PDA

View Full Version : Buying and Using Servos



macona
01-27-2012, 04:37 AM
I have been asked recently and in the past about acquiring servos and how to use them.

As many know I hate steppers. Noisy, slow, hot, and can loose position when you least expect it. There is a reason no commercial company uses them in anything other than desktop machine tools. Yes, servos are more complex and cost more, maybe around 50% if buying used servos vs new steppers. And there is a lot of used servo equipment out there. As the semiconductor industry evolves old equipment is decommissioned and taken apart. The brings a lot to ebay, especially out of Korea. There are several vendors on ebay that have sets from time to time.

Steppers are different from servos where they are open loop, meaning the control that drives the stepper does not know where it is at any given moment. It just sends pulses to the motor and the motor moves. If something binds, motor goes too fast, accelerates to fast, or for some reason there is an unexpected load the motor will stall or loose position. Sometimes it will be a step or two other times is just stops moving. The problem is the control never knows if this happens. There are some projects out there that have tried to close the loop, it has been kind of the holy grail of the cheap CNC market. Mariss of Gecko has worked with it but nothing has been successful.

Steppers also pull more current than servos regardless of load. This means more power dissipation which means more heat. Steppers have a high detent torque but they also loose torque the faster they run and if they do stall you must stop the signal before the motor catches again.

Steppers do have advantages. In general, they are easier to wire. Most drives now are Bipolar and only have 4 wires to drive the motor. They are cheaper and the motors can be more rugged since there is no feedback transducer that tends to be much more fragile than the rest of the encoder. Servos, especially older ones, can have a significant amount of wires coming from the encoder, one I used in the past had about 15 wires from the encoder alone. Steppers also have a high detent torque, meaning they can have a lot of torque required to move the shaft at a standstill. Servos, on the other hand, tend to be a little "squishy". You can move the shaft a little bit before the motor tries to push back, how much depends on the tuning.

But servos are not all fun and games. first you need to know how they work. Servo motors themselves are divided into primarily three basic categories, DC Permanent Magnet (DCPM), AC Brushless (BLAC) , and, linear. DCPM motors are pretty simple, they are very similar to generic brushed permanent magnet motors with a few design modifications for minimal cogging and lighter armatures. AC Brushless motors are similar to standard three phase motors except the laminated rotor has been replaced with a rotor with rare earth magnets. Linear motors are basically a AC Brushless that has been flattened out into a strip.

But where the real difference is that a servo motor has feedback of some kind. There are three feedback options, resolver, tachometer and hall effect, and encoder.

A resolver is a form of a rotary transformer. A signal is sent to it and the absolute position of the shaft is returned based on the relationship signals returned. Resolvers are very rugged, they are found a lot of older equipment and also motors that see rough service. They resolution is really only limited by the device that reads it, a R/D converter, a Resolver to Digital Converter - A fancy analog to digital converter. But they are very expensive, even small ones sell for $1500 new.

Tachometers are small generators that give velocity feedback to the drives to close the loop. They can be as simple as a small PM motor attached to the shaft or they can be a more modern electronic tachometer. The tach returns a signal that's voltage and polarity is proportional to the velocity motor. You often see these on PMDC servos with external position feedback. Hall effect transducers are seen on BLAC motors. The drives that control the motors need to know the basic position of the rotor so it can commutate the motor correctly, that is have the phases in the right sequence for the motor at start up. Typically, there are three hall effect sensors, each in relation to one set of poles on the motor. Hall effect sensors can have limited position feedback capability, but when they are used alone without any other form of feedback they will usually be used for velocity feedback.

Encoders are the most common feedback device you will see on modern servos. There are two basic types of encoder, absolute and incremental. Absolute encoder know the position of the encoder even if the power is turned off. Incremental encoders just measure the distance traveled arbitrarily. Think of a tape measure. An absolute encoder would be a tape measure with numbers you always know where you are from the end. An incremental encoder is like a tape measure with no number markings. You have to count from the beginning. If you loose count you have no idea where you are anymore.

There are three forms of encoder communication, serial, parallel data, and quadrature. Absolute encoders have serial or parallel interfaces. The parallel interface is a form of Gray code. It is decoded by the processor for the motors location. The serial encoders output a serial data stream for the position of the encoder. Since the drive knows the exact location of the motor at start up is knows what it needs to commutate the motor.

Incremental encoders have a quadrature output. Quadrature is two square waves 90 degrees out of phase of each other, phases A and B. Based on the leading of the phase you can determine the position with a counter. Some encoders also have an index track that sends out a pulse once per revolution usually labeled "Z". Incremental encoders on BLAC motors will also have commutation tracks, these will be three signals that simulate the hall encoders found on other motors. Quadrature encoders can have many forms out output. Differential, Single Ended, and open collector, to name a few. Usually you will see Differential and single ended. A single ended encoder will just send positive pulses to the controller, a differential encoder sends complementary signals for each channel. This makes the signal more immune to noise. It is usually possible to mix differential and single ended devices as long as the signal voltage is within the tolerated zone of the receiver. To connect a differential encoder to a single ended receiver you just ignore the inverted outputs from the encoder. To connect a single ended encoder to a differential receiver you can usually just connect the encoder signal wires to the non-inverted inputs and it will work. Some receivers will not work, in that case you need to use a line driver IC to convert the single ended signal to a differential signal. An IC such as the 26LS31 will do this. There is also quadrature sinusoidal output. These are found in higher end (read expensive) encoders. They output either a 1vp-p or 11uap-p signal. Interpolators can take this signal and interpolate the signal to give a TTL quadrature resolution anywhere from 1 to over 100 times the base period of the encoder.

Serial encoders are found in the most recent (Past 10 years) brushless motors. Every company seems to have their own protocol. EnDat, Mitsubishi, Yaskawa, etc. None of them are compatible. I have seen a board that will read some of the different formats as the formats are often documented. Serial has an advantage that the encoder can carry motor parameters as well as position and commutation data making set up much simpler.

Now we are down to the drives. Two basic kinds of drives are available. Analog amplifiers and digital drives. Analog amplifiers take a -10-0-10v analog voltage signal and drive the motor in one of two modes. Current and velocity. We really don't care about these, they require an external controller to handle the positioning and PID loop. There are a few controllers that will talk to these like Servo2Go on the EMC side and the DynoMotion on the Mach3 side. There are also a couple of people that make boards that will drive these drives from step and direction inputs like the Pixie and YAPSC:10V.

Digital drives are the ones we are most interested in. They have an on board processor that handled the position information that is being returned from the motor or machine and closes the loop. They drives will have different interfaces. Some will have basic PLC functionality where they are programmed and operate standalone from external inputs. There are also drives that communicate over an industrial network of some sort. We really can use either of these things. What we want are called positioning drives. These drives will have digital inputs that will move the motor based on the input pulses. Most are configurable for quadrature, CW/CCW, and step/direction (pulse/sign) inputs. We want step/direction.

There are several companies that make drives that will run a variety of different motors such as Allen Bradley/Reliance Electric (1398 and 2098 series), Elmo Motion, Pacific Scientific, and Aerotech. There are drives that will run both BLAC and DCPM with the same drive, ELMO drives do that.. If you have existing servos you way to drive you will need to find one that matches the feedback output of the motor and the electrical parameters of the motor itself. You can get them with different feedback options like Endat, Quadrature (sine or TTL), potentiometer, hall, resolver, etc. If possible you will need to know the basic electrical parameters of the motor. This can often be found in a data sheet. If not some drives can determine the data needed during the setup. Most of these modern drives connect to a PC via serial to handle the configuration via a GUI as well as handle tuning the drives.

macona
01-27-2012, 04:37 AM
If you are building a machine with no motors and have the option to start from scratch then it is best to get a matched set of motors and drives such as ones from Mitsubishi, Yaskawa/Omron, Panasonic, and others. You can buy drives and motors separately but I do not recommend it. A specific model of motor can be limited to once specific model of drives and they change over the generations. You can do it if you take time to study a specific brand and learn the difference. Probably the easiest to do this with is Mitsubishi. They have the simplest model number system and when it comes to the drives 400w and smaller all their motors will run on wither the 120v or 240v drives. Yaskawa has motors specific to 120v or 240v drives.

As mentioned above you need positioning drives. You need to learn the model numbers of the drives to know what is what.

With Mitsubishi you want to look for a letter "A" at the end of the model number. For example MR-H-500-AN or MR-J2S-350-A. Also small drives are available in 120v, they have an A1, ex. MR-J2-20A1.
Mitsubishi has several series of drives available.
MR-C - Cheapest drive, 400w max, the Economy Model
MR-H - Industrial Drive, you must have a pendant or the MRConfigurator software to set up the drive. Full time auto tune. 131072 line encoder option.
MR-J - First of the J line of drives. Has a hard to find interface connector, can be programmed on the front panel.
MR-J2 - Second series of the J line, Lots more options, better tuning options 16384 pulse per rev serial encoders.
MR-J2S - Better auto tune and 131072 line encoders
MR-J3 - Current generation, 262144 line encoders, on the fly adaptive tuning.

With Yaskawa you look for the last two letters in the model number. A or B is the voltage range. A is 240, B is 120. P or D means positioning drive. ex SGDA-02BP, 120V positioning drive. Yaskawa calls their drive series by the label "Sigma" Current generation is Sigma V, I believe. Omron relabels Yaskawa drives but changes the model number system. You have to look through the manuals to figure that out.

I have not messed with any other drives.

willmac
01-27-2012, 07:13 AM
Macona -

Thank you for posting this. It is a really useful summary, especially so since you have direct experience of re-using servos and drives. This makes it much more practical for people on a tighter budget (can't buy new!)to have a go at a servo based system.

amateur
01-27-2012, 10:29 AM
[QUOTE=macona].

. Steppers also have a high detent torque, meaning they can have a lot of torque required to move the shaft at a standstill. Servos, on the other hand, tend to be a little "squishy". You can move the shaft a little bit before the motor tries to push back, how much depends on the tuning.


Excellent tutorial on the comparison,so much to learn.
Now about the squishy part,how does this effect real world machining.Does it affect tolerance ?

MaxHeadRoom
01-27-2012, 11:37 AM
If I may make a couple of observations and additions.
If you are mixing and matching motor and drives, I have found the easiest is the DC brushed, followed by BLDC, the AC sinusoidal being the hardest due to the methods of feedback and commutation.
If using a closed loop controller such as the In-a-PC-Slot motion card or a USB control such as Dynomotion, the non-intelligent drive can be used in the simple torque mode Such as Aerotech and AMC, as the PID loop is closed back to the Controller, this makes such things as electronic gearing of one servo off another an easy task.
Mitsubishi have two pricing standards, one is for IA, Industrial Automation the other CNC control, the price is vastly different, I just priced a MR-J2 drive and 1Kw motor, under IA it costs $5248.00 for the pair.
A complete Mitsubishi system with spindle under the CNC label with 3 motors cost between $14,000 & $18,000.
To add to the encoder info, it is common for controls to increase the resolution of the encoder by using all four edges of the two quadrature pulses, for example a 1000/p rev encoder will be increased to 4000p/rev resolution.
Max.

gzig5
01-27-2012, 12:18 PM
Good Post, here's some more background on Allen-Bradley Servo Drives.

AB servo drives are typically not a "open" platform. I think the offshore brands make it easier to mix and match motors and drives from different vendors. I'm not sure how some of the other manufacturers handle third party motors but in the case of AB drives, you typically will need to use AB motors and cable sets, so buy them together if you can. Each drive will work with several different permanent magnet motor models, but do your home work first. Third partly motor files are available, but expensive to generate. The 1398 Ultra 100/200 is programed with UltraMaster and the 2098 Ultra3000 uses UltraWare, both of which should be available free. Both of these can be run as stand alone drives and can use presets, analogue, or step-direction/PTO inputs for control outside of a ControlLogix PLC platform and both con use serial host commands (yuck). Other models out there for a while that might be put into use are the Ultra 1500, and Kinetix 3. There is also a new model out, the 2097 K300 that uses Ethernet for communications and setup but probably won't be seen on ebay anytime soon.

MaxHeadRoom
01-27-2012, 01:10 PM
With AB motors I have found the BLDC are relatively easy as they have standard encoders with comm tracks and can be used with Advanced Motion and others.
It is the AC sinusoidal type with the resolver on that can get tricky.
Max.

John Stevenson
01-27-2012, 01:50 PM
Good post, very informative but a point on the side of the humble stepper motor.

Many people have got started with steppers and gone onto better things but might not have made that leap IF they had to understand servos from day one.

The Artful Bodger
01-27-2012, 01:57 PM
What about systems that might use simple DC motors and a control system that reads machine/tool/table position from existing machine DRO scales?

ikdor
01-27-2012, 02:11 PM
I think the control would not be as straightforward as for controlling a motor. There is a lot more flexible windup and backlash messing with the ideal linearity of the system.
It can be done of course, but will go beyond a simple PID control if you want any speed out of it.

Igor

The Artful Bodger
01-27-2012, 02:23 PM
I think the control would not be as straightforward as for controlling a motor. There is a lot more flexible windup and backlash messing with the ideal linearity of the system.
It can be done of course, but will go beyond a simple PID control if you want any speed out of it.

Igor


I am not thinking of speed and I would have thought that reading the machine position would have immediately taken care of any flexibility or backlash issues.

macona
01-27-2012, 02:25 PM
[QUOTE=macona].

. Steppers also have a high detent torque, meaning they can have a lot of torque required to move the shaft at a standstill. Servos, on the other hand, tend to be a little "squishy". You can move the shaft a little bit before the motor tries to push back, how much depends on the tuning.


Excellent tutorial on the comparison,so much to learn.
Now about the squishy part,how does this effect real world machining.Does it affect tolerance ?

It can if your tuning is not right. The squishiness can be also seen in the following error. When a servo motor is in motion it lags behind the commanded position, this is called follow error. There are a lot of parameters that effect how much of an error you have, mainly the Proportional part of the PID loop. If the drive has feed forward gain this can be used to help but can cause overshoot.

macona
01-27-2012, 02:30 PM
Good post, very informative but a point on the side of the humble stepper motor.

Many people have got started with steppers and gone onto better things but might not have made that leap IF they had to understand servos from day one.

Yes. I learned very quickly that if I didn't have to mess with steppers again I wouldn't.

What is nice by using the complete matched packages there is nothing more than plug in the motor, plug in the encoder, wire inputs and set a couple parameters, the auto tune on newer drives is really good.

It gets time consuming to tune manually.

macona
01-27-2012, 02:41 PM
I am not thinking of speed and I would have thought that reading the machine position would have immediately taken care of any flexibility or backlash issues.

It can be done. I am using this method on the laser cutter I am building. Brushed servos on the X and Y axis and linear encoders for feedback. The problem is, like ikdor mentioned, is any, and I mean ANY, lost motion in the system will treaty effect the performance of the system.

When I got my cnc mill it was that way, DC servos with heidenhain linear encoder on the axis. There was about .001" lash on the axis and while at rest the motors would dither back and forth between the backlash extents. To minimize this you have to tune the loop down and this means the whole thing runs slow.

Even on the laser cutter I have been building, simple flex in the system had caused tuning issues. it is still something I have to deal with.

The best way to handle this is to use a dual loop where feedback from the motor is used for gross positioning and then the final destination position is gathered from the scale on the axis. There are a couple of controllers on the low end that do this, one is the Dynomotion and the other is the almost vaporware cncbrain.

willmac
01-27-2012, 03:05 PM
Macona -

I have had a quick look at the Dynomotion documentation and it looks like a very interesting product. But I am not at all clear on HOW it handles dual loop control. I assume from what you say that it is capable of handling a glass scale encoder for fine (absolute) position and encoders or similar on a ball screw for coarse positioning, rather like the controllers used on (say) Deckel CNC mills.

Can you elaborate or point to where it is covered in the documentation?

The Artful Bodger
01-27-2012, 03:07 PM
Thats interesting as I was, perhaps naively, imagining that this would be a practical way for a home shopper to dabble in minimal CNC.

Evan
01-27-2012, 03:09 PM
You are being pretty hard on steppers. They can and do work well as long as the limitations are respected. There are plenty of steppers in use in industry when rugged, predictable and simple systems are required. Steppers can be closed loop just like servos and often are when very high precision is required. There is a reason that different versions of microprocessor masks are identified by the "stepping" number. The machines that make them are called "steppers" after the mechanisms used to drive them. They may no longer actually use steppers as drives but stepper motors were quite good enough to do the job.


Servos excel when high velocity motion is required. That is the primary reason for their use. Unlike stepper motors servos must always have some position error to generate an error signal that is used to determine where, how far and how fast to move. Unlike steppers there is always a small error band within which the servo stops since any closed loop system must have some hysteresis in order to be stable. Steppers have no such requirement as their position is determined by their construction.

macona
01-27-2012, 03:17 PM
Macona -

I have had a quick look at the Dynomotion documentation and it looks like a very interesting product. But I am not at all clear on HOW it handles dual loop control. I assume from what you say that it is capable of handling a glass scale encoder for fine (absolute) position and encoders or similar on a ball screw for coarse positioning, rather like the controllers used on (say) Deckel CNC mills.

Can you elaborate or point to where it is covered in the documentation?

I might be wrong. In the past week or so I have been looking at a bunch of different controllers I may have gotten them mixed up.

John Stevenson
01-27-2012, 03:21 PM
Yes. I learned very quickly that if I didn't have to mess with steppers again I wouldn't.

What is nice by using the complete matched packages there is nothing more than plug in the motor, plug in the encoder, wire inputs and set a couple parameters, the auto tune on newer drives is really good.

It gets time consuming to tune manually.

I am not rubbishing servo drives, they have a place and it's a very important one. However servos do not suit every application and I am not that blinkered that I can't see a use for both systems.

Throwing away the choice for steppers puts a beginner with a hefty $$ choice when it comes to built or convert a machine to further any learning curve.

By your blinkered attitude that means that possibly higher than 50% of newbies would never get a foothold into CNC ?

macona
01-27-2012, 03:25 PM
You are being pretty hard on steppers. They can and do work well as long as the limitations are respected. There are plenty of steppers in use in industry when rugged, predictable and simple systems are required. Steppers can be closed loop just like servos and often are when very high precision is required. There is a reason that different versions of microprocessor masks are identified by the "stepping" number. The machines that make them are called "steppers" after the mechanisms used to drive them. They may no longer actually use steppers as drives but stepper motors were quite good enough to do the job.


Servos excel when high velocity motion is required. That is the primary reason for their use. Unlike stepper motors servos must always have some position error to generate an error signal that is used to determine where, how far and how fast to move. Unlike steppers there is always a small error band within which the servo stops since any closed loop system must have some hysteresis in order to be stable. Steppers have no such requirement as their position is determined by their construction.

We use a lot of steppers at work, I mean a lot. From little things that move focus rings on lenses to a couple camera rigs about 14' tall. As long as you keep them slow and sluggish you are OK. With stop motion you can't afford to miss steps.

There have been a couple occasions where we have had to use servos for things like motion blur. A stepper just couldn't move the prop fast enough with enough acceleration. I brought in one of my 200W mitsubishi sets and it worked great. Where the same physical size stepper motor would stall out moving a 50lb prop my little servo would lift our work bench after I swapped motors on the linear actuator.

macona
01-27-2012, 03:34 PM
I am not rubbishing servo drives, they have a place and it's a very important one. However servos do not suit every application and I am not that blinkered that I can't see a use for both systems.

Throwing away the choice for steppers puts a beginner with a hefty $$ choice when it comes to built or convert a machine to further any learning curve.

By your blinkered attitude that means that possibly higher than 50% of newbies would never get a foothold into CNC ?


Not really, there a lot of lower end options for doing servos cheap. Drives like the Whale3 are cheaper than a Gecko stepper drive. New and old motors can be found many places and magnetic encoders are cheap. You can put a system together for almost the same price.

There is a lot of information out there and the low end drives get better and better. Servos should be a serious consideration now instead of just jumping on to stepper.

Of course there is always the case when the guy has virtually zero money available. In that cast used steppers and some cheap drivers like the Pololu are the way to go.

Evan
01-27-2012, 03:39 PM
You have access to a lot of high end equipment that is far beyond what the average hobby machinist has even seen. "Sluggish" is a very relative term. "Sluggish" steppers are all that most home shop machinists will ever need and do the job just fine. Your perspective is very skewed by what you are accustomed to working with.

macona
01-27-2012, 03:48 PM
You have access to a lot of high end equipment that is far beyond what the average hobby machinist has even seen. "Sluggish" is a very relative term. "Sluggish" steppers are all that most home shop machinists will ever need and do the job just fine. Your perspective is very skewed by what you are accustomed to working with.

I don't know about that. I dig and scrounge for what I want. It may take me a while to put something together. Just like the servo motor on the mill, it took me two years to find the matching drive for the motor. With ebay anyone has access to this stuff, though the prices seem to have risen as of late, probably due to mere people buying the motors and drives for home projects.

MaxHeadRoom
01-27-2012, 04:00 PM
Prices New are probably out of range of the average DIY'er, but if one is willing to be patient, there is some nice stuff that comes up on ebay, although I have noticed prices seem to have gone up over that past couple of years, probably to due to more and more getting into it as a hobby.
About a year ago I picked up 3 legacy Galil motion cards for $35.00 each!
I just wish ebay had been around 25 - 30yrs ago!.
Max.

dm1try
01-27-2012, 04:53 PM
You can buy drives and motors separately but I do not recommend it. A specific model of motor can be limited to once specific model of drives and they change over the generations
true, but if you don't mind some hacking, it can be a lot of fun. that restriction is a purely artificial. i've played with mitsubishi J2S and J3 and found that you can mix motors and drives of different capacity by just setting a parameter. it is also possible to set all motor parameters explicitly, so (theoretically) you can drive any motor with a suitable encoder..

macona
01-27-2012, 05:14 PM
I have also found I can use motors intended for the J2 on the J2S, you just now have 16384 steps pre rev instead of the 131072. I have never messes with the J3 series. I have not been looking. I have so many more motors and drive around I just don't need any.

macona
01-27-2012, 05:21 PM
Prices New are probably out of range of the average DIY'er, but if one is willing to be patient, there is some nice stuff that comes up on ebay, although I have noticed prices seem to have gone up over that past couple of years, probably to due to more and more getting into it as a hobby.
About a year ago I picked up 3 legacy Galil motion cards for $35.00 each!
I just wish ebay had been around 25 - 30yrs ago!.
Max.

Automation Direct also sells servos and drive. Significantly cheaper than Mitsubishi but still quite a bit more expensive than a stepper.

http://www.automationdirect.com/adc/Shopping/Catalog/Motion_Control/Servo_Systems

dm1try
01-27-2012, 05:32 PM
I have also found I can use motors intended for the J2 on the J2S, you just now have 16384 steps pre rev instead of the 131072. I have never messes with the J3 series. I have not been looking.
well, for me that started with a J2S-70B-U006 drive which refused to drive a KFS73 motor. it turned that they just locked in KFS710 motor constants and marked the drive with U006. not fair, imho :)
as for driving J2 motors with J2S - yeah, sure, it works. they also knows HC-PQ and HA-FF motors. every drive has the same table with all the motors, disregarding capacity. it just checks it is a "matched" pair.
also, i've found a custom J2S firmware which knows a HF-KP43 motor which is from J3 series. so mistubishi also does that if there's a need. i like J3 motors - they are much more compact. 400W is the same length as 200W from J2/J2S series

MaxHeadRoom
01-27-2012, 05:48 PM
Recent bulletin from Mitsubishi:
MR-J2-xxxA no longer available as replacement, is replaced by MR-J2S-xxxA,
Max.

macona
01-27-2012, 09:07 PM
Thats good to know. My little lathe has J2-20 and 40 drives in it. I have another set of J2-40A1's without motors. Been keeping an eye on ebay but have not seen a deal on motors for them.

Ill have to start looking at the J3 stuff, not like I really need it at this point. About the only thing left I could CNC is my surface grinder.

TexasTurnado
01-27-2012, 09:45 PM
Thats good to know. My little lathe has J2-20 and 40 drives in it. I have another set of J2-40A1's without motors. Been keeping an eye on ebay but have not seen a deal on motors for them.

Ill have to start looking at the J3 stuff, not like I really need it at this point. About the only thing left I could CNC is my surface grinder.

I don't know about CNCing my surface grinder, but it would be nice to put a servo on it to drive the table back and forth instead of hand cranking it....:D

How are you doing these days? Last I heard you were going to be unemployed.

djc
01-28-2012, 02:05 AM
While the closed loop aspect of servos is a good thing, perhaps it might be good to write a few words concerning how you can (or can't) use this to your advantage with current hobby-based controls.

So what if the servo drive tracks following error to 3/10 of a milicron*. What does it do with it? At the moment, the best it can do is fault the drive when the error exceeds a user-set value and send a signal to the controller (which then itself may fault).

Mach, for instance, cannot track the following error and absolutely cannot do any real time correction of it. I understand that EMC can incorporate this function but have no experience in using it.

This is not to knock Mach, EMC, servos or steppers just to say that in seeking performance improvements, you have to consider the whole system. We'd all like a Ferrari engine in the pick-up, but if it's running on cross-plys there's not much point.

* ISO measure of tolerance, applied to Asian-built machine tools.

macona
01-28-2012, 03:30 AM
I don't know about CNCing my surface grinder, but it would be nice to put a servo on it to drive the table back and forth instead of hand cranking it....:D

How are you doing these days? Last I heard you were going to be unemployed.

Don't really need a servo to do that, you could do that with a vfd driven gear motor and a couple limit switches. Maybe a little bit of control logic. My surface grinder was originally power fed but somewhere along the lines it was removed but a lot of it is still there.

Unemployed now. Been looking but not much out there. Hopefully filming will start on the next one soon.

macona
01-28-2012, 03:36 AM
While the closed loop aspect of servos is a good thing, perhaps it might be good to write a few words concerning how you can (or can't) use this to your advantage with current hobby-based controls.

So what if the servo drive tracks following error to 3/10 of a milicron*. What does it do with it? At the moment, the best it can do is fault the drive when the error exceeds a user-set value and send a signal to the controller (which then itself may fault).

Mach, for instance, cannot track the following error and absolutely cannot do any real time correction of it. I understand that EMC can incorporate this function but have no experience in using it.

This is not to knock Mach, EMC, servos or steppers just to say that in seeking performance improvements, you have to consider the whole system. We'd all like a Ferrari engine in the pick-up, but if it's running on cross-plys there's not much point.

* ISO measure of tolerance, applied to Asian-built machine tools.

Yes, EMC can close the loop but even then there will still be follow error. That is just the nature of servos. But I have found that as long as you size the motors correctly you won't be more than a few encoder counts out on follow error, and when you have a 16k line encoder it does not mean a whole lot, even less on new motors with 131k or 262k motors. The MR-H series drives I put on my X and Y axis have been working great and I have had no problems. Everything is as accurate as my iron will allow.

John Stevenson
01-28-2012, 05:48 AM
I don't know about CNCing my surface grinder, but it would be nice to put a servo on it to drive the table back and forth instead of hand cranking it....:D



TT on my old one I fitted a rod less air cylinder with two flip flop valves on an adjustable rail on the cylinder.
These set the stroke and the regulator set the speed.

Worked very well, cheap simple and quick to adapt.

John Stevenson
01-28-2012, 05:50 AM
Everything is as accurate as my iron will allow.

Aint that the truth and applies to all machines, no point trying to get microns if the machine or even the job doesn't warrant it.

PaulT
01-28-2012, 02:22 PM
Another comment on matching your drive capabilities to your iron.

I have a BP clone CNC, I got it stripped of the control but fitted with ball screws with big reasonably modern size 42 steppers.

So I put Gecko drives and Mach on it and its been a great machine for my business, we've made tons of parts for our products and lots of prototypes and one-offs on it.

But I've always had "servo envy" and had half-dreams of dumping the steppers and putting servos on it.

But then I started thinking about all the parts we make and realized we hardly ever cut at the full speed of the machine, which is 60 ipm on X and Y and 90 ipm on Z. The machine just doesn't have power and rigidity to go any faster than that in an actual cut.

So all my "servo envy" went away. The only way servo's would help us would be faster rapids but its not worth the trouble just for that.

If we made more truly 3D parts the faster acceleration might be beneficial but we really don't make many parts like that.

The fact of the matter is unless you are doing a lot of high speed contouring, servo's are overkill for most tool room sized and smaller CNC machines.

Paul T.
www.power-t.com

skunkworks
01-28-2012, 02:34 PM
Great info!

As far as analog drives.. Don't sell it too short. (I am an linuxcnc person)
You mentioned servo-to-go. There are a few other more popular/inexpensive interfaces.
http://wiki.linuxcnc.org/cgi-bin/wiki.pl?LinuxCNC_Supported_Hardware

mesa and pico systems both have great support and analog interface options (and step/dir)

for about $248 you can get a 6 axis analog + 48 isolated i/o pci setup (very expandable). This puts all the motion and pid within linuxcnc (plus you have access to ladder logic) Mesa does have servo amplifier cards but I have not really looked at them. (as does pico systems)

For retrofitting - this also allows you to sometimes be able to re-use the original drives and servos. Plus on ebay you seem to still be able to get inexpensive analog drives still.

Side note - you mention quadrature for control (instead of step and direction) - emc can do this in software and I am pretty sure that mesa hardware will also do it. this is less susceptible to noise and in software step generation - you can output higher rates (in emc)

macona
01-28-2012, 02:39 PM
Servos will make more of a difference than you think, especially when you are doing more complex profiles that require a lot of acceleration and deceleration. I have found on stepper set ups you never get close to full feed rate because it the machine has to start slowing down to change direction before it gets to full speed.

Also backlash comp is another place that benefits from servos. The servo can jump the backlash gap quickly and resume minimizing cutter dig in where steppers will tend to dwell at that point.

TexasTurnado
01-28-2012, 06:08 PM
TT on my old one I fitted a rod less air cylinder with two flip flop valves on an adjustable rail on the cylinder.
These set the stroke and the regulator set the speed.

Worked very well, cheap simple and quick to adapt.

Hi Sir John -

Now that's a great idea I had not considered... Do you have any photos of the setup? To me, this grinder drive might also be a good application for steppers....

Here's a pic of the grinder sans table taken right after I received it:

http://i288.photobucket.com/albums/ll168/TexasTurnado/P5110035a.jpg

John Stevenson
01-28-2012, 06:18 PM
TT, sorry no pictures, might have been pre digital camera and that grinder has long gone.

In defence of servo's I need to pick your brains over the servo hobbing setup.

I need to change the stepper driver unit on my mill as the black box isn't able to remember position once stopped and started again

TexasTurnado
01-28-2012, 07:17 PM
TT, sorry no pictures, might have been pre digital camera and that grinder has long gone.

In defence of servo's I need to pick your brains over the servo hobbing setup.

I need to change the stepper driver unit on my mill as the black box isn't able to remember position once stopped and started again

You won't have that trouble with a servo setup - unless you shut off the power. As long as the power is on to the servo controllers, the hobbers can be stopped or started anytime (or even be turned either forward or reverse) without losing position. It should be possible to design a servo hobber setup to remember where it was with power down, but the setup I used had incremental encoders without battery backup.

I will be happy to help you all I can, but macona and lazlo taught me almost all I know about servos....:D I had worked with vfds some, and the programming of the servos controller is similar, but of course, the parameters are different.

The company I used to work for made servo drives for huge parabolic dishes for satellite tracking, etc, but I was in a different department. I did discuss some of the application principles with their engineers, and so I had a basic understanding of what was going on - but their hardware was nothing like the modern AC servos.

If you like, send me a pm, as we best not hi-jack this thread....:D

BTW, a write-up of the electronic dividing head (EDH) I used for the gear hobber should appear in an upcoming issue of Digital Machinist...

macona
01-28-2012, 07:54 PM
The MR-H and MR-J2 drives on do have a battery backed up absolute encoder system, but it is anything but plug and play. The drives would have to be connected to a PLC or computer to read the position at start up through the drive's serial port. And then the computer would have to resynchronize the slave with the master. Not the simplest of projects, but it could be done.

It might be possible to do it with something like an Arduino Mega that has multiple hardware serial ports. I would have to look at the control specs but I am betting you could also control the ratio by changing parameters through the serial ports on the drive. With a small LCD and a couple buttons you could build a head unit.

dm1try
01-29-2012, 06:16 AM
The MR-H and MR-J2 drives on do have a battery backed up absolute encoder system, but it is anything but plug and play
absolute encoders shine with digital control, like SSCNET II. it also needs a bit of communication with the drive on startup, but then you just send a 32-bit target position every 0.88 ms. i've made a prototype module for emc2 talks thru ethernet my own simple SSCNET controller. now i need to learn how to make emc2 believe that all axes are homed right on startup.

zippo
05-29-2012, 12:50 PM
Hello,

I read the post regarding MITSUBISCHI MR-J2 servo drivers.

I brought for my BF20L witch now uses steppers two HA-FF-38-UE-S1 (8000 RPM, 110V, 2A, 300W, 3.1Kg) servo motors from eBay Germany new with 145Euro per item.
I had in the workshop from disassembling of packaging machines the next MR-J2S drivers:
MR-J2S-10B x3
MR-J2S-20B x1
MR-J2S-40B x1
MR-J2S-40CL x1
HC-MFS13-S13 x5
Neither one of this drivers does not want to match with HA-FF-38-UE-S1.
I have code 1A (motor- driver mismatch).

Is there any way to use the motors with MR-J2S-40B or MR-J2S-40CL somehow?
Is there a way to uses the SSCNET drivers them with MACH3 or EMC2, I no not have any SSCNET controllers.?


Thank you,
Razvan

lazlo
05-29-2012, 01:26 PM
Servos will make more of a difference than you think, especially when you are doing more complex profiles that require a lot of acceleration and deceleration.

Great article Jerry! I don't think I saw this back in January.

For the folks who aren't convinced that servos are suitable for small mills, take a look at this humble Sieg X3 benchtop mill that's been retrofitted with Mitsubishi servos:

http://www.youtube.com/v/ZxrikzOpnXM