Disassembling a Windows Wave File (.

wav)

The aim of this project was to create a “nice” sine wave as a test signal
to trace the function of the electronic circuit of a tube amplifier by
oscilloscope:

By adding two simple sine waves (red and green) a more complex test
sine curve is calculated (violet).

The plan was to reach this goal with a computer sound card equipment
and a Microsoft wave file, without the need of buying expensive wave
form generator hardware equipment. To be able to create my own
custom wave file containing a test signal, I had to understand how a
windows wave file is built. In this document I disassemble a sine tone
wave file and match the findings with a wave file documentation from the
web.

An excellent documentation for Windows wave files can be found here:

https://ccrma.stanford.edu/courses/422/projects/WaveFormat/

Before reading on, have a look at that documentation first. My document

very closely follows the information provided there, my contribution is the
disassembly of a file (screenshots) and some more detailed calculation
steps. In the following, a Windows example file is disassembled and
analyzed step by step for people less familiar with computer math (like
me). The analyzed example file ( a 5.3 second long 1khz 16 bit stereo
sine wave, 0 dBFS [ „0 decibel full scale“ meaning: using maximal
possible amplitude at sine wave peak or „as loud as technically possible
without causing clipping“] ) can be found here:

http://www.rme-audio.de/old/download/audtest.htm

The editor used to analyze the file is „HxD“

Before starting with this project I wasn’t very familiar with the terms little
and big endian. A good introduction about “endianess” is the
corresponding wikipedia article. Besides I can recommend to have a look
at the very entertaining, initially not quite seriously meant, 1980 first of
April rfc article of Danny Cohen that inadvertedly introduced the usage of
the terms „big / little endian“ to informatics. It is funny !

Please apologize the usage of German (weird language ) Windows

Vista (outdated, not very popular OS ) screenshots, this unfortunately
is the current setup of my laptop ().
Open the example sine wave testone.wav file with a hex editor (for
expamle HxD)

The first four hex figures designate the file container format used, which
was termed RIFF (Resource Interchange File Format) by Microsoft, this
field is called „ChunkID“ and it is four bytes long

ChunkID: field size of 4 Bytes, the four capital letters RIFF in

hexadecimal ASCII form

52 49 46 46 hex numbers representing the string

RIFF

R I F F corresponding ASCII Code representation

“ The ChunkID is followed by the ChunkSize field, this field indicates the
size of the following file in bytes. The field has a size of 4 bytes. It does
not includes the size of the two trailing fields ChunkID and ChunkSize,
thus the size it indicates is:

„full file size“ - „ChunkID fieldsize“ – „ChunkSize fieldsize“ which equals:

„full file size“ - 4 bytes - 4 bytes = „full file size“ – 8 bytes

Our example file has a size of 991‘284 bytes:

ChunkSize: Contains the size of the following chunk in bytes

It is a field of 4 Bytes in hexadecimal little endian form

ChunkSize = Filesize – 8 Bytes (size of field ChunkID and

field ChunkSize)

 Because of this 4 Bytes limitation in field length, maximal

file size of .wav files is limited to 4 GB (4 Bytes = 4 x 8
Bits = 32 Bits, 232 = 4‘294‘967‘296)

Example: ChunkSize of a file with a size of 991'284 bytes is

991284 bytes – 8 bytes = 991276 bytes

Dezimal to Hex conversion: 991276 (dec) = F202C (hex)

991‘276 decimal representation

00 0F 20 2C big endian hex representation

2C 20 0F 00 little endian representation

 This fits our example file screenshot:

The format field designates the format of data that follows, in our case
this is „WAVE“, other formats would also be possible in a RIFF file
container.

Format: field of 4 bytes, the four capital letters „WAVE“ in hexadecimal

ASCII form

57 41 56 45 hex numbers (string “WAVE”)

W A V E ASCII representation

The „WAVE“ format requires two format specific subchunks, the

describtive „fmt“ subchunk and the „data“ subchunk that contains the
actual audio data.
Subchunk1ID: field of 4 bytes, the three lower case letters „fmt“ followed

by the „space“ control command in hexadecimal ASCII

form

66 6D 74 20 Hex Numbers

f m t SPACE ASCII Representation

Subchunk1Size: field of 4 Bytes, this is the size of the rest of the

Subchunk which follows this number, in case of pulse

code modulation PCM, Subchunk1 size is 16 Bytes

16 (dec) = 10 (hex)

4 Bytes big endian hex: 00 00 00 10

4 Bytes little endian hex: 10 00 00 00

10 00 00 00 Hex Numbers (little endian)

00 00 00 10 Hex Numbers (big endian)

16 decimal representation
AudioFormat: little endian field of 2 Bytes, for PCM the value is 1 (i.e.

Linear quantization), values other than 1 indicate some

form of compression

1 (dec) = 1 (hex)

2 Bytes big endian hex: 00 01

2 Bytes little endian hex: 01 00

01 00 Hex Numbers (little endian)

00 01 Hex Numbers (big endian)

1 decimal representation

NumChannels: little endian field of 2 Bytes, designates the number of

channels used, Mono = 1, Stereo = 2, etc.

2 (dec) = 2 (hex)

2 Bytes big endian hex: 00 02

2 Bytes little endian hex: 02 00

02 00 Hex Numbers (little endian)

00 02 Hex Numbers (big endian)

2 decimal representation

SampleRate: little endian field of 4 Bytes, designates the sample rate

used, like 8000, 44100, 48000, etc.

44100 (dec) = AC44 (hex)

4 Bytes big endian hex: 00 00 AC 44

4 Bytes little endian hex: 44 AC 00 00

44 AC 00 00 Hex Numbers (little endian)

00 00 AC 44 Hex Numbers (big endian)

44100 decimal representation

ByteRate: little endian field of 4 Bytes, designates the byte rate

== SampleRate * NumChannels * BitsPerSample/8

== 44100 * 2 * 16/8 = 176‘400

176‘400 (dec) = 2B110 (hex)

4 Bytes big endian hex: 00 02 B1 10

4 Bytes little endian hex: 10 B1 02 00

10 B1 02 00 Hex Numbers (little endian)

00 02 B1 10 Hex Numbers (big endian)

176‘400 decimal representation

BlockAlign: little endian field of 2 Bytes, designates the The number of

bytes for one sample including all channels

== NumChannels * BitsPerSample/8

== 2 * 16 / 8 = 4

4 (dec) = 4 (hex)

2 Bytes big endian hex: 00 04

2 Bytes little endian hex: 04 00

04 00 Hex Numbers (little endian)

00 04 Hex Numbers (big endian)

4 decimal representation

BitsPerSample: little endian field of 2 Bytes, designates the sampling

Resolution, 8 bits = 8, 16 bits = 16, etc.

16 (dec) = 10 (hex)

2 Bytes big endian hex: 00 10

2 Bytes little endian hex: 10 00

10 00 Hex Numbers (little endian)

00 10 Hex Numbers (big endian)

16 decimal representation

The „data“ subchunks contains the size information oft he sound

information and the raw data

Subchunk2ID: field of 4 Bytes, the four lower case letters „data“ in

hexadecimal ASCII form (big endian)

64 61 74 61 Hex Numbers (big endian)

d a t a ASCII Representation

Subchunk2Size: field of 4 Bytes, this is the number of bytes in the data,

you can also think of this as the size of the read of the

subchunk following this number

== NumSamples * NumChannels * BitsPerSample/8

for one second of a 44100 sample rate this is:

= 44100 * 2 * 16 / 8 = 176400

176400 (hex) = 2B110 (dec)

4 Bytes big endian hex: 00 02 B1 10

4 Bytes little endian hex: 10 B1 02 00

10 B1 02 00 Hex Numbers (little endian)

00 02 B1 10 Hex Numbers (big endian)

176‘400 decimal representation

RawData: in case of 16 bit Stereo PCM this is a field of 4 bytes, the field

contains one sample for each channel, the first 2 bytes

contain the sample for the left channel, the second 2 bytes

contain the sample for the right cannel, each byte pair is in

little endian notation, in case of 16 bit PCM, 16-bit samples

are stored as 2's-complement unsigned integers, ranging from

0 to 65535
DB 9F 25 60 D8 AC 28 53 86 BB Hex Numbers (little endian)

9F DB 60 25 AC D8 53 28 BB 86 Hex Numbers (big endian)

40923 24613 44248 21288 48006 decimal (2's-complement signed integers)