Vertical Axis PID Tuning questions

The Wotan puts me at a disadvantage, but it should be close to a DeVlieg. The 1388's were rock solid and almost no trouble. With the mechanical balance ok and the drive balance zeroed, you should see near zero analog reference at axis stop. Make sure all drive train components are tight. Big machines like to hunt around zero position with analog drives and encoders at the far end of the screw from the motor or with scales. Screw windup or mechanical loss will drive you crazy on setup. I'm sorry I can't find my old 1388 manuals.

scotth wrote:

The Wotan puts me at a disadvantage, but it should be close to a DeVlieg. The 1388's were rock solid and almost no trouble. With the mechanical balance ok and the drive balance zeroed, you should see near zero analog reference at axis stop. Make sure all drive train components are tight. Big machines like to hunt around zero position with analog drives and encoders at the far end of the screw from the motor or with scales. Screw windup or mechanical loss will drive you crazy on setup. I'm sorry I can't find my old 1388 manuals.

Thanks for the insite. I do indeed see zero commanded reference with the head at rest. Verified with my o-scope. My X axis doesn't seem to hunt to bad for zero, I can't complain. 

For tuning in the drives themselves, did you just use a variable bench power supply for reference voltage and then switch it on and off with a switch to get the drive itself tuned in? I have a 1388 PDF from Allen Bradley and it shows how to tune the drive using an oscilloscope on the tach reference input. Maybe I need to tune that first, then go back to LinuxCNC. 

I prefer a battery box, a switched output of a 9 volt run through a trim pot to change the voltage.

By the way, what do you have for feedback. The DeVliegs normally have Farrand inductosyn scales.

scotth wrote:

By the way, what do you have for feedback. The DeVliegs normally have Farrand inductosyn scales.

This has Heidenhain Glass scales with EXE boxes. 1 micron resolution. This is on XYZ. I have Heidenhain rotary encoders on the W (quill) and B (rotary) axis and on the spindle.

I had a few hours today to work on this machine. I decided to hook my scope onto the X axis drive and tune that using the AB 1388 docs. I got it tuned so that there is less than .001 following error when running the table at 200 inches/min (max speed) with a spike at the beginning and small overshoot at the end (viewed on halscope) I had my physical o-scope hooked to the tacho and velocity command pins and they track very closely.

Then I moved onto the Y axis. This one is more tricky..I adjusted the offset potentiometer on the drive itself so that at idle, the voltage output of the mesa 7i77 is very close to zero, about 1-4 mV. Following error going UP is about .005" and stable with no spikes or overshoots. Going down, the commanded velocity is the same, but voltage spikes and sends the drive rocketing downward, followed immediately by a following error. I'm going to try the AA battery tomorrow which should be a 30 in/min jog on this system, and see how it behaves.

If you need to look at the tach feedback, I have used a 12 volt car battery to drive the axis motor. Then you can watch the tach output with your scope. Jumper cables are best so you can disconnect quickly.

Well, I believe I've narrowed down what is wrong with this axis. It's the encoder. I found that if I jog the axis up and down, I get a continuously higher and higher reading... If I jog up .01, it moves about .023. If I jog down .01 it travels about .1

This is a heidenhain 603 glass scale with EXE 602 box. I've swapped boxes with another axis and it made no difference, so I'm guessing it's the scale.

That said, has anyone had any experience with using the glass scales off of ebay that are for DROs? Most are 1um scales like this heidenhain and supposedly TTL output without a translator box..

I found the problem. I swapped the wiring plugs for the wires that go from the EXE box back to the mesa card and the problem followed the harness. I'm guessing there's a loose wire in a terminal strip somewhere... Now the axis works well (or, as well as an untuned axis will work.)