Audio data (unsigned char) that have been manipulated cannot be played - audio

Question

I have trouble with playing audio data after they have been manipulated.
The only API I use is the alsa lib API on Linux (Ubuntu) in C.
I get the data from a 16 bits integer wave file in a unsigned char array (called buffer1) using read() and buffer1 can be played properly. I want the data to be passed to another unsigned char array (called buffer2) of the same size. If I just make a loop with buffer2[i] = buffer1[i], it works : buffer2 can be played properly. But in order to manipulate the data, I convert it to a float array then back to unsigned char (Until now I do not manipulate the audio data; I just convert them to float then back to unsigned char to test how it works). But now buffer2 does not make sound although all of its values are strictly identical to the values of buffer1 (I made a printf of many values of buffer1 and buffer2; they are all identical) ... All I did is casting from unsigned to float and vice versa...
Please any idea of what's wrong?
Victor

Answer 1

The values in buffer1 and buffer2 cannot be identical or it would work. Perhaps the formatting that you use in your printf command is masking the differences (%i, %f etc.). Rather than use printf, try setting a breakpoint and looking at the values using your debugger. This might help reveal what is actually going wrong.
EDIT:
Given your comments about how you perform the cast, I think that I can now help. The raw data coming in is of type unsigned char. On most platforms, this will be an integer value between 0 and 255. You want to convert this value to a float to do your manipulation. To make the data meaningful as a floating point type for any manipulation, you want to scale this range between +/- 1.0. This is what the "scale" variable is for in the following code.
#include <iostream>
#include <math.h>
int main()
{
const int BUFFER_LEN = 6;
const unsigned char channelDataIN[] = {0,255, 1, 254, 2, 253};
unsigned char channelDataOUT[BUFFER_LEN];
float channelDataF[BUFFER_LEN];
std::cout.precision(5);
float scale = powf(2.f, 8.f*sizeof(unsigned char) ) - 1.f;
for (int mm = 0; mm < BUFFER_LEN; ++mm)
{
std::cout << "Original = " << (int)channelDataIN[mm] << std::endl;
channelDataF[mm] = (float)(channelDataIN[mm]) * 2.f/scale - 1.f; //Float cast
std::cout << "Float conversion = " << channelDataF[mm] << std::endl;
channelDataOUT[mm] = (unsigned char) ceil( ( 1.f+channelDataF[mm] ) * scale/2.f );
std::cout << "Recovered = " << (int)channelDataOUT[mm] << std::endl;
if (channelDataIN[mm] == channelDataOUT[mm])
std::cout << "The output precisely equals the input" << std::endl << std::endl;
else
std::cout << "The output != input" << std::endl << std::endl;
}
return 0;
}
The output array of unsigned chars after converting the values back is identical to the input array. This is the output from the code . . .
Original = 0
Float conversion = -1
Recovered = 0
The output precisely equals the input
Original = 255
Float conversion = 1
Recovered = 255
The output precisely equals the input
Original = 1
Float conversion = -0.99216
Recovered = 1
The output precisely equals the input
Original = 254
Float conversion = 0.99216
Recovered = 254
The output precisely equals the input
Original = 2
Float conversion = -0.98431
Recovered = 2
The output precisely equals the input
Original = 253
Float conversion = 0.98431
Recovered = 253
The output precisely equals the input