Using DOuble Float in STEP7

Floating point numbers are stored in the format REAL. A common data format

according to IEC754 used a 24 bit wide mantissa and an 8-bit exponent.

This results in an accuracy of 7-8 digits. Usually this is for applications

in control technology more than sufcient, but in certain cases can

lead to a problem. A typical case can which be solved with single-precision

only inadequate is a consumption meter. If you want to several Mwh (megawatt

hours) of total consumption adding up, taking a smallest power of

1 mW (milliwatt) at a distance of 10ms fairs and so you need a resolution

of 3.6 * 10^7 (equivalent 10MWs) and it would be a do add up 1* 10^-5

W's. To do this it requires a resolution of 12 digits.

The solution implemented by OSCAT is REAL Double precision and has a

resolution of about 15 digits. The implemented data type REAL2 consists

of R1 and RX, RX is here the value saved the first 7-8 points as Real and

the rest in one real R1. This data type has the advantage that no conversion

of REAL2 to REAL is needed, rather, the RX is rather part of single REAL

value.

Their code is available free of charge from oscat.de site. They are at version 3.22 at this time. In the attachment you will find a description of all the functions that can be used with the REAL2 datatype.
Hope this helps,
Daniel Chartier

Interesting implementation, although they broke IEEE rules, because 64bit float according to IEEE has additional bits for mantissa (10 instead of 8). But on implementing IEEE standard they won't achieve conversion-free handling between REAL and REAL2.

So i really doubt if double-float from PAC3200 would be "understood" correctly by OSCAT functions. On other other hand on using Siemens block you can work only with single-real. I've implemented 64bit math myself to solve such problem (+,-,*,/,shift and conversion of 64bit variables), but i'm unsure if i can place it here.