[Logix 5000] When should I worry about adding a small number to a large number?

[defcon.klaxon]

Hi all,

I'm familiar with the idea of an issue where at some point, adding a small number to a large number breaks down. I was wondering with an AB ControlLogix CPU, when should I start to consider that as an issue as far as numbers?

 

[L D[AR2P#0.0]]

Are referring to floating point numbers - if so how many bits?

 

[Nova5]

once it gets to exponetial point adding small numbers won't do anything. At that point you'll need to use a second one be it another float, an Dint, etc. when you max out the first one, increment the second by 1, zero the first, then you can just multiple the second by the max of the first and add the first for a display number. your HMI may be capable of a longer display number than the PLC's blocks.

A little here, Ken suggests hitting up Wiki http://www.plctalk.net/qanda/showthread.php?t=43932.

 

[drbitboy]

32-bit IEEE-754 floats* lose significance at one part in 16Mi (2**24 = 16,777,216). e.g. there is no fractional value (e.g. 0.5, 0.9, etc) at or above 8,388,608.0 i.e. the next possible value is 8,388,609.0, so if you add something less than 0.5 to the 8,388,608.0, the result will be the same i.e. 8,388,608.0.


Similarly, around eight billion, you cannot add anything smaller than one thousand.



* See here