This is from yesterday with the DI and the better mandrel.
Two things Michael.
1. That DTI should be mounted such that the DTI needle drags lengthwise on the mandrel, not sideways. I'm not saying that sideways is wrong, but it does load the DTI in a way that bends the needle and stresses the mechanism. I only load them the way you did when I have no choice.
2. The way that you describe these tests suggests that the mandrels were made previously in another operation and then remounted for evaluation. That works if everything is perfect. But it rarely is. Usually, the runout in the various parts stack up and conspire to bite you. Looking at your tests, one might conclude that one mandrel is better than the other. But how do you know that? Perhaps it is only mounted in a way that reduces the stack up. In effect, various runouts cancel. Have you tried re-indexing the mandrels in the collet? Have you tested the mandrel runout on a surface plate? If you have not evaluated this then how do you know where the runout is coming from? It might not be a bad mandrel, it might be a bad setup. Or it might be a bad setup when the mandrel was made.
There are many ways to reduce runout. But the most commonly used approach is a physical offset like that provided by 4 jaw chucks or SetTru chucks. In both cases, you might get alignment at one length, but not at another.
Most of my setups take very high advantage of indexing. Record or mark the position at the time the part was made and then reposition it so it is positioned that same way in use. But some parts need to be realigned each time they are used because they are centered only at one point in their axis when they were made. When this happens, they need to be
axially aligned each time they are used. Because this is so important in my own smithing hobby interests, I actually made several specialized chucks to deal with it. You might search on "align" in the members only subforum to get a sense of this and why it's important. Your mandrel is long enough that this could well be an issue for you.
