Float precision for Ultra-Precision applications

That sounds great!
So Linuxcnc is working with doubles internally but not outputting it to HAL in 2.9 right?

Hal floats are 64-bit doubles since 2.6 or so way back. But hal integers are only 32-bit, s32 u32. Not 64-bit.
Don't know how it is internally.

Oh great so i probably have no issue at all!
My drives only take 32-bit Integer inputs so that is fine.

Thank you!

since the core questions are answered:
I am building an air bearing Diamon Turning Lathe. The goal is to do single point diamond turning of telescope mirrors or similar.
Using ironless linear motors and high resolution glass scales. It is going to get calibrated with an Interferometer so i should get down to 100nm or so of straightness of the axes and linearity of the glass scales.

There will be watercooling of all relevant components down to about 100 millikelvin precision.
I am using Triamec servo drives.

The progress will be documented on my Instagram (Instagram.com/leo.wolf.the.engineer). Please follow if you are interested!