--- In DynoMotion@yahoogroups.com, Tom Kerekes <tk@...> wrote:
>
> Hi Darren,
>
> I think you may giving up on the Junus drives a bit early.  I don't agree that using D gain and the Low pass filter would necessarily result in bad performance.  You just might not have found the right tuning parameters.
>
> But there is a slight theoretical advantage to using the tachometer because it is measuring the motor velocity rather than the lead screw velocity.  If there is significant motor friction and belt backlash or compliance this can be a significant difference.  For example if the motor torque is gradually increased until the motor overcomes friction and starts to move, the tachometer will detect this immediately, but the lead screw will not start moving until later when enough belt tension builds up.
>
> AMC (a-m-c.com) drives are always a good choice.
>
> Regards
> TK
>
>
>
> ________________________________
> From: darrengladysz <darrengladysz@...>
> To: DynoMotion@yahoogroups.com
> Sent: Sunday, August 25, 2013 10:58 AM
> Subject: [DynoMotion] Re: Torque or velocity Mode
>
>
>
>  
> Hi Tom and thanks for the quick response.
>
> The motors are connected to the ballscrew via a timing belt.
>
> Correct me if I am wrong but from what I gather from your explanation is that it is confirming what I was observing, namely that the back emf is a poor feedback for the drive and not all that useful for a precision position sensing machine. I also experimented with current alone and had to crank the d gain way up to get a bit of smooth motion. Acceleration and velocity also had to be kept very low.
>
> If I am to use the junus drives I would be basically using d gain or a low pass filter to dampen the oscillations so as to get the system to stay in check with the commanded position but I would over and undershooting which will most likely result in vibrations and poor finish.
>
> In my cnc router where the drives are digital and have encoder feedback the performance is impressive. So, switching back to the old servo dynamics or to a dc drive with encoder feedback would be my best option as I doubt I can get what I want from the junus drives. Converting back to the old servo dynamics would let me use the tachometer on the motors which should help a lot with getting a more stable system.
>
> Alternatively, would you have a more modern option to consider than the servo dynamic drives?
>
> Regards,
>
> Darren
>
> --- In DynoMotion@yahoogroups.com, Tom Kerekes <tk@> wrote:
> >
> > Hi Darren,
> >
> > How is your motor connected to the lead screw?  A belt?
> >
> > I would expect that current mode making use of the encoder feedback would work best and give decent performance. 
> >
> >
> > It is very difficult to measure relatively small back emf voltage changes on a motor when the motor terminals are being switched to high voltages at 20KHz.  So I wouldn't expect back emf feedback to work well.
> >
> >
> > The Junus spec states the sample rates are 20khz for current and 4khz for velocity not the bandwidth.  To achieve a certain bandwidth the sample rate needs to be a much higher frequency than the bandwidth frequency.
> >
> > I would suggest switching to current mode and tuning the best you can then post your settings and some Move plots for us to look at.  A Bode plot would be very useful.  When making a Bode plot it is important to have the right stimulus magnitude and frequency.  Too little encoder movement and there will be no useful information, too much and the amplifier will be in a non-linear mode (saturation) not representative of normal servo operation.  So include a Time Domain plot so we can also have an idea if the Bode Plot should be meaningful.
> >
> > A current mode servo will always be nearly unstable without any phase lead compensation.  Simple P gain will result in a system like a pendulum.  Where the current/torque/acceleration reaches zero when the error reaches zero.  The only way it can be stable at all is if there is some parasitic friction.   So either D gain or a Lead/Lag Filter should be used to stabilize the system.  Either of these methods involves the derivative of the encoder position (delta position/delta time).  Because the encoder has finite resolution it is not smooth but actually consists of "steps".  The lower the resolution the larger the steps.  Taking the derivative of a signal that makes a sudden step results in a "spike" in the output.  Spikes in the output can be bad for various reasons.  So it is usually helpful to add a Low Pass filter to smooth the output.  Usually a 2nd order Low Pass filter at 500 or 1KHz with Q=1.4 works best. 
> >
> >
> > HTH
> > Regards
> > TK
> >
> >
> >
> >
> > ________________________________
> > From: darrengladysz <darrengladysz@>
> > To: DynoMotion@yahoogroups.com
> > Sent: Sunday, August 25, 2013 8:16 AM
> > Subject: [DynoMotion] Torque or velocity Mode
> >
> >
> >
> >  
> > Hi Tom,
> >
> > I am retrofitting kflop to an excello mill and replacing older servo dynamics (1985 or so vintage) brush dc drives that used a tachometer as feedback with Copley controls Junus dc drives that use back emf to measure velocity and can take 90 volts, 10 amps continuous and 20 amps peak. The motors are electrocraft brush dc motors E727 that can take up to 120 volts and 16 amps continuous and 80 amps peak. The motors are outfitted with a tachometer but I cannot use these with the junus servo drives.
> >
> > For the drives I have the option of configuring them to use only a current loop or a velocity loop that controls the current loop. The drives do not give me a pid option but only pi for either current or velocity mode. The current option also has an autotune function.
> >
> > I have the high resolution bei encoders fitted on the ballscrews for each axis that give me between 16000 and 40000 counts in quadrature along with a 5 tpi ballscrew.
> >
> > I have tried to configure the drives and kflop but am having issues with performance and configuration. The motion is quite sluggish, and the axis move in sluggish but jerky moves and have problems maintaining 60 inches per minute - the drives disable because of following errors even though for testing I increased these to more than 1 inch of encoder count. Previously, I would get speeds in excess of 200ipm no problem with the old acroloop controller and the servo dynamics drives.
> >
> > I am using transformer isolated power supply with a series of "soda can" sized transformers to remove most of the ripple in the dc power source - 9 of these with 3500 or so uf each.
> >
> > A few questions.
> >
> > 1. For kflop, is it best to configure the drives solely for the current loop or is it better to use the velocity loop as well.
> >
> > 2. I noted that the back emf works best at higher voltages and unfortunately since this is a metal milling I will be machining at slower speeds so possibly the back emf is not getting accurate readings. The bandwidth of the drives is 20khz for current and 4khz for velocity. Can it be that the drives velocity loop back emf measurement conflict with kflop analog commands?
> >
> > 3. Are the Copley controls junus drives the proper type of drives or should I revert to the old servo dynamics with the tachometer. Link for junus drives is http://www.copleycontrols.com/motion/pdf/Junus.pdf
> >
> > 4. Alternatively, should I switch to dc drives with encoder feedback on the drives or is this not necessary?
> >
> > 5. What is the best way to proceed to get half decent performance from this system. I am looking to machine to 1 thousand precision. I was able to do this with the acroloop / servo dynamics drives before without issue. I am using kflop in my cnc router and know that kflop can handle this without problem so believe the drives or tuning are mostly the issue.
> >
> > Thank you
> >
> > Darren
> >
>

--- In DynoMotion@yahoogroups.com, "darrengladysz" <darrengladysz@...> wrote:
>
> Hi Tom,
>
> Thanks again for the quick response.
>
> The machine is a mill converted to cnc so basically it runs on dovetails. As such there is much startup friction.
>
> I also had the junus in my previous cnc and was not all that pleased with performance. I had hoped that they would work better here but realize that tuning these will take some effort and again performance may suffer is some of the variables change. I also learned a lesson, it is better to have good feedback to both the drives and controller.
>
> I will try the servo dynamics. Should be about 2 hours to re-install them and will see if they do the trick.
>
> I will let you know.
>
> Regards and thanks again,
>
> --- In DynoMotion@yahoogroups.com, Tom Kerekes <tk@> wrote:
> >
> > Hi Darren,
> >
> > I think you may giving up on the Junus drives a bit early.  I don't agree that using D gain and the Low pass filter would necessarily result in bad performance.  You just might not have found the right tuning parameters.
> >
> > But there is a slight theoretical advantage to using the tachometer because it is measuring the motor velocity rather than the lead screw velocity.  If there is significant motor friction and belt backlash or compliance this can be a significant difference.  For example if the motor torque is gradually increased until the motor overcomes friction and starts to move, the tachometer will detect this immediately, but the lead screw will not start moving until later when enough belt tension builds up.
> >
> > AMC (a-m-c.com) drives are always a good choice.
> >
> > Regards
> > TK
> >
> >
> >
> > ________________________________
> > From: darrengladysz <darrengladysz@>
> > To: DynoMotion@yahoogroups.com
> > Sent: Sunday, August 25, 2013 10:58 AM
> > Subject: [DynoMotion] Re: Torque or velocity Mode
> >
> >
> >
> >  
> > Hi Tom and thanks for the quick response.
> >
> > The motors are connected to the ballscrew via a timing belt.
> >
> > Correct me if I am wrong but from what I gather from your explanation is that it is confirming what I was observing, namely that the back emf is a poor feedback for the drive and not all that useful for a precision position sensing machine. I also experimented with current alone and had to crank the d gain way up to get a bit of smooth motion. Acceleration and velocity also had to be kept very low.
> >
> > If I am to use the junus drives I would be basically using d gain or a low pass filter to dampen the oscillations so as to get the system to stay in check with the commanded position but I would over and undershooting which will most likely result in vibrations and poor finish.
> >
> > In my cnc router where the drives are digital and have encoder feedback the performance is impressive. So, switching back to the old servo dynamics or to a dc drive with encoder feedback would be my best option as I doubt I can get what I want from the junus drives. Converting back to the old servo dynamics would let me use the tachometer on the motors which should help a lot with getting a more stable system.
> >
> > Alternatively, would you have a more modern option to consider than the servo dynamic drives?
> >
> > Regards,
> >
> > Darren
> >
> > --- In DynoMotion@yahoogroups.com, Tom Kerekes <tk@> wrote:
> > >
> > > Hi Darren,
> > >
> > > How is your motor connected to the lead screw?  A belt?
> > >
> > > I would expect that current mode making use of the encoder feedback would work best and give decent performance. 
> > >
> > >
> > > It is very difficult to measure relatively small back emf voltage changes on a motor when the motor terminals are being switched to high voltages at 20KHz.  So I wouldn't expect back emf feedback to work well.
> > >
> > >
> > > The Junus spec states the sample rates are 20khz for current and 4khz for velocity not the bandwidth.  To achieve a certain bandwidth the sample rate needs to be a much higher frequency than the bandwidth frequency.
> > >
> > > I would suggest switching to current mode and tuning the best you can then post your settings and some Move plots for us to look at.  A Bode plot would be very useful.  When making a Bode plot it is important to have the right stimulus magnitude and frequency.  Too little encoder movement and there will be no useful information, too much and the amplifier will be in a non-linear mode (saturation) not representative of normal servo operation.  So include a Time Domain plot so we can also have an idea if the Bode Plot should be meaningful.
> > >
> > > A current mode servo will always be nearly unstable without any phase lead compensation.  Simple P gain will result in a system like a pendulum.  Where the current/torque/acceleration reaches zero when the error reaches zero.  The only way it can be stable at all is if there is some parasitic friction.   So either D gain or a Lead/Lag Filter should be used to stabilize the system.  Either of these methods involves the derivative of the encoder position (delta position/delta time).  Because the encoder has finite resolution it is not smooth but actually consists of "steps".  The lower the resolution the larger the steps.  Taking the derivative of a signal that makes a sudden step results in a "spike" in the output.  Spikes in the output can be bad for various reasons.  So it is usually helpful to add a Low Pass filter to smooth the output.  Usually a 2nd order Low Pass filter at 500 or 1KHz with Q=1.4 works best. 
> > >
> > >
> > > HTH
> > > Regards
> > > TK
> > >
> > >
> > >
> > >
> > > ________________________________
> > > From: darrengladysz <darrengladysz@>
> > > To: DynoMotion@yahoogroups.com
> > > Sent: Sunday, August 25, 2013 8:16 AM
> > > Subject: [DynoMotion] Torque or velocity Mode
> > >
> > >
> > >
> > >  
> > > Hi Tom,
> > >
> > > I am retrofitting kflop to an excello mill and replacing older servo dynamics (1985 or so vintage) brush dc drives that used a tachometer as feedback with Copley controls Junus dc drives that use back emf to measure velocity and can take 90 volts, 10 amps continuous and 20 amps peak. The motors are electrocraft brush dc motors E727 that can take up to 120 volts and 16 amps continuous and 80 amps peak. The motors are outfitted with a tachometer but I cannot use these with the junus servo drives.
> > >
> > > For the drives I have the option of configuring them to use only a current loop or a velocity loop that controls the current loop. The drives do not give me a pid option but only pi for either current or velocity mode. The current option also has an autotune function.
> > >
> > > I have the high resolution bei encoders fitted on the ballscrews for each axis that give me between 16000 and 40000 counts in quadrature along with a 5 tpi ballscrew.
> > >
> > > I have tried to configure the drives and kflop but am having issues with performance and configuration. The motion is quite sluggish, and the axis move in sluggish but jerky moves and have problems maintaining 60 inches per minute - the drives disable because of following errors even though for testing I increased these to more than 1 inch of encoder count. Previously, I would get speeds in excess of 200ipm no problem with the old acroloop controller and the servo dynamics drives.
> > >
> > > I am using transformer isolated power supply with a series of "soda can" sized transformers to remove most of the ripple in the dc power source - 9 of these with 3500 or so uf each.
> > >
> > > A few questions.
> > >
> > > 1. For kflop, is it best to configure the drives solely for the current loop or is it better to use the velocity loop as well.
> > >
> > > 2. I noted that the back emf works best at higher voltages and unfortunately since this is a metal milling I will be machining at slower speeds so possibly the back emf is not getting accurate readings. The bandwidth of the drives is 20khz for current and 4khz for velocity. Can it be that the drives velocity loop back emf measurement conflict with kflop analog commands?
> > >
> > > 3. Are the Copley controls junus drives the proper type of drives or should I revert to the old servo dynamics with the tachometer. Link for junus drives is http://www.copleycontrols.com/motion/pdf/Junus.pdf
> > >
> > > 4. Alternatively, should I switch to dc drives with encoder feedback on the drives or is this not necessary?
> > >
> > > 5. What is the best way to proceed to get half decent performance from this system. I am looking to machine to 1 thousand precision. I was able to do this with the acroloop / servo dynamics drives before without issue. I am using kflop in my cnc router and know that kflop can handle this without problem so believe the drives or tuning are mostly the issue.
> > >
> > > Thank you
> > >
> > > Darren
> > >
> >
>