Hello magintd,
The IQ samples are stored in the binary file as 16 bit signed integers, interleaved (I1, Q1, I2, Q2 … etc). The 16 bit integers are scaled to +/- 32767 with an additional scaleFactor to bring them into the 0-1 range.
The proper way to recover the IQ values are
float floatReal = ((float)int16Real / 32767.0) * (1.0 / scalefactor);
float floatImag = ((float)int16Imag / 32767.0) * (1.0 / scalefactor);
Where the scaleFactor is found in the XML file.
Let me know if you have additional questions.
Regards,
A.J.