Compact DAQ vs SCXI Noise

[DieselPower]

We have a large number of SCXI systems and they are very good about eliminating noise.  In software, I used a 4 Hz filter and take the average of 100 readings each second.  We are trying to move to CompactDAQ for several reasons, but it seems as though the voltage readings using a 9205 DSUB are much suceptible to noise. 

StDev for SCXI - 0.0089
StDev for cDAQ - 0.0206

The "average" lines in the attached chart are 10 second sliding window averages.

Why can't I use a 4 Hz filter in Compact DAQ?  Not supported?

Is this issue because SCXI channels are isolated? (SCXI-1125)

We are using an in house bias resistor board for the Compact DAQ and an SCXI-1328 for SXCI.

Thanks

[KyleBryson]

Michael,

The NI 9205 doesn't have a 4 Hz hardware filter like the SCXI-1125.  That hardware filter will certainly reduce high-frequency noise significantly, but you should be able to achieve similar (or hopefully, better) results using a digital filter on your CompactDAQ data.  Have you tried a digital lowpass filter?

Regards,
Kyle

[DieselPower]

No.  What sort of digital filter is available?

Thanks.

[KyleBryson]

Michael,

You're not programming in LabVIEW, are you?  I don't know what filtering tools (if any) are available in your programming environment.  Averaging is a very simple form of lowpass filter.  Am I interpreting your graph correctly - AverageLV is the 10 sec. running average of (1000 points of) the SCXI data, and Average is the 10 sec. running average of (1000 points of) the cDAQ data?  And you're reporting the standard deviation of those lines?

You could try taking more points per second and averaging.  Also, in terms of the measurement itself, the 9205 is also isolated (like the 1125), so it should be a similar situation, but if your front-end was completely floating, you might be susceptible to more noise.  Also, are you making single-ended or differential measurements?  Differential would be preferable.

Regards,
Kyle

[DieselPower]

I am using Measurement Studio and VB.NET.  I can try more readings.  I am using differential connections.  Any idea why I would have to read more readings on cDAQ to get the same quality of signal as SCXI if they are both isolated?

[KyleBryson]

Michael,

It's probably not really an issue of quality; the noise you are measuring on cDAQ is probably on the lines.  The SCXI-1125 can filter noise above 4 Hz, which is useful for your application, which is why its output suits your needs better.  The cDAQ system is measuring and returning this noise because it lacks the hardware filter, so to achieve the same results, we'll have to replace the hardware filter with software filtering.

What version and package of Measurement Studio are you using?  The professional and enterprise packages offer numerous types of digital filter:

http://www.ni.com/analysis/cwtools_analysis.htm

Regards,
Kyle

[DieselPower]

Do these filters work by reading in X samples, performing some analysis on those 100 samples and returning a better "average?"

How is the documentation?

Thanks for your quick responses.

[Abhinav T]

Michael,

The filter removes an undesirable frequency component. A low pass fitler removes all frequencies above a specific cut-off.

Here is the wiki page for filters: http://en.wikipedia.org/wiki/Low_pass

Hope this helps!

[DieselPower]

I know how filters work.  I'm trying to understand what capabilities the enterprise version of Measurement Studio has that can take in 100 readings of noisy data and give you a cleaner response through software.

[KyleBryson]

Michael,

I don't have Measurement Studio installed or have documentation available, so I can't give you specific advice.  From searching Developer Zone, it looks like you should consult the Measurement Studio Reference for the function(s) available.  I would start with a simple FIR filter; there should be legitimate options to choose either a filter that will operate on 100 readings at a time, or to filter your entire acquisition.  You'll want to read through the included documentation and/or consult some simple filter design pages on the web to consider how windowing and the selection of particular filter types will affect your results.

I suspect you'll find your results much better once you find the right settings.  Sorry for the extra steps to process your data, but you may find that you can improve the results of your data even more given knowledge of the frequency range of the signals you are measuring.  For example, you may find that a lower cutoff frequency than the fixed SCXI filter is better, or could improve measurements from both systems.

Regards,
Kyle