# Thread: Fanuc alpha servo tuning

1. ## Fanuc alpha servo tuning

I have a fanuc 0M-D on a hardinge vmc1000 and i'm sure that it could do better. In circular interpolation, even at low speed or with a great radius, error is abominable. If I cut different diameter with the same tool, I must put different offset for each diameter. Servo software version is 9046-09. I'm reading some fanuc manual about the subject, but all the options that could help me are only available on more recent software version. Any help??

2. I progress! I read about feed-forward function. It was turned off. I began to turn it on, following the procedure. Then, have to set the the feed foward coeficient FALPH according to these formula:

FALPH = α × 4096 ×8192/Position feedback pulses per revolution of the motor

But wath is α ?

After that it's easy, just to set the Velocity feed-forward coefficient (VFFLT)

VFFLT = (− PK2V) ×[(Load inertia +rotor inertia)/Rotor inertia] × 0.04 × 8000/(Position feedback pulses per revolution of the motor)

Where can I find the load inertia and the rotor inertia?

3. one more step. I found that load inertia ration (prm# 8x20) is 256*(load inertia +rotor inertia)/rotor inertia. But i don't know if the value entered in these prm is trustable. I have 64 for all 3 axes. Does it makes sense? I have the same motor on the 3 axes. So i assume it means that the head inertia is the same than the table inertia.

4. i begin to understand. Correct me if i'm wrong. α is Feed-forward coefficient (0 to 1). So if i want a coefficient of 0.7, according to the formula
FALPH = α × 4096 ×8192/Position feedback pulses per revolution of the motor
FALPH=0.7*4096*8192/12500=1879
To konw my Position feedback pulses per revolution of the motor, i check prm #8x24(Number of position detection feedback pulses (PPLS)) Is that o.k.?

5. Something is strange with the inertia ratio. If it is 256*(load inertia+rotor inertia)/rotor inertia i'm supposed to have a value grater than 256. It is impossible that load inertia + rotor inertia be smaller than just the rotor inertia. If anybody have the same machine, please can you give me your parameters back up.

6. if interpolation is bad, be sure exponential acc/dec for feed is zero- seen a lot of oems throw values in there and toolpath gets distorted really bad. I dont think load inertia will help with interpolation- i usually fudge the ldint values to calm a loose machine down, it can eliminate oscillation at rest if the screw preload is gone, etc...
be sure your gain (517 in a zero) is at default 3000. Weve had some bigger machines with lots of ldint in a compound axis due to excessive mass, it calms the motor, but never noticed any interpolation error changes with the ballbar except maybe quadrant protrusion at reversals.
Suggest programming a 45 degree move and make sure following error matches or is exactly multiplied or divided by a integer value...on the old analog stuff thats how i set tachs, if theyre off even 1% it messes with interpolation accuracy...radii become polygons as one axis will lead/pause- cant really see it, but shows on the ballbar

I know i'm not doing it right, but try to get by...

7. Yes i have a value in exponential acc/dec (40) i have to put 0 to use linear acc/dec but where and how i set linear acc/dec.

8. linear acc/dec is for rapid- expo is for feed...I always run expo for feed at 0. zero acc/dec for feed should work unless youre running crazy fast feedrates. linear for rapid usually 100msec is enough to keep the machine from shaking, too much linear can cause it to skip home, depending on length of decel dog.

9. Set exponential acc/dec to 0 help to reduce path error but it's not really smooth. There is a shock at each change of feed rate.

10. shock at block changes can be smoothed with increasing 'in position' window width too- but that also affects accuracy. try increasing to 20-30 counts of 'in position width' and leave zero expo, see if that helps or hurts.

we had a issue long ago where axis reversals and feedrate changes were clipping the gain- per a fanuc service guys suggestion, we went to 1/10 increment system and scaled parameters as needed...although the control wouldnt display .00001 increments, it still profiled the gain in ten steps on a .0001 move, smoothed it right out. really sucked as we had been looking at all kinds of mechanical stuff, and here it was some bull**** math thing in the servo software making our machine twitch on a 1/10 of a thou motion...