This Centroid setup is wrecking my machine!

[eng199]

Did you adjust the smoothing profiles or smoothing parameters 221-231?
Smoothing step P222 is larger than recommended for a mill.
Perhaps the combination of smoothing and machine parameters has resulted in the profile flat spots.

[Houseman303]

i have adjusted the profiles. I have 0.2mm corner rounding active and it works very well. Since I always provide corners of components with a chamfer of at least 0.5mm, the corner rounding cannot be seen. I always used this setting on a 4 axis milling center with Heidenhain control.

[ashesman]

I havent mucked around with the smoothing setup for a while. Just turned it off for the remainder of my testing.

Waiting for some official feedback from Centroid. As far as I am concerned, nothing should be able to create flat spots in the pulse train!

I have been short of time lately to do testing, hope to have another go tonight.

[ashesman]

OK, so I spent another night on this. Unfortunately I am no better off. The machine still thumps and bangs running a program.

I tried all sorts, the only thing that really helps is damping the hell out of the command signal in the servo drive. Using the low pass filter on the command signal in the delta drive. But then I introduce too much delay.

I got to the point where I softened all the gains, reduced the servo control bandwidth (from 100Hz to 25 Hz), set accelerations back to 0.5 and turned smoothing onto standard 'mill' mode. But still thumps and bangs during running g code.

I think the underlying problem is the servos are very fast and strong and do what they are told. So react quite aggressively. I can see the knocking is always associated with rapid fluctuation in motor RPM. Particularly when asked to accelerate then decelerate without reaching constant RPM.

With MDI step commands, they are always smooth and quiet at high feed rates, perfect I would say. Small steps are noisy due the sharp acceleration and sudden deceleration of the motor. If I wind the control bandwidth back and slow accelerations, the MDI commands get better. But still doesn't help when running G code.

I can see that when the g code runs, it seems that even though the axis appears to be on average meaning to travel at a constant velocity, it accelerates/decelerates at each g code. The motor RPM rockets up and down chasing the acceleration and deceleration in the position command. Unfortunately I don't have the exact gcode lines that match the picture. I need to do more work to isolate the offending lines.

I am starting to run out of ideas and could do with some help!

My gut feeling is that the controller is ramping the acceleration up too fast. If we could soften the acceleration ramp up rate, then the initial hit of a position change would be reduced but could still reach reasonable acceleration rates. But I don't see a way to tune the 's curve'?

[eng199]

I can see that when the g code runs, it seems that even though the axis appears to be on average meaning to travel at a constant velocity, it accelerates/decelerates at each g code. The motor RPM rockets up and down chasing the acceleration and deceleration in the position command. Unfortunately I don't have the exact gcode lines that match the picture. I need to do more work to isolate the offending lines.

If you provide the G code snippet, matching report, and video clip of the problem, we can try to reproduce the problem and work on it.

Your previous video is presumably rapid (G0) moves between holes. With the Centroid control, smoothing (AD2) is not active on the rapid moves. Rapid moves will use S curve profile. The rest of the tests in this thread seem to be about feed moves (G1), so I'm assuming that is the main problem area.

My gut feeling is that the controller is ramping the acceleration up too fast. If we could soften the acceleration ramp up rate, then the initial hit of a position change would be reduced but could still reach reasonable acceleration rates. But I don't see a way to tune the 's curve'?

The acceleration rate is determined by the accel time, which you mentioned you do not want to increase. There are no provisions to customize stages of the S curve profile. S curve adjustments would be relevant in rapid moves, or with smoothing off.
By default, smoothing (AD2) does not use S curve moves for feed. You could try adjusting parameter 228 from 0 to 1. This will enable S curve for AD2 feed moves. It is possible to adjust the degree of S curve applied with this parameter, which is the closest thing to what you are asking for.

[ashesman]

Thanks for the feedback. I isolated the g code last night. I wont get time to test for a few days now but will try make a video then. And test parameter 228.

[ashesman]

Ok, so another three hours of playing around tonight. I have got the machine a bit better now. A combination of motor drive tuning and smoothing settings. The annoying knock on slow speed moves is still there. I am still not really happy with it though.

I ran a few test programs and is always the same. If moving at say 500-3000 mm/min it knocks hard at the start and end of movement, particularly stop/start moves like moving between cutting areas. But if you go faster it doesn't knock as bad. Slowing controller acceleration right down only helps a bit. The sound is most likely caused by a combination of a tiny bit of backlash and mechanical "spring". But annoyingly it just cant be tuned out! I am convinced that being able to soften the acceleration increase rate (S curve) would fix this, especially from a dead start. I tried P228 set to 0 and 1 but couldn't notice any difference.

The following video shows the test file I am running and with smoothing on and off. The knocking sounds worse due to the machine doors vibrating,



Upon studying the gocde and the X axis machine motion together, I noticed a few things around where it knocks. The red circle shows that the generated tool path is not very nice, aggravating the situation. The blue circle shows a simple transition between two straight line sections, almost colinear. But there is a significant motor speed fluctuation between these lines. In this case the required velocity change in X is almost zero but the motor speed change is large (from 180RPM to 60RPM, then back to 180RM). Note this is with smoothing off.

Why does it change the speed so much between two straight lines when the ideal change in velocity is minimal?

It seems that getting motion as well tuned as possible without smoothing means minimal smoothing is required resulting in minimal loss in precision. Needless to say, smoothing does improve these excessive acceleration spikes. Many guys on this forum say they have never turned smoothing on. But I need it on heavy smoothing! The picture below shows the same motion with smoothing on.

But with or without smoothing on motion is still harsh in the MDI when doing simple start stop straight line moves, and exactly the same in a program where smoothing is not applied, like lift the tool, move to next hole, lower the tool. Smoothing works well on the continuous cutting parts, but not the step movements. In fact it seems to make it worse as per my previous post.

Thoughts nd suggestions?

[Houseman303]

I think the servos have too sharp an acceleration, both positive and negative. Write a program by hand, a rectangle with a radius at the corners. A smaller radius could still bang, but with a larger radius it should run smoothly with the G61.

your gcode file is created with a cam and consists of small line segments. Something like this can only be edited smoothly with G64. see circle programmed with G2/3 or circle made of line segments. G61 works with G2/3, with line segments only G64 works smoothly.

[tblough]

Part of the problem with your blue circle, is that Y is changing direction, so lash compensation is added to the move. That could account for the rapid accel/decel you are seeing at that point as the control tries to quickly remove the lash.

[ashesman]

I think the servos have too sharp an acceleration, both positive and negative. Write a program by hand, a rectangle with a radius at the corners. A smaller radius could still bang, but with a larger radius it should run smoothly with the G61.

I agree regarding servo acceleration. Unfortunately the Delga ASDA2 drives dont have an acceleration limit when in PT mode. They have a MPA and LP filter. Using the filter solves the problem but adds too much delay, resulting in the controller getting position errors. The LP filter knocks the sharp corner off the acceleration and really settles things down. This filter really needs to be in the control not the drive otherwise the actually motor position is not necessarily where the controller thinks it is.