Thoughts on Servo driver accell settings with Mach2

[Alaska Bob]

Anyone have experience (or know of a thread) with the best method to use the acceleration options on the servo driver itself when using Mach2?

I am using Emerson servo drives with step/direction inputs from Mach2. Currently it is setup as 2048 pulses per rev on the servo driver and I have some arbitrary acceleration constant input into the servo driver itself. As I understand it, it will move each 'step' with an acceleration and deceleration.

So If I put an infinite accelleration on the servo driver, then it will try to 'step' as fast as it can every time Mach tells it to step. ( I am not home now and cannot try things so I figured I would ask here) Mach will ramp up the stepping, but the driver is going to do all it can to get each step completed ASAP. I would be afraid that it will be too aggressive and cause excess heating.

I plan on upgrading to Mach3 and then increasing the pulse rate and change the pulse per rev to 4096 or 8192 which would make jumps smaller and maybe smoother. But if the servo driver circuit is fast enough it may have same issue.

Any thoughts or experience with this would be appreciated.
Thanks!
--bob

[Torchhead]

Bob: I doubt the acceleration (which is torque over time) is per step. I may be closer to a defined following error or a form of position feedback, but that is just a guess. At 2048 pulses per rev and 3000 RPM you have 50 RPS so it amounts the over 100,000 PPS. Putting accel and decel on each one would take some fancy circuitry. MACH has acceleration values that are over time (possibly thousands of pulses) They are in Units per second/sec. Acceleration has a time element. Perhaps the settings are to provide for control situations where the control does not offer acceleration or deceleration settings. Emerson builds drives for use in all kinds of applications.

TOM Caudle
www.CandCNC.com

[Alaska Bob]

Tom,
Thanks for the reply. After researching some more in the manual, I believe you are correct in that it does not have a true 'accell' 'decell' for each step. The specific tab where I input the acell/decell is only for analog input commands it seems. It does use a state-space algorithm.
It does also have specific decell limits that will trump the inputs if they are slower depending on if it is a limit switch or fault etc. I evidently need to study the tuning variables some more.

I am trying to get a handle on the frequency response also. It will accept a 2MHz pulse train for input, yet the encoder outputs 250kHz per channel. I imagine it counts the input train, yet runs it through some averaging algorithms to determine the appropriate velocity.


Thanks again for the reply.
--bob

[Torchhead]

Bob: Are you trying build a CNC machine or study to be a CNC engineer? Normally the default settings on most of that stuff will work fine for 99% of installs.

You are getting stuck on max specs and not what the unit will actually see. You won't get 250KHZ unless the motors will spin fast enough to output that from the encoders. Once again that is a MAX spec ont he encoders saying they COULD output that if you spin them fast enough. Even if they can you probably can't get there with most pulse engines. Then you have to ask your self: Is the motor rated to spin that fast and if it is can you use that much speed? Example: Hook a servo to a pinion 1" in diameter and let ti move a load down a rack spinning at 3000 RPM and you get rapids of 9,000 IPM . Gear it down with belts 10:1 and you STILL have 900 IPM rapids and gobs of torque and resolution.

What I am saying is I doubt you need anything close to 2 MHZ pulse freq or 250KHZ of encoder PPS unless you are building molecular centrifuges!

Never sacrifice speed you can't use for torque and resolution you can.

TOM Caudle
www.CandCNC.com

[Alaska Bob]

Tom,
A little bit of both. I'm an engineer who wants to understand it all. Got the NC machine running and just trying to dig a little deeper and understand the true potential.

My Mitsubishi drives/servos have 131072 pulses per rev and spin at 4500 rpm. That works out to 9,830,400 pulses per second. I need to pick up the book on encoders and do a little reading.

I completely agree it is all way more than I need for my old Tree knee mill, but I do have thoughts of a small precise and very quick machine even if it is just to see how fast it'll go.

Thanks for the replies, I appreciate the discussion.
--bob