[Winnfield]

That's the part that I find confusing. Given the limitations of stepper control, and the lack of positioning feedback - actual by linear scale, or theoretical by rotary encoder - how can Mach know how to "stack" movements. I guess I am not getting that concept.

[SwampDonkey]

I really need to achieve accuracy to within .001". Would the G540 with .2 pitch ballscrews be able to do this? Also, would 270oz steppers have enough torque for the LMS SX2?

Thanks guys, learning a ton!

[SCzEngrgGroup]

I really need to achieve accuracy to within .001". Would the G540 with .2 ballscrews be able to do this? Also, would 270oz steppers have enough torque for the LMS SX2?

Thanks guys, learning a ton!

Depends very much on the parts you plan to make. Unless you:

a) Are making very small parts, or:

b) You use high quality ground screws with zero-backlash ballnuts (which will cost more than the SX2), AND use servos (not steppers) with linear encoders mounted to the machine, AND spend a great deal of time, effort and money over-coming the many short-comings of the basic machine (basic geometry, poor quality ways and gibs, lack of overall rigidity, etc.) the answer is probably no.

Flex in an X2 can easily be several thou. Geometry errors, due to the poor machining, can add that much more. Backlash due to non-square, non-parallel ways can add that much more. Ballscrew lead errors can add that much more. And on, and on, and on.

You'd be FAR better off starting with a much higher quality base machine. In the end, it will cost you far more to try to do it "on the cheap".

Regards,
Ray L.

[SwampDonkey]

Depends very much on the parts you plan to make. Unless you:

a) Are making very small parts, or:

b) You use high quality ground screws with zero-backlash ballnuts (which will cost more than the SX2), AND use servos (not steppers) with linear encoders mounted to the machine, AND spend a great deal of time, effort and money over-coming the many short-comings of the basic machine (basic geometry, poor quality ways and gibs, lack of overall rigidity, etc.) the answer is probably no.

Flex in an X2 can easily be several thou. Geometry errors, due to the poor machining, can add that much more. Backlash due to non-square, non-parallel ways can add that much more. Ballscrew lead errors can add that much more. And on, and on, and on.

You'd be FAR better off starting with a much higher quality base machine. In the end, it will cost you far more to try to do it "on the cheap".

Regards,
Ray L.

Thanks Ray, but unfortunately a better/ bigger base machine is completely impossible. Money is tight on my end and im waiting paycheck to paycheck to buy what I can to make this mill perform at its very best.

Im aware of the flex in the frame of the LMS SX2, but plan to gusset and brace the hell out of it to minimize this. I have a very heavy (450lb) I6 tractor engine block with perfectly flat surface I plan to mount the column against to hold it steady and absorb vibrations. I also have some brass stock with which I will machine new gibs with closer tolerances. The whole thing will be shimmed and tuned to death. I also have a nice 3" thick slab of black marble to bolt it down to. I also plan to run it slowly with sharp bits to fight deflection, along with the 5000rpm spindle belt kit.

Im using Thompson .2 pitch ballscrews which seem to have good reviews. I also plan on using the Gecko G540 along with Gecko 270oz steppers. Any idea of what I can expect out of this setup?

Thanks again!

[Larken]

I really need to achieve accuracy to within .001". Would the G540 with .2 pitch ballscrews be able to do this? Also, would 270oz steppers have enough torque for the LMS SX2?

Thanks guys, learning a ton!

Yes , 2000 steps/turn and 5Tpi gives 10,000 steps/inch. So you should easily get a solid repeatable resolution of 0.001" (on the screw anyway, not counting linear backlash or friction)
But don't expect a 0.0001" resolution. Openloop microstepping beyond 1/8 step is not effective if there is any friction. The motor has more torque close to the poles than between the (50) poles.

BTW, regular china (chopper) drives are no where as smooth as a PWM drive like the gecko (but the motor does run alot cooler). I haven't tested a "Digital" drive though.

[SwampDonkey]

Yes , 2000 steps/turn and 5Tpi gives 10,000 steps/inch. So you should easily get a solid repeatable resolution of 0.001" (on the screw anyway, not counting linear backlash or friction)
But don't expect a 0.0001" resolution. Openloop microstepping beyond 1/8 step is not effective if there is any friction. The motor has more torque close to the poles than between the (50) poles.

BTW, regular china (chopper) drives are no where as smooth as a PWM drive like the gecko (but the motor does run alot cooler). I haven't tested a "Digital" drive though.

Yeah it looks like China is getting into the digital stepper driver game. Makes me wonder how these stack up against the keling:

nema23 Stepping motor driver DM542, digital,peak current4.2A ,18-50VDC CE | eBay

[SCzEngrgGroup]

"Yes , 2000 steps/turn and 5Tpi gives 10,000 steps/inch. So you should easily get a solid repeatable resolution of 0.001" (on the screw anyway, not counting linear backlash or friction)" - "easily"??

If you build a machine with rolled ballscrews with a maximum lead error of AT LEAST +/-0.003" per foot, how do you "easily get a solid repeatable resolution of 0.001"? This means, essentially, that the screw position can be off by +0.003" at one position, and off by -0.003" at another position an inch away from that first position. And ground screws are out of the question, as one of them would cost more than the whole machine. And without double ballnuts (which are hard to fit into an X2), or true zero backlash nuts and screws ($$$), the screw WILL have measurable backlash.

If you get an X2 where the X and Y ways in the saddle have been machined 0.005" out of square (which is not at all unusual) what are you going to do?

If you get a saddle, table, column or head where the ways are not milled parallel, or even straight, (both of which are VERY common on X2s) what are you going to do?

If you get a spindle with 0.0005" runout (which near the BEST you could hope for) what are you going to do?

A 10 degree change in the temperature of the machine will change the length of the table, the column and the screws by almost 0.001"! What are you going to do about that?

Even after it's been "shimmed and tuned to death", these problems WILL still exist.

Getting TRUE 0.001" accuracy AND repeatability (and one without the other is kinda useless...), over more than a very short distance, is NOT easy, and NOT cheap. Geometry errors in the basic mill components (which is almost ALWAYS present in these machines) cannot reasonably be corrected for. Lead error in the screws (which is ALWAYS present, even in the $$$ ones) , cannot reasonably be corrected for. Spindle runout cannot reasonably be corrected for. Getting these factors, and the many other important ones, individually down to the 0.001" range is difficult. Getting the TRUE overall system performance into that range is VERY difficult. On an X2, it's probably near impossible.

You need to set realistic expectations. If you REALLY need that kind of accuracy, then the X2 is not the place to start. OTOH, I'd bet you don't REALLY need that kind of accuracy, other than perhaps for a few specific operations, like boring bearing pockets.

Adding up all the theoretical capabilities of the machine based on individual component specs will give you numbers that are wildly optimistic compared to the machines actual, real-world capabilities. Commercial machines get their specs under 0.001" by using FAR more expensive components (a single VMC ballscrew can cost thousands of $), screw mapping, closed loop positioning, temperature controlled environments, and many other "tricks".

Regards,
Ray L.

[SwampDonkey]

"Yes , 2000 steps/turn and 5Tpi gives 10,000 steps/inch. So you should easily get a solid repeatable resolution of 0.001" (on the screw anyway, not counting linear backlash or friction)" - "easily"??

If you build a machine with rolled ballscrews with a maximum lead error of AT LEAST +/-0.003" per foot, how do you "easily get a solid repeatable resolution of 0.001"? This means, essentially, that the screw position can be off by +0.003" at one position, and off by -0.003" at another position an inch away from that first position. And ground screws are out of the question, as one of them would cost more than the whole machine. And without double ballnuts (which are hard to fit into an X2), or true zero backlash nuts and screws ($$$), the screw WILL have measurable backlash.

If you get an X2 where the X and Y ways in the saddle have been machined 0.005" out of square (which is not at all unusual) what are you going to do?

If you get a saddle, table, column or head where the ways are not milled parallel, or even straight, (both of which are VERY common on X2s) what are you going to do?

If you get a spindle with 0.0005" runout (which near the BEST you could hope for) what are you going to do?

A 10 degree change in the temperature of the machine will change the length of the table, the column and the screws by almost 0.001"! What are you going to do about that?

Even after it's been "shimmed and tuned to death", these problems WILL still exist.

Getting TRUE 0.001" accuracy AND repeatability (and one without the other is kinda useless...), over more than a very short distance, is NOT easy, and NOT cheap. Geometry errors in the basic mill components (which is almost ALWAYS present in these machines) cannot reasonably be corrected for. Lead error in the screws (which is ALWAYS present, even in the $$$ ones) , cannot reasonably be corrected for. Spindle runout cannot reasonably be corrected for. Getting these factors, and the many other important ones, individually down to the 0.001" range is difficult. Getting the TRUE overall system performance into that range is VERY difficult. On an X2, it's probably near impossible.

You need to set realistic expectations. If you REALLY need that kind of accuracy, then the X2 is not the place to start. OTOH, I'd bet you don't REALLY need that kind of accuracy, other than perhaps for a few specific operations, like boring bearing pockets.

Adding up all the theoretical capabilities of the machine based on individual component specs will give you numbers that are wildly optimistic compared to the machines actual, real-world capabilities. Commercial machines get their specs under 0.001" by using FAR more expensive components (a single VMC ballscrew can cost thousands of $), screw mapping, closed loop positioning, temperature controlled environments, and many other "tricks".

Regards,
Ray L.

Thanks for the explanation, though preferred my blind optimism. I suppose my expectations might be unrealistic, but this may be a good machine to get started with. I plan to build small (working) engines, RC, and Airsoft parts. These likely wont require the tolerances im after. Ive been able to manually mill parts to within a thou or two, but my cnc theory needs expanding. How effective would Mach3 be in compensating for the inherent backlash?

[SCzEngrgGroup]

Compensating for backlash is far from the same as having zero backlash. Machining forces will still push and pull the machine within the range of the backlash, degrading accuracy and finish quality, and backlash compensation will be of no help. All it can do is minimize the position error when the toolpath direction reverses.

The next step is being very aware of the backlash, and programming your toolpaths to minimize the effects. This is a major PITA, and, if you're using CAM, as you almost certainly will want to, it's really not possible to do this. And, if you have backlash, you want to avoid climb milling as much as possible.

Regards,
Ray L.

[SwampDonkey]

Compensating for backlash is far from the same as having zero backlash. Machining forces will still push and pull the machine within the range of the backlash, degrading accuracy and finish quality, and backlash compensation will be of no help. All it can do is minimize the position error when the toolpath direction reverses.

The next step is being very aware of the backlash, and programming your toolpaths to minimize the effects. This is a major PITA, and, if you're using CAM, as you almost certainly will want to, it's really not possible to do this. And, if you have backlash, you want to avoid climb milling as much as possible.

Regards,
Ray L.

Thats strange. Ive seen work done on CNC x2 and SX2 mills and the parts had good fit and finish. Bolt holes lined up, squares were square, rounds were round. Also tiny lettering and logos that certainly appeared flawless. It would be interesting to see what kind of tolerances they achieved. We have a few LMS mini mill conversions like mine here on the forum, I'll poke at them and see what kind of numbers they're getting.

[Fixittt]

swamp..... getting positional tolerances or near zero backlash is like was said, components in your machine. IMHO anything with dovetails is close to imposable to get zero backlash on (I could be wrong) its going to be a trade off on friction from the gibs or twisting of the slides.

holding tolerances on a part is a whole different animal. The machine is a tool. Someone with alot of experience and a crappy tool can still hit their mark. Someone with an expensive tool but lacks the experience may never hit the mark.

To give you an example. I have a project right now, cutting some acrylic parts. less then 2mms thick with about a 60mm diameter. I have to cut out a rectangle in a certain section of these parts. The section is small but has to be placed exactly right. within .05mm or .002. I have been working on a fixture and method for a week now as the parts are surface sensitive and bow very easy when clamped down. every time I change a tool I have to start all over with the toolpath calculating because no two tools cut exactly the same. Its a pain sometimes.

[SCzEngrgGroup]

"I'll poke at them and see what kind of numbers they're getting." - and when they give you some impressive numbers, ask them how they've actually *verified* those numbers, and how they've eliminated the error sources I've pointed out. I see people on here bragging all the time about how precise and accurate their machines are, and the claims are simply not credible. When someone claims to be getting 0.0001" precision on a machine with rolled ballscrews, they are clearly going by a simplistic calculated resolution, and have never actually measured. Or, they're basing their claim on measuring the ends of a 1-2-3 block or something similar, and calibrating the machine to the number they want to see, with no absolute reference. Getting precise positioning at two defined points on a single axis, a few inches apart is easy. Getting precise positioning at ANY point in 3D space is incredibly difficult. Few people here even have any equipment capable of measuring such small dimensions accurately. A $20 digital scale with *claimed* 0.0005" resolution will NOT do it!

Regards,
Ray L.

[waltpermenter]

We have a few LMS mini mill conversions like mine here on the forum, I'll poke at them and see what kind of numbers they're getting.

That's the best course you could take, get some FACTS about the machine you are using from people with ACTUAL experience with the SX2. djbird3 has a good thread. http://www.cnczone.com/forums/bencht...onversion.html and could tell you what to expect. If you were to listen to SOME people you'd be better off using a hammer and chisel than a benchtop mill but facts are people ARE getting good results though none claim .0001 on an SX2 so I don't know where that figure gets pulled out from.
walt

[Larken]

"Yes , 2000 steps/turn and 5Tpi gives 10,000 steps/inch. So you should easily get a solid repeatable resolution of 0.001" (on the screw anyway, not counting linear backlash or friction)" - "easily"??


Getting TRUE 0.001" accuracy AND repeatability (and one without the other is kinda useless...), over more than a very short distance, is NOT easy, and NOT cheap.
Regards,
Ray L.

Yes, not accuracy, just repeatability. Accuracy costs a lot more.

But repeatability is more important than accuracy IMO.
On any machine there will be cutter flex for a given speed. So most people cut a part, measure it and make adjustments to the file. But then they want all the rest to be exactly the same.

(The accuracy of your measuring tools is the most important.)

Yeah it looks like China is getting into the digital stepper driver game. Makes me wonder how these stack up against the keling:

Keling drives are made in china

Larry K

[Mad Welder]

Yeah it looks like China is getting into the digital stepper driver game. Makes me wonder how these stack up against the keling.........

Yup I was just about to suggest that the Keling Drivers are made in China too but Larken said it first and the Keling drivers I have are KL-8070D Digital drivers Digital Stepper Motor Driver and the funny thing is on the ebay link you posted the drives look similar to Kelings makes me think that one manufacturer is supplying to a lot of vendors....

and it really would be good advice to read up in the X2 threads as to the factual tolerences you can expect to achieve repeatedly

[SwampDonkey]

That's the best course you could take, get some FACTS about the machine you are using from people with ACTUAL experience with the SX2. djbird3 has a good thread. http://www.cnczone.com/forums/bencht...onversion.html and could tell you what to expect. If you were to listen to SOME people you'd be better off using a hammer and chisel than a benchtop mill but facts are people ARE getting good results though none claim .0001 on an SX2 so I don't know where that figure gets pulled out from.
walt

Djbird3's build is one of the things I studied before taking this on, great stuff! As long as I can get drivers and steppers that exceed the capabilities of my machine, I will feel like Im doing my best. Im sure the Gecko 540 can do that with its 10x microstepping, which works out to .0001 on my .2 pitch ballscrews, correct?
Eventually I will invest in a much more rigid machine and likely look into servos or a reduction drive, maybe higher-pitched screws. For now its going to be all about squeezing every ounce of performance out of this LMS SX2.

Thanks,

[Winnfield]

Yes, your electronics can certainly exploit every bit of your machine's potential. Your job will be to mitigate the mechanical limitations.

[SwampDonkey]

Yes, your electronics can certainly exploit every bit of your machine's potential. Your job will be to mitigate the mechanical limitations.

I can do that. I am now looking into those insanely expensive zero-backlash ball screws. Looks like they can be had for less than I thought. So how do I figure out the theoretical limits of the G540 and the steppers?

[Winnfield]

Truly zero-backlash ground ballscrews are extremely expensive - thousands, really, for a set of three the size you need. I think that is the variety Ray L. was referring to, and what would be necessary to make a machine as accurate as .001". His point, though, and others' as well, I think is that something like that would be completely wasted on a machine that has much larger problems to begin with that will prohibit ever reaching your accuracy requirements.

You can certainly use some of the better quality rolled ballscrews, which are made in China, ad can be found on ebay. They may be sold as "zero backlash", but I assure you, they are not.

[SwampDonkey]

Truly zero-backlash ground ballscrews are extremely expensive - thousands, really, for a set of three the size you need. I think that is the variety Ray L. was referring to, and what would be necessary to make a machine as accurate as .001". His point, though, and others' as well, I think is that something like that would be completely wasted on a machine that has much larger problems to begin with that will prohibit ever reaching your accuracy requirements.

You can certainly use some of the better quality rolled ballscrews, which are made in China, ad can be found on ebay. They may be sold as "zero backlash", but I assure you, they are not.

Right, someday I may upgrade everything. For right now I want to make sure the G540 is capable of driving steppers to .001 resolutions. I can potentially replace the dovetails with precision rails, modify the head/ spindle, upgrade the ball screws to precision ground, but I really dont want to have to replace the drivers. Ive found some amazing stuff from decommissioned and replaced equipment, much of it practically new.

I believe someone said that the theoretical resolution with .2 pitch ballscrews and the G540 was .001" and .0001" with 10X microstepping?