Dynomotion

Group: DynoMotion Message: 1624 From: Bengt Sjoelund Date: 7/31/2011
Subject: X axis with VFD
Hi,
Just uploaded 2 png files into Bengts CombiMill folder.
Is this OK or is something wrong? What next with this axis? Would like to get one axis at to work fully first so I can get a grip of what all necessary steps are.
Cheers Bengt
Group: DynoMotion Message: 1625 From: TK Date: 7/31/2011
Subject: Re: X axis with VFD
Not that good. Like I said before D gain (maybe somewhere between 2 - 50) is usually required when operating in torque mode. Increase D until the system goes unstable then decrease by 2 or 3x then try increasing P to get better performance. 


TK

On Jul 31, 2011, at 1:32 PM, "Bengt Sjoelund" <cnc@...> wrote:

 

Hi,
Just uploaded 2 png files into Bengts CombiMill folder.
Is this OK or is something wrong? What next with this axis? Would like to get one axis at to work fully first so I can get a grip of what all necessary steps are.
Cheers Bengt
Group: DynoMotion Message: 1626 From: Bengt Sjoelund Date: 8/1/2011
Subject: Re: X axis with VFD
Hi Tom,
Downloaded png's #3 to #6. Looks like I'm getting closer, what can be done to have the red position curve fully aligned with the commanded blue curve? Is this to be adjusted with acceleration?
As you can see the settings are far away from your suggestion, what can cause this kind of difference? In VSD I get what I believe to acceptable torque setting as you can see in the PDF file I downloaded.

What next?
Cheers Bengt

--- In DynoMotion@yahoogroups.com, TK <tk@...> wrote:
>
> Not that good. Like I said before D gain (maybe somewhere between 2 - 50) is usually required when operating in torque mode. Increase D until the system goes unstable then decrease by 2 or 3x then try increasing P to get better performance.
>
>
> TK
>
> On Jul 31, 2011, at 1:32 PM, "Bengt Sjoelund" <cnc@...> wrote:
>
> > Hi,
> > Just uploaded 2 png files into Bengts CombiMill folder.
> > Is this OK or is something wrong? What next with this axis? Would like to get one axis at to work fully first so I can get a grip of what all necessary steps are.
> > Cheers Bengt
> >
> >
>
Group: DynoMotion Message: 1627 From: Tom Kerekes Date: 8/1/2011
Subject: Re: X axis with VFD
Hi Bengt,
 
I think the final #6 plot looks pretty good. 
 
 
The error is only a few encoder counts.  And this is during high acceleration.  During lower speed cutting the errors would probably be smaller.  What is the resolution of your system?  How many counts per mm??  Is this error acceptable?  There will always be some error in a servo system because error is required before a correction can be made.
 
The P and D gains are relative to your system.  My suggestions were just starting guesses.
 
You could probably benefit by adding a low pass filter.  This is because we are trying to control velocity using the encoder which has finite resolution (quantization error) and has "steps".  When it changes by one count the system sees it as an instantaneous change.  A higher resolution encoder would reduce the effect.  A one count encoder change multiplied by a Derivative gain of 1000 causes the big 1000 count DAC spikes in the Output (Green plot).  The Amplifier is probably low pass filtering those to some degree, but it may help things to filter them first.  On the filter screen set Filter #2 to 2nd order low pass at a frequency of 500 Hz and Q of 1.4.  That should reduce the spikes in the output plot and possibly allow you to get a bit more performance.  You can experiment with lower and higher frequency filtering.  Lower frequency will provide more smoothing in the output but too much filtering will add lag to the response (overshoot) and is de-stabilizing (oscillations).
 
There also seems to be significant friction in your system which limits the performance.  Notice when the system begins to move that the output ramps ~100 DAC counts of torque before the system breaks loose and any motion begins at all. 
 
I hope this helps.
TK 


--- On Mon, 8/1/11, Bengt Sjoelund <cnc@...> wrote:

From: Bengt Sjoelund <cnc@...>
Subject: [DynoMotion] Re: X axis with VFD
To: DynoMotion@yahoogroups.com
Date: Monday, August 1, 2011, 5:04 AM

 
Hi Tom,
Downloaded png's #3 to #6. Looks like I'm getting closer, what can be done to have the red position curve fully aligned with the commanded blue curve? Is this to be adjusted with acceleration?
As you can see the settings are far away from your suggestion, what can cause this kind of difference? In VSD I get what I believe to acceptable torque setting as you can see in the PDF file I downloaded.

What next?
Cheers Bengt

--- In DynoMotion@yahoogroups.com, TK <tk@...> wrote:
>
> Not that good. Like I said before D gain (maybe somewhere between 2 - 50) is usually required when operating in torque mode. Increase D until the system goes unstable then decrease by 2 or 3x then try increasing P to get better performance.
>
>
> TK
>
> On Jul 31, 2011, at 1:32 PM, "Bengt Sjoelund" <cnc@...> wrote:
>
> > Hi,
> > Just uploaded 2 png files into Bengts CombiMill folder.
> > Is this OK or is something wrong? What next with this axis? Would like to get one axis at to work fully first so I can get a grip of what all necessary steps are.
> > Cheers Bengt
> >
> >
>
Group: DynoMotion Message: 1630 From: Bengt Sjoelund Date: 8/2/2011
Subject: Re: X axis with VFD
Hi Tom,
Thanks for your explanation, I think I have managed to set-up all 3 axis but will do further fine tuning.
As regards resolution, there are 1250ppr encoders on motors X/Y are geared via belt drive 2:1 and ball-screws are 5mm. Z axis is 4:1.
So if I have calculated correctly there should be 500 counts on X/Y and 1000 on Z. As you said earlier higher count encoders would be better but this will have to do for starters.

How do I get my axis up to speed in KMotionCNC so 1mm commanded will move 1mm?

Cheers Bengt

--- In DynoMotion@yahoogroups.com, Tom Kerekes <tk@...> wrote:
>
> Hi Bengt,
>  
> I think the final #6 plot looks pretty good. 
>  
> http://tech.groups.yahoo.com/group/DynoMotion/files/Bengt%27s%20CombiMill/xaxis%20vsd%20torq%206.PNG
>  
> The error is only a few encoder counts.  And this is during high acceleration.  During lower speed cutting the errors would probably be smaller.  What is the resolution of your system?  How many counts per mm??  Is this error acceptable?  There will always be some error in a servo system because error is required before a correction can be made.
>  
> The P and D gains are relative to your system.  My suggestions were just starting guesses.
>  
> You could probably benefit by adding a low pass filter.  This is because we are trying to control velocity using the encoder which has finite resolution (quantization error) and has "steps".  When it changes by one count the system sees it as an instantaneous change.  A higher resolution encoder would reduce the effect.  A one count encoder change multiplied by a Derivative gain of 1000 causes the big 1000 count DAC spikes in the Output (Green plot).  The Amplifier is probably low pass filtering those to some degree, but it may help things to filter them first.  On the filter screen set Filter #2 to 2nd order low pass at a frequency of 500 Hz and Q of 1.4.  That should reduce the spikes in the output plot and possibly allow you to get a bit more performance.  You can experiment with lower and higher frequency filtering.  Lower frequency will provide more smoothing in the output but too much filtering will add lag to the response (overshoot) and
> is de-stabilizing (oscillations).
>  
> There also seems to be significant friction in your system which limits the performance.  Notice when the system begins to move that the output ramps ~100 DAC counts of torque before the system breaks loose and any motion begins at all. 
>  
> I hope this helps.
> TK 
>
>
> --- On Mon, 8/1/11, Bengt Sjoelund <cnc@...> wrote:
>
>
> From: Bengt Sjoelund <cnc@...>
> Subject: [DynoMotion] Re: X axis with VFD
> To: DynoMotion@yahoogroups.com
> Date: Monday, August 1, 2011, 5:04 AM
>
>
>  
>
>
>
> Hi Tom,
> Downloaded png's #3 to #6. Looks like I'm getting closer, what can be done to have the red position curve fully aligned with the commanded blue curve? Is this to be adjusted with acceleration?
> As you can see the settings are far away from your suggestion, what can cause this kind of difference? In VSD I get what I believe to acceptable torque setting as you can see in the PDF file I downloaded.
>
> What next?
> Cheers Bengt
>
> --- In DynoMotion@yahoogroups.com, TK <tk@> wrote:
> >
> > Not that good. Like I said before D gain (maybe somewhere between 2 - 50) is usually required when operating in torque mode. Increase D until the system goes unstable then decrease by 2 or 3x then try increasing P to get better performance.
> >
> >
> > TK
> >
> > On Jul 31, 2011, at 1:32 PM, "Bengt Sjoelund" <cnc@> wrote:
> >
> > > Hi,
> > > Just uploaded 2 png files into Bengts CombiMill folder.
> > > Is this OK or is something wrong? What next with this axis? Would like to get one axis at to work fully first so I can get a grip of what all necessary steps are.
> > > Cheers Bengt
> > >
> > >
> >
>
Group: DynoMotion Message: 1631 From: Tom Kerekes Date: 8/2/2011
Subject: Re: X axis with VFD
Hi Bengt,
 
Usually when an encoder is rated in ppr (pulses per rev) then there are 4X as many quadrature counts per rev.  To be sure simply do a large move in counts and measure.
 
Assuming there are 2000 counts/mm enter the equivalent in inches on the
 
2000 x 25.4 = 50800 counts/inch
 
KMotionCNC | Tool Setup | Trajectory Planner Screen
 
Regards
TK

--- On Tue, 8/2/11, Bengt Sjoelund <cnc@...> wrote:

From: Bengt Sjoelund <cnc@...>
Subject: [DynoMotion] Re: X axis with VFD
To: DynoMotion@yahoogroups.com
Date: Tuesday, August 2, 2011, 5:03 AM

 
Hi Tom,
Thanks for your explanation, I think I have managed to set-up all 3 axis but will do further fine tuning.
As regards resolution, there are 1250ppr encoders on motors X/Y are geared via belt drive 2:1 and ball-screws are 5mm. Z axis is 4:1.
So if I have calculated correctly there should be 500 counts on X/Y and 1000 on Z. As you said earlier higher count encoders would be better but this will have to do for starters.

How do I get my axis up to speed in KMotionCNC so 1mm commanded will move 1mm?

Cheers Bengt

--- In DynoMotion@yahoogroups.com, Tom Kerekes <tk@...> wrote:
>
> Hi Bengt,
>  
> I think the final #6 plot looks pretty good. 
>  
> http://tech.groups.yahoo.com/group/DynoMotion/files/Bengt%27s%20CombiMill/xaxis%20vsd%20torq%206.PNG
>  
> The error is only a few encoder counts.  And this is during high acceleration.  During lower speed cutting the errors would probably be smaller.  What is the resolution of your system?  How many counts per mm??  Is this error acceptable?  There will always be some error in a servo system because error is required before a correction can be made.
>  
> The P and D gains are relative to your system.  My suggestions were just starting guesses.
>  
> You could probably benefit by adding a low pass filter.  This is because we are trying to control velocity using the encoder which has finite resolution (quantization error) and has "steps".  When it changes by one count the system sees it as an instantaneous change.  A higher resolution encoder would reduce the effect.  A one count encoder change multiplied by a Derivative gain of 1000 causes the big 1000 count DAC spikes in the Output (Green plot).  The Amplifier is probably low pass filtering those to some degree, but it may help things to filter them first.  On the filter screen set Filter #2 to 2nd order low pass at a frequency of 500 Hz and Q of 1.4.  That should reduce the spikes in the output plot and possibly allow you to get a bit more performance.  You can experiment with lower and higher frequency filtering.  Lower frequency will provide more smoothing in the output but too much filtering will add lag to the response (overshoot) and
> is de-stabilizing (oscillations).
>  
> There also seems to be significant friction in your system which limits the performance.  Notice when the system begins to move that the output ramps ~100 DAC counts of torque before the system breaks loose and any motion begins at all. 
>  
> I hope this helps.
> TK 
>
>
> --- On Mon, 8/1/11, Bengt Sjoelund <cnc@...> wrote:
>
>
> From: Bengt Sjoelund <cnc@...>
> Subject: [DynoMotion] Re: X axis with VFD
> To: DynoMotion@yahoogroups.com
> Date: Monday, August 1, 2011, 5:04 AM
>
>
>  
>
>
>
> Hi Tom,
> Downloaded png's #3 to #6. Looks like I'm getting closer, what can be done to have the red position curve fully aligned with the commanded blue curve? Is this to be adjusted with acceleration?
> As you can see the settings are far away from your suggestion, what can cause this kind of difference? In VSD I get what I believe to acceptable torque setting as you can see in the PDF file I downloaded.
>
> What next?
> Cheers Bengt
>
> --- In DynoMotion@yahoogroups.com, TK <tk@> wrote:
> >
> > Not that good. Like I said before D gain (maybe somewhere between 2 - 50) is usually required when operating in torque mode. Increase D until the system goes unstable then decrease by 2 or 3x then try increasing P to get better performance.
> >
> >
> > TK
> >
> > On Jul 31, 2011, at 1:32 PM, "Bengt Sjoelund" <cnc@> wrote:
> >
> > > Hi,
> > > Just uploaded 2 png files into Bengts CombiMill folder.
> > > Is this OK or is something wrong? What next with this axis? Would like to get one axis at to work fully first so I can get a grip of what all necessary steps are.
> > > Cheers Bengt
> > >
> > >
> >
>
Group: DynoMotion Message: 1633 From: Bengt Sjoelund Date: 8/2/2011
Subject: Re: X axis with VFD
Hi Tom,
I did manage to find out the trajectory planner so now I have movement on all axis with as it seems correct figures :)
Had some troubles wit Z axis that did cause over current in drivers. I do believe that I just wanted to much as regards speed, it is a heavy burden to lift Z+X so I have got something like 7mm/sec in fast jog speed. Time will tell if it is fast enough.
Next step is to figure out how G-code is run in KmotionCNC.
Thanks for you kind support.
Cheers Bengt

--- In DynoMotion@yahoogroups.com, Tom Kerekes <tk@...> wrote:
>
> Hi Bengt,
>  
> Usually when an encoder is rated in ppr (pulses per rev) then there are 4X as many quadrature counts per rev.  To be sure simply do a large move in counts and measure.
>  
> Assuming there are 2000 counts/mm enter the equivalent in inches on the
>  
> 2000 x 25.4 = 50800 counts/inch
>  
> KMotionCNC | Tool Setup | Trajectory Planner Screen
>  
> Regards
> TK
>
> --- On Tue, 8/2/11, Bengt Sjoelund <cnc@...> wrote:
>
>
> From: Bengt Sjoelund <cnc@...>
> Subject: [DynoMotion] Re: X axis with VFD
> To: DynoMotion@yahoogroups.com
> Date: Tuesday, August 2, 2011, 5:03 AM
>
>
>  
>
>
>
> Hi Tom,
> Thanks for your explanation, I think I have managed to set-up all 3 axis but will do further fine tuning.
> As regards resolution, there are 1250ppr encoders on motors X/Y are geared via belt drive 2:1 and ball-screws are 5mm. Z axis is 4:1.
> So if I have calculated correctly there should be 500 counts on X/Y and 1000 on Z. As you said earlier higher count encoders would be better but this will have to do for starters.
>
> How do I get my axis up to speed in KMotionCNC so 1mm commanded will move 1mm?
>
> Cheers Bengt
>
> --- In DynoMotion@yahoogroups.com, Tom Kerekes <tk@> wrote:
> >
> > Hi Bengt,
> >  
> > I think the final #6 plot looks pretty good. 
> >  
> > http://tech.groups.yahoo.com/group/DynoMotion/files/Bengt%27s%20CombiMill/xaxis%20vsd%20torq%206.PNG
> >  
> > The error is only a few encoder counts.  And this is during high acceleration.  During lower speed cutting the errors would probably be smaller.  What is the resolution of your system?  How many counts per mm??  Is this error acceptable?  There will always be some error in a servo system because error is required before a correction can be made.
> >  
> > The P and D gains are relative to your system.  My suggestions were just starting guesses.
> >  
> > You could probably benefit by adding a low pass filter.  This is because we are trying to control velocity using the encoder which has finite resolution (quantization error) and has "steps".  When it changes by one count the system sees it as an instantaneous change.  A higher resolution encoder would reduce the effect.  A one count encoder change multiplied by a Derivative gain of 1000 causes the big 1000 count DAC spikes in the Output (Green plot).  The Amplifier is probably low pass filtering those to some degree, but it may help things to filter them first.  On the filter screen set Filter #2 to 2nd order low pass at a frequency of 500 Hz and Q of 1.4.  That should reduce the spikes in the output plot and possibly allow you to get a bit more performance.  You can experiment with lower and higher frequency filtering.  Lower frequency will provide more smoothing in the output but too much filtering will add lag to the
> response (overshoot) and
> > is de-stabilizing (oscillations).
> >  
> > There also seems to be significant friction in your system which limits the performance.  Notice when the system begins to move that the output ramps ~100 DAC counts of torque before the system breaks loose and any motion begins at all. 
> >  
> > I hope this helps.
> > TK 
> >
> >
> > --- On Mon, 8/1/11, Bengt Sjoelund <cnc@> wrote:
> >
> >
> > From: Bengt Sjoelund <cnc@>
> > Subject: [DynoMotion] Re: X axis with VFD
> > To: DynoMotion@yahoogroups.com
> > Date: Monday, August 1, 2011, 5:04 AM
> >
> >
> >  
> >
> >
> >
> > Hi Tom,
> > Downloaded png's #3 to #6. Looks like I'm getting closer, what can be done to have the red position curve fully aligned with the commanded blue curve? Is this to be adjusted with acceleration?
> > As you can see the settings are far away from your suggestion, what can cause this kind of difference? In VSD I get what I believe to acceptable torque setting as you can see in the PDF file I downloaded.
> >
> > What next?
> > Cheers Bengt
> >
> > --- In DynoMotion@yahoogroups.com, TK <tk@> wrote:
> > >
> > > Not that good. Like I said before D gain (maybe somewhere between 2 - 50) is usually required when operating in torque mode. Increase D until the system goes unstable then decrease by 2 or 3x then try increasing P to get better performance.
> > >
> > >
> > > TK
> > >
> > > On Jul 31, 2011, at 1:32 PM, "Bengt Sjoelund" <cnc@> wrote:
> > >
> > > > Hi,
> > > > Just uploaded 2 png files into Bengts CombiMill folder.
> > > > Is this OK or is something wrong? What next with this axis? Would like to get one axis at to work fully first so I can get a grip of what all necessary steps are.
> > > > Cheers Bengt
> > > >
> > > >
> > >
> >
>