[CNC] CNC from sketch to part the way I do it

[Dan_S]

I am using both a stepper and servos on my mill, all with a single loop via magnetic scales. All of the velocity feedback that is required comes from the magnetic scales, there is no need for an encoder on the motor unless the servo drive requires it. There is no under or over shoot, it goes where it is told to go and is accurate to within plus/minus 1 pulse. In my case, 1 pulse is 0.000039 inches (1 micron).

ho are you dealing with screw flex and twist, and all the other forms of backlash, with a single loop sevo system? Are you deriving the velocity and acceleration inputs from the position inputs of the scales?

it's good to see more programmers showing interested in hobby cnc and I hope they can get something to market in an affordable price range, I

I don't think it's a software issue, as much as it is the nature of the beast. with all the possible hardware combinations I think it would be cost prohibitive to design and build a control that handles everything really well in a plug and play fashion. The number one thing the industry does, is buy matched drive and motor sets. They are essentially plug and play, because someone did the toning for you, but they are super expensive and only work well as a matched set.

 

[JimDawson]

ho are you dealing with screw flex and twist, and all the other forms of backlash, with a single loop sevo system? Are you deriving the velocity and acceleration inputs from the position inputs of the scales?

Yes, the only position inputs the controller sees is the feedback from the scales. The update rate is so fast (about 62 uS) that the motion profile is very smooth. Of course, you have to properly tune the system for the mass and motors to get things to work correctly. This compensates all of the backlash and other errors, the controller calculates everything by position over time.

I can usually manually tune a system in about 5 minutes.

 

[jumps4]

In this video I try to show how I adjust my gigs without over-tightening them
I hope this makes sense, I keep saying 1.8 or 1. something as an amount of travel, I mean 1.8 of a thousandth
sometimes I have problems expressing to others what I'm thinking. the brains running I just cant get my mouth in the right gear.
thanks for viewing
steve

[video=youtube;S6RELpadoTQ]

 

[angelfj1]

In this video I'm going to start showing what I have learned about gibb adjustment and backlash compensation. i'll have to break it up into a few videos to keep them short, that away I'm already set up to do what I want to show.
I'm not trying to be a know it all and any corrections are more than welcome.
hobby cnc machines are different in the fact they have to deal with a lot more friction and adhesion problems not encountered in factory machines with roller linear rails, guide ways and massive ball screws . there are ways to improve these faults but not eliminate them totally in a hobby cnc conversion.
I chased my own tail for years until I understood what was happening under my mill table.
In this first video I talk about what I have learned about adhesion, torsion, friction, flex and stored motion
thanks for viewing
steve

[video=youtube;CfsiZ-6D0W4]http://www.youtube.com/watch?v=CfsiZ-6D0W4&feature=youtu.be[/video]

Steve: Not to worry about having a college degree or not. You have demonstrated that you are a true professor regarding the operation of these hobby mills.
Two words come to mind; brilliant and practical.

Keep it coming.

Frank
)

 

[angelfj1]

I did some research about a year ago on direct reading scales in each axis directly reading table movement and their use with mach3 and emc2 and at that time the only controller card I found that would support linear scales with custom written software was kflop by Dynomotion. I spoke to the man that designed the card and wrote the software for it and he said he could write me the software but the best mach3 could do with the information was to stop if a preset error level was triggered. if mach3 thought it was at one reading and the scales showed a different reading it would stop mach3 from running. at this time mach3 would not support error correction by moving the axis to where it should be or stop the travel if it is going to overshoot. I figured I could do that by measuring the part after the software ran and didn't invest in it until better software comes around at a hobby price. maybe mach4 will be able to self correct but it's still in beta. when they finally figure out how to do that the sloppiest machine built will make great parts because it wont allow an error to take place. that's going to require a lot of processor speed and programing to look ahead and determine prior to it taking place there will be a problem.
steve

Steve: I'm ordering a G0704 this week, before Grizzly's 10% sale is over. Before I order the components for a CNC conversion, I intend to give Dan (Hoss) a call and ask about his new , not released yet (PhaseIII) CNC conversion for the 704 type mills. From what I have read so far, he intends to use double ball nuts. Since I don't really know what this means, I was wondering if double ball nuts are used to address some of the problems you have been discussing in these latest posts, e.g. error levels, etc. What do you think?

Frank
)

 

[jumps4]

hi frank
I have single ball-nuts in my machines and they all seem to have about .0015 thousands backlash. for my use and the fact mach3 compensates for this I could not see the expense of 2 ball-nuts (plus loss of axis travel to make room for them ) unless I was going to larger screws to eliminate as much torque in the screw as possible also. If I was to build another mill I'd put the screws outside the table instead of under so I could use a 25mm screw and then yes double ball-nuts. Just adding my 40lb vise to the table weight adds another .0015 of backlash due to the screw twisting.
I try to keep myself in the lower cost hobby use of cnc and chasing the last little bit of accuracy will have you pulling your hair out thinking about room temperature, material temperature, load on table, cutter force.
It's more knowing your machine than buying the best parts. I'm not making parts that require perfection in every area so if I really need a perfect circle in an exact spot I'll have the cnc cut it small and then I'll pull out the boring bar to finish it. I use my cnc to get the bulk of the work done and then I bring it into spec as required. It is easy to make certain sections of your part undersize by telling your cam software your cutter is a few thousands larger than it really is. this leaves material for you to remove later to come into spec. you can always take more off but putting it back on isn't that easy.
steve

 

[angelfj1]

hi frank
I have single ball-nuts in my machines and they all seem to have about .0015 thousands backlash. for my use and the fact mach3 compensates for this I could not see the expense of 2 ball-nuts (plus loss of axis travel to make room for them ) unless I was going to larger screws to eliminate as much torque in the screw as possible also. If I was to build another mill I'd put the screws outside the table instead of under so I could use a 25mm screw and then yes double ball-nuts. Just adding my 40lb vise to the table weight adds another .0015 of backlash due to the screw twisting.
I try to keep myself in the lower cost hobby use of cnc and chasing the last little bit of accuracy will have you pulling your hair out thinking about room temperature, material temperature, load on table, cutter force.
It's more knowing your machine than buying the best parts. I'm not making parts that require perfection in every area so if I really need a perfect circle in an exact spot I'll have the cnc cut it small and then I'll pull out the boring bar to finish it. I use my cnc to get the bulk of the work done and then I bring it into spec as required. It is easy to make certain sections of your part undersize by telling your cam software your cutterSteve is a few thousands larger than it really is. this leaves material for you to remove later to come into spec. you can always take more off but putting it back on isn't that easy.
steve

Thanks Steve. As usual you make perfect sense. Can you say a little more about why a double ball nut is used?

Frank

 

[jumps4]

sorry Frank
after reading my reply again I see I never answered your question. :nuts:
my single ball-nuts have .0015 backlash because of clearance between the balls and the housing and screw. the ball-nuts with circuit tubes you can remove. you purchase larger balls to remove this backlash. double ball-nuts have a spring in between the 2 nuts forcing them away from each other to take up this backlash. double circuit ball nuts are different also because they have two circuits of balls and are built for heavier loads.
steve

 

[jumps4]

while on the subject of backlash I want to mention these BK style preload bearings for the end of the ballscrew from china. All the parts are there to make a really good bearing but they come out of the factory defective ( I have installed probably 20 of them ) the inner and outer races of the opposing bearings are the same height so without a spacer installed inside between the two outer races you cannot add preload to the bearing by tightening the nut on the ballscrew. this only makes the two inner races contact each other before pressure is being applied to the outer races. a thin spacer installed inside the housing between the two outer races fixes this problem and removes all the backlash from the bearing.
steve

 

[angelfj1]

while on the subject of backlash I want to mention these BK style preload bearings for the end of the ballscrew from china. All the parts are there to make a really good bearing but they come out of the factory defective ( I have installed probably 20 of them ) the inner and outer races of the opposing bearings are the same height so without a spacer installed inside between the two outer races you cannot add preload to the bearing by tightening the nut on the ballscrew. this only makes the two inner races contact each other before pressure is being applied to the outer races. a thin spacer installed inside the housing between the two outer races fixes this problem and removes all the backlash from the bearing.
steve

Steve, this is very interesting but its a lot of trouble and $$$ for a hobby machine. You have already shown how Mach3 can compensate for backlash!!!

Frank