Weird Oscillation on Contouring

[the_canuck]

I'm fighting a weird oscillation on Contouring and circles. See pictures. Tool is traveling around the contour but I'm getting a very repeatable oscillation on every pass. I'm doing a finishing pass taking at most 0.02" with a 1/4" carbide ball mill. The pocket is approx. 6" x 12" x 4" and shaped like an elongated tear drop. X runs left to right on the picture, y up and down.

My machine is very rigid. Structural steel and Cast plate mainly. I'm running 1HP DMM-tech AC servos and using their older smaller servo drives. They are tuned as good as I can get them but I suspect it could be the root cause of the oscillation. The X-axis is dual driving 2 racks and pinions 3:1 belt drive reduction from CNC router parts. The Y-axis is 2:1 reduction. Spindle is 2.2KW water cooled Chinese.

I'm running Linux CNC in open loop and treating the servo and controller just like a stepper. I'm driving them with 2000 steps per revolution on each axis.

I've got the acceleration and max velocity set the same on both axis. I've tried different parameters but no change in cut quality.

Speed does affect the size and amplitude of the oscillation. Going slower is better. It's not noticeable by ear but the oscillation shows in the cut. Straight cuts are perfect. It's when I'm moving both axis at the same time I have issues. It does seem like both the x and y are contributing to the issue.

The loads on the x is much higher than the y and the x has more friction in the linear bearings. My gantry weighs approx 250 - 300 lbs. Just the y and x is probably 150 lbs. All axis are running Hiwin square linear rails and bearings. X I'm using 1" size and the y is using smaller size.

I'm using Rhino and MADcam for the rest of the software. The tool paths in MADcam are perfect and don't show and oscillation.

I've been fighting the circle issue for a while now and can't seem to find a final solution.

My suspicion is I have a velocity oscillation created by bad tuning job on the servos but I'm getting no other indication that my tuning is bad. I get very repeatable positioning, no audible noise. I can drive them fast and hard. My acceleration and deceleration are short but not unreasonable. The dual drive isn't making tuning any less complicated on the x either so I may have them fighting each other as another possibility. Another issue might be that my x and y are so different in servo tuning due to the different loads that I'm getting some sort of beat frequency in acceleration/velocity between the 2. A third option might be that I'm getting some sort of interaction between the rack and pinion and the servo tuning. The period of the oscillation is approx. the same as my rack tpi. I thought it would help by going from 2:1 to 3:1 on the belt reduction on the x. It did help my servo tuning but I get the same results before and after the hardware change.

Gantry deflection can't be it because the oscillation shows up even when just engraving with a tiny bit and small cut loads all the way up to hogging AL with a 1/2" bit at 150 ipm.

I really need to solve it very soon. I have a customer waiting on these molds and I would like to get a good surface finish right off the machine.

Any suggestions on what to try would be greatly appreciated! Going slower is really not an option as these pieces takes hours even at 150ipm.

Similar Threads:

• THC bounce/oscillation
• Oscillation
• Need Help!- X-axis oscillation VMC 4020
• servo oscillation
• Oscillation During Axis Acceleration

[keebler303]

Do your servo drives have a "scope" feature? Some I have used will allow you to capture the error vs. time for a short period. If you can do this, it might be helpful as it will tell you if the oscillation is due to the servo tuning. Normally a servo loop oscillation will be shaped like a sine wave. It seems like the marks you are seeing loop more like scallops. Have you verified that the gcode does not have these lines in it?

Do the lines repeat in the same places on the table? IE if you move the part to a different place on the table are the oscillations in the same exact place on the part? What if you rotate the part by 45 or 90 degrees?

[the_canuck]

No scope feature unfortunately. It would be very helpful for tuning. I'm familiar with the LinuxCNC scope for servo tuning but since I'm running open loop as far as LinuxCNC is concerned it's not available. With these servo drives I'm kind of shooting blind.

Here's the servo drive I'm using: DMM | DYN2 Servo Drive

The extra lines are not table location dependent. I get them both aligned with x and y axis when cutting arcs.

They are definitely not in the tool paths. They look nice and smooth on the screen and in the simulation.

Andrew

[Dragonfly]

What is your drive screw pitch? Check eventually whether this pattern occurs at the same step.

[keebler303]

Does the dmmdrv software have this functionality?
Do you have the proper IO to connect one of the encoders to linuxcnc to get the scope feature?

How about using the "On Position Range" output to see when the error exceeds a set amount? You can play with different values for that parameter and monitor the output with a meter or LED to see if the error gets large when you get to a certain spot.

It is possible you are exciting a natural frequency in your machine at certain speeds. Do you ever notice the issue on angled cuts or only on arcs/curves? If it is an interaction between the two axes you should be able to recreate it with an angled cut that sustains the oscillation longer which will help you diagnose what is going on.

[the_canuck]

good idea Keebler on the diagonal test. I'll mess around tonight and see if I can isolate the issue to one or the other axis. If it is an harmonic issue with the machine that will suck to trace down and fix.

I can't do closed loop with linuxcnc and these drives (or at least I haven't figured it out yet). Also I'm not running incremental encoders but absolute rotary encoders that came on the DMM servos made for these drives. They are designed for step/direction inputs driven just like a stepper. I'm driving them off a standard parallel port. I haven't upgraded to a Mesa card yet so am limited in I/O connections. I've worked on a Bridgeport knee mill conversion that was closed loop servos with linuxcnc and Mesa card so am familiar with that setup and tuning process.

I'm running rack and pinion setup the same as cnc router parts sells. Rack tooth count per inch is visually pretty close to the oscillation I'm seeing in the cut but I haven't actually measured to confirm.

It's not obvious from the pictures but the oscillation in the cut is sinusoidal for the most part.

Andrew

[keebler303]

If the oscillation matches the pitch of your rack, that is an obvious place to look.

Maybe your pinion is eccentric and causing some cyclic error?

If this is the case, I would move your part zero by 1/2 the rack pitch and see if the lines move or if they stay in the same place. If they remain aligned to the file rather than the machine I have to think it is a servo/tuning issue.

Did you get the bit about "On Position Range"? I think that could be quite helpful as it will allow you to know how big the servo error is.

Another thing you could try is to put an indicator in the machine and visualize the movement step by step. If there is something weird going on with your rack or pinion, you may see a stutter or a jump in the motion which could obviously look like oscillations. A 1" stroke indicator could be mounted near the end of the toolpath and you could see the axis movement while you were actually cutting a part.

What if you take an extra fine point sharpie or something and stick it in the spindle? Can you see oscillations in a traced path on a piece of paper? What if you draw a bunch of these paths offset a 1/4 or so from each other and steadily increase the feedrate for each? Does the oscillation get bigger and bigger?

[the_canuck]

I've cut a left and right set of these things and the lines are there on both in the same locations with the parts essentially mirrored on the table. I've given myself a couple hours tonight to get this figured out. Otherwise I'll deal with the quality I have now as I have a customer on a very tight deadline and calls everyday wanting progress updates. The final CNC pass will start on the first piece tonight. Machine can run without supervision fortunately.

Andew

[the_canuck]

I hooked up a fine tipped pen to the spindle and here's the results. There's a definite oscillation when traveling in an arc. It's worse traveling at 45.

[mactec54]

the_canuck

What do you have for your settings in the top row & in the 3rd row of the Dmm software, have you set the Drive ID numbers 0 for X axes 1 For Y axes & 2 for Z axes

[mactec54]

the_canuck

It looks like backlash is your Y axes a Rack as well or a Ball screw, looking at the photos you should be able to feel that in the machine when it is moving

[Devastator]

I suspect software. Does linuxcnc have a "look ahead" function? You might try adjusting that to look ahead more lines.

[the_canuck]

Here's my settings from the servo drives:

X1
Main Gain: 80
Speed Gain: 20
Integral Gain:15
Drive ID: 0
Gear #: 500

X2
Main Gain: 80
Speed Gain: 20
Integral Gain:15
Drive ID: 0 (was set to 3)
Gear #: 500

Y
Main Gain: 65
Speed Gain: 50
Integral Gain: 15
Drive ID: 1
Gear #: 875

I think the rest is set to defaults.

[the_canuck]

LinuxCNC only reads ahead one line. I've tried with and without G64 P0.0005 and I get the same results.

[UA_Iron]

the_canuck

It looks like backlash is your Y axes a Rack as well or a Ball screw, looking at the photos you should be able to feel that in the machine when it is moving

agreed, looks like that or actual pinion wear. It appears to be in the arc where the axes might start transitioning.

Might check any tensioners you have and ensure they're tight. Any vibration will make things loose over time.


On side note: I had Mitsubishi come in and give a demo of their servos, the melservo J4 series. The technology built into their servos is mindblowing, absolute encoders with 22bit resolution (4,194,304 pulses per revolution), auto-tuning that will match the servos to the machine mechanics - and some pretty intense dampening technologies... Even the braking regenerates power back into the system. I'm a stepper guy, and will be going with steppers for my CNC build, but now I am well aware of how much more advanced servo technology is over stepper technology.

[mactec54]

Here's my settings from the servo drives:

X1
Main Gain: 80
Speed Gain: 20
Integral Gain:15
Drive ID: 0
Gear #: 500

X2
Main Gain: 80
Speed Gain: 20
Integral Gain:15
Drive ID: 0 (was set to 3)
Gear #: 500

Y
Main Gain: 65
Speed Gain: 50
Integral Gain: 15
Drive ID: 1
Gear #: 875

I think the rest is set to defaults.

The Y axes Gear should be 500 the same as the other drives speed Gain seems to high should be more like the X axes setting & all the Integral Gains are to high yours should be in the 1 to 8 area 1 Being the best

Buy setting the Gear to 500 will give you the smoothest running with step/Dir , on position wants to be set to 127 on all the Drives By changing the Gear setting you may have to reset the step per for that axes

[the_canuck]

I'm leaning more and more to a mechanical issue with the belt reduction on the rack and pinion setup. I tried tuning numbers all over the place and I get the same oscillation. I setup and did 45 degree lines changing one parameter at a time on the y-axis to see if anything changed. No matter what I did the oscillation is showing up.

So I then did a once over on the machine looking for anything loose etc. Didn't really find anything. I tried messing with the pinion tension into the rack and that didn't help from too loose to really tight. I then tried messing with the belt tension on the belt reduction thinking maybe the belt tension was too low causing backlash. If anything higher belt tension made things worse. Which lead me to maybe the root cause. One issue I did see is that my bored out pulleys on the servo side of the belt reduction aren't perfectly centered on the servo shafts. I can feel the tension/friction change while pushing the gantry with the power off. Not good and is most likely a main contributor to the issue. It's also adding periodic error of my gear ratios as well as the belt tension and friction changing every servo revolution. I found some properly sized pulleys on ebay and ordered them. I'll swap them out and see if things get better. For now I'll use it as is.

BTW I have a lot more friction (bigger linear bearings, higher preload, more weight, etc.) on my x-axis then my y-axis as well I am dual driving so I'm expecting pretty different tuning numbers. I had the gear set to match the differences in the gear reductions between the x and y as well as dealing with a limit to the number of steps per second I can get out of a parallel port in LinuxCNC and still keep my rapids as high as possible and step size small enough to meet my accuracy needs.

I'll revisit the tuning again after getting the mechanic issue sorted out. As for the parts, a few swipes of some sand paper will fix the lines. They have to be sanded anyway. It will just add a bit more finish work.

[mactec54]

the_canuck

If you have enough speed with your X axes with a 3:1 gearing, then your Y axes at 2:1 would have not problems with speed, it should be set the same for the gearing 500 in all your drives this is the best setting for someone using only Parallel Port, if you were to run something like a smooth stepper then you can play & change the Gear setting, this has nothing to do with how you gear your machine, it will only affect the step pulses needed for that axes set at 500= 2000 steps per rev the bigger the Gear number number the more steps you need for that axes, so it will help to smooth that axes & use less steps to do it

[Khalid]

just check looseness... place a dial indicator and press one side with hand see any movement?

[Beachcues]

I know when we used mastercam to do our 3d passes it would generate the same issue. We now use camworks which has a bunch of pass options to combat this very thing. I don't think it is a machine issue just a programming software issue. Have you tried to decrease your step over amount?


smaller step over