General DAQ System Design

[CodeSDz]

I am designing a new DAQ system.  I just need a few pointers on how to design a system properly to acquire data accurately without too much overkill.

During the 100 ms event, we would capture four (single event, non-periodic) pressure curves and several voltage spikes (indicating specific event times).  We would use a charge amplifier to amplify the signal coming from the pressure transducers.

Can anyone point me in the direction of some useful books, advice, rules of thumb, websites, theorems, etc for designing a DAQ system including but not limited to the following topics?

Sample Rate:

What criteria should be used to determine sample rate?  Since the waveforms will not be periodic, I don't believe the Nyquist theorem applies.  Are there any rules of thumb in this case?

Voltage Range / Amplification:

How much do I amplify my signals?  Our specification defines an expected pressure, maximum acceptable value, maximum pressure the system should handle, and tolerances.  How much of the voltage range should be an average signal or the maximum signal?

Data Bits:

How do I determine how many data bits (and therefore distinct values) I need to acquire the characteristics of my pressure curves accurately?

[John_P1 · Accepted solution]

Hi Steven,

I have some PowerPoint slides (attached below) from the DAQ Hands On Session from last NI Week that give an overview of these exact three topics.  They were intended to be more of an auxilary to the presentation so they don't have too much information, but the images I think are still helpful (there are animations so you'll have to go through the slides in presentation mode).

Sample Rate:

This really depends on what you're trying to measure.  You're right that Nyquist frequency only applies to periodic signals (you won't have aliasing if the signal is not periodic), but sample rate is still very much important depending on how much you care about characterizing the shape of your signal. 

If the signal changes quickly, you'll want to make sure that you are sampling quickly enough to acquire enough samples along the transition to be able to characterize the edge (if it is important to you).  On the other side of things, if you're sampling too slowly, it would be possible to completely miss a pulse if its duration is less than the period of your sample clock.

Voltage Range / Amplification:

These definitely go together.  To maximize resolution of the ADC, you'll want to choose an amplification that fills the entire range of the ADC.  Many DAQ cards have built-in amplification and provide multiple available ranges (although in DAQmx the user just specifies which available range they want).

Data Bits (aka  Resolution):

This does tell you the number of available discrete levels that you can represent across the entire range of your DAQ card.  However, this value is different from accuracy due to gain/offset error, non-linearities in the ADC, and noise.  If you are interested in determining the accuracy of your measurement (which is pretty common), you should check the device specifications and look for the Absolute Accuracy specification.

So, in general I think you need to figure out what exactly your requirements are before choosing hardware.  What information are you trying to determing from your pressure curves?  Are the voltage spikes just digital signals that need to be timestamped?  What would this timestamp be relative to?

Best Regards,