Analog signal filter question

[GCone]

This sounds like a stupid question, but I can't find an answer anywhere. I realize that a lower filter freq on an analog input channel results in more smoothing, less noise, and less accurate representation of the signal, but my question is this: Is there any correlation between the filter frequency and the mains noise you might expect to be present. In other words, would you necessarily set a 50hz filter if you have 50hz power in the area, and a 60hz filter freq for 60hz supply power?

 

[James Mcquade]

What plc and analog module are you using?

james

 

[GCone]

rockwell Compactlogix 1769-L33ER, 1769-IF16C analog input module.

 

[Pete.S.]

You're mixing up two different filter settings.

Usually there is a notch filter that will filter out 50/60Hz from the analog signal, and nothing else.

Then there is a second filter that can smooth the signal which is basically a low-pass filter.

So the 50/60Hz filter is set to whatever mains you have.

And the other one depending on what type of signal/sensor you have and if it jumps around or not.

 

[GCone]

Thanks, that makes sense. I don't believe they have a notch filter though.

 

[Pete.S.]

Thanks, that makes sense. I don't believe they have a notch filter though.

I think today it's just software so no actual hardware present. But maybe band-stop filter is more accurate.

In old PLC systems they had actual RC components that could be changed.


PS. I had a look at the manual for the analog in you have.
They say it's a digital filter (p.39) but there is actually only one filter setting, not two different filters. Just as you said.
http://literature.rockwellautomation.com/idc/groups/literature/documents/um/1769-um018_-en-p.pdf
I guess you have no choice but pick 60Hz if the input signal is slow enough.

.

 

[GCone]

I think what makes it a little confusing is that two of the filter frequency settings are 50hz and 60hz, common in the US and Europe, which might make someone think those are suitable filters for when 50hz or 60hz power is in the area, while actually it's an entirely different thing than as you said a notch filter or band-reject. Thank you.

 

[Peter Nachtwey]

What is the frequency of the process?
What is the process.
What is the sample frequency?

Even most motion control applications have a frequency of motion less than 10 Hz.
Usually a low pass filter is all you need. The higher frequency components on the high side of a notch filter are too high to respond to anyway.

When using notch filters you usually want to make the notch very narrow. How do you adjust the width or Q of the notch filter?

 

[odwyerpw]

This is an interesting thread...

I'm seeing occasional spikes on analog inputs (temperature transmitters 4-20mA) on the Flex 5000 5094-IY8. It's on multiple channels so , it's not just 1 bad field termination. Something is inducing a spike. While we hunt down the source of the spike, I'd like to apply some sort of Filtering...

Notch filter for this channel is at 60Hz which is appropriate...but they also allow a digital filter... So, thinking of bumping that to 5ms or 10ms (I don't need speed... they are slow reacting temps... tons of process lag.. so a little measurement lag isn't going to hurt anything)..

Next step might be a low pass filter in the function block diagram that contains the effected PIDE ... (the spike is influencing the output which is undesireable.)..

 

[Mickey]

This might work...

 

[MaxK]

"Notch filter"... "low pass filter"...
Im redneck and dont understand what are you talking about

And I assume you are not interested in the redneck's ravings about adding a filter to a closed loop (sensor in a closed loop?).
I also think you shouldn’t bother building a model and bode plot of a closed-loop (although if there are tons of lag, the filter most likely will not have a significant effect).

But still, I’ll ask a redneck question:
If you have short spikes, then why not just add a reject-condition:

If abs( filtered value – measured value) < delta
Then filtered value := measured value
Else filtered value := filtered value

delta - physically the maximum permissible change in the measured value during the period of time between application of the reject-condition (apparently scan)

and another stupid question:
Are the concepts “filter” and “lag” somehow connected for you?