Hi,

I'm using Tia Portal V13 and an CPU 1214 DC/DC/DC with a analog input module.

In my application I'm using a load cell to weigh something (0-5 Kg), and I'm amplifying the output signal of the load cell from mv to V (+/- 0-10V). However, even with capacitors, the signal I'm getting in the AI module oscillates to much. Thus, I would like to know if  it would be possible to make an acquisition of "n" samples and calculate a average and show, or to calculate a median. Is it possible? Does anyone know how can I do something like that?

Thanks in advance 

hi

Check out the smoothing functionality of your Input. you find it in the properties of the AI in Device view.

See the picture 

hope this helps,

regards,

bruno_cardoso

Hi,

I'm using Tia Portal V13 and an CPU 1214 DC/DC/DC with a analog input module.

In my application I'm using a load cell to weigh something (0-5 Kg), and I'm amplifying the output signal of the load cell from mv to V (+/- 0-10V). However, even with capacitors, the signal I'm getting in the AI module oscillates to much. Thus, I would like to know if  it would be possible to make an acquisition of "n" samples and calculate a average and show, or to calculate a median. Is it possible? Does anyone know how can I do something like that?

Thanks in advance 

Bruno,

It will be a good idea to identify if the variation originates at the PLCs Analog to Digital conversion or from the device. Measure with you voltmeter and see what results you have. Also ensure you have a proper grounding.

If all else fails:    (PS - if you use it, you owe me a favour!)
(The following code is more effective than recording readings into an array and then calculating averages.)

FUNCTION_BLOCK "LowPassFilter"
TITLE = LOW PASS FILTER
{ S7_Optimized_Access := 'TRUE' }
AUTHOR : William_B
VERSION : 0.1
   VAR_INPUT
      READ  : Real := 0.0;   // [0.00 .. 1.00] Normalised reading
      FACTOR  : Real := 1.0;   // [0.01 .. 1.00] Adaption factor.. how big influence current reading is having
      ZONE  : Real := 1.0;   // [0.00 .. 1.00] A value bigger / smaller than this number will cause immediate jump
   END_VAR

   VAR_OUTPUT
      OUT  : Real := 0.0;
   END_VAR

   VAR
      Result  : Real := 0.0;
   END_VAR


BEGIN
    // L O W   P A S S   F I L T E R I N G   O F   A   R E A D I N G
    // READ   >> 0.00 .. 1.00
    // FACTOR >> 0.01 .. 1.00
    // ZONE   >> 0.00 .. 1.00
   
    IF #FACTOR < 1.0E-9 OR ABS(#READ - #Result) > #ZONE THEN
        #Result := #READ;
    ELSE
        #Result := #FACTOR * #READ + (1.0 - #FACTOR) * #Result;
    END_IF;
    #OUT := #Result;
END_FUNCTION_BLOCK

Helpful?
w

 Hi 

Thanks for the suggestion but I already did that before the problem I'm asking.

I used it and it helped reducing the variation but the problem I'm talking now is with this option active.

Thank you for the help 

Regards

mopS04

hi

Check out the smoothing functionality of your Input. you find it in the properties of the AI in Device view.

See the picture 

hope this helps,

regards,

 Hi,

First of all, thank you very much for all the help and the suggestions.

But I have a problem, I do not know SCL, I'm working with ladder because of that. Is there some tutorial where I can do the same (the code) with ladder? Sorry if I'm asking too much.

When I measure with the voltimeter of the device the variation exists, I think it seems not to be from the PLC.

Thanks for the help 

William_B

bruno_cardoso

Hi,

I'm using Tia Portal V13 and an CPU 1214 DC/DC/DC with a analog input module.

In my application I'm using a load cell to weigh something (0-5 Kg), and I'm amplifying the output signal of the load cell from mv to V (+/- 0-10V). However, even with capacitors, the signal I'm getting in the AI module oscillates to much. Thus, I would like to know if  it would be possible to make an acquisition of "n" samples and calculate a average and show, or to calculate a median. Is it possible? Does anyone know how can I do something like that?

Thanks in advance 

Bruno,

It will be a good idea to identify if the variation originates at the PLCs Analog to Digital conversion or from the device. Measure with you voltmeter and see what results you have. Also ensure you have a proper grounding.

If all else fails:    (PS - if you use it, you owe me a favour!)
(The following code is more effective than recording readings into an array and then calculating averages.)

FUNCTION_BLOCK "LowPassFilter"
TITLE = LOW PASS FILTER
{ S7_Optimized_Access := 'TRUE' }
AUTHOR : William_B
VERSION : 0.1
   VAR_INPUT
      READ  : Real := 0.0;   // [0.00 .. 1.00] Normalised reading
      FACTOR  : Real := 1.0;   // [0.01 .. 1.00] Adaption factor.. how big influence current reading is having
      ZONE  : Real := 1.0;   // [0.00 .. 1.00] A value bigger / smaller than this number will cause immediate jump
   END_VAR

   VAR_OUTPUT
      OUT  : Real := 0.0;
   END_VAR

   VAR
      Result  : Real := 0.0;
   END_VAR


BEGIN
    // L O W   P A S S   F I L T E R I N G   O F   A   R E A D I N G
    // READ   >> 0.00 .. 1.00
    // FACTOR >> 0.01 .. 1.00
    // ZONE   >> 0.00 .. 1.00
   
    IF #FACTOR < 1.0E-9 OR ABS(#READ - #Result) > #ZONE THEN
        #Result := #READ;
    ELSE
        #Result := #FACTOR * #READ + (1.0 - #FACTOR) * #Result;
    END_IF;
    #OUT := #Result;
END_FUNCTION_BLOCK

Helpful?
w

okay...

Let us take this to the next level.

1. Unzip the attached file - You will get LowPassFilter.scl.

2. Add the source to a project by selecting "add new external file". Use a new project - just to be on the safe side.

3. Right click the SCL file and ask for "generate blocks from source".

4. A new FB will be created with SCL code inside.

5. Study the code.. maybe you do not need LAD code equivalent anymore.

w

I did not thought this would be possible! 

Thanks a lot because now I know something new 

I did what you told but it's giving me an error when I do the step 3 (in the attachment).

I studied the code but I have some doubts: 
- I must normalize the output of the normalize function (0 to 1) before converting to Kg, must not?
- What if the factor for? I did not understand what the influence meant.
- The zone is a very small value that means the sensor is not working, isn't it?

Thanks again for all the help and all you are teaching me 

William_B

okay...

Let us take this to the next level.

1. Unzip the attached file - You will get LowPassFilter.scl.

2. Add the source to a project by selecting "add new external file". Use a new project - just to be on the safe side.

3. Right click the SCL file and ask for "generate blocks from source".

4. A new FB will be created with SCL code inside.

5. Study the code.. maybe you do not need LAD code equivalent anymore.

w

bruno_cardoso

I did not thought this would be possible! 

Thanks a lot because now I know something new 

I did what you told but it's giving me an error when I do the step 3 (in the attachment).

I studied the code but I have some doubts: 
- I must normalize the output of the normalize function (0 to 1) before converting to Kg, must not?
- What if the factor for? I did not understand what the influence meant.
- The zone is a very small value that means the sensor is not working, isn't it?

Thanks again for all the help and all you are teaching me 

Sorry about the error... in haste I changed the author name from my work-alias to the forum-alias without thinking of the typing rules they apply. The SCL source was replaced in the earlier post.

[Normalized]
When you bring in the analog value you are going to scale it to [kg]. If you use the standard S7-1x00 functions you will normalise (NORM_X) and then scale (SCALE_X) it. This FB is design to be after the NORM_X but before the SCALE_X.

This is not a rule. The function works without regard to the 0.0 - 1.0 range.

[#Factor]  
..see pseudo code on  [https://en.wikipedia.org/wiki/Low_pass_filter]

It tells the algorithm how much influence does the new reading have and how much influence does history have.

Definition of "Influence" - contribution / weight / value added
The size of the contribution of a variable to the end result.

With #Factor := 1.0, you will have what you have now... none of the previous readings will have any contribution to make to the output.

With #Factor := 0.1 the now input reading will contribute 10% to the output while the previous result (history) will contribute (100% - 10%) 90% to the output .
{RESULT} := 0.1 * {NOW READING} + 0.9 * {PREVIOUS RESULT}

[#Zone]

The problem with the low pass filter is that it is resistant to change. If the input value makes a jump change, then the output should do the same. So.. if the input stays within the zone (read: hysteresis / dead zone) the filtering algorithm is used. When the input value jumps outside the zone, that input value goes to the output as-is.

Now if you do not use Normalised values on the input, then you have to make the zone of the same scale. For example when reading up to 5000g, a jump of 50g should ignore the history values and change without any delay.

Does this help you?
w

 Hi,

I downloaded the file and now it's not giving erros. Thanks 

[Normalized]:

I will put it there where you told and give it a try. But like you said it can be put at the end (after the scale) - I'm asking this because the load cells that I'm using are not directly proportional all the way, and I'm thinking about doing a function (polynomial of 4th Order) to give me the Kg accordingly to tests with known weights. And it will still have all the variation, and I would like to know if it still works (more confirmation, because I think you already told that with changes it works).

[Factor]:

I understood your explanation about the Factor and what it is. But how do I know which value to use? It's achieved by experimenting, by trying different values, isn't it?

[#Zone]:

In this I have an huge issue unfortunately because my device/assembly is giving me an variation of at least 200g and everytime I use the liquid (in the 5Kg reservoir) it only reduce 30/40g. So the filter will not see this variation, I'm right?

Thanks again for all the help!

Regards