[arvidj]

I'll start with what I think is a general question and depending on the answers may refine the questions as I learn more.

My limited research suggests that most external servo drives have some form of PID parameters and algorithms embedded within them. This suggests that:

1. An encoder will be attached to the drive
2. The KFlop treats the drive as an open loop Step\Direction system
3. The KFlop is not involved in the PID process

Assuming this is correct then some of the questions that come to mind are:

Is this a reasonable system? It seems like it would be wasting much of the KFlop's potential.

If "position" information were made available to the KFlop either via incremental encoder information or the output from linear scales, would the KFlop be able to use this information and enhance its control of the drive? Would one source of information be preferred over the other?

If the drive accepted +10\-10 volt analog speed control could that be leveraged in any way, replacing the Step/Direction connection?

Thanks,
Arvid

[Al_The_Man]

The only time I use a drive with a Dsp to close the PID loop is with systems such as Mach. With a controller that possesses a Dsp that controls all axis, then I use non-Dsp drives such as A-M-C in the torque mode with ±10vdc analogue control, the encoder comes back to the controller not the drive, this could be done with Kanalog.
Al.

[arvidj]

The only time I use a drive with a Dsp to close the PID loop is with systems such as Mach. With a controller that possesses a Dsp that controls all axis, then I use non-Dsp drives such as A-M-C in the torque mode with ±10vdc analogue control, the encoder comes back to the controller not the drive, this could be done with Kanalog.
Al.

Would you care to speculate what would happen if I were to use a drive with a DSP ... in this case drives and motors from DMM Technologies ... with the KFlop?

I would be able send step\direction to the drive and use or ignore the encoder feedback or send +10\-10 analog to the drive for speed control using the encoder feedback.

But the thought of a smart controller and a smart drive might be too much intelegence.

[TomKerekes]

Hi Arvigj,

Both ways (step/dir or analog) will work. I think to a large extent it depends on personal preference and what you are familiar with. Do you prefer to do servo tuning in the drive or using KFLOP? Of course we prefer KFLOP

There might be some advantage to closing the position loop in the drive if you purchase a pre-tuned motor-drive set. Although the tuning will change when you install it in your system.

There is a clear advantage to have the feedback connected back to the controller. This allows the controller to track position during a fault or amplifier disable. Without feedback after a fault the drive might still know the axis position but the controller won't. So you will need to always re-home.

With feedback to the controller, the controller will be aware of the current following error. It can then fault on excessive following error without the need of bringing extra status in from the drive.

You can close the loop in the drive and also bring feedback back to the controller but that adds more wiring and complexity.

There is a slight advantage with the analog approach in that it eliminates "steps". The measured encoder position will still be in "counts". But the commanded trajectory doesn't need to be quantized.

HTH
Regards

[arvidj]

Hi Arvigj,

Both ways (step/dir or analog) will work. I think to a large extent it depends on personal preference and what you are familiar with. Do you prefer to do servo tuning in the drive or using KFLOP? Of course we prefer KFLOP

There might be some advantage to closing the position loop in the drive if you purchase a pre-tuned motor-drive set. Although the tuning will change when you install it in your system.

There is a clear advantage to have the feedback connected back to the controller. This allows the controller to track position during a fault or amplifier disable. Without feedback after a fault the drive might still know the axis position but the controller won't. So you will need to always re-home.

With feedback to the controller, the controller will be aware of the current following error. It can then fault on excessive following error without the need of bringing extra status in from the drive.

You can close the loop in the drive and also bring feedback back to the controller but that adds more wiring and complexity.

There is a slight advantage with the analog approach in that it eliminates "steps". The measured encoder position will still be in "counts". But the commanded trajectory doesn't need to be quantized.

HTH
Regards

Tom,

Thanks for the feedback.

If I understand this correctly, I can tune the drive and motor using the facilities in the drive. I can then attach the drive to the KFlop using the analog speed mode and the encoder information from the drive.

I am hoping that in this configuration the KFlop sees the drive\motor combination as a "very well behaved motor" because the KFlop does not know that the drive is "smart". The characteristics of the combination as a whole is all that the KFlop sees.

Any additional tuning that might be required or beneficial might be addressed by the KFlop. Even if no additional tuning would be required the KFlop would be able to control this "very well behaved motor" via the analog speed setting.

Does this make sense or have I simply arrived at the "knows enough to be incredibly dangerious" phase of my understanding?

[TomKerekes]

Hi Arvigj,

That sounds reasonable to me. I usually recommend that approach. If the drive does perfect velocity control then simple proportional (P) gain is all that is needed in KFLOP for a stable system. Commanded velocity that is proportional to error (distance to target) results in a nice exponential approach to the target (slow down as you get closer).

Whereas commanded torque (acceleration) is a bit more complex to understand and tune. Simple proportional gain will usually result in an unstable system. It would continuously accelerate and not even begin to slow down until it passes the target. So derivative+proportional gain is required at a minimum to have a stable system. Taking a pure derivative of an encoder position can introduce other issues. Not that torque mode will not function well, but rather it is more complex to understand and tune.

Regards
TK