TechShop has all sorts of nice parts to build a sweet workholding fixture (Picture 1). I took the 3-jaw chuck off of the 4th axis option and clamped it to the table. There was a feature on the back of the jaws that would have interfered with the clamping, so I stacked it on top of four precision 1-2-3 blocks.
In a previous prototype, I had mounted the entire 4th-axis (not just the jaws) facing upwards. This ended up being too tall for fitting the auto-tapping mechanism into the machine. So for my part, it was better with just the jaws.
I centered the part with an edge finder using a technique similar to the one from this two minute YouTube video
. However, this method is too inaccurate for this part -- it might leave some features as much as 0.005" off center, which on this piece would cause some hole chamfers to start to run into an o-ring groove.
So after rough-centering, I went around the groove with a dial indicator (Picture 2). During this step, I discovered that the Tormach PCNC 1100 has about one thou of backlash between the DRO (digital read out) and the table. In other words, depending on a few factors, the machine might be about 0.001" off from where it thinks it is.
In my experience, most machines with DRO's reference the table itself, so they don't have this kind of backlash behavior. My guess is that the positioning on this machine is based off of encoders on the stepper motors. Imperfections in the drive train then introduce the backlash.
To be fair, the Tormach PCNC 1100 is very good at being what it's supposed to be -- an affordable, hobby-grade CNC mill.