I use 6mm like whosa, works great, a bit of flexibility (which is a good thing), and all around better than 8mm. Wish I could find acme screws that weren’t so expensive, but its probably not worth the cost.
You don’t want acme because its an inch system. What you want is trapezoidal, which is what the metric equivalent is called. There’s an M5 thread with a 1mm pitch and a 2mm lead (2-start) that looks particularly nice. Robotdigg sells a nema14 with that on the shaft.
While we’re on the subject of threaded rods, do you have any recommendations as to what material they should be made of? finish (zinc plated?) straightness? Thanks!
@Whosa_whatsis
Have you done any empirical measurements on those M6 rods? I ask as I’ve never been convinced the pitch is exactly 1mm. I assume the tolerances would be quite low as they are not intended to be used for linear motion actuation. If this the case, you’d probably be no worse off with an imperial screw, as long as you were calculating layer heights to several dp.
I like stainless. They’re not usually rated for straightness, but it’s not a big deal if the rod isn’t over-constrained and the carriage isn’t under-constrained, though you should use the straightest ones available if you have a choice.
@Tim_Rastall I haven’t, but the issue in this case is less about dimensional accuracy than about consistency of layer height and its effect on surface finish.
My calculator can guess the errors from having imperial leadscrews 
I use M5 and I am happy. Few hundred printers out eith no problem
@Josef_Prusa but does it account for rounding errors introduced by converting the screw pitch to a steps/mm number?
The real trick to avoiding uneven layers is having a layer height that is a multiple of (microsteps / steps_per_mm), and making sure these numbers never get rounded. The math used in that calculator only works if the firmware accurately stores the number of steps/mm without rounding off a long decimal. This would be less of a problem if the firmware stored this value as mm/step instead of steps/mm. It’s not an issue with metric pitches because any conversion keeps the number of decimal places under control, but the conversion from inches combined with fractional nature of inch screw pitches (having prime factors other than 2 and 5 in the denominators) means that you can’t avoid getting a quotient with a nonterminating decimal.
WIth the 5/16 rod, the rounding error with .3mm layers comes out to something like .007mm. If there is a printer so accurate it can show those effects, the Solidoodle isn’t it. It seems to be hit and miss with Solidoodles whether you get a printer with a Z banding problem or not. The threaded rod is drilled out on the bottom and fitted over the motor shaft, and they don’t get that hole perfectly centered every time. A couple have been found that were way off, and could have been easily caught and rejected if they were doing proper QC in that area. Unfortunately there is no flex in the coupling, rod, or nut so the wobble all goes into the bed.
@Ian_Johnson Could you use the old I3 rubber hose coupler trick to solve that?
@Ian_Johnson All of my printers can easily show errors of that magnitude. the problem is not the difference in layer height, it’s the difference in the width of the threads being laid down due to the change in layer height because the volume of plastic being extruded into that area isn’t adjusted. This effect occurs at a regular interval producing visible banding. This is particularly visible on the parts of the Reprap Wallace prototype, as I was experimenting with a bad layer height and using opaque white plastic at the time that I printed it, and there’s also an image of a particularly striking example from @Jason_Gullickson in my article on Z axis artifacts.