Does CNC use a DRO

[AdrianRParticipant @adrianr18614]

[AdrianRParticipant @adrianr18614]

Just a curiosity question.

I have seen several articles about DIY CNC and they basically seem to be add stepper motors and count the steps to get the position. IE it is an open loop system.

I am wondering if any CNC machines use the position info from a DRO (or its linear sensors) to make a closed loop system. IE run each axis as a linear servo.

Adrian

[John HaineParticipant @johnhaine32865]

There are certainly "stepper servo" motors that have some electronics to count steps and work out how much to turn, then have a DC servo motor (or possibly a stepper) with an encoder, driving the motor to force the error to zero. They're a lot more expensive than standard steppers and drivers. These have the benefit of "knowing" where they should be so won't miss steps which normal steppers can do. I believe that high grade CNC will also/instead have axis encoders (like DROs) and do what you suggest but in this case the controller and drives are more integrated into the machine to get the control system dynamics optimised. DIY CNC tends to use steppers 'cos they are cheap and work well enough. Or, better, use a stepper servo which is guaranteed not to miss steps and can be treated just like a stepper. Using the DROs in a feedback loop is probably a step beyond most DIY conversions, and isn't I think supported by the usual run of CNC controllers like Mach3/4 or Linux CNC (and others).

Just found this which indicates that linear encoders are used in high grade achining centres.

Edited By John Haine on 18/04/2020 09:50:42

[Anonymous]

Most hobby stepper motor CNC mills and low end professional systems, even with servo motors and encoders, are open loop. So they rely on the inherent accuracy (or otherwise) of the ballscrews. High end professional CNC mills may well have independent measurement systems as an option and would therefore be closed loop.

Missed steps keep getting mentioned, but in over 10 years of CNC milling I've never knowingly missed steps; and I run my CNC mill fairly hard.

Andrew

[geoff adamsParticipant @geoffadams14047]

many years ago i programmed and set a Moog cnc mill the axis were positioned by hydralics and Heidenhain glass scales it worked well till i manged to rattle the z scale with a 20 mm carbide slot drill and it lost were was and took a dive through the job into the vice in rapid before i hit the o sh.. button

[Mike PooleParticipant @mikepoole82104]

Robots, which at the end of the day are a CNC machine usually have resolvers built into the motor so to some extent they are open loop as the end effector position is only guaranteed if the drive train is in good working order. The servo system only guarantees the motor has turned. Stepper drives in their basic form are totally open loop so in theory they can be stalled and lose position. A well specified drive should not lose steps in normal operation but a crash may cause a problem, but then rereferencing will be the least of your problems. Some steppers have an encoder built in to check the motor is in position but I think it just alarms if steps are lost. It is important to keep the machine in good mechanical order if steppers are used, any backlash in the drive train will not be compensated for but in a full feedback system backlash will cause problems like oscillation so an automatic machine must be setup properly and well maintained. Choosing the right motor is important to avoid problems.

Mike

[John HaineParticipant @johnhaine32865]

I agree with Andrew, I've never had a problem traceable to missed steps unless I've driven an axis into the end stop.

"Some steppers have an encoder built in to check the motor is in position but I think it just alarms if steps are lost."

The one I linked to above is fully closed loop though does use a stepper motor with encoder.

[Ex contributorParticipant @mgnbuk]

High end professional CNC mills may well have independent measurement systems as an option and would therefore be closed loop.

Not just high end – pretty well all industrial machines have been closed loop for the last 40+ years.

I was initially employed as a Technician Apprentice in 1977 specifically to learn to maintain the first two NC/CNC machines recently purchased by the company – the lathe was CNC (GE 1050) & the vertical machining centre NC (Plessey NC1l00). Both were closed loop systems, with DC servo motors & resolver feedback.

Moving forward, from 1983 I spent 27 years retrofitting machines. The only open loop, stepper motor systems, I encountered (i.e. replaced) were Posidata & some Bridgeport Boss systems on small turret mills. Posidata controls were typically used by Matchmaker (Shizuoka carcasses). Acton (Anayak carcasses) & Beaver.

Everyything else was closed loop, with either motor mounted encoder or scales. Both types of feedback were described as "closed loop" by most control manufacturers – the exception being Fanuc who refered to motor mounted encoders as "semi-closed loop" and scales as "closed loop". Some Fanuc controls could operate with both motor mounted encoders and scales simultanously, which IIRC they called a "hybrid" system. Only read about that in the manuals, though – didn't get to fit such a system.

Closed loop stepper motors have an encoder that connects to the drive, but it is only the drive system that is "closed loop" – the controls still run on step/direction pulses with no positional feedback, relying on the drive system to ensure that the motor moves by the number of steps sent out & stopping if the drive error output activates if the drive cannot keep up. So not "closed loop" at all in the sense that an industrial control is.

Nigel B.

[AdrianRParticipant @adrianr18614]

Thanks, that answered something that was bothering me.

[Anonymous]

Posted by mgnbuk on 18/04/2020 10:53:29:

Not just high end – pretty well all industrial machines have been closed loop for the last 40+ years.

In the context of the OP I wouldn't regard systems having an encoder or resolver on the motor as positional closed loop. They still rely on the accuracy of the ballscrews for position accuracy, even if the motor itself is closed loop control. My comment was based on having an independent method of measuring table position, and using that in a closed loop control.

Andrew

[blowlampParticipant @blowlamp]

Do these linear scales have the equivalent to an index pulse on a rotary encoder?

I thinking about fitting hybrid steppers & drivers to my milling machine and wonder if the rotary encoder could be disconnected and its function replaced by a linear scale to achieve a closed-loop system that actually relates to table movement – anyone think it worth persuing?

Martin.

[John HaineParticipant @johnhaine32865]

I'm not quite sure what you're suggesting, Martin. Would this be for CNC? The problem is that all the easily available controllers think in terms of steps out and don't accommodate positional feedback. Youd be on your own or needing a much more expensive controller.

[blowlampParticipant @blowlamp]

Posted by John Haine on 18/04/2020 14:40:12:

I'm not quite sure what you're suggesting, Martin. Would this be for CNC? The problem is that all the easily available controllers think in terms of steps out and don't accommodate positional feedback. Youd be on your own or needing a much more expensive controller.

John.

A hybrid stepper motor is fitted with a rotary encoder which sends its position back to its driver, but not the controller, so creating a feedback loop between motor & driver. It means we can be sure of motor position, but we can't be sure of table position.

I wanted to see if it were feasible to use a linear scale in place of the motor's encoder (1000 line) to feed back table position to the stepper driver (not controller).

To answer my own question:

Looking at the encoder wiring of a CL57T hybrid stepper driver, it seems that no index pulse is required, just the quadrature signals are necessary, which might mean it's something worth looking at further.

Martin.

[Mike PooleParticipant @mikepoole82104]

Haven’t heard about Posidata in a long time, we had six installed on an under frame welding machine to control the motion of the motorised weld guns, it was not a success, probably because weld guns can stick to the panel and this either stalls the drive or pulls the weld cap off which usually results in a flood of water even though these days flow and pressure sensors should detect this and turn the water off. Gemdrives replaced the Posidata controls and were far better. Time flies, that was nearly 40 years ago!

Mike

Edited By Mike Poole on 18/04/2020 19:01:03

[John HaineParticipant @johnhaine32865]

Martin, I was thinking further on this – I guess what you are thinking of is taking quadrature signals from the linear encoder back to the driver rather than from the rotary one on the motor shaft? Problem could be backlash – if you had been feeding in one direction to a given position, when you reversed the table wouldn't move until the play in the screw and nut was taken up. I suspect the controllers don't have backlash compensation since it isn't a problem with the encoder directly fitted to the shaft. You can compensate for backlash but it's quite tricky. I have it on my mill which uses ballscrews with just 20 microns or so of play, but I've tried it on my CNC lathe with a conventional leadscrew with ~0.25 mm and it doesn't work at all well.

[sam sokolikParticipant @samsokolik60334]

Posted by John Haine on 18/04/2020 14:40:12:

I'm not quite sure what you're suggesting, Martin. Would this be for CNC? The problem is that all the easily available controllers think in terms of steps out and don't accommodate positional feedback. Youd be on your own or needing a much more expensive controller.

Not quite… Linuxcnc is just as happy running open loop step/dir as it is running closed loop. You can even mix and max. (step/dir with feedback lets say.)

Crap explanation..

[John HaineParticipant @johnhaine32865]

Right! Thanks for that, I just assumed it was open loop. Interesting!

[Another JohnSParticipant @anotherjohns]

Posted by John Haine on 20/04/2020 13:58:39:

Right! Thanks for that, I just assumed it was open loop. Interesting!

John; this is one of these interesting conundrums:

a) LinuxCNC is light-years ahead of Mach3;

b) Mach3 does well enough for most current home workshops;

c) if Mach3 is good enough, there's no incentive to update, but you're stuck with essentially dead software.

What to do?

I most certainly understand people sticking with Mach3, but I decided to try LinuxCNC and do enjoy the utter reliability of it. I've currently got 3 machines running LinuxCNC, may be 4 soon if I get off my duff. I mean, these machines are tools, like comparing a Warco or Boxford to an Emco Maximat or something; all would do a fine job in my workshop, given my skill-set.

The late John Stevenson and I exchanged lots of email, and in-person discussions over pints about LinuxCNC and it's hard-to-configure compared to Mach3 setups, and he was of course correct; for me Linux is easy, as I've been developing software on it for over 2 decades; for me Windows is rarely used and seems awkward. That's just the way life is.

By the way, I picked up a KX1 over here (Canada) years ago, without stepper controllers, and John Stevenson and I had a real discussion on it, as he said "impossible". He called the factory in China to ask what's up with mine. It ends up, it was one of a batch for internal use in China, that somehow leaked out – he did not know this, as the KX1 was his baby.

[Author]

Posts