Machsupport Forum
Third party software and hardware support forums. => CS-Lab => Topic started by: jofriedl on July 22, 2016, 12:31:09 AM
-
Hi,
I have retrofit a 1979 matsuura mc500v with a cs labs csmio/ip-a ethernet analog motion controller. The machine was originally capable of accelerating at 40in/secsec and 400ipm rapids.
I currently have the acceleration and rapids set at 80% of their normal values.
I have look ahead set to 200 lines, but I have tried all combinations including 1 line and 1000 lines.
I have CV turned to on. All other settings related to CV are off but I have played with the settings ad nauseum.
I have CV feedrate turned off, although I have tried to run programs with it on.
The problem:
When running a toolpath of any complexity, The machine tends to jerk unpredictably at moments peppered throughout the program. It does not cut in a jerky movement through the whole program, just every few seconds and sometimes throughout entire sections of the the toolpaths. These jerks will always happen in the same places in the toolpaths. These jerks will trigger following errors if I do not have my max following error set somewhere around .05" These jerks do not happen when I'm running simple toolpaths, but only when I'm running complex toolpaths with lots of small arcs and linear movements.
My attempts to fix the problem involve:
What works is to run the program at much lower feeds and rapids. This results in a 40 to 50% longer cycle time. Since this is a production machine in high volume production shop, this is not an acceptable solution. I have also set the acceleration rate at half the value that the machine was designed to run at, but this just makes the jerks less prominent. This also results in a rougher finish when moving in arcs and contours. This is also unacceptable. I have also tuned the servos over and over, trying many different combinations in an effort to achieve the perfect tune. The tighter the servos are tuned, the more pronounced the jerks are. I have also uninstalled and reinstalled a different version of mach3. I am currently running V R3.043.062. I have also installed the lastest version of the firmware for the cs labs controller.
My theories:
I suspect that something is driving the servo drives far beyond their configured limit in brief pulses in order to move the machine along to the correct position. It is almost as if mach or the motion controller are ignoring the acceleration limits when moving in small increments along a contour.
Another possibility is that the look ahead function in mach is missing lines of code. Although this sees less likely since the program will be running along smoothly but then jerk in the same spot every time.
Response from CS labs:
I have recently made a video outlining the problem and posting it on my youtube channel. It can be found here.
https://www.youtube.com/watch?v=VEE7v3ewd3U
CS labs can't seem to figure out the problem.
-
Do you have backlash enabled?
Hood
-
Nope. No backlash enabled. What should my max following error be in terms of distance when I'm tuning the servos? When I use the autotune function I end up with a max error around 1000 pulses. Since I'm 84666.66 pulses per inch, thats about .011" according to my math. That seems high.
-
So yaskawa doesnt have a lot of data on these drives. they are cpcr mr-052 nb drives. I have long suspected that they were not a +-10v drive. I confirmed my suspicion today when I spoke to a company on the east coast here in the states that repairs them. They are actually a +-12V drive. I don't think this should be such a big issue though, right? Just lower acceleration rates and lower max velocities?
-
I just inadvertently discovered something.
After lowering my max acceleration by almost 30%, I saw an increase in the cycle time of my 13:25 program by 6 second. The program has lots of short lines in the parallel stepover portion of the program, so I would expect a more significant increase. This leads me to believe the machine can only accelerate at a little over half of what it could do when it came from the factory, regardless of what I have the acceleration set to in the mach configurations.
-
What should my max following error be in terms of distance
As little as possible I suppose is the answer :
My Chiron has rapids of 20m/min (787 IPM) and accel of (I think) 1200mm/s/s (47 I/S/S)
My encoders are 24,000 per rev, gearing 2:1 and ballscrews 10mm pitch so 4,800 steps per mm and I have a max error of about 200 if I recall correctly so that would equate to 0.04mm (0.0016")
I don't know anything about your drives but is there a way to configure the rpm for 12v? most drives I have seen you can do that.
Where you may have issues is that your drive thinks it should go at Xrpm for 10v but you have Mach set to Yrpm for 10v and thus there is a mismatch.
Another thing I found, autotune got reasonable results but doing some manual tuning got a lot better.
One setting that made a big difference was the KD
I think AutoTune left that at 0 but think I ended up with about 10,000 or 15,000.
Hood
-
Thanks, Hood.
There are three potentiometers on the front of the drives. Yaskawa doesn't have much info on them and the company that repairs them said they leave them alone in all circumstances. I'll attach a picture for clarification.
-
Here are the potentiometers I was referring to.
Thats an interesting theory about the mismatch of velocity.
-
Posted a reply in the other guys thread and never even realised it wasn't this one :D
Hood
-
hahaha I just did the same thing.
-
well I heard back from CS labs. They said the difference in voltage shouldnt be an issue. The also suggested I change some parameters to help with servo tuning. While the servos ended up with a better tune when I was done, I have exactly the same issue. Essentially, When running a program, I end up with a following error 10 TIMES the amount of error I get when jogging back and forth at full speed.
I'm wondering If I should take the plunge and set the machine to run in mach4. I need thing thing to be able to haul ass and cut molds and I just don't see it ever performing close to what it did before I installed mach3. I'm submitted a service ticket to machsupport weeks ago and never heard back from them. I even emailed Brian directly but haven't heard back. I suppose he's busy.
-
Just a wild thought and grasping at straws here but have you looked at your CAD/CAM generated toolpath - if it contains many short lines this could cause jerky motion (particularly with arcs and contours).
Tweakie.
(nothing to do with the problem but the diode, below the 3PR relay, appears to be connected incorrect polarity).
-
It definitely has lots of very short lines and arcs and contours. Shouldn't mach be able to handle that with look ahead and constant velocity?
-
Well yes it should but I recently changed the aspect ratio of a previously produced milled panel contour and regenerated the toolpath (in CAD/CAM) – this converted all my nice smooth arcs into many straight lines – the result was terribly rough motion from my mill.
Mach3 is good but it’s not perfect and I ended up redrawing the panel to the new dimensions, recreating the toolpath then everything was bliss.
Tweakie.
-
I don't know if you found a solution to this - but I'll tell you mine all the same!
I had the same issue. CS Support wasn't really much help, but I found a comment on line suggesting that a faster PC fixed it for them. I was running a new i3 with 8GB 2GHz - so it isn't slow by any means. My servos (Fanuc with Ultra3000) were perfectly tuned & lightning fast. Running in test mode from the Ultra3000, or even a pulse generator they moved smoothly.
Connected to the CSMIO however, they were lumpy / jerky as hell. Looking at the scope in the U3000, I could see the step pulses were coming in erratically.
I tried it with a i7 16Gb 4GHz - and now the axis are smooth.
I fail to understand why a motion controller needs a fast PC. The jerkyness appeared even on straight line moves G0 Z500 kind of thing where I imagine the PC would just send one coordinate to the controller which would deliver the pulses at the required frequency. However, it appears to send continuously - maybe so you can stop in the middle of a move? It seems even more daft when the old PC with Mach3 & a parallel port BoB could drive them just fine.
The upshot is - if your servos are jerky, use a faster PC.