[Todd Zuercher]

We have a couple of moving gantry routers that the servos need tuning (especially the Y axis and its slave A). The controls are OimB. The machines were built in 2005 and 2006. The drive-train for the y and a have been rebuilt a few times since new and the gear ratios are significantly different from what came from the factory, (hence why the tuning is needed). The machine tool builder has been dragging their feet getting it done, they say they are going to have their tech(s) trained to do it but they have not gotten to it yet (been getting this line for 2 years).

We got tired of waiting and called Fanuc direct and scheduled a tech to come tune em up. I was a bit surprised at his lack of success.

To start the tech was not able to interface his laptop with the controls in order to run Servo Guide. He only tried to use a PCMCIA either-net card (I don't think it was the special Fanuc one either, pictures of those I've seen on line looked different). Is this the only way to use Servo Guide on one of these controls (it has a PC front)?

After working on it all day (at nearly $200/hr), using some adapter boards and a scope, he tells me it is as good as he can get it, without doing a ball bar test, and that they don't have the equipment to do that.

All of the tuning he did was done jogging the axis back and forth. I don't think he did any programed cutting moves. Which surprises me since I thought different parameters are used for jogging and cutting. (I did not stand over him and watch the whole time so this might not have been the case)

It is a bit better, he got rid of most of the vibration, but it is still messing up the corners.

The little bit I've looked over the servo manuals, gave me the impression that the ball bar test probably wouldn't be necessary had he been able to use the Servo Guide software. And that it probably would have worked if it was installed on the front end PC, (or if he had the right card).

How hard is the Servo Guide software to get (searching on line got me nowhere).

[tc429]

I have zero experience with Fanucs servo tuning 'the right way', but have fumbled along setting them by ear and using a ballbar when accuracy is more important than cosmetics- that said, just wondered about your configuration, and what was changed:

is it using scale feedback and a high reduction on the servos? if so can picture a lot of windup at reversals...is the slave servo slaved to the master encoder, or external scale (just curious, never played with a slave setup either). if a windup/stickslip issue, I'd try running zero backlash comp, and maybe turn on 'n-pulse supression' to give a little deadband- just guessing being a router accuracy might not be as important as cosmetics, and a small flat at reversals looks way better than a step... we mill aluminum wheels, and cosmetics are very important compared to huge profile tolerances on cosmetic features... even a few thou flat on a quadrant of a radius is invisible compared to a few tenths step, and dont seem to vary with feedrate as much as backlash acceleration comps might... maybe try some test cuts at your typical feeds/radii and see what n-pulse supression does with from a few to maybe 100 counts of supression does... when playing with tuning for sticky or buzzy axes, the 'load inertia ratio' seems to be the easiest to tweak like a old gain pot, change I-P to P-I control bit often helps (almost all our machines run P-I)

ran into a few situations where reversal tuning was impossible to get rid of quadrant protrusions, we had a Fanuc guy in 20 yrs ago when we started buying the Zeroes with serial C encoders... he messed around a day, ran our ballbar, and he had a proprietary/prototype software that ran ballbar like, but used encoders for feedback thru a servo tuning board- looking at the two, he could see we were clipping the gain limits at the first .0001 reversal - although we couldnt even display it, we switched the hi-res bit on, did the times ten or divided by ten thing to our existing servo parameters, and it profiled that first .0001 in ten undisplayable steps and smoothed it right out... those machines are still running.

i'm still a huge fan of the old 6050 analogs...to this day, THE highest performance/tightest response Ive ever seen was on a 0-C with the analog axis interface running a set of 6050 amps...with a 0-C 486 CPU running the big gold, decades old, dedicated 7103 axis CPUs we could run the 517 parameter maxed out at 32767 without buzzing the servo, and near zero following error... I'm still not a fan of serial as too many things seem to be pretend...backlash acceleration compensation-3 types???, the only ac servos ive ever seen run away have had serial encoders (seems they can maintain commutation without position?), many other 'things' that simply didnt happen with analog... I know a old 0-A with its 80186 CPU could suffer accuracy issues at higher feeds (actually had to update some controls to get around a following error issue on some fast feed operations), but still, in my opinion the 0-Cs with the analog cards were tighter than even that factory tweaked/hocus-pocus, pretend ballbar *10 resolution whiz bang thing...even with drives/motors pulled from a '80s robot Ive still got two 0-C analogs squirrelled away intact waiting for the next time we need a robust/fast feed/tight response control to retrofit, till then we'll keep re-retrofitting our old analog 0-As/digital 0-Cs or cleaning up the faster 21s for a while longer...some of our zero A controls are being retrofitted for the 5th-6th time, those old things were like a F150 shop truck, they just keep on going... the C controls seem to have a lot more age related issues, but still have proven more durable (and repairable) than any of the newer higher performance controls - not a fan of any Fanuc control that used the PCR series honda plugs, we have issues almost daily with those, never EVER had a honda MR connector failure, even at near 30 years old on our last filthy old 6TB-2, they simply work... just touch a cable with PCR plugs on most of our machines at 5~10 yrs old it will shut down... plug/unplug a dozen times to scrape it off, or try cutting a flat eraser to scrub them off, theyre good for a year or two again, but still... miss the MR plugs

[Todd Zuercher]

It is a basic gantry router with a 5ft x10ft table rack and pinion on each side of the long direction Y-A axis. There is no external feed back, just the motor encoders. I am not exactly certain how the Y and A are slaved together.

The manufacturer updated the machine to larger pinion gears, and different timing belt pulley sizes. The net result was the flexible feed gear ratio went from about 1:2 to 3:4. We soldiered on for years battling a vibration in the Y-A axis. Manufacturer kept saying "Oh, it must be a mechanical issue." And when there was a mechanical issue It would get worse. I've lost count of how many times I've changed pinion bearings and pulley shafts, and their bearings (two belts for compound gear reduction).

Here is a drawing of the error we are seeing right now.
Attachment 249180

Any one with any suggestions where to start?

[tc429]

Id mess with the load/inertia parameter first - not saying its right, just what I'd mess with

try putting +/- 200 in there on both axes, see if it calms it down... you should never need negative comp, but with a loose machine it should soften the response. what is your distance per rev of the servo? rack and pinions tend to be pretty far, there is a exponential acc/dec for feed I hate to use, but putting say 50 milliseconds might help if youre moving fast- like > 1/4" per rev...careful with it if you do much circular interpolation, it can distort arcs... if expo helps, I'd bet the problem is too much mass in the y gantry having a hard time stopping on a square corner- perhaps try a smaller diameter cutter and interpolate the corner, the arc inherently gives a little acc/dec.

a quick test might show easiest tweaks without scrapping product- program a 200 ipm y move with a 1 second dwell, reverse/repeat, put indicator on the axis, see if its hunting as it stops/during the dwell...bet it is, and bet some expo will stop it, just keep adding time till it stops cleanly... just my 2 cents, doubt its worth that though

edit, still thinking- try that with two indicators, maybe gear lash is uneven from one side to the other- if one side is worse, try changing the mechanical lash by shimming or belt tension, see if it gets better or worse

good luck with it

[rambo77]

I don't come into the 'zone' much and am not sure how to do a private reply, but a friend saw your posts and suggested I contact you. I've tuned many routers and other machines during my time with GE Fanuc and can probably help you. Send me an email and I'll get back to you with my information and we can talk to see when our schedules will click. [email protected]