I think this is a interesting question. You might be a bit more specific. Are you referring to steppers or servos?
Obviously linear glass scales as the feedback would potentially provide the highest accuracy as all mechanical errors such as backlash, compliance, lead screw errors are all in the loop. The disadvantage with having more mechanical things in the loop is that it becomes more difficult to make the feedback loop respond quickly and to be stable. For example consider a servo system with lots of friction and backlash. If there is a
small error then the servo applies a torque. If the torque is not enough to overcome friction then nothing happens so the servo (integrator) applies more torque. Eventually the torque will overcome the friction and break away, but because of backlash the servo will still see no motion at the scales and so apply still more torque. By the time any motion (or any information) is detected by the scales it can be too late to avoid an overshoot. In such a system the only way to avoid increasing overshoot and oscillation is to use low gains such that the corrections are made slow and gradual enough to avoid excessive overshoot.
Most servos with linear scales utilize some form of feedback at the motor shaft to detect motion before the scales. Either a shaft encoder or an analog
tachometer. If an encoder is used it should be high resolution, not necessarily for accuracy, but rather to detect velocity with good resolution in a short period of time (dx/dt).
In some ways operating a Stepper Motor with closed loop linear scales is like a dual loop system. A stepper motor passively holds position much like a spring. So in the example above with friction and backlash, once the motor breaks loose and moves the motor torque will naturally reduce even before the scales have indicated any motion.
The next KMotion/KFlop software release will support a new Closed Loop Stepper Mode. See:
Currently to do true dual closed loop control with two active sources of feedback two KFlop Axis channels must be linked together in cascade by using a User Program. This allows a full set if Servo parameters for both loops and allows plotting the performance of each loop separately.
I hope this helps
TK
| Group: DynoMotion |
Message: 40 |
From: tmday7 |
Date: 8/31/2009 |
| Subject: Re: encoder type? |
Hey TK,
Had to chime in here. :) Are you saying that 1 Kflop is capable of handling both encoders and glass scales at the same time?Say.. for a 3 axis mill ;) For true Dual Closed Loop.
Thinking of a future project.
Thanks again,
Troy
> The next KMotion/KFlop software release will support a new Closed Loop Stepper Mode. See:
>
> http://dynomotion.com/Help/ClosedLoopStep/ClosedLoopStepper.htm
>
> Currently to do true dual closed loop control with two active sources of feedback two KFlop Axis channels must be linked together in cascade by using a User Program. This allows a full set if Servo parameters for both loops and allows plotting the performance of each loop separately.
>
> I hope this helps
> TK
>
>
>
>
>
>
>
> ________________________________
> From: mp006ltk <mp006ltk@...>
> To: DynoMotion@yahoogroups.com
> Sent: Friday, August 28, 2009 11:16:12 PM
> Subject: [DynoMotion] encoder type?
>
>
> Is it best to use rotary encoders on the screw or linear glass scale encoders on the actual table? The CNC Brain uses both and I wanted to see how this diferentiates.
> |
|
| Group: DynoMotion |
Message: 41 |
From: Tom Kerekes |
Date: 8/31/2009 |
| Subject: Re: encoder type? |
Hi Troy,
Yes it is looking like we can increase the number of encoder inputs from 4 to 8 which would be necessary also.
TK
| Group: DynoMotion |
Message: 42 |
From: mp006ltk |
Date: 8/31/2009 |
| Subject: Re: encoder type? |
I was thinking of using servos with a rotary encoder connected to the driver and then using glass scales through the motion controller. I have been reading up on the competetive product, the CNC brain, and they can handle both throught he controller but I'm told it runs really slow. I noticed dynomotion and now I was hoping the kflop has more processing power. Safeguard robotics and dynomotion need to merge and produce one kickass turnkey soloution. The brain seems to have better software but it has problems without optoisolators and needs you to make yout own breakout board. The Kflop seems to be much easier to install and implement.
--- In DynoMotion@yahoogroups.com, Tom Kerekes <tk@...> wrote:
>
>
>
> I think this is a interesting question. You might be a bit more specific. Are you referring to steppers or servos?
>
> Obviously linear glass scales as the feedback would potentially provide the highest accuracy as all mechanical errors such as backlash, compliance, lead screw errors are all in the loop. The disadvantage with having more mechanical things in the loop is that it becomes more difficult to make the feedback loop respond quickly and to be stable. For example consider a servo system with lots of friction and backlash. If there is a small error then the servo applies a torque. If the torque is not enough to overcome friction then nothing happens so the servo (integrator) applies more torque. Eventually the torque will overcome the friction and break away, but because of backlash the servo will still see no motion at the scales and so apply still more torque. By the time any motion (or any information) is detected by the scales it can be too late to avoid an overshoot. In such a system the only way to avoid increasing overshoot and oscillation is to use
> low gains such that the corrections are made slow and gradual enough to avoid excessive overshoot.
>
> Most servos with linear scales utilize some form of feedback at the motor shaft to detect motion before the scales. Either a shaft encoder or an analog tachometer. If an encoder is used it should be high resolution, not necessarily for accuracy, but rather to detect velocity with good resolution in a short period of time (dx/dt).
>
> In some ways operating a Stepper Motor with closed loop linear scales is like a dual loop system. A stepper motor passively holds position much like a spring. So in the example above with friction and backlash, once the motor breaks loose and moves the motor torque will naturally reduce even before the scales have indicated any motion.
>
> The next KMotion/KFlop software release will support a new Closed Loop Stepper Mode. See:
>
> http://dynomotion.com/Help/ClosedLoopStep/ClosedLoopStepper.htm
>
> Currently to do true dual closed loop control with two active sources of feedback two KFlop Axis channels must be linked together in cascade by using a User Program. This allows a full set if Servo parameters for both loops and allows plotting the performance of each loop separately.
>
> I hope this helps
> TK
>
>
>
>
>
>
>
> ________________________________
> From: mp006ltk <mp006ltk@...>
> To: DynoMotion@yahoogroups.com
> Sent: Friday, August 28, 2009 11:16:12 PM
> Subject: [DynoMotion] encoder type?
>
>
> Is it best to use rotary encoders on the screw or linear glass scale encoders on the actual table? The CNC Brain uses both and I wanted to see how this diferentiates.
> |
|
| |
