9205 noise

[Fbaird]

Hello all,  I'm having an issue balancing chassis slot space and ease of aquiring signals.  The cRIO systems that I have built in the past used the universal 9219 cards to aquire my 4-20mA analogue signals.  However I have found that at 4 channels per card the cost per channel is steep and frankly I'm running out of room in our applications.  So I have decided to experiment with using the 9205's in our compact daq chassis and have found that if I sample at 1khz then take an average i get not too bad of readings.  My systems are by no means high speed.  I update my display at a rate of 10hz and save data as slowly as 5min.  I tried taking a running average of 100 samples on 24 channels but I think that is eating up all my memory. 

In short I'm looking to get as stable readings that I was getting with the 9219 but using the 9205.  For my wiring I'm running both the chassis and the sensors on the same power supply, and then signals are being put through a shunt resistor (250Ohm precision resistor).  The 4-20mA cables are shielded, and the 1-5V cables to the card are only a foot or two long so there shouldn't be enough length to pick up much signal noise.  The mode of measurement is NSRE, as all my grounds are common so it doesn't seem that there should be any difference between taking a differential reading or the NSRE.

I would greatly appreciate any input that could be provided.

Thanks.

[RyanPoPo]

Hello Fbaird,

Your question seems to be more related to the specifics of the hardware acquisition for the 9205 versus the 9219 C-Series mdoules. As such, you would likely have more success with this question if you posted it on the Multifunction DAQ forum rather than Real-Time Measurement and Control, which is more for the Real-Time programming aspect of an application such as this.

Regards,

Ryan

[Fbaird]

Thanks RyanPoPo.  I think it is a bit of both.  If there is a way to acquire the signals and do some programatic filtering I would be all for that.  My question is manly to do with why does one card give me nice signals with little effort on my part while the other is so noisy that the data it generates is boarder line useless.  If there is a programatic solution I'm all ears.  I can't possibly be the first person to have this issue.

[RyanPoPo]

Hey Fbaird,

For looking into possible causes of noise on the hardware-side of things, I would definitely post on the forum I previously mentioned.

However, for software you definitely should be able to perform some type of filtering. It will really just depend on the characteristics of your signal to correctly configure the parameters of the filtering. For example, if you are seeing some higher frequency noise, but your signal is lower frequency (such as a thermocouple measurement) then we can simply pass your acquired data through a software lowpass filter to strip out the higher frequency noise.

What kind of signal are you trying to acquire? Also, what are your current acquisition parameters (i.e. sample rate)?

[Fbaird]

I'm just acquiring signals form pressure transmitter or level sensor.  They give very stable DC readings so extreemly low frequency signals.  I only want to read my samples at a frequency of 5-10hz. 

[RyanPoPo]

Hmmm, is there a way you could post some screenshots of a reading done with the 9219 in your system versus the 9205, just so we can view the type of noise you are seeing? A sampling rate of 5-10 Hz is pretty slow, so a filter may be less effective at those speeds.

[BruceAmmons · Accepted solution]

I am a fan of sampling at high rates, filtering, then downsampling and only outputting values at a low rate.  For your application, I would consider sampling at 1 kHz, using a low pass filter at 1 Hz, then recording the data at 10 Hz.  You get a nice, smooth signal even when there is a lot of high frequency noise.  This can usually all be done on the FPGA.  (oops, just noticed you are using cDAQ.  Still easy to do with PC.)

I would also investigate the hardware side of things.  I always get confused about which input mode to use, so I usually try several and figure out which one works best.  Maybe wire the signals up as differential to see if that works any better.  If it does, something is configured wrong for the single ended configuration.

Bruce

[Fbaird]

Thanks. That is what I ended up doing. But I am doing it on a cRio with part on the FPGA and part on the rt. I do have the cdaq with the same card and do sample at 1khz then just take an average. Seems to work alright.

[srm@viewpointusa.com]

if you are using a switching power supply, that may be a major source of your problem.

the NI 9219 has an anti-alias filter front end that the 9205 does not have.  switching power supplies typically have a high frequency components that will show up all over the place. you can filter the power supply or just use a linear supply to verify that is the issue first.

i have seen sensors that have a high frequency component as well (5 to 10 kHz).  you can see if your sensors have this issue by using a spectrum analyzer.

the 9205 is a multiplexed digitizer so you have to be careful of the output impedance of the circuit you are measuring. (see charge injection)

digitizing a signal at very high speed and doing your own spectral analysis might shed light on the source of the issue.