Closed Loop CNC

[mactec54]

Ha, well, high resolution encoders ... a bit mythical.

Only if you are uneducated on the subject

Should anyone have specs on encoders which do claim 24-bit ACCURACY, I would love to see them (and their price). Cheers Roger

I don't know of any Hobby users using the 24 Bit, there are a lot using 16 Bit and 20 Bit Encoder though

Go to any AC Servo Manufacture and you should be able to find the information you are looking for, when the Hobby builder can buy them and use them, the price you will find quite reasonable for what you are getting


Yaskawa
Mitsubishi
Siemens
Bosch
Fanuc

[samco]

Hmm - again with the dmm 16 bit encoder.. It is still only accurate to 4096 counts per rev (12bit)... Do you dispute their own documentation?

There isn't any magic there... BTW - mesa/linuxcnc will interface with ssi,biss (fanuc absolute encoders)

HOSTMOT2

sam

So - why do industrial controls like fanuc and such not use step/dir drives?

[samco]

(don't get me wrong - I use step/dir drives. Given the choice though - I would close the loop within the control)

sam

[Al_The_Man]

(don't get me wrong - I use step/dir drives. Given the choice though - I would close the loop within the control)

sam

Me too, never used steppers unless no other option.

Out of interest this is the resolution presently used by Fanuc on their Alpha motors since 2002.
Which is a mixture of absolute and the lower value incremental.
Prior to this, going back to 70's/80's they typically used 2000 pulse/rev incremental for the vast majority of applications.
Al.

[mactec54]

Hmm - again with the dmm 16 bit encoder.. It is still only accurate to 4096 counts per rev (12bit)... Do you dispute their own documentation?

There isn't any magic there... BTW - mesa/linuxcnc will interface with ssi,biss (fanuc absolute encoders)

HOSTMOT2

sam

So - why do industrial controls like fanuc and such not use step/dir drives?

I think you are looking at there old 14Bit Encoder not the new 16 Bit

[samco]

I think you are looking at there old 14Bit Encoder not the new 16 Bit

DMM | ABS-14-00 Absolute Encoder | AC SERVO DRIVE | AC SERVO MOTOR | ROTARY ENCODER

http://www.dmm-tech.com/Files/dmm_AB...chure_EN12.pdf

[mactec54]

Hmm - again with the dmm 16 bit encoder.. It is still only accurate to 4096 counts per rev (12bit)... Do you dispute their own documentation?

There isn't any magic there... BTW - mesa/linuxcnc will interface with ssi,biss (fanuc absolute encoders)

HOSTMOT2

sam

So - why do industrial controls like fanuc and such not use step/dir drives?

I don't think anyone said there was any magic with the control you are using, unless you are now getting delusional about it, I think most know that LinuxCNC is very versatile control

You should know that answer, Step / Dir is only supported for Position control, not many manufactures use Position Control, in an industrial setting, speed ( Velocity Torque ) control are better suited for CNC machine

[mactec54]

DMM | ABS-14-00 Absolute Encoder | AC SERVO DRIVE | AC SERVO MOTOR | ROTARY ENCODER

They say the same for the 14 Bit also, so not sure why, as I have run the 16Bit side by side in tests with Yaskawa 16Bit and it does not show any accuracy loss, both perform just as they should

[mactec54]

Linuxcnc can also do ethercat. (although i have no experience with that..)

https://forum.linuxcnc.org/24-hal-co...r?limitstart=0

sam

That's good if they can do Ethercat now, that would mean there are a lot of servo drives that could connect direct

[Al_The_Man]

I not many manufactures use Position Control, in an industrial setting, speed ( Velocity Torque ) control are better suited for CNC machine

I think now I have heard everything!!
Velocity/Torque speed control??
The older velocity drives were very much different from the current/torque mode.
I have used both extensively and know exactly how they react.

[RCaffin]

A lot of Kool-Aid sloshing around here. Yeah, myths.

With a 500 line or 2,000 count encoder on my system I am getting 0.8 microns resolution. And I have verified this with a touch probe.
Now, if you change that encoder for one with 1000 times the resolution, does that mean I would be getting 0.8 NANOmetres resolution? REALLY???? And to what point?

Somewhere along the line a little bit of engineering common sense needs to creep into this discussion.
First, what is the stability of the end bearings on the ball screw? 1 micron I can believe, but not 1 nanometre.
Next, what is the stability of the ball nut? Exactly the same comments apply.
Then, what is the flex in the system? I don't expect there to be much, below 1 micron for sure, but there will be some.

And then (roll of drums) what about temperature shifts and gradients? If you want to machine to the micron level you are going to have to temperature stabilise the entire machine and the room it is sitting in. Even so, local heating from the balls whizzing around the nut will introduce problems. Marginal temperature differences between the cast iron, hardened steel and aluminium parts will exist. What will you need at the nanometre level?

A problem I encountered on my machine was heating from the QI lighting! I had to replace the QIs with LEDs. How many of those older machines have LED lighting.

I suggest that we have strayed from verifiable engineering specifications into marketing one-up-manship. My system has 20 bit resolution. Well, mine has 24 bit resolution. Yada yada yada.

Incidentally, if you really want sub-micron accuracy, not only will you need serious temperature stabilisation of the room and the machine (embedded water cooling), you will also need to use laser interferometry on every axis. I think HP demonstrated this back in the 90s.

If your volume requirements are small, forget the ball screws completely. You will have to go to inchworms or similar.

An interesting perspective is to look at a modern semiconductor fabrication plant. They do have to position the masks to the few-nanometre level. One of those fabs will cost you at least one billion dollars, possibly more. (And the embedded processors they can spit out sell for 50 cents each...) Fun stuff.

Cheers
Roger

[RCaffin]

why do industrial controls like fanuc and such not use step/dir drives?
Because they are still wedded to their 30 and 40 year old technology?
Mind you, last I looked I thought Fanuc could handle Step/Dir, but I was not looking too closely. I might be wrong.

not many manufactures use Position Control, in an industrial setting, speed ( Velocity Torque ) control are better suited for CNC machine
Exactly the same problem: that is the technology they were using 40 years ago, they know it, and they don't want to change (yet). Change is both painful and expensive. My old VW Beetle was still using Kettering ignition in 2010 - same reason.

Cheers
Roger

[mactec54]

I think now I have heard everything!!
Velocity/Torque speed control??
The older velocity drives were very much different from the current/torque mode.
I have used both extensively and know exactly how they react.

May be you are so far behind, ( Older meaning 70's early 80's, for more than 20 years Yaskawa Ac servo Drives use Position Velocity or Torque control, all available in the same Servo Drive

[Al_The_Man]

Exactly I only use torque mode drives. They are not dated by any means.
±10vdc control command.

[mactec54]

Exactly I only use torque mode drives. They are not dated by any means.
±10vdc control command.

Here is a PDF worth reading that is more than 20 years old, servo Drives, you will see how Speed and Torque are used the same way as I wrote it, section 5

[Jim Dawson]

It seems to me that using torque control would be most appropriate for applications requiring maintaining a constant tension on a web material or something like that. The velocity control would seem the be most appropriate for positioning control, letting the drive apply the torque required to maintain minimum following error, like in a CNC machine application.

Is my thinking incorrect here?

[mactec54]

It seems to me that using torque control would be most appropriate for applications requiring maintaining a constant tension on a web material or something like that. The velocity control would seem the be most appropriate for positioning control, letting the drive apply the torque required to maintain minimum following error, like in a CNC machine application.

Is my thinking incorrect here?

No that is a correct way of thinking

[RCaffin]

I find it entertaining to compare the size and complexity of the older generation drive systems like the Yaskawa SGDM ones with a modern servo motor driver like a Gecko 320X unit. I suspect we would regard the older gen units as being outrageously expensive in comparison as well.

Cheers
Roger

[mactec54]

I find it entertaining to compare the size and complexity of the older generation drive systems like the Yaskawa SGDM ones with a modern servo motor driver like a Gecko 320X unit. I suspect we would regard the older gen units as being outrageously expensive in comparison as well.

Cheers
Roger

The Gecko does do not compare to a AC servo Drive, not even with the older ones, the PDF below will give you an insight of what you have to compare it with, this AC Servo Drive is very small and retail is only $138-00, and there is still not comparison

These can use Modbus as well

[samco]

The thing is - the magic in the closed loop drives - is the same magic in linuxcnc. It can control

step/dir (variants like - up/down, quadrature)
voltage
current
torque
velocity
smart serial
Ethercat
It can directly control 3 phase servos...
Others that I can't think of atm...

sam