Filtering Analog Signals

[nate]

I am currently using a PXI 6025E board to read an analog signal from a weigh station. The items that are being weighed are in motion so the signal contains some vibration from the item being in motion as it crosses the scale. I would like to be able to filter the analog signal so that I just get the pure weight and not the noise from the vibration. What is the best way to filter the signal using LabView?

[preston_johnson]

One method is to use a "median filter" this is a statistical median of the data you have collected over a brief period of time. Presumably, you will have a good number of cycles from the vibratory motion. Use the statistical function for median. This will sort the data from most to least and produce a number in the middle of the list.

A second method is to use a low-pass filter to filter all data above say 1hz.

A thid method is to use the DC and RMS function and take the DC output of the function.

Try all three to see which produces the best result.

Sincerely,

Preston Johnson

[urs boegli]

Another method is the store a number of results and calculate the
average of these points.
Do sample fast to get more data and you will get a better result.

The method with the best results depends on the noise expected, on its
spectrum and amplitude distribution.
Median is fine if you have a lot of good data and a few wrong ones far
away from the 'real' result..
AVG and Lowpass is fine for symetrical noise.
DC / AC is fine in the same case, but the parameters cannot be chosen by
yourself.

Good Luck
Urs Bögli

preston johnson schrieb:

>One method is to use a "median filter" this is a statistical median
>of the data you have collected over a brief period of time.
>Presumably, you will have a good number of cycles from the vibratory
>motion. Use th
e statistical function for median. This will sort the
>data from most to least and produce a number in the middle of the
>list.
>
>A second method is to use a low-pass filter to filter all data above
>say 1hz.
>
>A thid method is to use the DC and RMS function and take the DC output
>of the function.
>
>Try all three to see which produces the best result.
>
>Sincerely,
>
>Preston Johnson
>
>

[urs boegli]

Another method is the store a number of results and calculate the
average of these points.
Do sample fast to get more data and you will get a better result.

The method with the best results depends on the noise expected, on its
spectrum and amplitude distribution.
Median is fine if you have a lot of good data and a few wrong ones far
away from the 'real' result..
AVG and Lowpass is fine for symetrical noise.
DC / AC is fine in the same case, but the parameters cannot be chosen by
yourself.

Good Luck
Urs Bögli

preston johnson schrieb:

>One method is to use a "median filter" this is a statistical median
>of the data you have collected over a brief period of time.
>Presumably, you will have a good number of cycles from the vibratory
>motion. Use th
e statistical function for median. This will sort the
>data from most to least and produce a number in the middle of the
>list.
>
>A second method is to use a low-pass filter to filter all data above
>say 1hz.
>
>A thid method is to use the DC and RMS function and take the DC output
>of the function.
>
>Try all three to see which produces the best result.
>
>Sincerely,
>
>Preston Johnson
>
>