This is directed mostly at people who have done their own retrofit project, anyone with experience on Rutex drives your input is especially appreciated.
I'm working on my Lagun FTV-3 retrofit, I'm at the point now where I'm tuning drives and (hopefully) getting closer to making chips.
Since a good lot of the parts from when this machine was originally a Hasbach/Bandit driven machine were gone before I took delivery, I've done a lot of research and built my own power supply, I went with an Antek 1500VA transformer that I dewound to output ~57 VAC (= 75VDC rectified), filtered thru a 112000 uF cap bank, and lower power secondaries outputting ~15 VDC rectified each with a 56000 uF cap.
Because this was originally a Tach/Resolver feedback machine and the servo drives were all gone, I fitted 500 line Avago HEDS 5500 series encoders to the X servo so I could use much newer (and cheaper!) servo drives. The Y servo was gone when i got the machine but the replacement servo has also been fitted with a 500 resolution encoder. I then found a set of 3 Rutex R2010 drives on ebay for dirt cheap, I also have a Gecko G320 that I was originally planning on using to drive my servos before I found the R2010's that is lying on the bench waiting for a purpose (possibly a 4th axis? Someday...). Right now the X and Y servos are hooked up electrically to their drives but disconnected mechanically from the mill, I've done some preliminary messing about with PID tuning and now I'm running into some questions I don't know how to answer
- Rutex posted a document here: http://www.rutex.com/files/R2000setup.pdf detailing how to set things up, so I've made the SPI tuning cable, verified correct motor/encoder hookup, and done some preliminary PID tuning with the R2xtuneVB6 program. As an aside, I wish Rutex would have specified that the VB6 program has a lot more features and better GUI than the other R2xtune program.
I get that I want to turn up P to get a fast response time, when I get excessive hunting add more D to get more stability. I understand that when I have an overdamped step response graph with no overshoot, I need to increase my P to get more acceleration out of my servo, and when I induce too much I get oscillation either from disturbing the position of the motor or commanding steps, so I add more D to keep hunting/ringing away.
If my servo isn't ringing at a standstill, after a step command or when I disturb the shaft of the motor, is there such a thing as too much P or D? Or should I aim for the maximum P that I can figure out how to tune, with enough D to keep the system stable?
So say for the sake of argument now that my P and D are where they should be for my graph to be "optimally damped", and my mechanically detached servo zips from where its at to its commanded position with little to no overshoot and no hunting for position at the end of its commanded run.
What I don't get is how much "I" that I need to add, as per the R2000setup PDF I need to add some to make sure that the graph shows some overshoot once "I" is added, but doesn't specify how much I need to have, or what is optimal. I get that P, I, D numbers are all dimensionless and that their values in the context of the Rutex drives may be different from what P,I,D are expressed as for other servo tuning systems, but I feel like I'm stabbing in the dark on this one and while Rutex has given me some pointers on tuning I think it made me less confident of what I *think* I'm trying to achieve with my servo tuning exercises.
If I understand "I" correctly its like adding a flywheel, in that it takes more "push" to move the servo but once moving the motion is more stable. So now I'm wondering, how does "I" interact with the weight of the table once the servo is mechanically connected to it? I imagine moving a heavy (12.5x54") milling table with a 5 TPI ballscrew, even well lubricated, still pushes back against the motion the servo is trying to produce when starting an axis move, and again when decelerating it to a stop?
How much should I tune the servo with the table disconnected before I go ahead and connect it? Or is the suggestion to start with the table disconnected just to keep me from doing anything exceptionally dumb while I'm still learning, and hopefully keep from making any life or machine threatening mistakes?
Am I overthinking this too much?
Thanks for the help
I'm working on my Lagun FTV-3 retrofit, I'm at the point now where I'm tuning drives and (hopefully) getting closer to making chips.
Since a good lot of the parts from when this machine was originally a Hasbach/Bandit driven machine were gone before I took delivery, I've done a lot of research and built my own power supply, I went with an Antek 1500VA transformer that I dewound to output ~57 VAC (= 75VDC rectified), filtered thru a 112000 uF cap bank, and lower power secondaries outputting ~15 VDC rectified each with a 56000 uF cap.
Because this was originally a Tach/Resolver feedback machine and the servo drives were all gone, I fitted 500 line Avago HEDS 5500 series encoders to the X servo so I could use much newer (and cheaper!) servo drives. The Y servo was gone when i got the machine but the replacement servo has also been fitted with a 500 resolution encoder. I then found a set of 3 Rutex R2010 drives on ebay for dirt cheap, I also have a Gecko G320 that I was originally planning on using to drive my servos before I found the R2010's that is lying on the bench waiting for a purpose (possibly a 4th axis? Someday...). Right now the X and Y servos are hooked up electrically to their drives but disconnected mechanically from the mill, I've done some preliminary messing about with PID tuning and now I'm running into some questions I don't know how to answer
- Rutex posted a document here: http://www.rutex.com/files/R2000setup.pdf detailing how to set things up, so I've made the SPI tuning cable, verified correct motor/encoder hookup, and done some preliminary PID tuning with the R2xtuneVB6 program. As an aside, I wish Rutex would have specified that the VB6 program has a lot more features and better GUI than the other R2xtune program.
I get that I want to turn up P to get a fast response time, when I get excessive hunting add more D to get more stability. I understand that when I have an overdamped step response graph with no overshoot, I need to increase my P to get more acceleration out of my servo, and when I induce too much I get oscillation either from disturbing the position of the motor or commanding steps, so I add more D to keep hunting/ringing away.
If my servo isn't ringing at a standstill, after a step command or when I disturb the shaft of the motor, is there such a thing as too much P or D? Or should I aim for the maximum P that I can figure out how to tune, with enough D to keep the system stable?
So say for the sake of argument now that my P and D are where they should be for my graph to be "optimally damped", and my mechanically detached servo zips from where its at to its commanded position with little to no overshoot and no hunting for position at the end of its commanded run.
What I don't get is how much "I" that I need to add, as per the R2000setup PDF I need to add some to make sure that the graph shows some overshoot once "I" is added, but doesn't specify how much I need to have, or what is optimal. I get that P, I, D numbers are all dimensionless and that their values in the context of the Rutex drives may be different from what P,I,D are expressed as for other servo tuning systems, but I feel like I'm stabbing in the dark on this one and while Rutex has given me some pointers on tuning I think it made me less confident of what I *think* I'm trying to achieve with my servo tuning exercises.
If I understand "I" correctly its like adding a flywheel, in that it takes more "push" to move the servo but once moving the motion is more stable. So now I'm wondering, how does "I" interact with the weight of the table once the servo is mechanically connected to it? I imagine moving a heavy (12.5x54") milling table with a 5 TPI ballscrew, even well lubricated, still pushes back against the motion the servo is trying to produce when starting an axis move, and again when decelerating it to a stop?
How much should I tune the servo with the table disconnected before I go ahead and connect it? Or is the suggestion to start with the table disconnected just to keep me from doing anything exceptionally dumb while I'm still learning, and hopefully keep from making any life or machine threatening mistakes?
Am I overthinking this too much?
Thanks for the help