Data conversion

The ADC of the NI PCI-6052E provides us with the voltages of each channel digitized in 16 bits. Now a call to the function DAQ_VScale() with appropriate parameters does the conversion back to volts for us. But these will be of type double (8 bytes) instead of short (2 bytes). So in order to make the .sm files as small as possible, the 2 byte unconverted binary data will be stored, rather than the 8 byte double-precision converted voltages. In order to retrieve the voltages later, we need to save the conversion factors in a structure ai_settings which will be stored along with our data. The members of this structure are

short Polarity (0 = bipolar, 1 = unipolar)

short VoltRange (will usually be 10, to signify -5 V to 5 V in bipolar mode or 0 V to 10 V in unipolar mode)

long DigitalRange (will be $2^{16}=65536$ for 16 bit binary data)

double Offset (can correct for offset in binary; this will usually be a value of 0.0)

short Gain (in case we want to measure lower voltage range at higher resolution). Unfortunatly this parameter is expressed as a fraction of VoltRange. We will usually use a value of 1.

double GainAdjust (will usually not be used, i.e., be 1.0)

For unipolar mode, the binary data of type short will be treated as unsigned short, which makes a type cast necessary. The general formula used is:

$$\mbox{Voltage}=\frac{\mbox{Binary-\texttt{Offset}}} {\mbox... ...\frac{\mbox{\texttt{VoltRange}}}{\mbox{\texttt{DigitalRange}}}$$

In C++, this formula usually leads to the same result as a call to DAQ_VScale(). In some cases, however, the results differ by a number in the order of $10^{-15}$ or smaller. The source of this error is as yet unknown (presumably it is due to some internal rounding effect), but its magnitude is of no concern to us.