Click here to Skip to main content
15,882,209 members
Articles / Desktop Programming / MFC

MPEG Audio Frame Header

Rate me:
Please Sign up or sign in to vote.
4.89/5 (95 votes)
12 Apr 2007LGPL317 min read 895.7K   20.8K   245   140
An article about the MPEG audio frame header.

Screenshot - MPAInfo.gif


  1. Introduction
  2. MPEG Audio Frame
    1. MPEG Audio Frame Header
    2. CRC Checksum
    3. VBR Headers
      1. XING Header
      2. VBRI Header
  3. Additional Tags
  4. Using The Code
  5. Links And Miscellaneous
  6. History

1. Introduction

This article is about the structure of the MPEG audio frame header including the XING and VBRI headers. The aim is to estimate the duration of the MPEG audio file as exact and fast as possible. The article does not include any hints on how to decode/encode the actual audio data. MPEG audio files exist in different layers. The most common is the MPEG 1 Layer III (also known as MP3), as it has the most sophisticated compression technology.

I know that there are other articles about the MPEG audio frame header (even on CodeProject), but I will go into it a little bit deeper.

2. MPEG Audio Frame

An MPEG audio file consists out of frames. Each frame contains a header at its beginning followed by the audio data. This audio data always contains a fixed number of samples. There currently exists three layers of MPEG audio, which differ in how the audio data is encoded in the frame, although they all have the same header format. The frame itself consists of slots. In Layer I, a slot is always 4 byte long, in all other the layers a slot is 1 byte long.

Additional to the layers there are also three versions of MPEG audio, which differ in the sampling rate they can handle (see table 2.1.2). MPEG 1 (ISO/IEC 13818-3) and MPEG 2 (ISO/IEC 11172-3) are ISO standards. MPEG 2.5 is an unofficial extension of MPEG 2 to support even lower sampling rates. MPEG 2/2.5 is also known under the abbreviation LSF, which stands for Lower Sampling Frequencies. Each version can handle the three layers. If you want to know more about the technical details of an MPEG audio file please have a look at the specifications. You can find them and many other useful information about MPEG at

A file can be encoded either with a constant bitrate (CBR) or with a variable bitrate (VBR), which means that each frame can have a different bitrate. Therefore, the quality of those files is often higher than files encoded in constant a bitrate mode, because they can use higher bitrates where the music needs it.

2.1. MPEG Audio Frame Header

The header at the beginning of each frame is 32 bits long and has the following format. The bit 0 in the header is the most significant bit (MSB) of the complete header. Note that the position is zero-based; position, length and example are all in bit-format.

2.1.1 MPEG Audio Frame Header
011Frame sync to find the header (all bits are always set)1111 1111 111
112Audio version ID (see table 3.2 also)

00 - MPEG Version 2.5 (unofficial extension of MPEG 2)
01 - reserved
10 - MPEG Version 2 (ISO/IEC 13818-3)
11 - MPEG Version 1 (ISO/IEC 11172-3)

132Layer index

00 - reserved
01 - Layer III
10 - Layer II
11 - Layer I

151Protection bit

0 - protected by 16 bit CRC following header
1 - no CRC

164Bitrate index (see table 2.1.3)1001
202Sampling rate index (see table 2.1.2)11
221Padding bit

If it is set, data is padded with with one slot (important for frame size calculation)

231Private bit (only informative)1
242Channel mode

00 - Stereo
01 - Joint Stereo (Stereo)
10 - Dual channel (Two mono channels)
11 - Single channel (Mono)

Note: Dual channel files are made of two independent mono channels. Each one uses exactly half the bitrate of the file. Most decoders output them as stereo, but it might not always be the case.


Mode extension (Only used in Joint Stereo)

(see table 2.1.6)

281Copyright bit (only informative)1
291Original bit (only informative)1

00 - none
01 - 50/15 ms
10 - reserved
11 - CCIT J.17

The emphasis indication is here to tell the decoder that the file must be de-emphasized, that means the decoder must 're-equalize' the sound after a Dolby-like noise suppression. It is rarely used.


The sampling rate specifies how many samples per second are recorded. Each MPEG version can handle different sampling rates.

2.1.2 MPEG Versions and Sampling Rates
Sampling Rate IndexMPEG 1MPEG 2 (LSF)MPEG 2.5 (LSF)
0044100 Hz22050 Hz11025 Hz
0148000 Hz24000 Hz12000 Hz
1032000 Hz16000 Hz8000 Hz

The bitrates are always displayed in kilobits per second. Note that the prefix kilo (abbreviated with the small 'k') doesn't mean 1024 but 1000 bits per second! The bitrate index 1111 is reserved and should never be used. In the MPEG audio standard there is a free format described. This free format means that the file is encoded with a constant bitrate, which is not one of the predefined bitrates. Only very few decoders can handle those files.

2.1.3 Bitrates (in kilobits per second)
Bitrate IndexMPEG 1MPEG 2, 2.5 (LSF)
Layer ILayer IILayer IIILayer ILayer II & III

In MPEG 1 Layer II, there are only some combinations of bitrates and modes allowed. In MPEG 2/2.5, there is no such restriction.

2.1.4 Allowed bitrate and mode combinations
BitrateAllowed modes
32single channel
48single channel
56single channel
80single channel
224stereo, intensity stereo, dual channel
256stereo, intensity stereo, dual channel
320stereo, intensity stereo, dual channel
384stereo, intensity stereo, dual channel

For the calculation of the frame size, you need the number of samples per MPEG audio frame. Therefore, you can use the following table:

2.1.5 Samples Per Frame
Layer I384384384
Layer II115211521152
Layer III1152576576

Then you can calculate the frame size like this:

Frame Size = ( (Samples Per Frame / 8 * Bitrate) / Sampling Rate) + Padding Size

Because of rounding errors, the official formula to calculate the frame size is a little bit different. According to the ISO standards, you have to calculate the frame size in slots (see 2. MPEG Audio Format), then truncate this number to an integer, and after that multiply it with the slot size. You can find the correct way of calculating the frame size in the class CMPAHeader in my code.

You get the duration of the file in seconds by applying the following formula:

Duration = File Size / Bitrate * 8

The method of getting the first frame header in the file and then calculating the duration by the above formula works correctly only for CBR files.

The mode extension is used to join information that is of no use for the stereo effect, thus reducing the needed bits. These bits are dynamically determined by an encoder in the Joint Stereo mode, and Joint Stereo can be changed from one frame to another, or even switched on or off. For all other channel modes, the mode extension field is invalid.

The complete frequency range of MPEG audio files is divided into subbands. There are 32 subbands. For Layers I & II, the two bits in the header determine the frequency range (bands) where the intensity stereo is applied. Within this frequency range, only one channel is stored. All other bands contain information in two separate channels. For Layer III, these two bits determine which type of joint stereo is used (intensity stereo and/or M/S stereo).

2.1.6 Mode Extension
ValueLayer I & IILayer III
M/S stereoIntensity stereo
00bands 4 to 31offoff
01bands 8 to 31offon
10bands 12 to 31onoff
11bands 16 to 31onon

2.2. Verifying CRC

If the protection bit in the header is not set, the frame contains a 16 bit CRC (Cyclic Redundancy Checksum). This checksum directly follows the frame header and is a big-endian WORD. To verify this checksum you have to calculate it for the frame and compare the calculated CRC with the stored CRC. If they aren't equal probably a transfer error has appeared. It is also helpful to check the CRC to verify that you really found the beginning of a frame, because the sync bits do in same cases also occur within the data section of a frame.

The CRC is calculated by applying the CRC-16 algorithm (with the generator polynom 0x8005) to a part of the frame. The following data is considered for the CRC: the last two bytes of the header and a number of bits from the audio data which follows the checksum after the header. The checksum itself must be skipped for CRC calculation. Unfortunately there is no easy way to compute the number of frames which are necessary for the checksum calculation in Layer II. Therefore I left it out in the code. You would need other information apart from the header to calculate the necessary bits. However it is possible to compute the number of protected bits in Layer I and Layer III only with the information from the header.

For Layer III, you consider the complete side information for the CRC calculation. The side information follows the header or the CRC in Layer III files. It contains information about the general decoding of the frame, but doesn't contain the actual encoded audio samples. The following table shows the size of the side information for all Layer III files.

2.2.1 Layer III side information size (in bytes)
MPEG 1MPEG 2/2.5 (LSF)
Stereo, Joint Stereo, Dual Channel3217

For Layer I files, you must consider the mode extension (see table 2.1.6) from the header. Then you can calculate the number of bits which are necessary for CRC calculation by applying the following formula:

4 * (number of channels * bound of intensity stereo + 
                   (32 - bound of intensity stereo));

This can be read as two times the number of stereo subbands plus the number of mono subbands and the result multiplied with 4. For simple mono frames, this equals 128, because the number of channels is one and the bound of intensity stereo is 32, meaning that there is no intensity stereo. For stereo frames this is 256. For more information have a look at the CRC code in the class CMPAFrame.

2.3. VBR Headers

Some files are encoded with variable bitrate mode (VBR). To estimate the duration of those files, you have to know the average bitrate of the whole file. It often differs a lot from the bitrate of the first frame, because the lowest bitrate available is used for silence in music titles (especially at the beginning). To get this average bitrate, you must go through all the frames in the file and calculate it, by summarizing the bitrates of each frame and dividing it through the number of frames. Because this isn't a good practice (very slow), there exists additional VBR headers within the data section of the first frame (after the frame header). They contain the total number of frames in the file from which you can calculate the duration in seconds with the following formula:

Duration = Number of Frames * Samples Per Frame / Sampling Rate

Additional to that, the VBR header often contains a table which is necessary to seek positions within the file.

2.3.1 XING Header

This header is often (but unfortunately not always) added to files which are encoded with variable bitrate mode. This header stands after the first MPEG audio header at a specific position. The whole first frame which contains the XING header is a valid but empty audio frame, so even decoders which don't consider this header can decode the file. The XING header stands after the side information in Layer III files. So you can calculate the beginning of a XING header relative to the beginning of the frame by adding 4 bytes (for the MPEG audio header) to the values from the table 2.2.1. The offset calculation doesn't consider the 16 bit CRC following the header and is equal for all Layers, although only Layer III has a side information.

For reading out this header, you have to find the first MPEG audio header and then go to this specific position within the frame. The XING header itself has the following format. (Note that the position is zero-based; position, length and example are each in byte-format.) XING Header


4VBR header ID in 4 ASCII chars, either 'Xing' or 'Info', not NULL-terminated'Xing'



Flags which indicate what fields are present, flags are combined with a logical OR. Field is mandatory.

0x0001 - Frames field is present
0x0002 - Bytes field is present
0x0004 - TOC field is present
0x0008 - Quality indicator field is present

(means Frames, Bytes & TOC valid)
84Number of Frames as Big-Endian DWORD (optional)7344
8 or 124Number of Bytes in file as Big-Endian DWORD (optional)45000
8, 12 or 16100100 TOC entries for seeking as integral BYTE (optional)
8, 12, 16, 108, 112 or 1164Quality indicator as Big-Endian DWORD
from 0 - best quality to 100 - worst quality (optional)

According to this format, a XING header must only contain the ID and the flags. All other fields are optional and depend on the flags which are set. Sometimes this header is also added to CBR files. It then often has the ID 'Info' instead of 'Xing'.

There exists the LAME extension to this header, which is used by the common LAME Encoder, but I didn't take it into account because it isn't necessary for duration estimation. Nonetheless, here is the link for the documentation of the MP3 Info Tag.

2.3.2 VBRI Header

This header is only used by MPEG audio files encoded with the Fraunhofer Encoder as far as I know. It is different from the XING header. You find it exactly 32 bytes after the end of the first MPEG audio header in the file. (Note that the position is zero-based; position, length and example are each in byte-format.) VBRI Header


4VBR header ID in 4 ASCII chars, always 'VBRI', not NULL-terminated'VBRI'


2Version ID as Big-Endian WORD1
62Delay as Big-Endian float7344
82Quality indicator75
104Number of Bytes as Big-Endian DWORD45000
144Number of Frames as Big-Endian DWORD7344
182Number of entries within TOC table as Big-Endian WORD100
202Scale factor of TOC table entries as Big-Endian DWORD1
222Size per table entry in bytes (max 4) as Big-Endian WORD2
242Frames per table entry as Big-Endian WORD845
26 TOC entries for seeking as Big-Endian integral. From size per table entry and number of entries, you can calculate the length of this field.

3. Additional Tags

Please consider that at the end or at the beginning of the file, there might be additional data which is not part of the MPEG audio frames. This data is called a tag, because it contains metadata about the file, like title, artist, track, years etc. You must consider these tags, because only the MPEG audio data count for the duration estimation.. At the end of a file there might be an ID3V1 tag, a Lyrics3 tag or a Musicmatch tag. At the beginning or at the end of the file there might be an ID3V2 tag and/or an APE tag. You find information about ID3 and Lyrics3 tags at APE was originally developed for lossless compressed audio files in APE (Monkey's Audio) format. But now this tag is also used for MPEG audio files. The Musicmatch tag was used by older version of the Musicmatch Encoder. Unfortunately there is very little information on the web about this tag, since it is a proprietary format of Musicmatch.

4. Using The Code

I wrote some C++ classes to handle the MPEG audio frame header and the VBR headers. The class CMPAFile represents the whole file, and provides methods for accessing a specific frame (CMPAFrame class). This class instantiates a CMPAHeader class which represents the MPEG audio frame header of the frame itself. All fields from the header can be accessed via the class variables. Additionally, there is the CVBRHeader class which is the generalization of CXINGHeader and of CVBRIHeader. All tags are derived from the class CTag. These classes are completely independent of any libraries like MFC or ATL. They only use the Win32 API. All classes will throw exceptions in case of errors. All exceptions are of the type CMPAException. You can display information about the exception by using the ShowError() method. All classes use CMPAStream for file-access. This class includes a simple buffer.

Here is a code snippet which demonstrates how you could use these classes. Note that you must at least include the header mpafile.h:

try {
    // open file and look for first header
    CMPAFile MPAFile(_T("C:\\test.mp3")) ;
    cout << MPAFile.GetLengthSec() << _T(" seconds");
catch(CMPAException& Exc)
    // show error message

For more information, have a look at the source of the demo project MPEG Audio Info, which is a simple MFC Dialog Application, which uses the class CMPAFile for getting information about a MPEG audio file. It also can perform a check of a whole file for errors within the frame structure.

5. Links And Miscellaneous

Here, you can find the sources which I used for this article:

If you find any mistakes in the code or in the article, or have suggestions for improvements, just post them to the forum below or write me an e-mail to

6. History

  • 2004-09-15
    • Version 1.0 of MPEG Audio Info released.
    • First version of the article published.
  • 2004-11-01
    • Version 2.0 of MPEG Audio Info released.
      • added possibility to get information about arbitrary frames (not just the first one).
      • added checking of MPEG audio files.
      • better handling of faulty MPEG files (tolerance range +/-3 bytes to look for the next frame).
      • padding is detected correctly.
      • improved buffer management.
      • fixed some other small issues.
      • added a readme file.
    • Article updated:
      • added information about mode extension field.
      • XING header offsets fixed (Mono and Stereo inverted).
      • frames and bytes in VBRI fixed (they were inverted).
      • position of quality indicator in VBRI header fixed.
      • corrected some typos.
  • 2005-11-17
    • Version 2.1 of MPEG Audio Info released.
      • cleaned up class architecture.
      • better navigation through frames.
      • improved checking of files.
      • fixed original bit checking.
      • fixed frame size calculation for all layers (correct truncation).
      • fixed skipping of small frames in VBR files.
      • new cache management.
      • added CRC check.
      • added detection of ID3V1/V2, APE and Lyrics3 tags.
    • Article updated:
      • added information about CRC.
      • fixed sampling rate table.
      • some minor enhancements.
  • 2007-04-09
    • Version 2.2 of MPEG Audio Info released.
      • added Musicmatch-Tag detection
      • better exception handling
      • improved design of CMPAStream
      • added Drag&Drop functionality to dialog
      • fixed memory leaks
      • updated solution to Visual Studio 2005
      • better performance for finding frames
    • Article updated:
      • fixed broken links
      • added link with MPEG Audio Specifications
      • added link with informations about Musicmatch


This article, along with any associated source code and files, is licensed under The GNU Lesser General Public License (LGPLv3)

Written By
Web Developer
Germany Germany
Author of the shareware WinCD.

Comments and Discussions

QuestionGreat article Pin
Member 1558701021-Mar-24 19:31
Member 1558701021-Mar-24 19:31 
QuestionZero division exception occured on a file less than 1 second. Pin
Member 1364806929-Aug-22 21:25
Member 1364806929-Aug-22 21:25 
QuestionDuration Pin
Member 1515191622-Apr-21 0:25
Member 1515191622-Apr-21 0:25 
QuestionPosition of CRC before of after XIng Header Pin
Member 1515191620-Apr-21 0:57
Member 1515191620-Apr-21 0:57 
Generalanother look at frame length in bytes Pin
Midnight48927-Oct-19 10:42
Midnight48927-Oct-19 10:42 
QuestionNEED HELP!!! Pin
Member 1148883810-May-15 8:21
Member 1148883810-May-15 8:21 
BugFIX for CompilerERR: This file requires _WIN32_WINNT to be #defined at least to 0x0403. Value 0x0501 or higher is recommended. Pin
qwe45629-Dec-14 12:49
qwe45629-Dec-14 12:49 
QuestionBit Rate Index for MPEG4? Pin
tom ballister22-Oct-14 13:48
tom ballister22-Oct-14 13:48 
QuestionCalculating Correct Frame Size Pin
Neil Siegfried28-Jul-12 8:57
Neil Siegfried28-Jul-12 8:57 
AnswerRe: Calculating Correct Frame Size Pin
Jim Balter8-Oct-23 19:30
Jim Balter8-Oct-23 19:30 
GeneralMy vote of 5 Pin
Rick Parsons10-Jan-12 4:17
Rick Parsons10-Jan-12 4:17 
QuestionNew Win 7 sample music files Pin
Rick Parsons10-Jan-12 4:10
Rick Parsons10-Jan-12 4:10 
AnswerRe: New Win 7 sample music files Pin
Mark Heath29-Jan-13 1:16
Mark Heath29-Jan-13 1:16 
Generalexcellent code!! Pin
kingwrcy13-Sep-11 5:50
kingwrcy13-Sep-11 5:50 
GeneralMy vote of 5 Pin
tczq9-Jun-11 23:35
tczq9-Jun-11 23:35 
GeneralCalcFrameSize incorrect? Pin
jMeson20-Apr-11 5:40
jMeson20-Apr-11 5:40 
GeneralRe: CalcFrameSize incorrect? Pin
Jim Balter8-Oct-23 19:28
Jim Balter8-Oct-23 19:28 
GeneralMy vote of 5 Pin
Mark Heath7-Nov-10 9:57
Mark Heath7-Nov-10 9:57 
GeneralCalcCRC16 Pin
Snuff Daddy27-Oct-10 8:56
Snuff Daddy27-Oct-10 8:56 
GeneralCommand-line version and .NET assemblies Pin
Onur Guzel20-Jun-10 5:28
Onur Guzel20-Jun-10 5:28 
GeneralBug in ID3 Tag v1.2 [modified] Pin
k06a26-Apr-10 10:08
k06a26-Apr-10 10:08 
NewsBug in ID3 Tag v1.1 [modified] Pin
k06a26-Apr-10 9:52
k06a26-Apr-10 9:52 
QuestionWant C# classes? [modified] Pin
k06a26-Apr-10 7:22
k06a26-Apr-10 7:22 
AnswerRe: Want C# classes? [modified] Pin
shengcheng_jin12-Jul-12 0:03
shengcheng_jin12-Jul-12 0:03 
GeneralDetecting Frame Header Pin
Helium9927-Mar-10 22:48
Helium9927-Mar-10 22:48 

General General    News News    Suggestion Suggestion    Question Question    Bug Bug    Answer Answer    Joke Joke    Praise Praise    Rant Rant    Admin Admin   

Use Ctrl+Left/Right to switch messages, Ctrl+Up/Down to switch threads, Ctrl+Shift+Left/Right to switch pages.