Interpolation with Floating-point Numbers

This Documentation will explain how to interpolate Floating-point Numbers with Breakpoint Tables.

The Problem

Interpolation is a type of Linearization that uses the RVAL field of a record with which to interpolate. The RVAL field can only be a LONG value, i.e. integers. Floating-point Numbers are however DOUBLE values and are only saved in the VAL field, the RVAL value stays 0. No Interpolation is performed.
The workaround of simply forcing a DOUBLE value into the RVAL field, e.g. by writing it directly into the RVAL field, will lose all decimal digits, therefor lose accuracy.

The Solution

To circumvent this, an intermediate output record is needed to scale the value. This record will do the following:

  1. gets the value of a record that has the value with which to interpolate.

  2. multiplies it by a factor of 10n, n depending of the needed accuracy, i.e. decimal digits needed (for this example, n will be 3).

  3. writes the new value to the RVAL field of a record which then interpolates with that breakpoint table.

To give an example, lets say a tank fill level is calculated in millimeters with three decimal digits and needs to be interpolated to get a fill level in liters. This record gets that fill level in mm written into by a calcout record. Note that since all records here are Passive, the calcout record should force the scaling record to process, for example via the FLNK field.

record(ai, FillLevel:filllevelMM)
{
	field (DESC, "Fill level of a helium tank in mm")
	field (DTYP, "Soft Channel")
	field (SCAN, "Passive")
	field (EGU, "mm")
}

The scaling record is an analog output record, that should look like this:

record(ao, FillLevel:scaling)
{
	field (DTYP, "Raw Soft Channel")
	field (OMSL, "closed_loop")
	field (SCAN, "Passive")
	field (OIF, "Full")
	field (DOL, "FillLevel:filllevelMM.VAL")
	field (OUT, "FillLevel:filllevelL.RVAL PP")
	field (LINR, "LINEAR")
	field (ASLO, 0.001)
}

  • DTYP needs to be Raw Soft Channel, so the RVAL value is written to the record in the OUT field.

  • OMSL set to closed_loop enables the input field DOL.

  • OIF is set to Full because otherwise the VAL field is incremented and not used as a new value.

  • DOL is where the value is taken from.

  • OUT is where is written to. Its important to write to the RVAL field to enable interpolation in the target record. The PP flag is set to process the receiving record.

  • LINR needs to be set so ASLO is used.

  • ASLO is set to 10-3. This is because the way this record works is by reading the field specified by the DOL field and saving that value in its VAL field. Then, due to being a Raw Soft Channel, calculates the outgoing RVAL. Since VAL = RVAL * ASLO, RVAL = VAL / ASLO.

The receiving record then can interpolate with a breakpoint table, in this example a custom one called filllevel.

record(ai, FillLevel:filllevelL)
{
	field (DESC, "Filllevel of the Heliumtank in L")
	field (DTYP, "Raw Soft Channel")
	field (SCAN, "Passive")
	field (LINR, "filllevel")
	field (EGU, "l")
}

  • LINR set to filllevel tells the record to use the breakpoint table filllevel

  • DTYP needs to be Raw Soft Channel again to use RVAL.

Due to the input value of the breakpoint table now being scaled up by a factor of 103, the breakpoint table also needs to be adjusted by this factor. For this example, while the (abbreviated) original breakpoint table looked like this:

breaktable(filllevel) {	
	0.0	0.0
	10.0	1.0
	20.0	3.0
	30.0	6.0
...
...
	1470.0	4231.0
	1480.0	4237.0
	1487.0	4239.8
}

It now has to be scaled to look like this:

breaktable(filllevel) {
	0.0     	0.0
	10000.0		1.0
	20000.0		3.0
	30000.0		6.0
...
...
	1470000.0	4231.0
	1480000.0	4237.0
	1487000.0	4239.8
}