Running servos in position mode instead of analog

More
22 Jan 2021 05:17 #196179 by Michael
I have been retrofitting an old Bridgeport VMC and currently have a 7i92, 7i77 and 7i84 powering DMM 1.3 and 1.8kw servos in analog mode. I have been hitting a wall with their setup and it mostly revolves around not being able to turn off drive features that allow Linuxcnc to effectively close the position loop. In speaking with tech support I believe I am out of options to address my latest issues (5ms lag in motor response and overcurrent at high speed). The language from the manual leads me to believe that the servos are designed more for operation in position mode. I also have not found any examples of these motor/drives being successfully setup in analog.

So the question is what should I use to achieve this? Looking at adding a 7i78 to run XYZ and bring the encoders back to the 7i77 which would also run the spindle and IO.

What becomes of the encoder feedback? I can set the driver to alarm at different out of position ranges but could I use the encoders to monitor the position loop and /or modify it if necessary? I have read some advice that you don't really need to do that but seems kind of pointless to not use them somehow.

Any advice on pros or cons of going to this setup? Seems to be the cheaper route versus buying a different servo brand.

Please Log in or Create an account to join the conversation.

More
22 Jan 2021 16:09 #196234 by Roiki
What servos are you running? The common dyn4 should have velocity mode on it. That's what you want to use. I believe that should be available in both step/dir and +-10V signal on that drive(by analog do you mean this?). The pid loop on the Linuxcnc outputs a velocity that the drive has to follow. From what I've read in the manual, setting encoder feedback on those drives should be no problem.

Please Log in or Create an account to join the conversation.

More
22 Jan 2021 16:21 #196237 by Michael
Already hashed this out in other threads. The DNY4 motor/drive has a 5ms delay in response to linuxcnc commands. This causes a huge spike in error as the motor rushes to catch up to the command. Happens at various acceleration settings even extremely low ones. In the end they fail to run at more than 2/3 speed without overcurrent alarm when accelerating.
Yes the manual states they can do it but the reality is they don't seem to be tunable to achieve it. The position control via +-10v is not accurate enough from what I see to be used on an axis. I have been back and forth with the manufacturer and there last and only suggestion to correct the issue didn't.
Another thing to note, I can't find a single post of someone who was successful in running the DMM in velocity mode with analog control. Lots of posts of people trying but most seemed to give up. Lots of info out there from people using centroid controls and DMM but all of those are in position mode with step/sir.

Please Log in or Create an account to join the conversation.

More
22 Jan 2021 16:52 - 22 Jan 2021 16:54 #196242 by PCW
An interesting question is whether the 5 ms delay is present in position mode as well.
Connecting the encoders could answer this. This delay might cause issues with spindle
synced motion like rigid tapping.

This brings up a general observation, your errors are not huge, in fact you may have
never noticed them when machining and similar errors may also exist in step/dir
position mode.

A similar situation happens when people add linear encoders and discover all
the backlash and stick-slip errors that they had never noticed before.

Does the overcurrent alarm go away if you reduce the acceleration?
(200 IPS/S is likely beyond what the motors can do with a high inertial load)
Last edit: 22 Jan 2021 16:54 by PCW.

Please Log in or Create an account to join the conversation.

More
22 Jan 2021 17:10 - 22 Jan 2021 17:21 #196244 by Michael

PCW wrote: An interesting question is whether the 5 ms delay is present in position mode as well.
Connecting the encoders could answer this. This delay might cause issues with spindle
synced motion like rigid tapping.


This is my concern and why I would really like Linuxcnc to be in control of everything. That way it is all synced, kind of the purpose of a control. I can live with it being out a bit, as long as its known. I already have a machine that functions in position mode, and is fine for low tolerance work but my work has progressed to tighter tolerances mostly because of reduced cutter size. I could have these issues in position mode but the drive would monitor itself and probably wouldn't get a fault. I would be curious to know if they remained and maybe $100 for a new board and a couple days time would be worth it. The only other option I see is changing brands of servo motors. In the end I think its pretty telling not finding information about successful application of this drive and motor in analog mode.

As for the over current alarm, I have been tuning at 50 ips/s. The over current happens even when I go down to 20 ips/s. Looking at the F error at 400 ipm the line shoots straight up at the start of a jog and the drive faults. Another interesting note is that when I get the position loop tuned at 270 ipm within ~.002" and back the speed off to say 60 ipm then I get an overshoot at the start and under shoot at the end of motion that I didn't have at the higher speed (around .005"), also FF1 is off again.

All in all this post is just to explore some options if the current setup isnt working. The project has been eating up a bit more time then I expected and is cutting into getting other things done not to mention eating up what could have been a decent down payment on an off the shelf machine, but I am enjoying the process.
Last edit: 22 Jan 2021 17:21 by Michael.

Please Log in or Create an account to join the conversation.

More
22 Jan 2021 18:45 #196257 by Michael
Another option would be to try tuning these in torque mode. Any thoughts on that as well?

Please Log in or Create an account to join the conversation.

More
22 Jan 2021 20:51 #196272 by tommylight
Torque mode is harder than velocity to tune and in some cases requires two separate PID loops.
Never used DMM stuff, but from what i read online they are very good, so:
-are the motors mounted on the machine and how are they coupled?
-after mounting them in place, make sure they have at least space for +-5 revolutions forward and backward, meaning position them manually at the middle of the table
-use the DMM software to auto tune the drives, during this the motors will start to move slowly back and forth and the speed will gradually rise till it is tuned
-repeat that procedure for each axis
-set the drives to velocity mode, check if there are timings that affect the input lag, not sure how they are named in the DMM software
-only after all the above, set the drives to velocity mode and proceed to tuning in LinuxCNC.
Same applies for position mode, only tuning in LinuxCNC should be much easier in the position mode.

Please Log in or Create an account to join the conversation.

More
22 Jan 2021 21:19 #196282 by Michael
The servos are mounted to the machine and are coupled via htd8 20mm belts. Originally 40 to 40 but reduced that to 40 to 24(screw to motor) as I didn't need the extra speed. Motors are 1.3kw x/y and 1.8kw z which brings all the torque numbers to about double what the originals were. Also keeping the speeds to what the original machine was designed to so I am not trying to run a ballscrew or linear rail faster than it was designed.

Everything else you mentioned was done. Which basically means of the 5 parameters the drive has, 3 are turned off completely (torque filter, acceleration filter and integration gain). The remaining main gain and speed gain were adjusted with auto tune and then reverified with help from DMM tech support that they are correct. When asked about the 5ms delay DMM said it was because of the acceleration filter and advised to turn it off, which it already was. These settings also cause the drives to fault accelerating at 400ipm which is about 2/3 of rated speed.
One last thing I will be trying is running the motors under rs232 up to max speed, which I am sure they can do on there own since the drive will be using all of its filtering to create smooth motion.

Please Log in or Create an account to join the conversation.

More
18 Feb 2021 03:35 - 18 Feb 2021 03:36 #199216 by Dbsharp
Hey this is a bit alarming to read, I have had a projects in the works with dmm servos and Mesa analog control.... not running yet! Wondering if I should consider trying to sell these and buy another brand? What is known to work well in analog mode that doesn't break the bank?
Last edit: 18 Feb 2021 03:36 by Dbsharp.

Please Log in or Create an account to join the conversation.

More
18 Feb 2021 06:09 #199219 by Michael
From a personal standpoint I would like to see if yours have similar issues in speed mode as mine. I can walk you through the parameters and setup in the drives, got really good at that. The DMM manual has some typos and inconsistency, there are a couple settings that are easy to miss.

The bulk of problems revolved around the delayed response of the encoder feedback. I set them up in position mode open loop the other day and they work great and are accurate. Running a 6000lb vmc at 500ipm and an acceleration setting of 45ips/s. By Friday I should have them in position mode and closed loop with linuxcnc. See if the encoders still have delay.

As far as options I settled on yaskawa if I end up changing out from DMM. About $5000.... I have also heard good things about delta as well. I have used a Chinese servo as a spindle and that was no fun with little support.

In my research across many forums I didn't find any posts where someone had success with DMM in speed mode. I found a couple in torque mode but at least on I read had an encoder delay issue. Hope this helps.

Please Log in or Create an account to join the conversation.

Moderators: PCWjmelson
Time to create page: 0.118 seconds
Powered by Kunena Forum