Hi,
I don't know whats changed but something has. The norm when setting up PID is to go for the fastest response with least following error.
Those conditions also put you right on the verge of instability. Any small change will often see low level oscillatory motion. When you set up
you tuned for absolute maximum performance now you need to de-tune for reliability.
If you were modelling this with Matlab you would tune for desired performance and then do a Monte Carlo analysis. The Monte Carlo analysis
varies all the parameters in a usually small but statistically random manner while observing one or more performance indicators. Requires a lot
of computing power. What is does do is show up elements in your system which with minor variance either manufacturing tolerance or wear
will cause a large deviation in desired performance. You can make tuning allowances or redesign that component or whatever...
The upshot is that it could be the aging of a capacitor or resistor inside the drive nut in the feedback loop. It may also be about the mechanical load,
either the weight has increased, you know that added dial gauge or extra thick vice jaws etc or a change in frictional forces due to wear or lack of
lubrication.
A while ago I bought an Allen Bradley servo and drive off Ebay for use as a spindle motor. I ended up buying the Ultraware software that is used
to program the drive. Must say I didn't appreciate having to pay for it but I needed it and bit the bullet. The advantage of the software is that
the EXACT model of my servo was alredy in the memory, including phase resistance and inductance, magnetic losses at various currents, rotational
inertia and more. There is more info in the software than I can find in the specs sheet and certainly way more than I could reasonably be expected
to measure or estimate. All that info meant I just set up how I wanted the servo to behave, max speeds, torques etc and that's it! All the PID stuff
was already done! Amongst the facilities of the software is an oscilloscope. I've used it to track my actual servo response and its bloody good, I could
make it worse by fiddling with the PID settings but doubt I could make it better.
Have to commend the idea that if you buy a servo buy the matching drive from the same manufacturer with that manufacturers software will result
in optimum performance with the least frustration.
Craig