I want to generate numbers with a range of 0.625 to 1 with increment of 0.001

In reply to dave_59:

But Dave, the problem comes from inability to represent certain decimal numbers exactly in binary(IEEE 754). for eg, approximating 0.2 with 1/8+1/16…etc is going to have round-off errors.
whether we scale or not, don’t you think we may eventually have this issue ?