rounding/truncation fixed-point

[mkenobi]

Hi, how to round off or truncate a fixed-point number? for example 32-bit to 16-bit. any help (docs or code) will be appreciated. thanks.

 

[eda_wiz]

it depends on the type of data in the 32 bit number.. if all bits contain data truncate the 16 LSBs .. if say some of MSBs are repeated sign bits truncate them except one. or if the data content varies from 0 to 2^32 select a variable truncation point and adjust the bit postions in the next block..

 

[mkenobi]

i'm doing 16-bit fixed-point [SIII.FFFFFFFFFFFF (s:sign, i:integer, f: fraction)] multiplication. 16-bit product is needed

 

[eda_wiz]

ok in your case the result is 31 bit.. (16+16-1).. so you can remove 1 MSB and 15 fractional LSBs to get 16 bit. the result would be SIIIIII.FFFFFFFFF