A little note on some of the test results, and a movie of the process.

As I mentioned earlier, I built a tinyg driver for firenodejs and am using that to test the precision of my design using various rail options. I first mounted the cheap chinese rails I sourced from here: http://www.robotdigg.com/product/493/MGN12-1H-L600+Linear+Rail+and+Carriage+for+Kossel+XL

I have had these rails on for quite a while. I can start by saying that these things are cheap in every single way, and I dont mean "kind of cheap". The literally clank their way down the track, even after cleaning and lubing them up real good. Of the 3 I got 1 was VERY bad. I put that one on the X becuase I figured as long as it was directly driven by a motor it might be OK. I took the best of the 3 and put it on the passive rail. Hoping to make some lemonade.

Based on my tests the machine would move maybe 8-9 times nearly perfect. Below is a video of the cheap rails in motion, as you can see, pixel perfect precision (each pixel is about 50 microns) As you will see, motion is perfect during this leg of the test. BUT, inevitably, and eventually, that catching, binding comes back to bite you, and when you bind in a movement, it might throw you off as much as 1mm. The binding only occurs in the Y axis, that is the one with that passive rail that I thought might bind on me, it does. The good news is that is the rail with the worst rail/slide combo I had, and believe me, it is TERRIBLE. Yet in all the test it NEVER was off more than 50 microns, pretty impressive.

This leads me to my conclusion on the cheap chinese rails... they will work, IF you actively drive them. If you actively drive them they will be accurate too. Be aware that you will need to crank up the pots on your stepper drivers to deal with the friction, but they can work. That said, after I began the rail swap to the second set I will test, I found the rails/slides to be very dirty, a few metal shavings here and there. I'm not convinced that these will last, so even if I could make them work, I'm probably not inclined to every build anything with them again. $70 bucks down the drain all in the name of science :)

Moving along to the Hiwins... for now here is a video of the testing procedure: