2 Replies Latest reply on Feb 1, 2018 1:01 AM by jyi

    Data Limits



      sometimes because of noise I'm getting ill logical data either very big/small is there any way to put filters/ boundaries to accept the data withing some range

      for example

      i have temperature reading from 0:50 what are the settings i have to use to neglect the values beyond this range


      Best Regards;

        • Re: Data Limits

          Hey, Waly.


          Unfortunately, it is not possible to trim incoming data inside a specified range; this is documented by Known Issue 18617OSI8. However, you could create an analysis in Asset Analytics that takes the raw data (with noise) as input and then only stores data within a specified range in another output tag. See below for a sample analysis and results:

          As you can see, the output tag trimmed_data only stores values within the range 190 < x < 210.


          However, if you are not concerned with what is archived and only would like to filter values outside of a range when viewing data, you could use a Filter Expression. In Excel, for example, you could use the Compressed Data PI DataLink function using the filter expression 'tag' > [lower_bound] and 'tag' < [upper_bound].


          I hope this helps!



          3 of 3 people found this helpful
          • Re: Data Limits

            If you are ok with float16 data resolution, then you can do it by setting your tag type to float16. Below only works for float16 and will not work for float32.

            Below is exerpt from KB00472 - "Under Range"/"Over Range" Behavior in float16 and float32 tags.


            Float16 tags require the data value to be within the Zero and Span attributes (range = zero + span). For example, if you have Zero set to -50 and your Span set to 200, your acceptable range of values will be -50 to 150 . Any values under -50 will be "Under Range" and any values over 150 will be "Over Range."


            Please note that setting zero and span attribute right in the first place is critical in doing so. For float32 and float64, Mike's method will work.