Back to Back Motor Control

[Eoghann]

We are building a high inertia robot that requires very high precision on several rotary axis. We need a a system to negate the backlash and have decided that Back To Back motors would be best. Because one motor is "dragging" on the other Indramat and Siemens solutions only allow one motor to drive in one direction ie. both motors must be capable of producing the full torque required for the axis. Does anyone know of a controller that would allow torque "sharing" (so that we could have two smaller motors) between the two drives?

any help gratefully received!

 

[RadarRay]

I seem to be interpreting your post in two entirely diffent directions. Do you wish to have two motors where one motors supplies forward torque while the other supplies reverse torque, which is what you could do to reduce backlash OR do you just want to have one load driven by multiple motors? Which can be done in several ways.

 

[jbartos]

Suggestion: Visit

http://www.engineeringtalk.com/news/len/len120.html

http://www.nasatech.com/Briefs/June01/NPO20904.html

http://161.58.192.251/_fe_files/PDF/Fine_Cyclo/FineCyclo-004_Overview.pdf

etc. for more info
A motor controller can be developed precise enough to produce the negligible backlash

 

[SERVOCAM]

I read this post last night and I thought about this more this morning. It sounds like the motors go in the same direction. It sounds like what they are doing is using 2 motors not neccessarily to eliminate gear boxes, but to eliminate backlash buy having one of the motors drag behind a bit.

I have done applications like this where the customer took 1 motor into 2 gearboxes, then on the output of the gearbox, the preloaded the two against each other. Think of a rack and pinion, then having two pinions on it. Preload the 2 against each other to eliminate backlash.

In this case, we want to load the motors against each other. For control, I think using more like a cam table to do this rather than gearing because we want position based, not speed based. Some how the system would have to on power up, rotate the motors in oppisite direction to preload them, then enable camming and keep one motor always behind the other, keeping the mechanics stiff, but also still both of them are moving the load.

Can you define more in detail what your trying to do exactly.

Thank you,

Cameron Anderson - Sales & Applications Engineer
Aerotech, Inc. -

http://www.aerotech.com

"Dedicated to the Science of Motion"

 

[jbartos]

Comment: Even if there is a preload or two motors working against each other, there will always be some small or negligible backlash.

 

[SERVOCAM]

Yes, jbartos is correct...also spungy tunning will give some play in the system.

I think that possibly in this application, that 3D error mapping may be very benificial. Mapping out the positions of the end-effoctor and createing a calibration table to compensate for errors in mechanics...still can't full compensate for backlash...so large direct-drive motors may be the way to go.

Maybe look at how GE Fanuc builds their robots. Now when you say robot, are you talking about Gantry/Cartesian, SCARA, or Articulated? Or is this a non-traditional robot system that you just call a robot?

More details when you can please.


So we don't get on a "Robot" argument....

ro·bot n. (Courtesy of Dictonary.com)
A mechanical device that sometimes resembles a human and is capable of performing a variety of often complex human tasks on command or by being programmed in advance.
A machine or device that operates automatically or by remote control.
A person who works mechanically without original thought, especially one who responds automatically to the commands of others.



Cameron Anderson - Sales & Applications Engineer
Aerotech, Inc. -

http://www.aerotech.com

"Dedicated to the Science of Motion"

 

[Eoghann]

Thanks for the response!
The Dual Motor system came from my boss who has used them in the past. The controller that he used previously meant that the combined power of the motors was equal to running torque required. Indramat and Siemens systems on offer today do not support this configuration meaning that you must spec each motor to the full running torque thus putting the price up.

The robot structure is a little unorthodox. It has a central column with linear vertical movement (rack and pinion) and a carriage that a set of grippers will be mounted upon that is moved horizontally on another rack and pinion. The gripper head will have a "yaw" movement (double drive system), -45degree to +15degree pitch, 180degree roll movement and the grippers themselves. The complete structure will rotate at the central column by a double drive Slew. The maximum distance from the center of the robot to the gripper tips is 6m with a max payload of 500kg and an accuracy of +/- 1mm.

You can see why we need to remove as much of the backlash as possible but as an OEM we also need to keep the costs down. There will have to be interpolation between the slew (main rotation), horizontal and the yaw axis to keep the center line of the grippers perpendicular to the 3m long work piece as the robot rotates left and right. The pitch, roll and vertical functions will only be for special manipulations and will be in one position for most of the working life. The robot is completely automated with the option of an operator taking over if something goes wrong.

So, we are basically looking for a motion controller/double drive controller that can interpolate between 3 axis, control a further single axis (vertical) and be happy with an Siemens S7 PLC doing the sequencing

Thanks again for the help and ideas!

 

[GusD]

Hi foghann

I believe that "gears boxes" or gears when meshing,have to have a certain amount of Backlash.There is no getting away from it.As far as the motor torque being equal to load torque,just like backlash ,motor TQ has to be > load Tq otherwise we have a Tie.No "Work" can be produced.

Good roboting

GusD

 

[SERVOCAM]

Do distributed feedback also known as dual-loop feedback. Do not use the encoders on the motors for position. Put linear encoders on the linear axis and rotary encoder on the rotary axis and feed them back to the controller to close the position loops.

So any backlash in the mechanics will be compensated for, you will strictly position to the encoder on the load.




Cameron Anderson - Sales & Applications Engineer
Aerotech, Inc. -

http://www.aerotech.com

"Dedicated to the Science of Motion"

 

[Eoghann]

The only problem with using the external feedback encoder (ie. not on the motor shaft) is if there is a lot of backlash the controller can "hunt" for position. The controller will continually oscillate back and forth through the backlash producing a very unstable machine. As the rotary slew ring is 2m in diameter it is very difficult to get the backlash low enough to prevent this.

 

[RadarRay]

I haven't drawn out you app as you described it, so I may not have full appreciation yet, but I don't understand why you say you can not split the torque.
If you had the servo vendor supply half the torque on each motor with a separate power unit, you should be able to run one as lead and the second as loadshare. The load drawn by the lead would limit the torque on the slave to the same value. The limitation of this comes when you are close to zero torque or the torque is reversing. You have to do some fancy software maniputlation to guarantee the slave torque is the same direction and magnitude as the master.
GusD is correct when he says all gear boxes have backlash. Even those that are sold as zero backlash have some. I guess I will have to go over your description again to see why two motors is helping this.

ON another issue, Servocam, I agree with Eoghann about the encoder mounted on the machine instead of on the motor. I have never been successful in getting the performance I wanted out of doing this for the exact reasons he mentioned, although all servo manufactureres say you can do this. Have you ever been successful with it and was the application demanding?

 

[SERVOCAM]

ON another issue, Servocam, I agree with Eoghann about the encoder mounted on the machine instead of on the motor. I have never been successful in getting the performance I wanted out of doing this for the exact reasons he mentioned, although all servo manufactureres say you can do this. Have you ever been successful with it and was the application demanding?

Yes, we do it all the time with our controllers on our stages. Rotary encoder on motor & linear encoder in the stage. Of course we are using a preloaded nut, but there is still a degree of backlash. I have done it in other applications on belt drives using an rotary encoder on the other end, or even a linear encoder. We have also put linear/ring encoders on worm-drive rotary stages to close the position loop.



Cameron Anderson - Sales & Applications Engineer
Aerotech, Inc. -

http://www.aerotech.com

"Dedicated to the Science of Motion"

 

[Eoghann]

On another note Servocam, I'll probably see your UK counterparts in a week or so at the Drive and Controls Exhibition in England. Do you guys do a lot of business in Europe?

 

[robotboy]

Sorry... but it appears your described "two" motor approach is a kludge, and an expensive one at that. The inserted motor "drag" merely serves as a system damper, but playing two motors off one another robs overall efficiency (and again elevates costs).

I believe if you can get a handle on describing your system mathematically, you should be able to design a control loop which monitors motor current and thereby control torque. Just remember if/when the system changes direction, that kinetic energy has to go somewhere (regardless of the control system), and protection on the drive electronics also needs to be implemented.

Good Luck!

 

[Eoghann]

As a follow up if anyone is still watching this thread.
We have found two companies who have implemented this system: NUM-Schneider and SEW Eurodrive, applied mainly to large CNC machinery.
Initial figures for the motors show that in single-direction motor movement we need a little over 100NM (at the motor) to accelerate but with the two drives running in tandem this is dropped to 38NM!! Cost difference of about £2000 per axis.
As far as I understand it (still deep in talks with their Apps guys) the motion controller keeps both motors driving until the very last part of the decelerate-to-position movement then switches one of the motors to lag the other thus pulling the robot consitently to one side of the backlash and letting us hit our repeatability targets
Anyone intrested drop me an email and I'll send out information/results once the project is completed/moving (soon-ish).
Cheers for the ideas!

 

[RadarRay]

Thanks for the update Egghann. Ask them when the load is being accelerated, one motor is in lead and I assume position controlled. What is the other motor controlled by? Torque, Loadshare or also position control. Then ask the same thing about when it up to speed. Is the second motor lagging in Torque mode, speed mode or position mode.

Just curious.

 

[jOmega]

JBartos & GusD,

If you're ever in California, I invite you to stop by the Point Magu Naval Air Station and have a look at the antennae drives for the pedestal mounted dishes.

They have precision gearboxes and back-to-back motors controlled by a dc servo drive system that results in .....ZERO backlash..... and hold position in varying wind conditions.

 

[jbartos]

Suggestion to the previous posting: They probably have some small or negligible backlash as I mentioned in my previous posting. There is no such thing in manufacturing as the zero error or accuracy. I learned this more than 40 years ago.

 

[jOmega]

Sorry to disappoint you, JB, but there is NO discernable backlash. The system is electro-mechanical consisting of a PRECISION gearbox, and back-to-back motors in a very stiff servo system. Tracking antennaes cannot have ANY backlash.

p.s. I was there for the commissioning of the drive system and witnessed the measurements....a low frequency bi-polar squarewave was applied to the input of the servo loop to test the system response and check for backlash.


p.p.s. Perhaps you were absent the day this lesson was taught.

p.p.p.s. There has been much written on ZERO BACKLASH systems over the last 50 years.... A lot of it in the Servo Drive community. Perhaps you'd be inclined to crank up your search engine and probe for Antennae Drives. Makes for some fascinating reading.

 

[jbartos]

Suggestion to the previous posting: Many things have absolute accuracy, zero errors, etc., however, on paper only, not in practice. Even, the digital displays, reading, for example 1.0001, have an error behind the last digit. I have not seen a digital display with infinite number of decimals. It would surely need some commissioners to commission.