You don't say anything about the DAQ hardware you are using. There are (at least) two dimensions to the concept of "channel calibration" -- "gain" (or the "Y dimension" of the signal (you might also add "offset" here) and "time" (or the "X dimension" of the signal.
Consider taking "vibration" (or "movement") signals from a 3-D Accelerometer. Suppose you want to take data (continuously) at 1 kHz, and the specification for the three axes were that the "gain" of each channel (which are arranged to be accurately spatially arranged to be orthogonal to each other) is 1 V/g with an offset of 0 V, with gain and offset being ±0.3 V/g in gain and ±0.3 V offset. So if you read a signal of 0V from the Z channel, the "true" acceleration can range from -0.3 g to +0.3g. [I actually asked some BME grad students how they were going to handle this question -- seems they never learned about "calibrating the Instrument"). Hint -- for a simple triaxial accelerometer, designing a "calibration routine" is really simple, and only depends on being able to hold the accelerometer "relatively motionless" for a second and take 6 readings at 6 diffeent (spatial) orientations).
But what about "time" calibration? How do you ensure your PC can acquire data at 1 kHz with (relatively) high precision and accuracy? Here the "good news" is most good DAQ hardware. You need to acquire the data using the hardware clock in the DAQ hardware and use its internal buffer to hold, say, 1000 samples (which, at a sampling rate of 1 kHz takes 1.000 seconds) and, whicn acquired, dumps it into PC memory when the code does a "DAQ Read" request. The timing accuracy is managed by the timing chip on the DAQ device, and isn't affected by what is happening on the PC (such as a Virus scan, display update, or other activities).
So good DAQmx code handles Time calibration for you, and all you need to do is to take measurements using known inputs to calibrate the "non-time" dimensions of your signal.
Bob Schor
