Documentations

To go further

Wav format

A very complete and clear description of the file format can be found at

http://www.borg.com/~jglatt/tech/wave.htm

PCM format is the "standard" format for sound samples. Data is not compressed and can be easily accessed. There is some possible variations about PCM/Wave format but the most popular of them has a 44 bytes header. This header contains very important data about the rest of the file and will tell you the number of channels, the frequency, etc...

The first 12 bytes give you the.

• 1 - FORMAT

• You find at first the « RIFF » mention ($52, $49, $42, $42 in hexadecimal. These are the ASCII values for R,I,F,F letters)

• You find then the TOTAL size of the file coded as a long integer (4 bytes)

• You find then the « WAVE» mention (4 lettres): $57,$41,$56,$45

This first part is followed by the

• 2- Sound desciption : sampling frequency, number of bytes per sample, etc. Those datas are exactly ordered as in the WAVEFORMAT structure, (Microsoft) descripted in Microsoft APIs documentation.

• zone descriptor : « fmt » (4 characters) take care! the last character is a "space" : $66,$6D,$74,$20. « fmt » is an abbreviation for « format ». This is to tell us that the following is a format description (as you will see)

• size of the WAVEFORMAT structure as a long integer (4 bytes). WAVEFORMAT is 16 bytes long, so you will find the $10,00,00,00 value here. ($10 in hexa equal to 16 in decimal)

• wFormatTag.w (2 octets) : for PCM files, this value must be 1

• nChannels.w (2 octets) : number of channels. 1 for mono, 2 for stereo. In can be more.

• nSamplesPerSec.l (4 octets) : number of samples per seconde.

• nAvgBytesPerSec.l (4 octets) : Average bytes per seconde. This value is a combination of other values recorded in this chunck. You HAVE to complete it if you generate a wave file and want to be compatible with all softwares. The formula to compute it is: nAvgBytesPerSec= nSamplesPerSec* nBitsPerSample/8* nChannels

(nBitsPerSample is the number of BITS per seconde. One byte equal to 8 bits)

• nBlockAlign.w (2 octets) : total size of a sample. This value is a combination of other values recorded in this chunck. You HAVE to complete it if you generate a wave file and want to be compatible with all softwares. The formula to compute it is: nBlockAlign = nBitsPerSample/8* nChannels*

(nBitsPerSample is the number of BITS per seconde. One byte equals to 8 bits)

• nBitsPerSample.w (2 octets) : number of bits per sample.

The following of the fil contains the.

• 3 - DATAS

• zone descriptor : «data» (4 characters) : $64,$61,$74,$61.

• Data size as a long integer (4 bytes). Duration of the sound can be computed uppon this value :
Duration = DataSize/ nAvgBytesPerSec

CAUTION : DataSize must ALLWAYS be a multiple of nBlockAlign. You can be sure of that by doing :

Newsize.l = Newsize/ nBlockAlign ; It will "round" newsize

Newsize = Newsize* nBlockAlign ; Restore the good rounded value

Summary:

Offset (decimal)	offset (hexa)	name	length (byte)	description
0	00h	rID	4	contains "RIFF"
4	04h	rLen	4	size of the file
8	08h	wID	4	contains "WAVE"

Format Chunk:

Offset (decimal)	offset (hexa)	name	length (byte)	description
12	0Ch	fId	4	contains "fmt " ("fmt space ")
16	10h	fLen	4	Chunck length
20	14h	wFormatTag	2	format (1 = Microsoft Pulse Code Modulation PCM)
22	16h	nChannels	2	number of channels (1=mono, 2=stereo)
24	18h	nSamplesPerSec	4	Sampling frequency (in Hz)
28	1Ch	nAvgBytesPerSec	4	= nChannels * nSamplesPerSec * (nBitsPerSample/8)
32	20h	nBlockAlign	2	= nChannels * (nBitsPerSample / 8)
34	22h	nBitsPerSample	2	length of one sample in bits (8, 16, 24 ou 32)

WAVE Data Chunk:

Offset (decimal)	offset (hexa)	name	length (octet)	description
36	24h	dId	4	contain"data"
40	28h	dLen	4	length of dData chunck(in bytes)
44 et plus	2Ch	dData	dLen	Sound datas

NOTE FOR PUREBASIC

There is not any problem to read, transform and write 16, or 32 bits data.

Data.l = PeekW (*adress) or PeekL(*adress)
.
. your job here
.
PokeW(*adress,Data) or PokeL(*adress,Data)

It's a bit less simple with 8 bits data because PureBasic always considere that datas are signed and it's not the case with 8 bits datas. The mid point for a 8 bits sample is 127, it's not zero. So, you have to do that :

Read 8 bits datas

Data.l = PeekB(*adress)

If Data<0 : Data + 256 : EndIf ; transform signed to unsigned

Data - 127 ; set mid point to zero

Store 8 bits datas

If Data>127 : Data = 127 : Endif

Data + 127

PokeB (*adress,Data)

Convert 8 bits to 16 bits :

Data = PeekB(*Sourceadress)

If Data<0 : Data + 256 : EndIf

Data - 127

Data * 256

PokeW(*Destinationadress,Data)