Using a capacitor to reduce noise on AI signal

[Jmc006]

I have an application where we are reading 30+ AI signals which are 4-20 ma, converted to 1-5 volts using a 249 ohm resistor.  This is done for channel count reasons to stay within a PXI 8 card chassis.  Anyway - we have used this setup numerous times and always get noise on the signal, even at steady state.  Without getting specific on sample rates, I can tell you we ready the signal very slowly, and there is not a fast rate of change on the signals.  Typically I would average 15-20 samples and would get by.  But recently it was recommended to use a capacitor in parallel across the terminals for smoothing of the signal.  It seems fairly common, but I am struggling to determine a size of the cap.  So far I have come up with a 1uf, 100V, ceramic cap.  Can anyone tell me if this is ok?  And I would like to buy a size up, and a size down, can anyone recommend those?  like perhaps  .5uf, and 2uf, etc?  So I can try each one and see the effect on the signal.  I would imagine a cap WAY too large would kill the response of the signal, and one too small would have little to no effect?

Thanks 

[Minions]

Have you ever looked at the signal with a scope or meter to determine the frequency elements along with their associated Vac?  If you know the frequency, then it is easier to choose the proper filter.

[Rhys_CKTC]

Just a suggestion, but maybe you should try to look at the source of the noise and try to eliminate that?

Is the noise being picked up through mains? Are the signal cables shielded?

Rhys

[Bob_Schor]

Have you ever heard of a "low-pass filter"?  Do you know what a Time Constant is (and means)?  Have you ever heard of an "RC Circuit"?  If "R" = Resistance (in ohms), and "C" = Capacitance (in Farads), what dimension(s) does RC (implicit multipication) have?

This is very basic Electrical Engineering, or Electronics.  If you are contemplating measuring voltages and currents, you should know at least as much as a high school senior taking Physics ...

Bob Schor

[Henrik_Volkers]

You really should look at the noise sources, line? or switched power supplies or ??

Hook up a scope and have a look.

For a quick filter solution I made a graph for you

For common capacities the frequency response is shown. OK, the 470n is most rigth and the 22µ is most left 😉 

And  ~36V rated caps are fine (assuming a 24V sensor supply)  10V should work but ....

What is slow?  And how fast are you sampling?

[Henrik_Volkers]

Small update:

The cap will result in an error, so here is a detail with different scaling (100mA is 100% )

With a 470nF cap you get an 1% error for frequencies >(~)200Hz

for the 22µF the 1% error is reached at ~4Hz

[Jmc006]

Ok, I will try to simulate on the bench and measure with a scope.  

For some of the other questions:

*Yes the cables are shielded (Belden 2 wire + shield, foil wrapped)

*I am not an electrical engineer.  And telling me I should know the answer from high school does not help me.  

[Borjomy]

In this case, the capacitor is part of the LCR filter. L and R components are determined by the characteristics of the cable and the signal source (which are unknown). To create a filter with controlled characteristics of a single capacitor is not enough. At least one more resistor is needed to produce the RC filter. If the signal is differential, then the resistors must be two (one for each line)

[Henrik_Volkers]

---

@Borjomy wrote:

In this case, the capacitor is part of the LCR filter. L and R components are determined by the characteristics of the cable and the signal source (which are unknown). To create a filter with controlled characteristics of a single capacitor is not enough. At least one more resistor is needed to produce the RC filter. If the signal is differential, then the resistors must be two (one for each line)

---

We are dealing with a current loop  😉  the resistor is already build in the current source 😄

[protocoled]

If you'd like to do a little learning on this subject matter, I'd give the following white paper a read: http://www.ni.com/white-paper/3344/en/#toc4

Section 6 has some general recommendations on tracking down sources of noise and eliminating them. As others have said, a low pass filter is a quick and easy method to eliminate high frequency noise in your measurement setup. You could also try implementing a smoothing digital filter that performs a moving average on your data in post processing. Again, however, filtering techniques should be secondary to ensuring a proper system connection in the first place.