Re: [Emc-users] DRO scales and EMC2
Kirk Wallace wrote: I played with the deadband a little at 1e-06 and 1e-05, but apparently these numbers are way too small for my .0005 scales. Well, of course. ,0005 is 2000 counts/inch or 5e-4. Probably the deadband should be set to about 7.5e-4 to establish a mechanical zone where there will be no effort from the drive. You don't want it any larger as it will cut into machine accuracy. Jon - Check out the new SourceForge.net Marketplace. It's the best place to buy or sell services for just about anything Open Source. http://ad.doubleclick.net/clk;164216239;13503038;w?http://sf.net/marketplace ___ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users
[Emc-users] DRO scales and EMC2
I am a newbie to EMC2 but not to machining. I am thinking about creating a CNC Mill with EMC2 however, my problem is with backlash compensation. Is it possible to use DRO scales with EMC2 to have the feed back that would allow you to know exactly the tool position? Thanks WGB - Check out the new SourceForge.net Marketplace. It's the best place to buy or sell services for just about anything Open Source. http://ad.doubleclick.net/clk;164216239;13503038;w?http://sf.net/marketplace___ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users
Re: [Emc-users] DRO scales and EMC2
Hi, It is indeed possible to use linear scales with emc. With backlash they do not tune as well as encoders on the ballscrew. If the ballscrews are very tight the performance should be excellent. Recently there was a proposal by someone on the list (JMK?) to use encoders on the ballscrew/servo motor for velocity and rough positioning and then use the linear scale to drive final position. AFIK there is no software at present to do implement this but it is certainly a cute idea. Many years ago when I was just getting started with CNC I retrofitted a Cincinnati tracermill with servos and glass scales. It worked but because of backlash in the 0.003 region it was jerky in tuning. When I replaced the glass scales with encoders on the ballscrew then tuning became very smooth. One still has to be aware of the backlash and approach drilled holes, etc from a consistent direction but it will get the job done. HTH Dave On Jan 8, 2008, at 9:16 AM, Witek GB wrote: I am a newbie to EMC2 but not to machining. I am thinking about creating a CNC Mill with EMC2 however, my problem is with backlash compensation. Is it possible to use DRO scales with EMC2 to have the feed back that would allow you to know exactly the tool position? Thanks WGB -- --- Check out the new SourceForge.net Marketplace. It's the best place to buy or sell services for just about anything Open Source. http://ad.doubleclick.net/clk;164216239;13503038;w?http://sf.net/ marketplace___ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users - Check out the new SourceForge.net Marketplace. It's the best place to buy or sell services for just about anything Open Source. http://ad.doubleclick.net/clk;164216239;13503038;w?http://sf.net/marketplace ___ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users
Re: [Emc-users] DRO scales and EMC2
Dave Engvall wrote: Hi, It is indeed possible to use linear scales with emc. With backlash they do not tune as well as encoders on the ballscrew. If the ballscrews are very tight the performance should be excellent. Recently there was a proposal by someone on the list (JMK?) to use encoders on the ballscrew/servo motor for velocity and rough positioning and then use the linear scale to drive final position. AFIK there is no software at present to do implement this but it is certainly a cute idea. I believe you could probably mock this up with HAL components (possibly might need to add a new math function along the way, but the set of components is getting quite robust) to merge the two encoder inputs. The math shouldn't be very complex. But, a caution: Knowing position via the glass scales doesn't eliminate the backlash problem. The real problem is that the position of the table is not constrained in both directions by the servo. it is only constrained in one direction at a time, and the motor cannot hop from one side of the backlash to the other instantly, therefore cutting forces and inertia can flip it from one side to the other faster than the motor can compensate. This can lead to messed-up parts, broken tools and general foul language around the shop. So, don't think the glass scales are some kind of panacea that allows you to do precision work with sloppy leadscrews. Jon - Check out the new SourceForge.net Marketplace. It's the best place to buy or sell services for just about anything Open Source. http://ad.doubleclick.net/clk;164216239;13503038;w?http://sf.net/marketplace ___ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users
Re: [Emc-users] DRO scales and EMC2
On Tue, 2008-01-08 at 12:30 -0600, Jon Elson wrote: Dave Engvall wrote: Hi, It is indeed possible to use linear scales with emc. With backlash they do not tune as well as encoders on the ballscrew. If the ballscrews are very tight the performance should be excellent. Recently there was a proposal by someone on the list (JMK?) to use encoders on the ballscrew/servo motor for velocity and rough positioning and then use the linear scale to drive final position. AFIK there is no software at present to do implement this but it is certainly a cute idea. I believe you could probably mock this up with HAL components (possibly might need to add a new math function along the way, but the set of components is getting quite robust) to merge the two encoder inputs. The math shouldn't be very complex. But, a caution: Knowing position via the glass scales doesn't eliminate the backlash problem. The real problem is that the position of the table is not constrained in both directions by the servo. it is only constrained in one direction at a time, and the motor cannot hop from one side of the backlash to the other instantly, therefore cutting forces and inertia can flip it from one side to the other faster than the motor can compensate. This can lead to messed-up parts, broken tools and general foul language around the shop. So, don't think the glass scales are some kind of panacea that allows you to do precision work with sloppy leadscrews. Jon We have (safely) assumed so far, that WGB isn't using stepper motors. Because a fairly stiff mechanical system is necessary for a good finish, and a rotary encoder can easily put a ballscrew well within the positional uncertainty region, which affordable scales can't do, I find it hard to justify the effort to be able to mix feedback. It would be interesting to find the least invasive ways to eliminate the backlash. I don't know what it's called, but I have seen a system that utilizes matching sections of a leadscrew as caged rollers arranged around the circumference of the leadscrew, similar to a planetary gear. Two caged assemblies in tension would take up the backlash. I like this design because it looks like it could be made by most home shops at reasonable cost. I tend to like solutions that trade time and talent for cash. -- Kirk Wallace (California, USA http://www.wallacecompany.com/machine_shop/ Hardinge HNC lathe, Bridgeport mill conversion, doing XY now, Zubal lathe conversion pending) - Check out the new SourceForge.net Marketplace. It's the best place to buy or sell services for just about anything Open Source. http://ad.doubleclick.net/clk;164216239;13503038;w?http://sf.net/marketplace ___ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users
Re: [Emc-users] DRO scales and EMC2
The most satisfactory tuning I've gotten on loose ballscrews , i.e. 0.003, is with a 2500 cpr encoder on the end of the ball screw. In order of increasing ease of tuning, smoothness, etc. the linear scale was worst, an encoder on the servo motor shaft was next and the best results were obtained with the encoder coupled to the end of the ball screw. I've not tried other resolution encoders nor attempted to tighten up the backlash between the servo motor and the ball screw ( the drive is a 2:1 timing belt into a worm gear ). Sorry I can't be more helpful. $$$ spent in the right place can be really effective. Good Luck. Dave On Jan 8, 2008, at 10:45 AM, Kirk Wallace wrote: On Tue, 2008-01-08 at 11:16 -0600, Witek GB wrote: I am a newbie to EMC2 but not to machining. I am thinking about creating a CNC Mill with EMC2 however, my problem is with backlash compensation. Is it possible to use DRO scales with EMC2 to have the feed back that would allow you to know exactly the tool position? Thanks WGB I recently set up my Bridgeport with linear scales, new ballscrews with .002 backlash and EMC. It was basically un-tunable. For my DC servo system at least, at the beginning of a move, feedback would indicate no movement, so EMC would crank up the signal to the servo until the backlash was taken up. Then the feedback would indicate too much movement and pullback the signal, creating an oscillation. Kind of like trying to tow a car with a rope. I could tune in reasonable performance, which always stopped within the accuracy of my scales (.0005), but the accuracy during the path was awful, and slow speeds were un-tunable. Currently, the only practical way to live with backlash, is to have an encoder connected to the servo motor in a way that has no backlash (on the motor shaft, or by belt). Then tune out the backlash with EMC's backlash compensation feature. It would be nice to have a system that could mix the motor feedback and the linear feedback, but so far, the current solution works well enough to not justify the effort to change it. I am guessing, but I think the current system works because backlash is fairly predictable, and encoders generally have a much higher resolution than linear scales. You can see my project at the link below. -- Kirk Wallace (California, USA http://www.wallacecompany.com/machine_shop/ Hardinge HNC lathe, Bridgeport mill conversion, doing XY now, Zubal lathe conversion pending) -- --- Check out the new SourceForge.net Marketplace. It's the best place to buy or sell services for just about anything Open Source. http://ad.doubleclick.net/clk;164216239;13503038;w?http://sf.net/ marketplace ___ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users - Check out the new SourceForge.net Marketplace. It's the best place to buy or sell services for just about anything Open Source. http://ad.doubleclick.net/clk;164216239;13503038;w?http://sf.net/marketplace ___ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users
Re: [Emc-users] DRO scales and EMC2
for the rest of you, i think kirk is talking about this: http://www.hobbycnc.hu/CNC/Otletek/Otletek.htm anyone speak hungarian? On Tue, 8 Jan 2008, Kirk Wallace wrote: It would be interesting to find the least invasive ways to eliminate the backlash. I don't know what it's called, but I have seen a system that utilizes matching sections of a leadscrew as caged rollers arranged around the circumference of the leadscrew, similar to a planetary gear. Two caged assemblies in tension would take up the backlash. I like this design because it looks like it could be made by most home shops at reasonable cost. I tend to like solutions that trade time and talent for cash. - Check out the new SourceForge.net Marketplace. It's the best place to buy or sell services for just about anything Open Source. http://ad.doubleclick.net/clk;164216239;13503038;w?http://sf.net/marketplace ___ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users
Re: [Emc-users] DRO scales and EMC2
Hey Kirk what kind of linear encoders were u using? Do you know what kind of signal they produce ? square edge (quad) or sin ? Sin wave signal encoders can be subdivided to produce more counts. .0005 inches per count is too course for pid tuning. Also I found that when i stopped the slide would very slowly oscillate. eventually I decided it was because The PID component which has a deadband setting didn't work as I thought it would. If I remember right it kills the P but not the I or P. so the I and D would slowly more the motor though the backlash till it moved the slide past the deadzone. But my backlash was something like .005-.007 , probably more because the encoder head was not bolted down very well. This was just a test-I wanted to see something move! When I finish the machine The rotary encoders will be on the motor (Thats the oem way), and I will fix the backlash, which I'm pretty sure is a bearing. I will also toy with dual encoders since the mounts are there using t he idea talked about - setting P with the linear scale (at least .0001 inches per count I would like more) and I and D with the rotary. At the very least i will try mapping the back lash with the linear scale. Chris Morley From: [EMAIL PROTECTED] Date: Tue, 8 Jan 2008 11:33:36 -0800 To: emc-users@lists.sourceforge.net Subject: Re: [Emc-users] DRO scales and EMC2 The most satisfactory tuning I've gotten on loose ballscrews , i.e. 0.003, is with a 2500 cpr encoder on the end of the ball screw. In order of increasing ease of tuning, smoothness, etc. the linear scale was worst, an encoder on the servo motor shaft was next and the best results were obtained with the encoder coupled to the end of the ball screw. I've not tried other resolution encoders nor attempted to tighten up the backlash between the servo motor and the ball screw ( the drive is a 2:1 timing belt into a worm gear ). Sorry I can't be more helpful. $$$ spent in the right place can be really effective. Good Luck. Dave On Jan 8, 2008, at 10:45 AM, Kirk Wallace wrote: On Tue, 2008-01-08 at 11:16 -0600, Witek GB wrote: I am a newbie to EMC2 but not to machining. I am thinking about creating a CNC Mill with EMC2 however, my problem is with backlash compensation. Is it possible to use DRO scales with EMC2 to have the feed back that would allow you to know exactly the tool position? Thanks WGB I recently set up my Bridgeport with linear scales, new ballscrews with .002 backlash and EMC. It was basically un-tunable. For my DC servo system at least, at the beginning of a move, feedback would indicate no movement, so EMC would crank up the signal to the servo until the backlash was taken up. Then the feedback would indicate too much movement and pullback the signal, creating an oscillation. Kind of like trying to tow a car with a rope. I could tune in reasonable performance, which always stopped within the accuracy of my scales (.0005), but the accuracy during the path was awful, and slow speeds were un-tunable. Currently, the only practical way to live with backlash, is to have an encoder connected to the servo motor in a way that has no backlash (on the motor shaft, or by belt). Then tune out the backlash with EMC's backlash compensation feature. It would be nice to have a system that could mix the motor feedback and the linear feedback, but so far, the current solution works well enough to not justify the effort to change it. I am guessing, but I think the current system works because backlash is fairly predictable, and encoders generally have a much higher resolution than linear scales. You can see my project at the link below. -- Kirk Wallace (California, USA http://www.wallacecompany.com/machine_shop/ Hardinge HNC lathe, Bridgeport mill conversion, doing XY now, Zubal lathe conversion pending) -- --- Check out the new SourceForge.net Marketplace. It's the best place to buy or sell services for just about anything Open Source. http://ad.doubleclick.net/clk;164216239;13503038;w?http://sf.net/ marketplace ___ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users - Check out the new SourceForge.net Marketplace. It's the best place to buy or sell services for just about anything Open Source. http://ad.doubleclick.net/clk;164216239;13503038;w?http://sf.net/marketplace ___ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users _ Read what Santa`s been up to! For all the latest, visit
Re: [Emc-users] DRO scales and EMC2
On Wed, 2008-01-09 at 03:23 +, Chris Morley wrote: Hey Kirk what kind of linear encoders were u using? Do you know what kind of signal they produce ? square edge (quad) or sin ? My scales have quadrature output. Sin wave signal encoders can be subdivided to produce more counts. .0005 inches per count is too course for pid tuning. I agree. Also I found that when i stopped the slide would very slowly oscillate. eventually I decided it was because The PID component which has a deadband setting didn't work as I thought it would. If I remember right it kills the P but not the I or P. so the I and D would slowly more the motor though the backlash till it moved the slide past the deadzone. I played with the deadband a little at 1e-06 and 1e-05, but apparently these numbers are way too small for my .0005 scales. But my backlash was something like .005-.007 , probably more because the encoder head was not bolted down very well. This was just a test-I wanted to see something move! Same here. When I finish the machine The rotary encoders will be on the motor (Thats the oem way), and I will fix the backlash, which I'm pretty sure is a bearing. My Hardinge lathe had resolvers and tachometers on the ballscrew ends, which I replaced with with US Digital 10k pulses/rev. I am using the same arrangement on the Bridgeport. I see it as a compromise between the servo motor shaft and the real table position that the linear scale measures. After fussing with the linear scale without much success, I jury-rigged an encoder on the servo motor shaft to see if my tuning problems were from the linear scale, and it tuned well. (see fourth picture down, here: http://www.wallacecompany.com/machine_shop/bridgeport/ ) I will also toy with dual encoders since the mounts are there using the idea talked about - setting P with the linear scale (at least .0001 inches per count I would like more) and I and D with the rotary. At the very least i will try mapping the back lash with the linear scale. Chris Morley I was hoping to do more with the linear scales, but at this point, it is much more important for me to get the mill into production. Though it looks like you will be finishing the work for me. Please keep us informed. I wonder how linear scales and measuring equipment are checked or calibrated? It would be nice to have an independent means to classify and measure ballscrew error and backlash. -- Kirk Wallace (California, USA http://www.wallacecompany.com/machine_shop/ Hardinge HNC lathe, Bridgeport mill conversion, doing XY now, Zubal lathe conversion pending) - Check out the new SourceForge.net Marketplace. It's the best place to buy or sell services for just about anything Open Source. http://ad.doubleclick.net/clk;164216239;13503038;w?http://sf.net/marketplace ___ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users