Help with Filtering

[Tim Ganz]

I have a question on filtering. I am working with a center winder that is ran by a compact logix brick processor and the program is filtering a laser sensr like this

Expression Local:3:I.Ch0Data/40+(Diameter_Input_Filtered*39/40

Destination Diameter_Input_Filtered

Can someone help me understand how this works? I am really struggling and math is not my strongest subject.

 

[rdrast]

Write it out and look. You basically have:

Output = Input/40 + (Output * 39/40)

That is just a slightly strange form of an averager, without having to devote a bunch of memory locations.
It integrates every scan, but might never actually converge.

Just being lazy and dropping that into Excel, even after 250 scans with a raw input of 20, it only hits 19.9661.

 

[Tim Ganz]

Is there a better way to average it in compact logix? It still is not filtered enough using what is currently there.

 

[Peter Nachtwey]

There is nothing strange about the filter. It is a low pass IIR fillter.
Y=A*Y(n-1)+B*X
Where A=0..1 and B=1-A
If you want smooth the data even more then make A bigger. Right now A is 39/40. Try 0.99 and B=0.01
There is a lot information about IIR low pass filters on the web. One can calculate A and B in terms of a response time constant or cut off frequency.

 

[Mickey]

Here's another .

FV=FV+C(NV-FV)
Where:
FV= Filtered value
C= constant (range 0-1) The smaller the number the more dampening.
NV= New Value

 

[Tim Ganz]

So i would do

Output = Input/0.01 + (Output * 0.99/0.01)

Is that correct?

 

[Mickey]

So i would do

Output = Input/0.01 + (Output * 0.99/0.01)

Is that correct?

No, that would give you a very large number. Run the math to see for
yourself.

 

[jjh1982]

I have a question on filtering. I am working with a center winder that is ran by a compact logix brick processor and the program is filtering a laser sensr like this

Expression Local:3:I.Ch0Data/40+(Diameter_Input_Filtered*39/40

Destination Diameter_Input_Filtered

Can someone help me understand how this works? I am really struggling and math is not my strongest subject.

What type of laser sensor? Is it a "noisy" signal, or a "noisy" medium which the laser is sensing? There are some great laser sensors out there which I have personally used in high noise environments(material being sensed and high freq.) for dynamic height control. They filter the "environment's noise" for you!

 

[creativepaper]

I've attached a spreadsheet which graphically shows these 2 filters that they are mentioning.
I put in process change and noise amplitude.
I set the sample as twice the time so a small step will be seen for unfiltered.

Playing with the constants on the filters always helped me.

 

[Tim Ganz]

Thank you all for trying but it looks like i am in over my head with this one. This is some of my first plc work and i thought i would try to fix the problem but it will have to wait till the plc guru is back from vacation.

 

[bernie_carlton]

In your expression provided:

Local:3:I.Ch0Data/40+(Diameter_Input_Filtered*39/40

To make MORE filtering (and slower response) then Increase 39 and 40 by 1 at a time. The number at the '39' position will always be 1 less than the number at the '40' positions.

So for twice the filtering you could change it to:

Local:3:I.Ch0Data/80+(Diameter_Input_Filtered*79/80

And, as was being pointed out, more filtering will decrease the effect of noise and very short terms stuff but will make it slower to react to longer term changes.