On Wed, 2010-08-11 at 20:02 -0500, Jon Elson wrote: > James Louis wrote: > > Hello everyone. I had this on the forum for a while: This question goes > > out to any "servo loop" experts out there in EMC2 world. > > I understand that EMC2 uses PID to form a control loop when using > > servomotors, but . . . > > what if you use a drive that already has its own loop? For example, my > > machine uses Yaskawa SGM-0xA31x series servos, Yaskawa SGDA-0xAS series > > analog drives, and a Pico Systems PPMC (ENC, DAC, DIO) interface. > > > > These drives receive the encoder signals and redirect them to the PPMC ENC > > card, but have their own PID terms which are set during Autotuning. So is > > it loop or loops? > > > > What effect do the PID terms in the EMC2 *.ini file have? I am using just > > the P term right now and it works. The motors move OK, but I would like to > > know HOW. Like what is an optimal value for the P term? Is HALscope the > > right tool for this? If I use values in the other terms will this form a > > second nested loop? Is that better? What about other INI entries like > > DEFAULT_ACCELERATION, etc? > > > Yes, lots of people use a variety of servo amps with EMC2's PID. In > fact, this is the classic scheme. The amps I use on my Bridgeport have > P and D, but no I. But, it is more complicated, as they have an inner > torque loop (with its own P and D) and an outer velocity loop, with ITS > own P and D! If the amp has gain, then you need less gain in EMC's PID > to get tight control. I suspect your Yaskawa is much the same. > > Assuming the Yaskawa amp performs perfect control, then all you need in > EMC is P, so that the velocity command to the amp exactly matches the > voltage it needs to provide that velocity. The I and D terms > accommodate less than perfect control of velocity and help keep the > error smaller, > > You can't put servo loops in a nutshell, it is a very big topic, with > many ramifications. There are resonances all over the place, finite > gain and dynamic stability concerns. For instance, on a machine with a > leadscrew, the resonant frequency of the torsional moment of the screw > varies as the nut runs down the screw from one end to the other, so you > can't just damp out one resonant frequency! > > Jon
I agree that this is a complex subject. In my setup my drives can be configured to be controlled in current or in velocity. I'm using the current configuration. There is a constant that relates analog output volts to intensity in the windings, that more or less relates trough the motor constant to torque. The current configurations uses a control loop with tunable PID gains. Also, if used, the velocity control is itself a tunable PID control loop, that internally uses the current PID loop. Current loop works at 8KHz, while switching frequency of the power electronics at 20KHz. I have also a figure of the inductance of the motor and the characteristic time for settling a given current is really low. This (although not exact) allows me to assume that the current loop is almost instantaneous, and to develop a model of my system based on that. Obviously if I try to impose a given velocity due to inertia, my machine will take not a so negligible effect to reach the same velocity. That is important, because whatever the controller tuning, internal or emc provided, that delay is unavoidable. So in my opinion, if following an analytical approach to tune the emc PID, that is to use a model of your machine, the model of your plant is different is using current or velocity configurations in the drive. Ideally the motor + drive should be modelled and model parameters identified doing some sort of experiment. I think this experiment is doable within EMC (processing using external tools like Scilab or Octave). I HAVE PROBLEM. Even if I use automatic configuration of the PID for current loop within the drive, when I switch on my machine, (emc off or pid not enabled), the machine vibrates, and the noise is audible. In my opinion the noise is due to the current PID tunning within the drive. I would like to remove that noise, because I think it is somehow responsible of some nasty artifacts (nonlinear vibrations). In fact I tried - not scientifically - to lower the gain and the noise lower, so it seems that there is some room for improvement, but I don't want to play in that arena without some criteria to tune it. Can somebody give some guidelines or point to some references on this?. Thanks in advance, Javier ------------------------------------------------------------------------------ This SF.net email is sponsored by Make an app they can't live without Enter the BlackBerry Developer Challenge http://p.sf.net/sfu/RIM-dev2dev _______________________________________________ Emc-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/emc-users
