Data fluctuations (noise filters?)

[Mikey2]

Hey everyone,

I'm running into some problems with data fluctuations under steady state conditions. I don't know if its just noise, or hardware starting to go bad. I've always noticed this problem but it seems to be getting worse over time... although our equipement is well maintained. I've attached a pdf file with just a sample time vs. temp plot to illistrate whats hapening. I get these "noise" spikes that can be as great as 2 degrees. I've also attached a simple VI that i used. Maybe i'm missing a noise filter or some sort of scan rate averaging? If it is a hardware problem, is there anything i can include in the VI that may atleast minimize what i'm seeing? 

Thanks guys, any help would be appreciated!

-Mike

[johnsold]

Mike,

That data looks like a typical 1 LSB variation. Are you very near the resolution limit for your data acquisition device? What kind of sensor and signal conditioning are you using?

I looked at your VI with a much newer version of LV than yours, so many of the subVIs are not available.

A moving average filter is a simple way to smooth the data. I don't recall what filter options were available in LV5, but you can easily build a moving average filter with a shift register.

Lynn

[Mikey2]

Thanks Lynn,

I'm fairly new at this, so the term 1 LSB variation doesn't ring a bell. The setup i have is a 32 channel TC 2095 to an SCXI 1102. I'm not very familliar with filters, let alone a moving average filter, but i will look into it and see if i can figure it out. It just appears to me that the spikes are getting a little more applified from what i'm use to seeing.

Thanks again,

Mike

[johnsold]

Mike,

LSB stands for Least Significant Bit. If the Analog to Digital (A/D) converter converts the input votage to a number containing 8-bits, it is called a 8-bit converter. An 8-bit number can have one of 256 possible values. The least significant bit is the bit representing 0 or 1. If the average value was 45 and had a 1 LSB variation, the values would be 44 and 45 or 45 and 46. Looking at the flat part of the middle of your graph (lower trace) I see values of about 86 and 88 degrees but nothing between those values. And all the steps in the waveform are approximately the same size. These are the characteristics which make me think of a discrete sampling issue. Some A/D boards have resolution (or LSB) equivalent to about a millivolt. Some temperature transducers have outputs of about a millivolt per degree.

If your board has an amplifier, try setting it to a higher gain or, equivalently, reduce the voltage range. If the range is +10 to -10 V, change it to +1 to -1 V. That will make the steps much smaller and appear to smooth your data.

Lynn

[Mikey2]

Thanks a lot! Very nice explanation.