Step/Dir or Analog?

New user looking for some insight. Decided to retrofit my mill and after much reading I'm still unclear about a few things.

I decided to go with AC servos but don't know what control method to use. The drivers support both position (step/dir) and analog (velocity/torque). I'm planning on using the MESA 5i25 card and either the 7i76 or 7i77 depending. Is one more preferable than the other in terms of performance?

Also, my mill currently has a DRO with scales that output quadrature signals. Can LinuxCNC use these to any benefit?

Thanks

I would recommend to use the 7i77 as you can directly connect the encoder signals to linuxcnc and therefor display the actual machine position instead of the "should be position".

One other option is step/dir with encoder feedback.

If you drives can accept step/dir _and_ output quadrature
encoder signals, you can make a closed loop step/dir system.

With Mesa hardware, this can be done with a 5I25+7I85S combination
(4 encoder inputs, 4 step dir outputs +sserial expansion) with perhaps a 7I84 for misc I/O

Thanks for the input. I wavered for a moment & considered using steppers with my scales to implement a hybrid system. However, I'm sticking with the servos & decided to use some driver/motors from DMM (Dynamic Motor Motion).

The MESA 5i25/7i77 combo looks like the ticket using the analog/encoder interface. Also, there are plenty of encoder inputs to hook-up the scales with.

Are there any examples using LinuxCNC with both encoder/scale inputs. Could you toggle between the two or use both at the same time to improve accuracy? Just wondering if it's worth while hooking up the scales.

I can be wrong (never did it), but I think that the best way is:

Connect the encoders of the motors to the servo drives
Connect the glass scale encoders to the 7i77

This would perform best as the glass scales would measure real displacement (what matters most for accuracy), and the encoder motors will measure motor rotation (that are also needed to the servo drives internal PID or PIV), that due to the mechanics of the transmission can loose some real displacement (backlash for example).

The only problem, is often glass scales are not high enough resolution for good pid tuning by themselves. There seem to be a lot of topics on using linear scales with a servo set up. Most of the more successful ones I have read about use both a rotary encoder on the motor (for velocity feed back) and the linear scale for positioning.

This is only what I've read as well. I have 0 practical experience.

emcPT

This is an interesting point & worth considering. With the hardware I've selected I can swing either way so I'll probably experiment with both.

My scales (magnetic) have a 2 tenths (5 um) resolution and are good for 10m/s – should handle any speed I throw at it.

I think there might be a potential for more slop using the scales vs motor encoders. Using the scales includes the pulley drive, ball-screw backlash (both should be very small) and way stiction into the PID. My feeling is that PID's don't like things like this maybe necessitating a deadband. I have not yet gone through all what LinuxCNC offers so I don't know how flexible it will be.

I do have a little experience, although my testing used steppers in velocity mode instead of servos and a 5u linear scale. Not that much difference; counting steps also provides a 'virtual encoder'.

I used a PID loop with only FF1 and P paramaters on the 'motor encoder'. For the glass scale I used a second PID loop, with little P, a fairly slow I and a deadband of +/- 2 counts.

This works well unless you have backlash in your system. Then you ned to increase the deadband, which renders the scales useless (except for creating a screw map)


My new mill will use analog controlled AC servos and later on I will add the scales. But I will go for 1u scales; 5u is a bit low resolution though useable.

Yeah, now that I think about it the scale resolution is not going to be enough for motor control.
When I get my hardware setup it will be fun to experiment around & learn some more.

Thanks to everyone for being my sounding board..