in shuffle mode. The .m4b extension is usually used on long-playing, spoken-word, audiobook files.
  
  
• .m4p — An Apple-proprietary invention, the .m4p extension indicates a protected MPEG-4 file. Protected in this sense means use of Apple's proprietary FairPlay DRM (Digital Rights Management) technology within the file. Such files are usually purchased in Apple's iTunes Music Store (iTMS).
  
  
• .m4r — Another Apple invention, this extension is used to designate an audio-only MPEG-4 AAC-LC file that's intended for use as a custom ringtone on the iPhone.
  
  
• .m4v — A video-only MPEG-4 file, although this extension is also sometimes used to indicate an audio+video file, in which case the intention is to convey the fact that the file is not audio-only. The video stream within an .m4v file can conform to either the older MPEG-4 Part 2 standard (ISO/IEC International Standard 14496-2) or the newer, more modern and more efficient MPEG-4 Part 10 AVC standard (ISO/IEC International Standard 14496-10), also published by the ITU-T as Recommendation H.264. The video encoding method utilized by the widely deployed and very popular Panasonic/Sony AVCHD video format (including Panasonic's AVCCAM format and Sony's NXCAM format) is based upon the AVC/H.264 standard, although the AVCHD format itself is not formalized by any standards body. In contrast, the video encoding method utilized by Sony's high-end HDCAM SR video format is based upon MPEG-4 Studio Profile, part of the older MPEG-4 Part 2 standard (ISO/IEC International Standard 14496-2).

It's important to remember that in order to successfully play .m4a, .m4b, and/or .m4v files in certain programs and/or on certain devices, it's necessary to first change (rename) the file extension to .mp4. The Sony PSP is an example of such a device.