Loading ...
Sorry, an error occurred while loading the content.

Servo Chatter

Expand Messages
  • Mark Vaughan
    Hi guy s I have a problem that is technically due to too low a gain bandwidth in my servo control and was wondering if anyone has experienced this before and
    Message 1 of 11 , Sep 1, 2007
    • 0 Attachment
      Hi guy's

      I have a problem that is technically due to too low a gain bandwidth in my
      servo control and was wondering if anyone has experienced this before and
      what solutions you tried.

      I have a large mill being retrofitted with mach, it uses large 4.5Nm SEM
      servo's and many of the DIY servo controllers have been unsuitable for it. I
      have gone back to three new large analogue servo drives from a reputable
      commercial equipment manufacturer. I am using Pixie cards to drive them,
      these are standard PID control algorithm cards, no feed forward, no velocity
      mode just basic PID.

      The Y axis (the heaviest) and Z axis (the lightest) are running perfectly.

      They are set up in the normal manner, adjust P until it starts to go
      critical with a little decaying ringing, raise D to stop it, and I to
      correct any error. This will give me a good step response, however during a
      movement I have high frequency chatter on the servo position.

      This can sometimes happen due to a loose ball screw, but the ball nut is a
      double preloaded nut, and the thrust race is now correctly preloaded, there
      is no measurable play on the bed screws with a 0.1 thou DTI, the slightest
      manual turn of the 5tpi screw results in a change on the DTI so it isn't
      this.

      Think of it that the motor is far too responsive to the command and
      overshoots substantially before the pixie PID control can see it, resulting
      in a commanded move the shuffles the screw forwards and backwards at hi
      frequency during the move.

      I can reduce the gain and get rid of the high frequency portion, but then it
      is over damped and doesn't follow up very well.

      I can reduce the gain and damping, i.e. set it up a lot less aggressive, but
      acceleration is so poor the machine is unusable, the Y axis will do
      1500mm/sec/sec without problem, X is down to 100mm/sec/sec without the HF
      chatter.



      It is possible the motor has lost part of it's magnetism so it will actually
      have a higher target rpm/volt than it should, but I don't think this is the
      issue, I am sure it's a control system issue.

      There are techniques to apply filters between a PID servo control, and servo
      but these can also affect the transient response. (Remember them from
      college days, never ever needed them)

      There are professional ways to correct the issue, adding feed forward can
      help, dropping the I term during input commands another, but with a pixie we
      don't have access to this. It is also why professional machines work in
      velocity mode, a PIV drive doesn't have overshoot in the same way as a PID
      drive and isn't supposed to suffer from these sorts of problems, again in
      the DIY field this isn't an option.

      The other two axis are belt drive and it is possible the belt gives the
      needed damping, the X being direct on the screw. It is also curious when I
      commented to XYZ a few days ago why they belt drive their X when there is no
      need, they responded that it helps the damping! Perhaps I should get a
      flexible rubber coupling?



      Has anyone come across this, and did you find any solution or have a
      recommendation please, before I delve in deeply?



      Regs Mark





      Dr. Mark Vaughan Ph'D. B.Eng. M0VAU

      Managing Director

      Vaughan Industries Ltd, reg in UK no 2561068

      Water Care Technology Ltd, reg in UK no 4129351

      Addr Unit3, Sydney House, Blackwater, Truro, Cornwall, TR4 8HH, UK.
      Phone/Fax 44 1872 561288

      RSGB DRM111(Cornwall)





      [Non-text portions of this message have been removed]
    • James Connerton
      Hi Mark, I think I would try switching components one at a time between the axis with the problem and a good axis to eliminate any faulty components. If you
      Message 2 of 11 , Sep 1, 2007
      • 0 Attachment
        Hi Mark,

        I think I would try switching components one at a time between the axis with the problem and a good axis to eliminate any faulty components. If you could mount all the good Y axis components including the signal path to the troublesome X axis, you will know if it is a mechanical resonance issue or a problem with the X axis screw or motor. Could that screw be bent? Could an encoder be flaky, or the signals being sent from your controller? I know it's a pain but I think you need to start swapping components to verify their condition.
        Please do keep us informed as to your progress.
        Jim C.

        Mark Vaughan <mark@...> wrote:
        Hi guy's

        I have a problem that is technically due to too low a gain bandwidth in my
        servo control and was wondering if anyone has experienced this before and
        what solutions you tried.

        I have a large mill being retrofitted with mach, it uses large 4.5Nm SEM
        servo's and many of the DIY servo controllers have been unsuitable for it. I
        have gone back to three new large analogue servo drives from a reputable
        commercial equipment manufacturer. I am using Pixie cards to drive them,
        these are standard PID control algorithm cards, no feed forward, no velocity
        mode just basic PID.

        The Y axis (the heaviest) and Z axis (the lightest) are running perfectly.

        They are set up in the normal manner, adjust P until it starts to go
        critical with a little decaying ringing, raise D to stop it, and I to
        correct any error. This will give me a good step response, however during a
        movement I have high frequency chatter on the servo position.

        This can sometimes happen due to a loose ball screw, but the ball nut is a
        double preloaded nut, and the thrust race is now correctly preloaded, there
        is no measurable play on the bed screws with a 0.1 thou DTI, the slightest
        manual turn of the 5tpi screw results in a change on the DTI so it isn't
        this.

        Think of it that the motor is far too responsive to the command and
        overshoots substantially before the pixie PID control can see it, resulting
        in a commanded move the shuffles the screw forwards and backwards at hi
        frequency during the move.

        I can reduce the gain and get rid of the high frequency portion, but then it
        is over damped and doesn't follow up very well.

        I can reduce the gain and damping, i.e. set it up a lot less aggressive, but
        acceleration is so poor the machine is unusable, the Y axis will do
        1500mm/sec/sec without problem, X is down to 100mm/sec/sec without the HF
        chatter.

        It is possible the motor has lost part of it's magnetism so it will actually
        have a higher target rpm/volt than it should, but I don't think this is the
        issue, I am sure it's a control system issue.

        There are techniques to apply filters between a PID servo control, and servo
        but these can also affect the transient response. (Remember them from
        college days, never ever needed them)

        There are professional ways to correct the issue, adding feed forward can
        help, dropping the I term during input commands another, but with a pixie we
        don't have access to this. It is also why professional machines work in
        velocity mode, a PIV drive doesn't have overshoot in the same way as a PID
        drive and isn't supposed to suffer from these sorts of problems, again in
        the DIY field this isn't an option.

        The other two axis are belt drive and it is possible the belt gives the
        needed damping, the X being direct on the screw. It is also curious when I
        commented to XYZ a few days ago why they belt drive their X when there is no
        need, they responded that it helps the damping! Perhaps I should get a
        flexible rubber coupling?

        Has anyone come across this, and did you find any solution or have a
        recommendation please, before I delve in deeply?

        Regs Mark

        Dr. Mark Vaughan Ph'D. B.Eng. M0VAU

        Managing Director

        Vaughan Industries Ltd, reg in UK no 2561068

        Water Care Technology Ltd, reg in UK no 4129351

        Addr Unit3, Sydney House, Blackwater, Truro, Cornwall, TR4 8HH, UK.
        Phone/Fax 44 1872 561288

        RSGB DRM111(Cornwall)

        [Non-text portions of this message have been removed]






        ---------------------------------
        Boardwalk for $500? In 2007? Ha!
        Play Monopoly Here and Now (it's updated for today's economy) at Yahoo! Games.

        [Non-text portions of this message have been removed]
      • art
        Mark: Cant say Im any expert in this one, but I like th eidea of the rubber damper or at least the split anti-resonance coupler, it may do wonders for that
        Message 3 of 11 , Sep 1, 2007
        • 0 Attachment
          Mark:

          Cant say Im any expert in this one, but I like th eidea of the rubber damper or at least the split anti-resonance coupler, it may do wonders for that kind of dithering , perhaps it will speed up the dither by reducing the torque required for the dithering, and thus increase its frequency and reduce its amplitude to be acceptable..

          Thanks,
          Art
          www.artofcnc.ca

          Videos And Support Forums http://www.machsupport.com
          Users Map: http://www.frappr.com/machsupport
          ----- Original Message -----
          From: Mark Vaughan
          To: mach1mach2cnc@yahoogroups.com
          Sent: Saturday, September 01, 2007 5:25 AM
          Subject: [mach1mach2cnc] Servo Chatter


          Hi guy's

          I have a problem that is technically due to too low a gain bandwidth in my
          servo control and was wondering if anyone has experienced this before and
          what solutions you tried.

          I have a large mill being retrofitted with mach, it uses large 4.5Nm SEM
          servo's and many of the DIY servo controllers have been unsuitable for it. I
          have gone back to three new large analogue servo drives from a reputable
          commercial equipment manufacturer. I am using Pixie cards to drive them,
          these are standard PID control algorithm cards, no feed forward, no velocity
          mode just basic PID.

          The Y axis (the heaviest) and Z axis (the lightest) are running perfectly.

          They are set up in the normal manner, adjust P until it starts to go
          critical with a little decaying ringing, raise D to stop it, and I to
          correct any error. This will give me a good step response, however during a
          movement I have high frequency chatter on the servo position.

          This can sometimes happen due to a loose ball screw, but the ball nut is a
          double preloaded nut, and the thrust race is now correctly preloaded, there
          is no measurable play on the bed screws with a 0.1 thou DTI, the slightest
          manual turn of the 5tpi screw results in a change on the DTI so it isn't
          this.

          Think of it that the motor is far too responsive to the command and
          overshoots substantially before the pixie PID control can see it, resulting
          in a commanded move the shuffles the screw forwards and backwards at hi
          frequency during the move.

          I can reduce the gain and get rid of the high frequency portion, but then it
          is over damped and doesn't follow up very well.

          I can reduce the gain and damping, i.e. set it up a lot less aggressive, but
          acceleration is so poor the machine is unusable, the Y axis will do
          1500mm/sec/sec without problem, X is down to 100mm/sec/sec without the HF
          chatter.

          It is possible the motor has lost part of it's magnetism so it will actually
          have a higher target rpm/volt than it should, but I don't think this is the
          issue, I am sure it's a control system issue.

          There are techniques to apply filters between a PID servo control, and servo
          but these can also affect the transient response. (Remember them from
          college days, never ever needed them)

          There are professional ways to correct the issue, adding feed forward can
          help, dropping the I term during input commands another, but with a pixie we
          don't have access to this. It is also why professional machines work in
          velocity mode, a PIV drive doesn't have overshoot in the same way as a PID
          drive and isn't supposed to suffer from these sorts of problems, again in
          the DIY field this isn't an option.

          The other two axis are belt drive and it is possible the belt gives the
          needed damping, the X being direct on the screw. It is also curious when I
          commented to XYZ a few days ago why they belt drive their X when there is no
          need, they responded that it helps the damping! Perhaps I should get a
          flexible rubber coupling?

          Has anyone come across this, and did you find any solution or have a
          recommendation please, before I delve in deeply?

          Regs Mark

          Dr. Mark Vaughan Ph'D. B.Eng. M0VAU

          Managing Director

          Vaughan Industries Ltd, reg in UK no 2561068

          Water Care Technology Ltd, reg in UK no 4129351

          Addr Unit3, Sydney House, Blackwater, Truro, Cornwall, TR4 8HH, UK.
          Phone/Fax 44 1872 561288

          RSGB DRM111(Cornwall)

          [Non-text portions of this message have been removed]






          [Non-text portions of this message have been removed]
        • Mark Vaughan
          Thanks for your thoughts Jim, Done all that. Encoders are new, expensive £200 IVO, 2000 count differential encoders, signals taken as differential back to
          Message 4 of 11 , Sep 1, 2007
          • 0 Attachment
            Thanks for your thoughts Jim, Done all that.

            Encoders are new, expensive £200 IVO, 2000 count differential encoders,
            signals taken as differential back to RS422 decoder that sits immediately
            beneath pixies. Encoder signals are beautifully clean.

            Have swapped pixies, and analogue servo cards.

            Have renewed all wiring between motor and controls, all cables screened both
            encoder and motor.

            Lead screw is straight, about 1 inch diameter, has no backlash at the ball
            nut which is a double nut preloaded to remove backlash, and not end float
            now I have properly set the bearings. It isn’t a cheap screw (no pun
            intended)

            I haven’t swapped the motor and I will try that with one of the other axis
            just in case, but strangely a Rutex card will run this axis OK as long as
            you don’t push it too hard when it pops fets so I don’t think it’s a demag
            motor. I suppose the other test I could do is measure motor RPM with
            constant voltage applied, if it has been demagged it will run faster than it
            should.



            I have done various tests and it appears to be a control system problem, the
            pixie claims a control loop time of 10KHz I think, and the resonance is less
            than this, I presently doubt their 10KHz figure which probably doesn’t allow
            for nyquist sampling problems.



            I also tried reducing the dead band in the analogue control, in case that
            was generating the overshoot. Another possibility is to make the PID control
            signal non linear, but this often leads to other resonance issue that need
            mathematical modelling to eradicate, and we don’t really have enough data to
            put into the equation. If a rubber drive doesn’t cure it, I think I will
            have to filter the control signal, and hope that doesn’t effect the impulse
            response too much.



            Thanks for your thought though, this is one of those issues when something
            brain stormed that seems totally stupid might actually provide a solution.



            Regs Mark







            Dr. Mark Vaughan Ph'D. B.Eng. M0VAU

            Managing Director

            Vaughan Industries Ltd, reg in UK no 2561068

            Water Care Technology Ltd, reg in UK no 4129351

            Addr Unit3, Sydney House, Blackwater, Truro, Cornwall, TR4 8HH, UK.
            Phone/Fax 44 1872 561288

            RSGB DRM111(Cornwall)

            _____

            From: mach1mach2cnc@yahoogroups.com [mailto:mach1mach2cnc@yahoogroups.com]
            On Behalf Of James Connerton
            Sent: 01 September 2007 13:15
            To: mach1mach2cnc@yahoogroups.com
            Subject: Re: [mach1mach2cnc] Servo Chatter



            Hi Mark,

            I think I would try switching components one at a time between the axis with
            the problem and a good axis to eliminate any faulty components. If you could
            mount all the good Y axis components including the signal path to the
            troublesome X axis, you will know if it is a mechanical resonance issue or a
            problem with the X axis screw or motor. Could that screw be bent? Could an
            encoder be flaky, or the signals being sent from your controller? I know
            it's a pain but I think you need to start swapping components to verify
            their condition.
            Please do keep us informed as to your progress.
            Jim C.

            Mark Vaughan <mark@.... <mailto:mark%40vil.uk.com> com> wrote:
            Hi guy's

            I have a problem that is technically due to too low a gain bandwidth in my
            servo control and was wondering if anyone has experienced this before and
            what solutions you tried.

            I have a large mill being retrofitted with mach, it uses large 4.5Nm SEM
            servo's and many of the DIY servo controllers have been unsuitable for it. I
            have gone back to three new large analogue servo drives from a reputable
            commercial equipment manufacturer. I am using Pixie cards to drive them,
            these are standard PID control algorithm cards, no feed forward, no velocity
            mode just basic PID.

            The Y axis (the heaviest) and Z axis (the lightest) are running perfectly.

            They are set up in the normal manner, adjust P until it starts to go
            critical with a little decaying ringing, raise D to stop it, and I to
            correct any error. This will give me a good step response, however during a
            movement I have high frequency chatter on the servo position.

            This can sometimes happen due to a loose ball screw, but the ball nut is a
            double preloaded nut, and the thrust race is now correctly preloaded, there
            is no measurable play on the bed screws with a 0.1 thou DTI, the slightest
            manual turn of the 5tpi screw results in a change on the DTI so it isn't
            this.

            Think of it that the motor is far too responsive to the command and
            overshoots substantially before the pixie PID control can see it, resulting
            in a commanded move the shuffles the screw forwards and backwards at hi
            frequency during the move.

            I can reduce the gain and get rid of the high frequency portion, but then it
            is over damped and doesn't follow up very well.

            I can reduce the gain and damping, i.e. set it up a lot less aggressive, but
            acceleration is so poor the machine is unusable, the Y axis will do
            1500mm/sec/sec without problem, X is down to 100mm/sec/sec without the HF
            chatter.

            It is possible the motor has lost part of it's magnetism so it will actually
            have a higher target rpm/volt than it should, but I don't think this is the
            issue, I am sure it's a control system issue.

            There are techniques to apply filters between a PID servo control, and servo
            but these can also affect the transient response. (Remember them from
            college days, never ever needed them)

            There are professional ways to correct the issue, adding feed forward can
            help, dropping the I term during input commands another, but with a pixie we
            don't have access to this. It is also why professional machines work in
            velocity mode, a PIV drive doesn't have overshoot in the same way as a PID
            drive and isn't supposed to suffer from these sorts of problems, again in
            the DIY field this isn't an option.

            The other two axis are belt drive and it is possible the belt gives the
            needed damping, the X being direct on the screw. It is also curious when I
            commented to XYZ a few days ago why they belt drive their X when there is no
            need, they responded that it helps the damping! Perhaps I should get a
            flexible rubber coupling?

            Has anyone come across this, and did you find any solution or have a
            recommendation please, before I delve in deeply?

            Regs Mark

            Dr. Mark Vaughan Ph'D. B.Eng. M0VAU

            Managing Director

            Vaughan Industries Ltd, reg in UK no 2561068

            Water Care Technology Ltd, reg in UK no 4129351

            Addr Unit3, Sydney House, Blackwater, Truro, Cornwall, TR4 8HH, UK.
            Phone/Fax 44 1872 561288

            RSGB DRM111(Cornwall)

            [Non-text portions of this message have been removed]

            ---------------------------------
            Boardwalk for $500? In 2007? Ha!
            Play Monopoly Here and Now (it's updated for today's economy) at Yahoo!
            Games.

            [Non-text portions of this message have been removed]





            [Non-text portions of this message have been removed]
          • Mark Vaughan
            Thanks Art It s funny really when we normally strive to keep things as rigid as possible, that the opposite might be better (or worse) Regs Mark Dr. Mark
            Message 5 of 11 , Sep 1, 2007
            • 0 Attachment
              Thanks Art

              It's funny really when we normally strive to keep things as rigid as
              possible, that the opposite might be better (or worse)



              Regs Mark



              Dr. Mark Vaughan Ph'D. B.Eng. M0VAU

              Managing Director

              Vaughan Industries Ltd, reg in UK no 2561068

              Water Care Technology Ltd, reg in UK no 4129351

              Addr Unit3, Sydney House, Blackwater, Truro, Cornwall, TR4 8HH, UK.
              Phone/Fax 44 1872 561288

              RSGB DRM111(Cornwall)

              _____

              From: mach1mach2cnc@yahoogroups.com [mailto:mach1mach2cnc@yahoogroups.com]
              On Behalf Of art
              Sent: 01 September 2007 15:03
              To: mach1mach2cnc@yahoogroups.com
              Subject: Re: [mach1mach2cnc] Servo Chatter



              Mark:

              Cant say Im any expert in this one, but I like th eidea of the rubber damper
              or at least the split anti-resonance coupler, it may do wonders for that
              kind of dithering , perhaps it will speed up the dither by reducing the
              torque required for the dithering, and thus increase its frequency and
              reduce its amplitude to be acceptable..

              Thanks,
              Art
              www.artofcnc.ca

              Videos And Support Forums http://www.machsupp <http://www.machsupport.com>
              ort.com
              Users Map: http://www.frappr <http://www.frappr.com/machsupport>
              com/machsupport
              ----- Original Message -----
              From: Mark Vaughan
              To: mach1mach2cnc@ <mailto:mach1mach2cnc%40yahoogroups.com> yahoogroups.com
              Sent: Saturday, September 01, 2007 5:25 AM
              Subject: [mach1mach2cnc] Servo Chatter

              Hi guy's

              I have a problem that is technically due to too low a gain bandwidth in my
              servo control and was wondering if anyone has experienced this before and
              what solutions you tried.

              I have a large mill being retrofitted with mach, it uses large 4.5Nm SEM
              servo's and many of the DIY servo controllers have been unsuitable for it. I
              have gone back to three new large analogue servo drives from a reputable
              commercial equipment manufacturer. I am using Pixie cards to drive them,
              these are standard PID control algorithm cards, no feed forward, no velocity
              mode just basic PID.

              The Y axis (the heaviest) and Z axis (the lightest) are running perfectly.

              They are set up in the normal manner, adjust P until it starts to go
              critical with a little decaying ringing, raise D to stop it, and I to
              correct any error. This will give me a good step response, however during a
              movement I have high frequency chatter on the servo position.

              This can sometimes happen due to a loose ball screw, but the ball nut is a
              double preloaded nut, and the thrust race is now correctly preloaded, there
              is no measurable play on the bed screws with a 0.1 thou DTI, the slightest
              manual turn of the 5tpi screw results in a change on the DTI so it isn't
              this.

              Think of it that the motor is far too responsive to the command and
              overshoots substantially before the pixie PID control can see it, resulting
              in a commanded move the shuffles the screw forwards and backwards at hi
              frequency during the move.

              I can reduce the gain and get rid of the high frequency portion, but then it
              is over damped and doesn't follow up very well.

              I can reduce the gain and damping, i.e. set it up a lot less aggressive, but
              acceleration is so poor the machine is unusable, the Y axis will do
              1500mm/sec/sec without problem, X is down to 100mm/sec/sec without the HF
              chatter.

              It is possible the motor has lost part of it's magnetism so it will actually
              have a higher target rpm/volt than it should, but I don't think this is the
              issue, I am sure it's a control system issue.

              There are techniques to apply filters between a PID servo control, and servo
              but these can also affect the transient response. (Remember them from
              college days, never ever needed them)

              There are professional ways to correct the issue, adding feed forward can
              help, dropping the I term during input commands another, but with a pixie we
              don't have access to this. It is also why professional machines work in
              velocity mode, a PIV drive doesn't have overshoot in the same way as a PID
              drive and isn't supposed to suffer from these sorts of problems, again in
              the DIY field this isn't an option.

              The other two axis are belt drive and it is possible the belt gives the
              needed damping, the X being direct on the screw. It is also curious when I
              commented to XYZ a few days ago why they belt drive their X when there is no
              need, they responded that it helps the damping! Perhaps I should get a
              flexible rubber coupling?

              Has anyone come across this, and did you find any solution or have a
              recommendation please, before I delve in deeply?

              Regs Mark

              Dr. Mark Vaughan Ph'D. B.Eng. M0VAU

              Managing Director

              Vaughan Industries Ltd, reg in UK no 2561068

              Water Care Technology Ltd, reg in UK no 4129351

              Addr Unit3, Sydney House, Blackwater, Truro, Cornwall, TR4 8HH, UK.
              Phone/Fax 44 1872 561288

              RSGB DRM111(Cornwall)

              [Non-text portions of this message have been removed]

              [Non-text portions of this message have been removed]





              [Non-text portions of this message have been removed]
            • Eirik Karlsen
              Well... It could be that this axis has a particular mass/inertia that the drive can t handle. It could be that the problem disappears if you introduce some
              Message 6 of 11 , Sep 1, 2007
              • 0 Attachment
                Well... It could be that this axis has a particular mass/inertia that
                the drive can't handle.
                It could be that the problem disappears if you introduce some play in
                the system, e.g.. in the thrust bearing...
                Will obviously not solve any problems but could point you in the right
                direction.

                I have had hunting problems in servo systems that just could not be
                tuned out... changing the inertia
                up or down a little bit solved the problem.

                Mark Vaughan wrote:

                > Thanks Art
                >
                > It's funny really when we normally strive to keep things as rigid as
                > possible, that the opposite might be better (or worse)
                >
                > Regs Mark
                >
                > Dr. Mark Vaughan Ph'D. B.Eng. M0VAU
                >
                > Managing Director
                >
                > Vaughan Industries Ltd, reg in UK no 2561068
                >
                > Water Care Technology Ltd, reg in UK no 4129351
                >
                > Addr Unit3, Sydney House, Blackwater, Truro, Cornwall, TR4 8HH, UK.
                > Phone/Fax 44 1872 561288
                >
                > RSGB DRM111(Cornwall)
                >
                --
                *******************************************
                VISIT MY HOME PAGE:
                <http://home.online.no/~eikarlse/index.htm>
                LAST UPDATED: 23/08/2003
                *******************************************
                Regards
                Eirik Karlsen



                [Non-text portions of this message have been removed]
              • Mark Vaughan
                Yes that s what I am thinking Eirik, it is very much like motor inertia verses bed inertia resonance problems, however I ve only ever read about them, or done
                Message 7 of 11 , Sep 1, 2007
                • 0 Attachment
                  Yes that's what I am thinking Eirik, it is very much like motor inertia
                  verses bed inertia resonance problems, however I've only ever read about
                  them, or done the plots, never actually experienced them.

                  The strange thing is normally you set up the inertia balance between bed and
                  motor, and commutation bounce to ensure resonant frequencies are above 500
                  to 600Hz where it doesn't usually become a problem The resonance I get is
                  couple of KHz like gain bandwidth issues, but it shouldn't be a gain
                  bandwidth problem since the pixie claims a 10KHz loop time, hence it might
                  be a software loop sampling issue or something of that nature. I think
                  there's something slightly a miss in the pixie maths routine, the integral
                  parameter behaves slightly odd compared with other PID systems.



                  The other two axis with belt drive work well, the belt would provide some
                  rigidity, possibly more than a rigid coupling, but would also provide some
                  damping.

                  I'm presently thinking of removing the rigid coupling between motor and
                  screw, and putting a spider coupling with polyurethane connecting web in to
                  see if that damps it, it would theoretically make commutation resonance
                  worse, but may help inertial resonance. It's a cheap direct fit solution and
                  worth a try, if not I am going to have to add some sort of notch filter at
                  the resonant frequency, most text books say add a low pass filter but that
                  would effect the transient response, however I think a notch filter would be
                  OK, and the resonance whilst varying in amplitude is always the same
                  frequency (for now).

                  It's a shame really we don't have any slightly more advanced step/dir
                  control cards to the pixie. Removing the D term and using a PI + feed
                  forward command structure would probably eliminate the problem, or having a
                  PIV control rather than torque PID control which would give a much better
                  response anyhow.



                  Thanks for your thoughts Eirik



                  Regs Mark.



                  Dr. Mark Vaughan Ph'D. B.Eng. M0VAU

                  Managing Director

                  Vaughan Industries Ltd, reg in UK no 2561068

                  Water Care Technology Ltd, reg in UK no 4129351

                  Addr Unit3, Sydney House, Blackwater, Truro, Cornwall, TR4 8HH, UK.
                  Phone/Fax 44 1872 561288

                  RSGB DRM111(Cornwall)

                  _____

                  From: mach1mach2cnc@yahoogroups.com [mailto:mach1mach2cnc@yahoogroups.com]
                  On Behalf Of Eirik Karlsen
                  Sent: 01 September 2007 19:57
                  To: mach1mach2cnc@yahoogroups.com
                  Subject: Re: [mach1mach2cnc] Servo Chatter



                  Well... It could be that this axis has a particular mass/inertia that
                  the drive can't handle.
                  It could be that the problem disappears if you introduce some play in
                  the system, e.g.. in the thrust bearing...
                  Will obviously not solve any problems but could point you in the right
                  direction.

                  I have had hunting problems in servo systems that just could not be
                  tuned out... changing the inertia
                  up or down a little bit solved the problem.

                  Mark Vaughan wrote:

                  > Thanks Art
                  >
                  > It's funny really when we normally strive to keep things as rigid as
                  > possible, that the opposite might be better (or worse)
                  >
                  > Regs Mark
                  >
                  > Dr. Mark Vaughan Ph'D. B.Eng. M0VAU
                  >
                  > Managing Director
                  >
                  > Vaughan Industries Ltd, reg in UK no 2561068
                  >
                  > Water Care Technology Ltd, reg in UK no 4129351
                  >
                  > Addr Unit3, Sydney House, Blackwater, Truro, Cornwall, TR4 8HH, UK.
                  > Phone/Fax 44 1872 561288
                  >
                  > RSGB DRM111(Cornwall)
                  >
                  --
                  *******************************************
                  VISIT MY HOME PAGE:
                  <http://home. <http://home.online.no/~eikarlse/index.htm>
                  online.no/~eikarlse/index.htm>
                  LAST UPDATED: 23/08/2003
                  *******************************************
                  Regards
                  Eirik Karlsen

                  [Non-text portions of this message have been removed]





                  [Non-text portions of this message have been removed]
                • Mariss Freimanis
                  A symptom of an over-constrained system is it exhibits doubling behavior (near 100% PWM for 1 cycle, near 0% PWM for the next). See if you can scope out the
                  Message 8 of 11 , Sep 1, 2007
                  • 0 Attachment
                    A symptom of an over-constrained system is it exhibits "doubling"
                    behavior (near 100% PWM for 1 cycle, near 0% PWM for the next). See if
                    you can 'scope out the offending frequency and find out if it's
                    phase-locked to the loop update rate. Feed-forward compensation is the
                    usual standard cure. Just a thought.

                    Mariss



                    --- In mach1mach2cnc@yahoogroups.com, "Mark Vaughan" <mark@...> wrote:
                    >
                    > Yes that's what I am thinking Eirik, it is very much like motor inertia
                    > verses bed inertia resonance problems, however I've only ever read about
                    > them, or done the plots, never actually experienced them.
                    >
                    > The strange thing is normally you set up the inertia balance between
                    bed and
                    > motor, and commutation bounce to ensure resonant frequencies are
                    above 500
                    > to 600Hz where it doesn't usually become a problem The resonance I
                    get is
                    > couple of KHz like gain bandwidth issues, but it shouldn't be a gain
                    > bandwidth problem since the pixie claims a 10KHz loop time, hence it
                    might
                    > be a software loop sampling issue or something of that nature. I think
                    > there's something slightly a miss in the pixie maths routine, the
                    integral
                    > parameter behaves slightly odd compared with other PID systems.
                    >
                    >
                    >
                    > The other two axis with belt drive work well, the belt would provide
                    some
                    > rigidity, possibly more than a rigid coupling, but would also
                    provide some
                    > damping.
                    >
                    > I'm presently thinking of removing the rigid coupling between motor and
                    > screw, and putting a spider coupling with polyurethane connecting
                    web in to
                    > see if that damps it, it would theoretically make commutation resonance
                    > worse, but may help inertial resonance. It's a cheap direct fit
                    solution and
                    > worth a try, if not I am going to have to add some sort of notch
                    filter at
                    > the resonant frequency, most text books say add a low pass filter
                    but that
                    > would effect the transient response, however I think a notch filter
                    would be
                    > OK, and the resonance whilst varying in amplitude is always the same
                    > frequency (for now).
                    >
                    > It's a shame really we don't have any slightly more advanced step/dir
                    > control cards to the pixie. Removing the D term and using a PI + feed
                    > forward command structure would probably eliminate the problem, or
                    having a
                    > PIV control rather than torque PID control which would give a much
                    better
                    > response anyhow.
                    >
                    >
                    >
                    > Thanks for your thoughts Eirik
                    >
                    >
                    >
                    > Regs Mark.
                    >
                    >
                    >
                    > Dr. Mark Vaughan Ph'D. B.Eng. M0VAU
                    >
                    > Managing Director
                    >
                    > Vaughan Industries Ltd, reg in UK no 2561068
                    >
                    > Water Care Technology Ltd, reg in UK no 4129351
                    >
                    > Addr Unit3, Sydney House, Blackwater, Truro, Cornwall, TR4 8HH, UK.
                    > Phone/Fax 44 1872 561288
                    >
                    > RSGB DRM111(Cornwall)
                    >
                    > _____
                    >
                    > From: mach1mach2cnc@yahoogroups.com
                    [mailto:mach1mach2cnc@yahoogroups.com]
                    > On Behalf Of Eirik Karlsen
                    > Sent: 01 September 2007 19:57
                    > To: mach1mach2cnc@yahoogroups.com
                    > Subject: Re: [mach1mach2cnc] Servo Chatter
                    >
                    >
                    >
                    > Well... It could be that this axis has a particular mass/inertia that
                    > the drive can't handle.
                    > It could be that the problem disappears if you introduce some play in
                    > the system, e.g.. in the thrust bearing...
                    > Will obviously not solve any problems but could point you in the right
                    > direction.
                    >
                    > I have had hunting problems in servo systems that just could not be
                    > tuned out... changing the inertia
                    > up or down a little bit solved the problem.
                    >
                    > Mark Vaughan wrote:
                    >
                    > > Thanks Art
                    > >
                    > > It's funny really when we normally strive to keep things as rigid as
                    > > possible, that the opposite might be better (or worse)
                    > >
                    > > Regs Mark
                    > >
                    > > Dr. Mark Vaughan Ph'D. B.Eng. M0VAU
                    > >
                    > > Managing Director
                    > >
                    > > Vaughan Industries Ltd, reg in UK no 2561068
                    > >
                    > > Water Care Technology Ltd, reg in UK no 4129351
                    > >
                    > > Addr Unit3, Sydney House, Blackwater, Truro, Cornwall, TR4 8HH, UK.
                    > > Phone/Fax 44 1872 561288
                    > >
                    > > RSGB DRM111(Cornwall)
                    > >
                    > --
                    > *******************************************
                    > VISIT MY HOME PAGE:
                    > <http://home. <http://home.online.no/~eikarlse/index.htm>
                    > online.no/~eikarlse/index.htm>
                    > LAST UPDATED: 23/08/2003
                    > *******************************************
                    > Regards
                    > Eirik Karlsen
                    >
                    > [Non-text portions of this message have been removed]
                    >
                    >
                    >
                    >
                    >
                    > [Non-text portions of this message have been removed]
                    >
                  • Mark Vaughan
                    Thanks Mariss, I thought you d have some theories. Without scoping it to access the frequency exactly, that does appear to be the effect, the PID output
                    Message 9 of 11 , Sep 2, 2007
                    • 0 Attachment
                      Thanks Mariss, I thought you'd have some theories.

                      Without scoping it to access the frequency exactly, that does appear to be
                      the effect, the PID output analogue command is pushing hard on, hard off at
                      this resonant frequency and the more I try to constrain the impulse response
                      the higher the amplitude of oscillation, but the axis isn't really
                      constrained and the mechanical system isn't really loose. The impulse
                      response where this occurs is quite sloppy with two to three times the
                      settling time of other axis than run OK. I cannot constrain this axis enough
                      to even manage 1/10 of the acceleration of the others, which is as expected
                      with gain values many times smaller than the other axis.

                      The oscillation is also not just in the PID output command, but occurs in
                      the encoder position, so the system is in mechanical resonance at this
                      point.



                      Whilst feed forward is the text book cure, we don't have this luxury with a
                      pixie card, just simple PID.

                      Simply thinking about it, one would think by making the system stiffer, this
                      overshoot at the resonant frequency could not occur, so the control could
                      not oscillate, but resonance like this is usually down to commutation
                      problems, wind up of the lead screw even when it's an inch diameter, such
                      resonance though is also effected by lead screw length and that is not
                      happening.

                      A notch filter is looking like the best compensation I can add, short of
                      designing my own PID + feed forward control.



                      Thanks for your thoughts, any more would be appreciated.



                      Regs Mark





                      Dr. Mark Vaughan Ph'D. B.Eng. M0VAU

                      Managing Director

                      Vaughan Industries Ltd, reg in UK no 2561068

                      Water Care Technology Ltd, reg in UK no 4129351

                      Addr Unit3, Sydney House, Blackwater, Truro, Cornwall, TR4 8HH, UK.
                      Phone/Fax 44 1872 561288

                      RSGB DRM111(Cornwall)

                      _____

                      From: mach1mach2cnc@yahoogroups.com [mailto:mach1mach2cnc@yahoogroups.com]
                      On Behalf Of Mariss Freimanis
                      Sent: 01 September 2007 22:35
                      To: mach1mach2cnc@yahoogroups.com
                      Subject: [mach1mach2cnc] Re: Servo Chatter



                      A symptom of an over-constrained system is it exhibits "doubling"
                      behavior (near 100% PWM for 1 cycle, near 0% PWM for the next). See if
                      you can 'scope out the offending frequency and find out if it's
                      phase-locked to the loop update rate. Feed-forward compensation is the
                      usual standard cure. Just a thought.

                      Mariss

                      --- In mach1mach2cnc@ <mailto:mach1mach2cnc%40yahoogroups.com>
                      yahoogroups.com, "Mark Vaughan" <mark@...> wrote:
                      >
                      > Yes that's what I am thinking Eirik, it is very much like motor inertia
                      > verses bed inertia resonance problems, however I've only ever read about
                      > them, or done the plots, never actually experienced them.
                      >
                      > The strange thing is normally you set up the inertia balance between
                      bed and
                      > motor, and commutation bounce to ensure resonant frequencies are
                      above 500
                      > to 600Hz where it doesn't usually become a problem The resonance I
                      get is
                      > couple of KHz like gain bandwidth issues, but it shouldn't be a gain
                      > bandwidth problem since the pixie claims a 10KHz loop time, hence it
                      might
                      > be a software loop sampling issue or something of that nature. I think
                      > there's something slightly a miss in the pixie maths routine, the
                      integral
                      > parameter behaves slightly odd compared with other PID systems.
                      >
                      >
                      >
                      > The other two axis with belt drive work well, the belt would provide
                      some
                      > rigidity, possibly more than a rigid coupling, but would also
                      provide some
                      > damping.
                      >
                      > I'm presently thinking of removing the rigid coupling between motor and
                      > screw, and putting a spider coupling with polyurethane connecting
                      web in to
                      > see if that damps it, it would theoretically make commutation resonance
                      > worse, but may help inertial resonance. It's a cheap direct fit
                      solution and
                      > worth a try, if not I am going to have to add some sort of notch
                      filter at
                      > the resonant frequency, most text books say add a low pass filter
                      but that
                      > would effect the transient response, however I think a notch filter
                      would be
                      > OK, and the resonance whilst varying in amplitude is always the same
                      > frequency (for now).
                      >
                      > It's a shame really we don't have any slightly more advanced step/dir
                      > control cards to the pixie. Removing the D term and using a PI + feed
                      > forward command structure would probably eliminate the problem, or
                      having a
                      > PIV control rather than torque PID control which would give a much
                      better
                      > response anyhow.
                      >
                      >
                      >
                      > Thanks for your thoughts Eirik
                      >
                      >
                      >
                      > Regs Mark.
                      >
                      >
                      >
                      > Dr. Mark Vaughan Ph'D. B.Eng. M0VAU
                      >
                      > Managing Director
                      >
                      > Vaughan Industries Ltd, reg in UK no 2561068
                      >
                      > Water Care Technology Ltd, reg in UK no 4129351
                      >
                      > Addr Unit3, Sydney House, Blackwater, Truro, Cornwall, TR4 8HH, UK.
                      > Phone/Fax 44 1872 561288
                      >
                      > RSGB DRM111(Cornwall)
                      >
                      > _____
                      >
                      > From: mach1mach2cnc@ <mailto:mach1mach2cnc%40yahoogroups.com>
                      yahoogroups.com
                      [mailto:mach1mach2cnc@ <mailto:mach1mach2cnc%40yahoogroups.com>
                      yahoogroups.com]
                      > On Behalf Of Eirik Karlsen
                      > Sent: 01 September 2007 19:57
                      > To: mach1mach2cnc@ <mailto:mach1mach2cnc%40yahoogroups.com>
                      yahoogroups.com
                      > Subject: Re: [mach1mach2cnc] Servo Chatter
                      >
                      >
                      >
                      > Well... It could be that this axis has a particular mass/inertia that
                      > the drive can't handle.
                      > It could be that the problem disappears if you introduce some play in
                      > the system, e.g.. in the thrust bearing...
                      > Will obviously not solve any problems but could point you in the right
                      > direction.
                      >
                      > I have had hunting problems in servo systems that just could not be
                      > tuned out... changing the inertia
                      > up or down a little bit solved the problem.
                      >
                      > Mark Vaughan wrote:
                      >
                      > > Thanks Art
                      > >
                      > > It's funny really when we normally strive to keep things as rigid as
                      > > possible, that the opposite might be better (or worse)
                      > >
                      > > Regs Mark
                      > >
                      > > Dr. Mark Vaughan Ph'D. B.Eng. M0VAU
                      > >
                      > > Managing Director
                      > >
                      > > Vaughan Industries Ltd, reg in UK no 2561068
                      > >
                      > > Water Care Technology Ltd, reg in UK no 4129351
                      > >
                      > > Addr Unit3, Sydney House, Blackwater, Truro, Cornwall, TR4 8HH, UK.
                      > > Phone/Fax 44 1872 561288
                      > >
                      > > RSGB DRM111(Cornwall)
                      > >
                      > --
                      > *******************************************
                      > VISIT MY HOME PAGE:
                      > <http://home. <http://home. <http://home.online.no/~eikarlse/index.htm>
                      online.no/~eikarlse/index.htm>
                      > online.no/~eikarlse/index.htm>
                      > LAST UPDATED: 23/08/2003
                      > *******************************************
                      > Regards
                      > Eirik Karlsen
                      >
                      > [Non-text portions of this message have been removed]
                      >
                      >
                      >
                      >
                      >
                      > [Non-text portions of this message have been removed]
                      >





                      [Non-text portions of this message have been removed]
                    • Lee K. White
                      Mark, Thanks for the update. I now have my pixies but no time to do the upgrade just yet. My other machine went down on Saturday, it is not Mach, so I cannot
                      Message 10 of 11 , Sep 5, 2007
                      • 0 Attachment
                        Mark,

                        Thanks for the update.

                        I now have my pixies but no time to do the upgrade just yet.

                        My other machine went down on Saturday, it is not Mach, so I cannot fix it.

                        Ouch. It's gunna run me 5000 CDN to get it running.

                        Keep me posted.


                        Thanks.

                        Lee K. White
                        Premier Pattern Equipment Ltd
                        9725 - 60 Ave
                        Edmonton. AB
                        T6E 0C4
                        Ph: 780-430-1549
                        Fx: 780-430-1526
                        ----- Original Message -----
                        From: Mark Vaughan
                        To: mach1mach2cnc@yahoogroups.com
                        Sent: Wednesday, September 05, 2007 11:55 AM
                        Subject: RE: [mach1mach2cnc] Re: RE: Re: Servo Chatter


                        Hi Guys

                        Just thought I'd let anyone that is following this know how we are going.

                        I have now measured the frequency of the vibration which I thought was
                        around several KHz, but turns out to be at 380Hz, still well above the
                        control loop resonance which is only at a few Hz.

                        The parts arrive today to make up a notch filter, I have used a twin T,
                        Salen key activated style notch filter at 380Hz which relies on the
                        principle of adding in phase at out of phase components at the notch
                        frequency. As always with these when put into practice the notch does have
                        the infinite attenuation one would want at the notch frequency, but the
                        notch is pretty good and reduces the gain by a significant amount. The phase
                        shift is also pretty low.

                        I fitted the filter up between the pixie and analogue card, and it works
                        very well. I can almost treble the gain before any resonance occurs, and I
                        can get a transient response that is pretty good.

                        I would like however to stiffen and constrain the system some more so am
                        going to increase the number of stages in the notch filter and see if we can
                        get some more gain into the system, there will come a time when to many
                        stages will give too great a phase delay and the transient response will
                        collapse, but I think we can get a little better.

                        I'll let you know how we get on, but if anyone is experiencing a hi
                        frequency audible noise during servo motor moves, which I know is rare,
                        especially on small low inertia machines, I can recommend this easy to build
                        filter approach.

                        Regs Mark

                        Dr. Mark Vaughan Ph'D. B.Eng. M0VAU

                        Managing Director

                        Vaughan Industries Ltd, reg in UK no 2561068

                        Water Care Technology Ltd, reg in UK no 4129351

                        Addr Unit3, Sydney House, Blackwater, Truro, Cornwall, TR4 8HH, UK.
                        Phone/Fax 44 1872 561288

                        RSGB DRM111(Cornwall)

                        [Non-text portions of this message have been removed]





                        [Non-text portions of this message have been removed]
                      • Mark Vaughan
                        Ouch Lee Your not the only one that has experienced this sort of thing though, I hope the machinery will be worth it in the end. I have discussed the resonant
                        Message 11 of 11 , Sep 5, 2007
                        • 0 Attachment
                          Ouch Lee



                          Your not the only one that has experienced this sort of thing though, I hope
                          the machinery will be worth it in the end.



                          I have discussed the resonant frequency issue with Rick of Pixie, the only
                          thing so far I can see that a professional PID card has is many have an
                          inbuilt variable frequency, variable width filter.

                          Rick is going to see if there is room to add one to the code, if successful
                          we can all have it as a simple firmware upgrade.



                          Regs Mark



                          Dr. Mark Vaughan Ph'D. B.Eng. M0VAU

                          Managing Director

                          Vaughan Industries Ltd, reg in UK no 2561068

                          Water Care Technology Ltd, reg in UK no 4129351

                          Addr Unit3, Sydney House, Blackwater, Truro, Cornwall, TR4 8HH, UK.
                          Phone/Fax 44 1872 561288

                          RSGB DRM111(Cornwall)

                          _____

                          From: mach1mach2cnc@yahoogroups.com [mailto:mach1mach2cnc@yahoogroups.com]
                          On Behalf Of Lee K. White
                          Sent: 05 September 2007 19:28
                          To: mach1mach2cnc@yahoogroups.com
                          Subject: Re: [mach1mach2cnc] Re: RE: Re: Servo Chatter



                          Mark,

                          Thanks for the update.

                          I now have my pixies but no time to do the upgrade just yet.

                          My other machine went down on Saturday, it is not Mach, so I cannot fix it.

                          Ouch. It's gunna run me 5000 CDN to get it running.

                          Keep me posted.

                          Thanks.

                          Lee K. White
                          Premier Pattern Equipment Ltd
                          9725 - 60 Ave
                          Edmonton. AB
                          T6E 0C4
                          Ph: 780-430-1549
                          Fx: 780-430-1526
                          ----- Original Message -----
                          From: Mark Vaughan
                          To: mach1mach2cnc@ <mailto:mach1mach2cnc%40yahoogroups.com> yahoogroups.com
                          Sent: Wednesday, September 05, 2007 11:55 AM
                          Subject: RE: [mach1mach2cnc] Re: RE: Re: Servo Chatter

                          Hi Guys

                          Just thought I'd let anyone that is following this know how we are going.

                          I have now measured the frequency of the vibration which I thought was
                          around several KHz, but turns out to be at 380Hz, still well above the
                          control loop resonance which is only at a few Hz.

                          The parts arrive today to make up a notch filter, I have used a twin T,
                          Salen key activated style notch filter at 380Hz which relies on the
                          principle of adding in phase at out of phase components at the notch
                          frequency. As always with these when put into practice the notch does have
                          the infinite attenuation one would want at the notch frequency, but the
                          notch is pretty good and reduces the gain by a significant amount. The phase
                          shift is also pretty low.

                          I fitted the filter up between the pixie and analogue card, and it works
                          very well. I can almost treble the gain before any resonance occurs, and I
                          can get a transient response that is pretty good.

                          I would like however to stiffen and constrain the system some more so am
                          going to increase the number of stages in the notch filter and see if we can
                          get some more gain into the system, there will come a time when to many
                          stages will give too great a phase delay and the transient response will
                          collapse, but I think we can get a little better.

                          I'll let you know how we get on, but if anyone is experiencing a hi
                          frequency audible noise during servo motor moves, which I know is rare,
                          especially on small low inertia machines, I can recommend this easy to build
                          filter approach.

                          Regs Mark

                          Dr. Mark Vaughan Ph'D. B.Eng. M0VAU

                          Managing Director

                          Vaughan Industries Ltd, reg in UK no 2561068

                          Water Care Technology Ltd, reg in UK no 4129351

                          Addr Unit3, Sydney House, Blackwater, Truro, Cornwall, TR4 8HH, UK.
                          Phone/Fax 44 1872 561288

                          RSGB DRM111(Cornwall)

                          [Non-text portions of this message have been removed]

                          [Non-text portions of this message have been removed]





                          [Non-text portions of this message have been removed]
                        Your message has been successfully submitted and would be delivered to recipients shortly.