Moving Average in S7-300

Hal9000 · Jul 4, 2010

Does anyone know if there is a standard block to perform Moving Average in S7-300?

I can't see one in the library, even though Siemens suggest MOV_AVG

My CPU is S7-315 PN/DP

KalleOlsen · Jul 4, 2010

You could take a look in the oscat site. Check the library and f.ex. the block FILTER_MAV_DW.

K

Hal9000 · Jul 4, 2010

Hi, Thanks for the information.

I've downloaded the OSCAT311_Step7.zip file from the link and I can see the FILTER_MAV_DW in the .pdf document.

I've not used this material before, so I would be grateful if you could tell me what do I do with the zip file? Do I simply retrieve like a normal S7 archile file and drop it into S7_LIB?

Regards,

Justin

Hal9000 · Jul 4, 2010

It's ok, I've just answered my own question.

Paul B · Jul 4, 2010

If you have any problems with what you have, I'll be glad to share a rolling average function that takes X samples, throws out the high & low, and averages the rest. I have it for INT and for REAL inputs.

Hal9000 · Jul 5, 2010

Hi Paul,

I would be very grateful if you could, I think your solution sounds more in line with what I need.

Basically I need to average my PV over say 10s and chop out some noise. My PV is a REAL.

Correct me if I'm wrong, but when you throw out highs and lows is that classed as "weighted", or am I completely on the wronmg track?

Cheers for now,

Justin

Paul B · Jul 5, 2010

Not sure about the weighted question. It probably wouldn't be too hard for someone to remove the "throwing out high and low" part.

Here is a library with the requested rolling average code.

Thomas_v2 · Jul 5, 2010

Hal9000 said:
Correct me if I'm wrong, but when you throw out highs and lows is that classed as "weighted", or am I completely on the wronmg track?

I think this would be called a "Winsorized mean":

http://en.wikipedia.org/wiki/Winsorized_mean

If you know that your signal has more outliers than noise in it, a median filter will give you a better signal than a simple first degree low-pass filter.

Hal9000 · Jul 5, 2010

Thanks Paul,

I don't have S7 simulator on my PC, so I won't be able to test it until I get to site.

If I want to sample say 60 values at 1/value/sec. Should I simply call the block once a second to do this?

Justin

Peter Nachtwey · Jul 5, 2010

This has me wondering

It was mentioned that the PV was being filtered, is this a PV for a PID? If so then has anybody considered what adverse affects this will have on the response on the PID? Does anybody know how this will affect the bandwidth of the feedback device?

Excuse me for being an old fuddy duddy engineer but tried reducing the source of the noise first?

The sample time and the number of samples affect the bandwidth of the filter. Do you know how? If not, then why 10 samples as opposed to 20 or 30 or even 2?

Paul B · Jul 5, 2010

Hal9000 said:
Thanks Paul,

I don't have S7 simulator on my PC, so I won't be able to test it until I get to site.

If I want to sample say 60 values at 1/value/sec. Should I simply call the block once a second to do this?

Justin

Exactly. I put examples in the library in OB1. Right now they are being called every 100ms. Change the timer preset to 1 second and make the length "60" and then run it in the simulator.

This code uses an FC and a global DB.

Paul B · Jul 5, 2010

Peter Nachtwey said:
It was mentioned that the PV was being filtered, is this a PV for a PID? If so then has anybody considered what adverse affects this will have on the response on the PID? Does anybody know how this will affect the bandwidth of the feedback device?

Excuse me for being an old fuddy duddy engineer but tried reducing the source of the noise first?

The sample time and the number of samples affect the bandwidth of the filter. Do you know how? If not, then why 10 samples as opposed to 20 or 30 or even 2?

These are excellent question for the original poster.

I use a rolling average for display purposes. For example on low-end systems with low margins it is common to use inexpensive load-cell amplifiers. This signal always fluctuates, making the displayed weight on the HMI go up & down. It bothers the operators. So I filter the display tag with a 5-second rolling average. In the PLC code I use the value from the amplifier.

I have used a 3-second rolling average for the PV of a PID loop controlling temperature in a room the size of two football fields. The loop is called once per second and tuned S L O W.

But you are concerned about loops that must react fast...

Hal9000 · Jul 5, 2010

The PV I mentioned is not being used as a PV in my PID loop. Instead I'm reading two Integers over modbus, one is a whole part the other the fraction part. I convert these two integers to a real and display the "inlet water temp" on my MP377.

The issue is my modbus values are a little erratic, I've proved it's not my code because I get the same effect in Kepware. So I suspect it's a feature in the slave device. This is why I would like to chop off the high and low and then say average over a minute.

The averaged value will then (hopefully) be more stable and I will use this in a fallback routine to switch my PID loop to manual if the temp goes above 35 deg C > 30sec.

If it doesn't work I will arrange to have a temp transmitter installed and wired back as a 4-20mA.

L D[AR2P#0.0] · Jul 5, 2010

Are you reading the data from the modbus device data "consistantly" as you would for a Profibus device ?

Moving Average in S7-300

Hal9000

Member

KalleOlsen

Member

Hal9000

Member

Hal9000

Member

Paul B

Member

Hal9000

Member

Paul B

Member

Thomas_v2

Member

Hal9000

Member

Peter Nachtwey

Member

Paul B

Member

Paul B

Member

Hal9000

Member

L D[AR2P#0.0]

Lifetime Supporting Member

Similar Topics