martinav (Customer) asked a question.

Edited April 18, 2022 at 2:56 PM
Pressure transducer stability/accuracy

I was hoping for a little better accuracy with stability with this device. Please have a look:

 

https://youtube.com/shorts/p0h8lQ2RpuM?feature=share

 

This is a SPTD25-20-1000H transducer, and I was hoping for a pretty solid result to one decimal place. Seems that I would be hard pressed to get a reading stable to the nearest whole number.

 

There are two of these shown here. They are on the same pipe too. Surprised to see such a different number too. Do they need calibrated?

 

Maybe I got the wrong idea?? Is hysteresys what I need to look for for a more solid number at an accuracy?

 

I dont know if I'm even saying this right. Please be gentle.

  • 8 answers
  • 333 views

  • martinav (Customer)

    I think I can manage with this... 100ms looks reasonalbe, and gets me a nice firm xx.x number when rounded.

     

    image

    Selected as Best

  • RBPLC (Customer)

    You're using a 1000 psi sensor, to measure ~ 3 psi. You can't expect sub psi level accuracy with this big a range. If you want to smooth out the readings, you probably need to filter the input reading as well. Accuracy on these sensors is listed as <=0.5% of full range. Being conservative, you shouldn't expect better than +/- 5 psi readings.

  • martinav (Customer)

    Pressures will be up to 550psi. I'd been happy with a whole number accuracy. One decimal would be nice. For my purpose, this sensor is needed to detect a sudden pressure drop. It may not be quick enough if I have to wait for a 10psi change, lets say. Anyway, thanks.

  • Todd Dice (Customer)

    I think you have to code the filtering of the analog signal to get the result you wish to achieve.

     

    My evidence of this is purely anecdotal, but with a Click PLC using an analog ultrasonic sensor reading grain height, I used a block copy triggered at 100ms intervals taking 50 readings and into a math instruction those 50 readings were summed and divided by 50 to obtain a stable level workable for the remainder of the code.

    Expand Post

  • martinav (Customer)

    @Todd Dice (Customer)​ 

     

    Would you mind sharing your code blocks for your filter? I'll search here too, but I'd like to see what you have thats actually working for you. The Math block is no problem. But how you took the 50 pressure points, and summed them together efficiently I would like to see.

    • Todd Dice (Customer)

      I think you're using Do-More Designer (DMD), so how I coded in Click won't match. However, there is a FILTER instruction in DMD under the Analog/Process tab.

      • martinav (Customer)

        Ok, ok. Cool. Not used it before. I'll check it out.

  • martinav (Customer)

    I think I can manage with this... 100ms looks reasonalbe, and gets me a nice firm xx.x number when rounded.

     

    image

    Selected as Best

    • HOST_franji1 (HOST Engineering)

      Gotta love that embedded Trend View in the FILTER instruction status (in addition to the current value status of the In/Out parameters)! 😁

 

Did this information resolve your issue?