I think @Julien uses blue tape to help get the belt position in the clip consistent.
However, I don’t think you need very much tension in the Shapeoko belts, certainly no more than the holding force of the steppers - 60N, or less, that’s breakaway torque, not actual usable during motion torque.
My motivation in trying to measure the tension quantitatively was to figure out what sort of forces we were actually achieving and running at so that we could;
- Set the tension to a known value
- Check if these values are “reasonable” compared to other machines, the Gates spec book, stepper motor specs etc.
- Be able to go back and test if it had changed due to belt wear, stretch or some other effect.
I only realised part way through as I did the math(s) that matching Y tensions was important for squaring & calibration.
It currently seems to me that the slipping of the belt clips may well be either smart design of happy accident and that slippage is there to avoid us cranking the belt tensions up to the point where we fatigue the stepper shafts and break stuff. I’m considering going back to mine and backing off the tension, when I’ve done a few more measurements.
I have found, since measuring the tension with my phone that I can get a couple of teeth worth of extension in the clip with finger pressure holding the clip down, if it jumps before the clip gets cammed down by the bolt I’m probably over-tightening it.
I agree with you that the belts are a pain, but they’re much cheaper than ballscrews (which need more motor torque and therefore bigger stepper drivers in the controller etc. etc. ), they’re at least the equal of a leadscrew for backlash, on the regular size Shapeoko probably better and their material is actually pretty good at vibration damping, this is all by design at Gates who designed the GT2 & GT3 belt profiles.
That’s not to say I’m not trying to work out ways to improve them but I think I’ll be focusing my dial gauge on the X rail to Z carriage V wheels first as many others have before me (thanks for posting the CAD Dan).