Testing and small batch production
This month has involved fine-tuning and testing production versions of our light-duty and heavy-duty antibacklash worm blocks and ramping component stocks. In parallel we have been documenting the installation and ‘breaking-in’ process.
It appears that the level of backlash control we see is comparable to what we can obtain using our older antibacklash gearing solution, but this new one is end-user installable. 200-300 millisecs (3-5 arcsecs) is typical.
We also strongly suspect that some of this residual backlash is not really backlash at all, but is in fact stepper motor positioning error due to the low detent torque that is a consequence of microstepping. We hope to start to explore this later this year, in an AC servo-driven EQ8 prototype with 16 bit motor encoders.
I’ll comment about worm periodic error.
We source our worms from a specialist gear manufacturing company with a strong capability in thread-grinding and heat treatment processes. As we have yet to find a commercial supplier of worm gearing who grind GB1089-2018 Class 6 or lower worms and wheels at a reasonable cost, we buy good quality ones and refine them in-house via corrective hobbing and diamond lapping to bring them within spec. We monitor wheel and worm run-out error. In the case of worm gears, this amounts to controlling gearset transmission error. In the lexicon of the astronomical equipment specification, this term translates to periodic error. For precision telescope tracking applications, the actual Pk-Pk value of periodic error is far less important than achieving a smooth periodic error curve. So we focus on doing the latter. Gearsets under test are fitted to a ‘mule’ EQ8 in our workshop used for testing. We use a laser displacement sensor and tangent arm with datalogging at 100Hz to follow tracking motion over several worm cycles. This is processed into plots of following error vs time, i.e., the difference between commanded sidereal tracking motion and observed motion. The resolution of our setup is about 0.15 arcsecs, similar to that of a 23 bit rotary encoder. Known rotary 23 bit and greater encoders can be very difficult and time consuming to install to and remove from a telescope mount. Our test equipment, by contrast, can be installed or de-installed in about 10 minutes.
The high bandwidth and sensitivity of this type of non-contact motion logging means we can see very dynamic phenomena such as stick-slip, microvibration, the effects of viscoelastic and structural damping, motor microstepping errors, resonance and more, as the screenshot above indicates. It is possible to see directly the effects of lapping gearsets in the before vs after plots in the figure above.
One interesting discovery we made to our cost was the profound effect that timing pulley tolerances can have on periodic error of timing belt-driven worm gears. This goes far beyond simple defects such as run-out error. Tooth spacing error, and errors of tooth form can confound attempts to assemble a a telescope drive with low transmission error. Supply chain quality management and an in-house means of inspection proved very important to us. Despite this, periodic error of a worm block assembly can sometimes be improved on the test bench using phase shifting of timing pulleys relative to worm gears. All our timing pulleys allow for phase-shifting for this reason.
