sample rate can affect the NI 9205 reading noise?

[ping2002]

I am using NI 9205 to measure voltage signal 10mv-200mv, when using the default sample rate 1000HZ to take 100 samples and average it to get the mean value, the reading is flickering, with a standard deviation 1% of the mean value, that is close to the NI 9205 specification.

In order to reduce this noise, I experimented different sample rate and sample number, I find that if the sample rate and sample number are multiples of 60 and 100, the reading noise can be reduce by 100 times, there is much less flickering, the standard deviation is about 0.01% of the mean value. For example , if you choose sample rate 1000, or 4000, or 100000, with sample number 100, 200, 400, 1000, the result is not a good as you choose 36000, 72000, 144000, 216000 sample rate, with sample number 3600, 7200, or 10800.

I do not know the exact reason, it seemed that the random electronic noise is reduced a lot. This trick can help to reduce the reading noise of NI 9205. While it does not necessarily mean the accuracy of NI 9205 is improved so much,   since this trick might not be able to reduce the offset error, temperature drift error, etc.

 I am not sure if this trick is a valid method to reduce noise? please advise.

ping

[Doug_L._Bear]

Hey Ping,

I tried to reproduce your result using a 9205 by generating a square wave on a 9263 module and then reading it in the 9205.  Using a max range of ±150m (which would use the ±200m board range) I was consistently getting an noise amplitude of 1.5 mV on a 20 mV amplitude wave, regardless of my sample rate or number of samples.

Because you were able to target in on multiples of 60 Hz, this leads me to believe that a large portion of your noise is due to external sources, such as lighting or power supplies nearby, as this is the frequency of standard AC current.  By sampling at a multiple of this frequency you are always sampling at the same point on the AC wave and reducing its noise influence, but will still be seeing its offset.

Doug F.

[IshanG]

hi doug

I am facing the same problem as ping. I am encountering a lot of noise in my signal while calibrating multiple channels in signal express using 9205. (i am using the in-built acquire temperature express).  therefore while i am calibrating it, the values does not stabilize leading to erroneous calibration.

please suggest few troubleshooters... ( i am assuming that the 'reference', 'uncalibrated' and 'difference' columns in the channel calibration refer to reference temp and corresponding voltage required for calibration of a thermocouple; though i have no clue what difference means)... should i use an ice bath in this case and check the difference?

Ishan 

[Mark_SAI]

Hello all,

To verify Doug's findings I present the same noisy 50hz signal sampled at 48hz and 50hz:

48hz:

50hz:

I hope this helps anyone who is thinking of reading noisy measurements without signal conditioning.

[Mark_SAI]

Furthermore, notice that the 50hz sampled signal did not average to 2.5mV, there was an offset which depends on where you sample the signal in the 50hz wave. So this doesn't make a good alternative to averaging in software.