How to calculate or estimate floating point rounding error for a specific number

Question

I\'m converting 64 bit numbers to 32 bit numbers and in doing the conversion I want to ensure that my numbers only get bigger in the conversion and not smaller. I know that