Mill produces faceted finishes

[jake2465]

My mill uses Servo Dynamics SD1525's to run all three of its axis. Originally the system was set up to receive encoder pulses from rotary encoders off the back of the servo shafts. It was found that keeping the noise away from the lines was difficult at best and random pulses would cause the servos to drift. The finish the servos would give was smooth with no faceting. The system was set to receive 40,000 pulses per inch of travel. With the rogue noise being hard to fix, I decided that magnetic scales would be the fix for that. So, magnetic scales were installed on all three axis and the rotary encoders were disconnected. These scales move at 25,400 pulses per inch, so much less counts than the rotary encoder setup. After the mod was completed, I tried out some parts and noticed that my servos would now produce a faceted finish. The interesting thing is that it does not really happen if only moving parallel to the X or y axis, but rather when making some angled move. At a 45deg movement (y=x), the faceting is at its worst. So, if one were to machine a circle on my mill, it would be seen that the quadrantal angles are the best in finish and as the angle advances toward where both servos start moving closer to an equal speed with each other (advancing toward a 45deg move) the faceting starts to show itself and become apparent. Then, once the move gets past the 45deg and begins to return to a quadrantal angle, then the faceting subsides and ultimately goes away. I can only assume this is a servo tuning issue, since I can simply jog the mill and reproduce this phenomenon without the assistance of a CNC program. if I jog in just X or Y, then there is no faceting to speak of. But, if I jog with both X and y, the faceting is there. I have never performed any servo tuning, so I am not really sure what should be done here. I can only guess that I would need to adjust the pots on the drives to fix this, but I am not sure. The one other thing I noticed is that my Y servo produces a ringing noise when it runs. This goes away once the servo is speed up to something like a rapid move, but returns when doing feed moves.

Similar Threads:

• toolpath is too faceted.
• Problem- Mazak QT20 Gear change to 1 produces spindle to stop and go
• Markon generator only produces 5.2 volts
• Need Help!- FACETED ELLIPSE
• Problem- PCMCIA CARD PRODUCES NO MOVEMENT IN KT2514

[mactec54]

My mill uses Servo Dynamics SD1525's to run all three of its axis. Originally the system was set up to receive encoder pulses from rotary encoders off the back of the servo shafts. It was found that keeping the noise away from the lines was difficult at best and random pulses would cause the servos to drift. The finish the servos would give was smooth with no faceting. The system was set to receive 40,000 pulses per inch of travel. With the rogue noise being hard to fix, I decided that magnetic scales would be the fix for that. So, magnetic scales were installed on all three axis and the rotary encoders were disconnected. These scales move at 25,400 pulses per inch, so much less counts than the rotary encoder setup. After the mod was completed, I tried out some parts and noticed that my servos would now produce a faceted finish. The interesting thing is that it does not really happen if only moving parallel to the X or y axis, but rather when making some angled move. At a 45deg movement (y=x), the faceting is at its worst. So, if one were to machine a circle on my mill, it would be seen that the quadrantal angles are the best in finish and as the angle advances toward where both servos start moving closer to an equal speed with each other (advancing toward a 45deg move) the faceting starts to show itself and become apparent. Then, once the move gets past the 45deg and begins to return to a quadrantal angle, then the faceting subsides and ultimately goes away. I can only assume this is a servo tuning issue, since I can simply jog the mill and reproduce this phenomenon without the assistance of a CNC program. if I jog in just X or Y, then there is no faceting to speak of. But, if I jog with both X and y, the faceting is there. I have never performed any servo tuning, so I am not really sure what should be done here. I can only guess that I would need to adjust the pots on the drives to fix this, but I am not sure. The one other thing I noticed is that my Y servo produces a ringing noise when it runs. This goes away once the servo is speed up to something like a rapid move, but returns when doing feed moves.

This is one of the main problems of low encoder CPR, it happens to everyone that goes below a 14 Bit Encoder, this is around the lowest you can go the higher you go the better it gets 16 Bit ( 65536 CPR ) is the minimum I go on any build, anything below this is a waste, and of not much use for machining, over the last 10 year machine builder have realized this also and now most cnc machines that are built have 1 million CPR or more

[Jim Dawson]

The interesting thing about this is that my machine does not have this problem. The control systems are nearly identical. The one difference is that Jake's machine is using a Galil DMC-1840 card and mine has the current generation DMC-1846 card, which is a bit faster. Other than that, same motors, drives, encoders, and software.

As of right now, I'm of the opinion that the tuning on the servo drives needs a bit of tweaking. We've had the PID parameters all over the map in the Galil card and have played with the other software parameters.

[mactec54]

The interesting thing about this is that my machine does not have this problem. The control systems are nearly identical. The one difference is that Jake's machine is using a Galil DMC-1840 card and mine has the current generation DMC-1846 card, which is a bit faster. Other than that, same motors, drives, encoders, and software.

As of right now, I'm of the opinion that the tuning on the servo drives needs a bit of tweaking. We've had the PID parameters all over the map in the Galil card and have played with the other software parameters.

25,400 pulses per inch is pathetic no matter what way you look at it, or what control you use

[Jim Dawson]

25,400 pulses per inch is pathetic no matter what way you look at it, or what control you use

Maybe on the next build I'll use 0.1 micron encoders

[jake2465]

Even if 25,400 pulses per inch is considered on the low end, that's still less than 40 millionths travel distance per pulse in resolution. The faceting I am experiencing can have peaks and valleys that span over .020" distance from one to the other. If a lack of pulses are to blame, then to get something that bad, I would need encoders that could only give just over 500 pulses per inch to be able to definitively say its gotta be those scales no question about it. Unless there is something I am not aware of that would require a machine to have orders of magnitude more pulses than what the advertised resolution is in order to have an ok surface finish. If it works, then it works. I do find it interesting that Jim's machine has no visible faceting with those micron scales, but mine does. That tells me that the scales are not to blame. I am fairly convinced what is going on here is that I am getting servo oscillation for some reason. I do know that my Y axis servo is making a whining noise when it operates and it did not do that before the retrofit was performed.

[mactec54]

Even if 25,400 pulses per inch is considered on the low end, that's still less than 40 millionths travel distance per pulse in resolution. The faceting I am experiencing can have peaks and valleys that span over .020" distance from one to the other. If a lack of pulses are to blame, then to get something that bad, I would need encoders that could only give just over 500 pulses per inch to be able to definitively say its gotta be those scales no question about it. Unless there is something I am not aware of that would require a machine to have orders of magnitude more pulses than what the advertised resolution is in order to have an ok surface finish. If it works, then it works. I do find it interesting that Jim's machine has no visible faceting with those micron scales, but mine does. That tells me that the scales are not to blame. I am fairly convinced what is going on here is that I am getting servo oscillation for some reason. I do know that my Y axis servo is making a whining noise when it operates and it did not do that before the retrofit was performed.

Machine resolution, does not have anything to do with how your machine is going to run, that is your misconception, you have been lead up the garden path, if you believe that, the benefit of high count Encoders is smooth and fast servo response, to any change, that's what makes a cnc machine

500 pulses per inch, you would be better off with a Stepper Motor system

Without using your servo motors encoders, you won't be able to tune them correctly

You have a lot of things going on with your machine not only the feed back system, you are using, if you have solid ways you will be getting stick-slip, if there is any back lash this will show up as well, so it's not just one thing you have to contend with

[Jim Dawson]

Machine resolution, does not have anything to do with how your machine is going to run, that is your misconception, you have been lead up the garden path, if you believe that, the benefit of high count Encoders is smooth and fast servo response, to any change, that's what makes a cnc machine

500 pulses per inch, you would be better off with a Stepper Motor system

Without using your servo motors encoders, you won't be able to tune them correctly

You have a lot of things going on with your machine not only the feed back system, you are using, if you have solid ways you will be getting stick-slip, if there is any back lash this will show up as well, so it's not just one thing you have to contend with

So you are saying that the servo motor encoders should be connected to the drives?

[jake2465]

I don't believe I have been lied to. I would rather stick to what the symptoms are with the mill. My machine worked fine and produced good finishes with its rotary encoders. I would have kept it that way, but noise was too much of an issue and I needed a solution. That was provided by the micron scales. The problems associated with noise and drift have been taken care of, but a new problem arose and that was faceting. This faceting is quite noticeable. I have not taken precise measurements on the surface finish, but the roughness is impressive. Now, something I have noticed and tends to be repeatable in this scenario; the faceting is at its greatest when y=x. I would have thought that I would have seen this faceting at any angle, but this is not the case. Also, the faceting is not comprised of small straight sections that sum up to a circle, but rather take on the appearance of cusps. Another interesting comparison is that Jim's mill has the exact same scales, servos and drives as mine, but his mill produces finishes that are totally acceptable for a mill. So, where would the real problem then lie? If it was simply a case of lacking pulse counts, then Jim would not be able to escape this faceting problem either and finishes would also be lacking. To me this would imply that something else is at play. Again, my mill worked fine and produced good finishes before the scales were installed. This means that mechanical problems with my mill can be effectively ruled out. On a bad day backlash for my mill might be .0002". Once the scales were installed, the faceting surfaced. But, one critical argument in favor of the scales is that Jim's machine works great with identical scales. So, I must ask myself what had changed. If the scales are not directly to blame for my faceting, then what could it be? One thing that came to mind was that the PID values had totally changed between the old control and the new one. So, is it possible to produce faceting from improperly tuned servos? Can oscillation occur from something like that?

[Bdrangdo]

Are the servos oscillating when stationary or making any kind of humming noise which would indicate hunting for position ?. If not you may be able to turn up the gain, it will not necessarily remove the facets but it could make the period between them smaller. Do you have any way of monitoring the following error, if you could put a scope on it then you would see what is happening on the screen. Another thing to check is the "dead band" setting, maybe that is too high which can cause dithering. Changing from encoders to scales can introduce some unexpected problems due to slackness in the machine and also springiness and movement in the machine structure. As the servos try to keep position there will often be unexpected movement from them. No machine is perfectly rigid and often machines will use both encoder and scales at the same time to prevent this happening using a complex algorithm.

[jake2465]

The servos do oscillate back and fourth when the spindle is under power. Not in a violent manner, but perhaps around three times per second with something like a 3 deg movement. My impression is that the servos are doing this even as the machine is working, not just in a stationary hold. With an indicator on the table and allowing the plunger to rest on the quill, the needle movement is obvious. I do have a scope on the way, but I have never performed a servo tune so this is new territory for me.

[Bdrangdo]

I have just retrofitted a Bridgeport with a Fagor control. Initially I used the encoders a which were perfectly stable. Out of curiosity I switched yo the scales which are just Chinese clones of Sino scales and the "X" axis now slowly oscillates. The screw is noticeably moving yet the DRO only shows 0.01 mm movement. I know for a fact there is backlash on the screw. The "Y" axis which has preloaded ball nuts is stable. All servos of this type have a small following error. They are always trying to catch up and there lies the problem,, because of flexing or backlash the servo will always try to maintain position but the slight time lag means there will often be a small amount of overshoot that the servo tries to correct and ends up oscillating. This oscillating can be very capricious, it is a bit like a tool bit chattering. One small thing can stop it or make it worse. Many controls will have some means of displaying the following error but often the movement (or lack of it) is too fast to see that is why you really need a scope. I am not sure what type of servos you are using. If they are brushed DC you could monitor the demand signal and then compare it to the tacho output.

[jake2465]

The servos I am using are Baldor brushed DC. The drives I run are Servo Dynamics SD-1525's. The attachment shows what pots I have to work with and test points which are not labeled, but I can only assume they are matched with each pot respectively.

[jake2465]

Here are a couple pictures that show the faceting. The picture with the part on the vice shows it clearly. I do find it odd that this really only shows up when both X and Y servos need to work together to make a cut. The closer to the same speed they move at, the more noticeable the faceting. looking at the forks on that part, there is no faceting to really speak of. The mill was moving at a quadrantal angle.

[Bdrangdo]

Unless I am missing something you don't say what the control system is. When I mentioned turning up the gain I was referring to the PID filter settings. P - Proportional gain, I - Integration, D - Derivative. Different controls can have different terms for these. These are usually adjusted via the control parameter settings which vary from one control to another. If P is too low there will be an large following error which can lead to faceting, ironically if it too high it can lead to oscillation which gives the same effect. Adjusting the PID filter settings can be something of a black art if you are not used to it and to do it as close as possible you need to have some form of visual feedback ie an oscilloscope. Finding the tacho connections to the amp should be easy. You should have 3 pairs of wires coming from the motor. The thick ones will be power, the other two will be tacho and thermal overload. Usually there will be a label on the inside of the connection cover. All three will go to the amp. the other wire connections will be command, and enable/disable. You should try and see if there is any backlash with a clock and try and move the bed with a lever of some kind. You will need rather more than just hand pressure. You may be surprised how much movement you can get with very little pressure, mostly it will be springing somewhere but wherever it is remember the servo will be constantly trying to overcome it and losing the battle. It will also help if there is no undue drag in the slideways, check the jibs are not too tight.

[mactec54]

Unless I am missing something you don't say what the control system is. When I mentioned turning up the gain I was referring to the PID filter settings. P - Proportional gain, I - Integration, D - Derivative. Different controls can have different terms for these. These are usually adjusted via the control parameter settings which vary from one control to another. If P is too low there will be an large following error which can lead to faceting, ironically if it too high it can lead to oscillation which gives the same effect. Adjusting the PID filter settings can be something of a black art if you are not used to it and to do it as close as possible you need to have some form of visual feedback ie an oscilloscope. Finding the tacho connections to the amp should be easy. You should have 3 pairs of wires coming from the motor. The thick ones will be power, the other two will be tacho and thermal overload. Usually there will be a label on the inside of the connection cover. All three will go to the amp. the other wire connections will be command, and enable/disable. You should try and see if there is any backlash with a clock and try and move the bed with a lever of some kind. You will need rather more than just hand pressure. You may be surprised how much movement you can get with very little pressure, mostly it will be springing somewhere but wherever it is remember the servo will be constantly trying to overcome it and losing the battle. It will also help if there is no undue drag in the slideways, check the jibs are not too tight.

You are missing something, The main problem is they don't have any control connection between the Motor and Drive

[vmax549]

One of the things to be aware of is you went from a rigidly mounted encoder mounted on teh motor shaft to a scale which is NOT rigidly mounted to the motor. ANY lost motion between the 2 (scale, motor) will lead to the drive hunting for position IF you do not change the motor tuning to compensate. Even then accuracy could suffer.

I would simply replace the original encoders and use teh scales as a backup reference.

Just a thought, (;-) TP

[Bdrangdo]

I have just been playing with my Bridgeport which I only just finished fitting the Fagor control to and checked the following error. Using scales as feedback the lowest I can get is just under .2mm at a slow jog speed. Using the motor encoders it was less than .08mm, in fact with time I am sure I could get it less than that. Using scales I had to use a much lower gain to stop oscillation hence the high following error. My Bridgeport is well worn and has less than perfect ball screws. On a new machine it may not be much of a problem. On all machines the scales will be offset some distance from the screw and the cutter center and the slightest twist on the bedways/dovetails will produce inevitable inaccuracy's and slop which lead to the scales not accurately reflecting the true bed movement. Just by pulling on the bed I could make the servos visibly move yet the DRO which is showing 0.001mm resolution only moved a few thousands of a millimeter. On a typical Bridgeport set up the cutter is offset from the scale by around 150 mm so there could easily be a large disparity between the actual cutter position and the theoretical position. I will be going back to the encoders. It seemed a good idea at the time to use scales but it seems to cause more problems than it solves. As Mactec wrote the lack of resolution can cause problems unless you are using scales specially designed for cnc feedback.

[Jim Dawson]

You are missing something, The main problem is they don't have any control connection between the Motor and Drive

You are the one missing something. There is no provision for an encoder connection on a SD1525 drive. There is however a tach feedback connection, and they are connected. In the original system, the encoder went straight to the controller, not the drive.

[mactec54]

You are the one missing something. There is no provision for an encoder connection on a SD1525 drive. There is however a tach feedback connection, and they are connected. In the original system, the encoder went straight to the controller, not the drive.

Don't think so, so you have dumb useless drives, life is to short to be messing with that junk

There has been many try to do this, but no one to date has succeed, just relying the a linear scale, there is always going to be some problems, even though you have yours running, dual feed back is the only way to use both, Linear and Encoder feed back to the control, if you want to do something like this