reducing noise of incoming data

[sayndesyn]

I have to say first off that I don't particulary have any experience with Lab View, but I have been given a "project" that requires me to jump in quickly.  I have a summer internship at a drilling company and they use an ultrasound transducer to get wall thickness readings off pretty thin stock that is in the order of .5 diameter.  Typical wall thickness that they need a reading on is in the range of .1 to .08.  The problem is that the stock is often long and while almost everything has been done to reduce vibration it is still considerable when getting a reading which is getting inputed into a labview .vi program.  It would be easy if the drilling could stop to get a reading, but this is not an option.  As such it can be hard to get an accurate reading of the wall when there is considerable noise.  Is there an area that anyone could point me towards in terms of filtering out the spikes so I could perhaps get something more resembling a sine wave instead of the jungfrau mountain range?  I was thinking from a non-technical point of view that there must be some sort of a statiscal analysis of incoming signals that allows the best fitting line to be plotted or something along those lines to smooth things out and make it easier for technicians to locate thin or thick areas.  Thank you for any help.  Sam

[Ray.R]

Can the filter be hardware?

Do you really want to filter those peaks?  Could they provide some sort of indication to what is going on?

R

[sayndesyn]

The spikes are not something that I want in the graph being that it is not indicitave of thousands of jumps in wall thickness, but rather in chatter from the tool that is cutting.  When you say filtering through hardware, exactly what do you meen?  Setup of the physical process?

[Ray.R]

How is the signal acquiired at the source?  It is an electrical (analog) signal going to a DAQ, right?  If so, then add a low-pass filter to cut out the spikes.

[unclebump]

What are the specs for the transducer that you are using?? Maybe it isn't the correct device to be using. Is there any other external hardware connected to the device. Maybe you could do some averaging in the A/D conversion.

[Cory_K]

One way to smooth noisy data is to calculate a moving average.

This is a very straight forward method to handle smoothing, and relatively simple.

[JÞB]

Another option would be to use a FIR bandpass filter on the sampled waveform.  Set the filter to accept the ultrasound xducer's frequency and reject out of band signals. 

Isn't math fun;)

[sayndesyn]

I'm going to be getting "into" this matter shortly and will have more info.  They have me multitasking you might say at the momen.  thanks for all the tips guys.  I will be trying out some methods soon and then send an update.

Message Edited by sayndesyn on 06-16-2009 07:55 AM

[sayndesyn]

This is the .vi file I am dealing with for future reference.

Message Edited by Support on 07-07-2009 03:55 PM

[Coq_Rouge]

Are you using labview 5.1? It will be quite hard for you to receive any help other than pictures. Just as you know it. Do you not have any newer Labview version?

But I think Labview 5.1 has all the functions you need.....