Stereo Wave Data

On a sunny day (Fri, 07 Mar 2008 15:24:57 GMT) it happened Marco Trapanese wrote in :

That is actually the sign bit.

Here is a piece of C code to get those low and high bytes and make an unsigned number from it.

int cm; double dm; unsigned long p = 1

Jan Panteltje ha scritto:

Ok, that was just an example. Anyway, I found the data is stored as unsigned int.

Thanks.

I solved quickly inverting the most significant bit. It works!

Marco / iw2nzm

On a sunny day (Fri, 07 Mar 2008 17:01:50 GMT) it happened Marco Trapanese wrote in :

Yes, possible, but normally in wave files sound is stored as 16 bit signed

Actually the C code I showed, was code that converted the 16 bit sample to a double, with range from -1.0 to + 1.0. That allows all sorts of math to be done with it. You can then convert it back with: /* double data to output buffer */ ll = floor(dout * p); if( dout < 0) ll += 2 * p; //fprintf(stderr, "new ll=%lu\\n", ll);

If you want to look up that code, it is from 'substract_wave', a GPL program I wrote, it can substract 2 wave files (to cancel background noise).

Be very careful, sometimes audio PCM may SEEM to work as unsigned, but of course it will be distorted and not usable.

Jan Panteltje ha scritto:

You're right, it's a 16 bit signed 2 complement's. I was wrong when I wrote the post. However this short code works fine:

ad5547_output(audio[idxR], audio[idxR+1] ^ _BV(7)); ad5547_output(audio[idxR+2], audio[idxR+3] ^ _BV(7));

The first line is for left channel, the second for the right one. The first parameter is the low byte, the second the high byte with MSB flipped.

audioB[] is the 1024-bytes buffer loaded by audio data.

Using a 8 bit MCU (ATmega644 @ 20 MHz) I can't do some math in the interrupt routine so I just send out the values to the DAC.

Bye! Marco / iw2nzm

now that you have that under your belt..

just remember that 16 bit cross over point is

for 8 bit. the cross over of the sine is at $7F.

Not that i am want to sound overly critical but, in standard "C" double is a 64-bit floating point number. I see no value in doing the conversions.

Wav files on X86 machines are int 16, little endian. The MSB of the

On a sunny day (Sat, 08 Mar 2008 19:07:40 GMT) it happened JosephKK wrote in :

Actually, as I pointed out in the next posting as correction, dm is range -1.0 to +1.0. The conversion was done in my program to be able to work with -1.0 to +1.0 signals, and do all math in type double. This code is part of a program I wrote to cancel background sound in a translation channel.

p = 1