PCB Milling Feed Rate issue

[jpbarad7]

Tom,

My mill is working very well now. It's not perfect, but for a 40+ year old machine with steppers and belt driven axes, operated by a man with limited knowledge and experience, I think it's as good as it can be.

Thank you again for your help through the process.

In answer to some of the questions/comments from your last email:

During my tuning sessions with ChatGPT, I provided it with as much information as possible about the machine, conroller, software, etc. and my 'requirements' pertaining to accuracy and most common feed rates used. When it told me 'tuning was complete for an axis', it meant that the changes it was recommending were no longer resulting in measurable benefit (as determined by its analysis of the raw data from the step response screen test runs, as well as my the feed back I was providing concerning motor sound and smoothness of motion).

During the tuning process, ChatGPT had me run the same sample PCB milling program and compare it to previous tests, to make sure we were moving in the right direction. The final tweaking involved trying to eliminate/reduce the 'points/peaks' that were occurring at approximately 10 and 4 o'clock on the 0.076" diameter circles (these can be seen in the photos at the beginning of this thread). ChatGPT said Mach3's circular interpolation was known to 'have issues' than other better/newer software packages. When I ran the same program with KmotionCNC, software known to not have the same circular interpolation issues as Mach3, and reported idential results, it replied that the 'points' where less likely do to a software issue, but were most likely mechanical. Turns out, the imperfections in the circle were located where the axes were reversing direction, making the belt drives or a small amount of slop probable. The step count of my axes (233500 per inch for X and Y) and the resolution of my encoders made it impossible for the system to correct the small errors - unless I reduce the feed rate to 1 inch per minute. There is now no noticible difference in appearance/result between 2 ipm and 20 ipm (which is the most obvious measure of improvement - refer to the photos again to see what I mean)

Concerning PID not being effective at reduced velocity and acceleration values, what I meant is, through experimentation I found that my machine (and all others, I suspect) require certain velocity and acceleration settings sufficient to initiate and sustain smooth motion. That is why, again I suspect, before PID tuning is undertaken, the max velocity and acceleration settings are determined (and then reduced to 80%). In my reading and conversations with others (including ChatGPT), when problems are encountered during the tuning process, 'reducing the acceleration and/or velocity' is often recommended. However, if these values are reduced below the point required to 'initiate and sustain smooth motion', many other, albiet different, tuning difficulties arise. When in the throws of a process as confusing and challenging as PID tuning, it's easy to lose sight of the forrest for all the trees. I wasted a lot of time before increasing velocity and acceleration to a level where PID corrections were possible. I do understand that once tuning is optimized, PID works well for all speeds and feeds.

Jim