[SCzEngrgGroup]

You don't lose steps with a servo if you aren't using drivers that utilize steps! Sorry i don't use Mach. I use LinuxCNC with a PID loop taken direct off of an encoder. I CAN'T lose steps, because I don't USE steps. I'm not messing with a stone axe this time.

OK..... Then you also don't understand how Mach3, EMC, and digital PID loops work. Whether you're using Mach3 or EMC, you ARE doing digital PID, the only difference being whether it's done on the PC or in an external drivers. Just because you don't see a discrete STEP signal coming out of the PC does not mean there isn't one internally. If you're using ANY kind of digital encoder, you ARE using steps, with the step size dictated by the encoder resolution! If you're using an analog system (like the old 0-10V true analog systems) with Mach3 or EMC, you're STILL using steps, with the step size dictated by the D/A converter resolution. ANY digital PID system will still use steps, and will still have a following error limit, which, if exceeded (by, for instance, accelerating too quickly, or running into a hard stop), WILL result in position loss. There is NO way around this.

You really should do a lot more research on how this stuff REALLY works, because much of what you think you know is simply not true.

Regards,
Ray L.

[wizard]

You don't lose steps with a servo if you aren't using drivers that utilize steps! Sorry i don't use Mach. I use LinuxCNC with a PID loop taken direct off of an encoder. I CAN'T lose steps, because I don't USE steps. I'm not messing with a stone axe this time.

At best you would have a system using some sort of encoder feedback that the controller maintains as a count someplace in the system. Upon a crash and assuming the counter can maintain counts, when that happens you might be able to recover your position. Even with relatively solid CNC systems though you will often loose position for any number of mechanical or electrical reasons. Frankly after a crash you are only somewhat certain that everything is in position when you start back up, an improvement over steppers most likely but you can still loose position.

It is also a mistake to say you can't loose steps because you don't use them. Steps are counts and as such they can be lost, miscounted or otherwise corrupted. The difference is encoder steps are counted on the return from the process. This does make encoder based feed back more robust, as long as the encoder is good, but it isn't infallible. In this regard at work we always rehome servo based systems after a crash.

[Al_The_Man]

This is where the advantage of a single central controller that calculates and performs the motion trajectory of all axis and control of the PID loop and can recover or retain the axis position in the worst of conditions.
With this controller, lets say an axis comes up against a stop while performing a move preventing it from completion, with a torque mode drive, the motor current will increase rapidly with the increasing following error, ultimately either the controller or the drive itself will sense this and take the necessary steps to turn the drive off and in most cases perform an E-stop before motor damage occurs.
Lets say it requires manual intervention of manually rotating the axis motor shaft/ballscrew etc to take it off the inadvertent stop and allow a reset of the E-stop circuit and enabling the drive.
Because the motor feedback is returned to the motion controller, the manual movement is constantly recorded and the motor position immediately updated to its new position when the drive is re-enabled therefore preventing any motor 'jump' or sense of error, or even the need for re-referencing the axis zero.
Al.

[samco]

I guess it depends on what you mean by position loss.. With linuxcnc setup as a true closed loop controller - and tuned with the following error set sanely.. If you push the system past the point that it can correct and the following error goes above its set limit - the machine will estop. Now - linuxcnc still knows where it is by the encoder feedback. So - if you correct the problem and take linuxcnc out of estop - you can take back off without re-homing. (you could even move the axis around manually without loosing position)

Steppers work. For me though - for the extra cost (which really isn't that much anymore) I would go servos. The steppers torque rating is at 0 rpm. It drops from there. A servo is going to have a pretty flat torque through its full rpm range. Plus servos usually have a peak rating allowing for 'reserve' torque for a short period of time. For some situations - this my get you out of a jam.

sam

OK..... Then you also don't understand how Mach3, EMC, and digital PID loops work. Whether you're using Mach3 or EMC, you ARE doing digital PID, the only difference being whether it's done on the PC or in an external drivers. Just because you don't see a discrete STEP signal coming out of the PC does not mean there isn't one internally. If you're using ANY kind of digital encoder, you ARE using steps, with the step size dictated by the encoder resolution! If you're using an analog system (like the old 0-10V true analog systems) with Mach3 or EMC, you're STILL using steps, with the step size dictated by the D/A converter resolution. ANY digital PID system will still use steps, and will still have a following error limit, which, if exceeded (by, for instance, accelerating too quickly, or running into a hard stop), WILL result in position loss. There is NO way around this.

You really should do a lot more research on how this stuff REALLY works, because much of what you think you know is simply not true.

Regards,
Ray L.

[MrWild]

Stepper systems have one fatal flaw. They are open loop. BUT, BUT, BUT.... "if the system is designed properly, it will not have a problem and work quite well." No, it will work as long as everything is perfect. What happens if you adjust the gibs too tight? In a servo system, the closed loop saves you. What happens if your cutter gets dull? What happens if you're a newbee and sure as newbees make mistakes they get the feed wrong? They might not get a catastrophic failure, and with an open loop system they won't know until they check the work. A good stepper system will work great almost always. A good servo system will be even more fool proof.

[wizard]

Stepper systems have one fatal flaw. They are open loop. BUT, BUT, BUT.... "if the system is designed properly, it will not have a problem and work quite well."

Yes happens every day on an endless number of devices.

No, it will work as long as everything is perfect. What happens if you adjust the gibs too tight? In a servo system, the closed loop saves you.

No it doesn't. Given equal power both a stepper based system and a servo based system will fail when overloaded.

I think most of these old wives tales come from guys that upgrade their CNC systems with servos and when doing so substantially upgrade the power of those servos.

This isn't even debatable, I've enough experience here with precision machinery that I know exactly what happens when the gibbs, ways or other parts of the machine are not working properly. The machine will faile to work properly even if. You have state of the art servos installed. You may or may not loose position but you will fail to cut your parts either way.

What happens if your cutter gets dull? What happens if you're a newbee and sure as newbees make mistakes they get the feed wrong? They might not get a catastrophic failure, and with an open loop system they won't know until they check the work.

So you are saying excessive following error won't show up in a part? Especially in the context of the type of CNC controllers talked about here.

A good stepper system will work great almost always. A good servo system will be even more fool proof.

I'm not one to dismiss the advantages of servos! What gets me going is the ideas expressed here that imply that servos can make up for mechanical issues. For the most part they can't. Sure you might get some capability to overcome a certain amount of friction but a servo won't solve the following types of issues:

1. Ways that have not been undercut causing issues over the length of travel.
2. Leadscrew thrust bearings going bad.
3. Gibbs too tight.
4. Worn ways with a dip in them causing binding at the extremes of travel.
5. Lubrication starvation.

There are others but these are off the top of my head.

It isn't that servos don't have advantages, rather it is the idea that they can somehow solve mechanical problems in a way that steppers can't that I object to. Given two systems with the same problem the stepper and servo system may fail differently but they both will fail given the same performance levels.

[Hans_G]

I'm a servo guy. I have years of experience with many motion control systems thanks to my day job.

For made in China hobby mills and lathes, servos are (for the most part) lipstick on pig.

That said, I bought BLDC motors, AMC drives on ebay, and a PCI control card on ebay.... and made a good closed loop system for my mill. Does it make my little mill a bridgeport? Not even close.

Advantages: style points. And to a much lesser degree IMHO... if there is a motion control issue, the system will let me know with *hopefully* minimal damage to the stock. With an open loop system you learn a problem exists in QA (measuring after part is done), w/a closed loop system you learn as soon as it occurs.

Closed loop motion control is a neat topic to learn about and has many applications that can lead to lucrative careers in all sorts of automation endeavors.

Disadvantages: for me, none. It was cheap. 5 axis control cost me about $300 for motors and drives and got the control card for free, BUT, I knew what I was looking for and knew what to avoid (connectors and cables can kill motion control costs!).

[Al_The_Man]

I think most of these old wives tales come from guys that upgrade their CNC systems with servos and when doing so substantially upgrade the power of those servos.

.

There is also a myth that servo's are best used with gearing and steppers suited to direct drive systems.
The fact is that most motors, servo's and steppers alike possess a maximum torque value at zero rpm.
However, although NEMA size does not indicate a certain torque value, steppers are generally higher torque for a given or equal frame size.
So when a system is upgraded to servo, it is either necessary to use a physically larger motor to achieve the same torque value or use gearing if the same frame size is used.
Al.

[SCzEngrgGroup]

There is also a myth that servo's are best used with gearing and steppers suited to direct drive systems.
The fact is that most motors, servo's and steppers alike possess a maximum torque value at zero rpm.
However, although NEMA size does not indicate a certain torque value, steppers are generally higher torque for a given or equal frame size.
So when a system is upgraded to servo, it is either necessary to use a physically larger motor to achieve the same torque value or use gearing if the same frame size is used.
Al.

Servos are also virtually always capable of much higher RPM than steppers (2-3X is typical), which necessitates the use of gearing to make use of the full dynamic range of the motor. Gearing both reduces max RPM to a more usable value, AND increases torque. That combination allows, for a given application, a servo motor to be roughly the same frame size as the comparable stepper motor. You certainly CAN design a servo system that is direct drive, but, except for machines MUCH larger than those being considered here, it's a wasteful way to go, as you'll end up with a grossly over-sized motor, with half, or more, of its total capability never being used.

I never cease to be amazed at the HUGE amount of folk-lore and outright mis-information that continues to be espoused about the relative benefits of steppers vs. servos. It seems that in many, if not most, cases, this misinformation is adamantly defended by people with no servo experience, and clearly little or no understanding of the technical characteristics of either. Mere facts, based on deep technical knowledge and years of hands-on experience, are no match for the certainty borne of misinformation and ignorance. Much like politics these days, where one truly does get to make up ones own "facts" to suit ones position.

Regards,
Ray L.

[Al_The_Man]

I never questioned that servo's are capable of higher rpm and can be sized to an economical level when geared. I DID say larger physical size, not larger torque!
Frame size does not guarantee a torque value per-se.
As to 'Servo Experience' I have been in the Industrial Electronic industry for over 50+yrs and have used and designed many servo and CNC systems over the years, if this qualifies?
I still stand by the observation and remarks used in the last quote by me.
Al.

[waltpermenter]

How many benchtop mills have you built Ray? One stepper driven x2 right? Boy that's experience under your belt. This is just another useless he said/she said argument that repeats itself over and over. Isn't there a forum to argue in for you guys that have to be right about everything?
walt

[SCzEngrgGroup]

I never questioned that servo's are capable of higher rpm and can be sized to an economical level when geared. I DID say larger physical size, not larger torque!
Frame size does not guarantee a torque value per-se.
As to 'Servo Experience' I have been in the Industrial Electronic industry for over 50+yrs and have used and designed many servo and CNC systems over the years, if this qualifies?
I still stand by the observation and remarks used in the last quote by me.
Al.

Al,

Sorry, I wasn't disagreeing with you at all, just pointing out that, for a given frame size, the lower torque rating and higher RPM of a servo, when coupled with a belt drive, result in performance comparable to a direct-drive stepper. You highlighted to lower torque, but didn't mention the benefit of the higher RPM in making it possible to easily mitigate that lower torque.

Regards,
Ray L.

[jviss]

I read this thread with great interest, as I have the same question as the OP. There is clearly a lot of knowledge and experience among the participants in this thread.

The idea of a closed-loop, or even absolute-positioning control system appeals to me, but at the same time, I would like to build a balanced, economical, efficient system, so steppers are the obvious answer.

I don't think anyone has mentioned hybrid stepper systems. I saw one from Automation Technology, Inc.. They have a NEMA 23 kit with motor, digital driver and cables for $250; $440 for the NEMA 34 version.

(I have no affiliation with the vendor).

Any comments on a hybrid solution for this mill?

[wizard]

I've pretty much seen this case. A guy is having problems with his stepper driven system so he says screw it I'm putting in a servo based system. While doing so he doubles the output of the servo relative to the stepper. The machine runs much better so he declares servos far superior.

You can see the problem here. There is a need to compare Apples to Apples here not Apples to Peaches. There really is very little in the way of advantages for servos when driven by simple control systems like Mach. I'm not dinging Mach here at all it is just a reality that it can't take advantage of servos the way some controllers can.

As a side note I was on the road today from Charlotte to Greenville today and happened to see the largest Peach of my life. I had to look multiple times as it had my mind confused, thought it was something else. That is why we aren't comparing Apples to Oranges.

There is also a myth that servo's are best used with gearing and steppers suited to direct drive systems.
The fact is that most motors, servo's and steppers alike possess a maximum torque value at zero rpm.
However, although NEMA size does not indicate a certain torque value, steppers are generally higher torque for a given or equal frame size.
So when a system is upgraded to servo, it is either necessary to use a physically larger motor to achieve the same torque value or use gearing if the same frame size is used.
Al.

[bjones]

don't you just love the "experts" that saw something once. must have had peaches at a holiday inn express last night.

[SCzEngrgGroup]

"Closed loop steppers" are inherently incapable of "fixing" the errors caused by over-driving a stepper motor. Unlike servos, stepper motors ALWAYS operate at full torque capability when moving. So, if the torque requirement exceeds the torque available, position loss WILL result. Servos operate most of the time WELL below their maximum torque. So, when more torque is required, they can dramatically increase their torque output, far above their normal running torque, to maintain correct position. Exactly as with steppers, if you exceed that maximum peak torque for long enough to exceed the max allowable following error, position loss will occur.

A closed-loop stepper controller generally cannot do much of anything but halt all motion when a following error is detected, because it cannot correct the error as a servo can. The only way to MAKE it "fault-proof" would be to sense when the torque limit is being approached, and REDUCE the torque requirement BEFORE it gets into trouble. This, however, is at odds with the way both Mach3 and EMC function. In both systems, the trajectory is defined up-front, then passed off to the low-level controller outputs. This means that the velocities and accelerations for the trajectory is effectively carved in stone long before the motor control receives the first step pulse of a given path. You CANNOT change velocity and acceleration on-the-fly, as that, in itself, will create position errors. And the amount of computational power required to re-calculate on-the-fly trajectory changes is WAY beyond what can be handled in real-time on a PC.

Ask yourself this simple question: If a closed-loop stepper control can really eliminate position errors, and provide most of the benefits of a servo control at a much lower cost, why are there so VERY few closed-loop stepper controls out there? Could it be because it doesn't really work as well as you think?

Never mind that the actual incremental cost between a closed-loop stepper system and servos of comparable performance is really not that large.... A servo controller may cost, worst case, slightly more than a comparable stepper controller, and the price difference for the motors is typically even less. The encoder, which is the only other major component, is the same for both. Servos do typically require belt or gear reduction, but that only amounts to a small amount per axis (perhaps $30 or so).

Regards,
Ray L.

[109jb]

If you want to use an encoder to a stepper to detect following errors (ie: Lost steps), and halt operation then you can do it in LinuxCNC by just adding the encoder to the stepper and getting the encoder output to an LinuxCNC input. You will probably need to add a second parallel port. Here is a link about that.

LinuxCNC Documentation Wiki: Steppers With Encoders

I plan to do this on my G0704 conversion. As stated, if you are losing steps then you are already exceeding the torque capability of the stepper or it wouldn't have lost that step in the first place. I see the encode as a nice thing to have for verification that steps weren't lost, or to save a part before the lost steps stack up enough to scrap it. Then you can slow down feedrate, or whatever and save the part. Not much more than that; however, I have read some stuff on the LinuxCNC site where an encoder is used in a closed loop with a step/dir output. As I understand it, most times this is done by people using a servo controller that uses step/dir pulses (Gecko servo controllers come to mind), but it has also been used with steppers as in this link

Closed loop with steppers - LinuxCNC Support Forum

The advantage here is you could use linear encoders and not have to worry about screw lead error, heating of screws causing screw growth, etc. So, there is some benefit that can be had from operating steppers in closed loop, but catching up from lost steps isn't one of them.

[otto_pjm]

If you want to use an encoder to a stepper to detect following errors (ie: Lost steps), and halt operation then you can do it in LinuxCNC by just adding the encoder to the stepper and getting the encoder output to an LinuxCNC input. You will probably need to add a second parallel port. Here is a link about that.

LinuxCNC Documentation Wiki: Steppers With Encoders

I plan to do this on my G0704 conversion. As stated, if you are losing steps then you are already exceeding the torque capability of the stepper or it wouldn't have lost that step in the first place. I see the encode as a nice thing to have for verification that steps weren't lost, or to save a part before the lost steps stack up enough to scrap it. Then you can slow down feedrate, or whatever and save the part. Not much more than that; however, I have read some stuff on the LinuxCNC site where an encoder is used in a closed loop with a step/dir output. As I understand it, most times this is done by people using a servo controller that uses step/dir pulses (Gecko servo controllers come to mind), but it has also been used with steppers as in this link

Closed loop with steppers - LinuxCNC Support Forum

The advantage here is you could use linear encoders and not have to worry about screw lead error, heating of screws causing screw growth, etc. So, there is some benefit that can be had from operating steppers in closed loop, but catching up from lost steps isn't one of them.

Thanks for posting that. I was wondering about similar things. If one wanted to find out the limits of accuracy on a G0704 you could do a servo setup with linear scales. I was concerned how bad the cheapy scales are, at Grizzly and the like .001" repeatability, but then I found these glass scales, which are more than the cheapy C7 ballscrews, but not terrible for a G0704 sized machine, maybe $600. Made in the US too, who knew.

Scales

I know that LinuxCNC can also use multiple means of gauging distance, and combine them in a weighted fashion, there was an article on this being done on a huge Giddings & Lewis machine converted to LinuxCNC where thermal expansion of the screws etc was a factor. It's all overkill for a G0704, but it would be interesting if scales would help with flex correction etc, spindle run out wouldn't be helped by this.

I'm planning to stick with steppers for my G0704, to be ordered soon. I'll be curious how the feedback with steppers will work for you, do you have a thread going?

Pete

[Fastest1]

Ok so today I tuned up my G0704. Like I should have all aong. I could rapid repeatedly on the Y axis at 331 ipm with no positional loss. Btw what is everyone trying to build that requires this accuracy or feedback? My indicator scale is .0005 it wasnt off even a half of a tick after 100's of full rapids. I could worry about adding a 6000hp engine to my car but my tires are still only 12" wide. If you want to machine something the path to success has been layed out pretty well. If you just want a challenge that is fine. Btw I do believe Mach will recognize a loss of positional error if encoders are installed, it will fault and not correct.

[SCzEngrgGroup]

I know that LinuxCNC can also use multiple means of gauging distance, and combine them in a weighted fashion, there was an article on this being done on a huge Giddings & Lewis machine converted to LinuxCNC where thermal expansion of the screws etc was a factor. It's all overkill for a G0704, but it would be interesting if scales would help with flex correction etc, spindle run out wouldn't be helped by this.

Linear scales do nothing to correct for flex in the machine. They DO correct for thermal expansion of the machine, and, especially the screws. That is precisely WHY they are used on commercial machines. Thermal expansion of the machine itself, even on benchtop machines, is one of the largest sources of error. On a high-speed machine, thermal expansion of the screws can be a major source of error, even on a very rigid machine.

Regards,
Ray L.