Importance of the simultaneous sampling function?

[Zilong]

Hi,

I am looking for a DAQ to analyze data from 3 different channels. The sampling rate is about 100 KS/s (for each one), and the signal level is between 2mV to 2V. I found NI USB-6281 may work for us, but noticed that the DAQ does not have the simultaneous sampling function.

In theory, we need to process data from different channels at the same time. But as a matter of fact, our signal is expected to be a square wave with the frequency of 10 KHz. In other words, we plan to use 10 points in each cycle to describe the entire square wave at 10 KHz. I wonder whether the simultaneous sampling function is necessary or not in this case, if the data starting and ending moments are available for post process.

We have no experience on working with DAQ. Can anyone with experience make some comments? If necessary, I can explain more about my project.

Thanks.

[crossrulz]

Simultaneous sampling is really needed when very tight synchronization is needed.  This is typically in a Sound & Vibration.

From what you quickly described, it sounds like simultaneous sampling is not necessary.

[Zilong]

Thanks for your reply.

As a matter of fact, the signals in three channels are dependent on each other. We will need to analyze the signal from two channels, and use the result to "fit" the third curve. In this case, is it still not necessary?

To make my question more clear:

1 - are the starting and ending moments of each channel available from a DAQ?

2 - in this case, we do need a synchronization of all channels, but I don't know if it is "tight" enough or not. The signal (amplitude of the square wave) is expected to change in each 1-ms, but the frequency of the square wave is 10 KHz, and the sampling rate is 100 KS/s. If the signals of three channels are obtained in the sequence of 1-2-3-1-2-3..., we basically will miss timing about 0.01ms between neighbor channels, is it correct?

Thanks again.

[Kevin_Price · Accepted solution]

100 kS/sec means that you sample *all* channels at 100 kHz.  There's a little skew from channel to channel *within* that 0.01 msec interval, and that skew is programmable within the board's limits.

However, considering that your signals are square waves, your timing quantization error in detecting edges is gonna be larger than your channel-to-channel skew anyway.  A simultaneous sampling board would only address the smaller skew error and still leave you with the quantization error.

EDIT: just reread the original post and am reminded you're new to DAQ.  Let me summarize in simpler terms:

- the board you're using is very likely sufficient

- the skew between channels is programmable and queryable, and can be taken into account for your curve fitting if you really want to.  

- In my opinion, both simultaneous sampling and skew correction will be of dubious value as they compensate only for the smaller skew error and can't address the larger quantization error.

-Kevin P

[Zilong]

I see. Thank you for your detailed explanation!