Data from wav file are between -1 and 1, c++, sndfile - c++

Question

I am trying to read data from .wav and put it to fft.
To read wav file I am using sndfile library.
SNDFILE* infile;
SF_INFO sfinfo ;
memset (&sfinfo, 0, sizeof (sfinfo)) ;
infile = sf_open ("sound.wav", SFM_READ, &sfinfo);
double data [BUF_SIZE];
while (readcount = (int)sf_readf_double (infile, data, BUF_SIZE))
{
for (int i = 0; i < readcount; i++)
{
cout << data[i] << " ";
}
}
But every values in this (and other files) are between (-1 ; 1).
Is this correct? Why every values are so small? I was expected to read amplitude in time domain (volume of sound).

Answer 1

This is the canonical format of floating point samples. With float values, you get full 32-bit precision. Clipping is also easy to represent. If a sample value is higher than 1 or lower than -1, it means the sample clipped. With integer values, there's no way to know that.
Floating point is also an easy sample format to apply operations to. Mixing for example is trivial (you just add the sample values together.)
So even if it looks weird at first, it is the best format for audio sample representation. Once you applied the operations you need to the float values, you then convert them to the format you want for output (like 16-bit integers.) This operation is trivial. Here's a function that converts and clips float samples to any known integer sample format in use today:
#include <limits>
/* Convert and clip a float sample to an integer sample. This works for
* all usual integer sample types (8-bit, 16-bit, 32-bit, signed or
* unsigned.)
*/
template <typename T>
T floatSampleToInt(float src) noexcept
{
if (src >= 1.f)
return std::numeric_limits<T>::max();
if (src < -1.f)
return std::numeric_limits<T>::min();
return src * (float)(1UL << (sizeof(T) * 8 - 1))
+ ((float)(1UL << (sizeof(T) * 8 - 1))
+ (float)std::numeric_limits<T>::min());
}
If you want to convert a float sample to a signed 16-bit integer sample for example, you do:
int16_t intSample = floatSampleToInt<int16_t>(floatSample);
Note that 24-bit integer samples are covered by 32-bit. A 32-bit sample is also a valid 24-bit sample; its lower 8 bits are just truncated.