[bgolash]

Hi Group

I`ve just got a few of the parts to start my gantry cnc wood router.
I have this book as a reference CNC Robotics by Williams. I like
the design with the lead screw placed close to one of the lineal
motion rail assemblies. This design seems easier to construct with
less materials then a gantry router with the lead screw centered underneath.
I am wondering if the CNC Robotics design with one lead screw off-centered
will track the far edge of the gantry as it moves. I`ve provided a few links
to clarify what I`m attempting to convey.
If this makes a difference the footprint of the machine will be 36 X 67.
Thanks for the advice.
Regards Barry

http://www.durhamrobotics.com/
http://www.elx.com.au/images/product...95cb7728779736

[ESjaavik]

Ideally the lead screw should pass through the center of gravity which again should coincide with the center of cutting forces. Try to get as close to this as you can within budget and other constraints. Having the lead screw underneath could easily be as bad as on one side if that brings it as far from the mentioned position. You might just swap unwanted horizontal forces with vertical. For a gantry two screws is the universal remedy. But it costs money and other headaches.

But the one you referred to does not even try to compensate for having only one screw! It should have 2 bearings on the rail on each side, preferrably with some distance between them. And then a fairly stiff frame keeping them rigidly connected. Then even with one screw it will withstand the twisting forces much better. In all probability good enough for your wood router.

So I would go for one screw, longer rails and closer to a square pattern of 4 bearings.

Vertically it looks OK, as there is not a high "tower bridge". It's easier to add lateral stiffness in the non-moving frame than the moving gantry. Both because you don't add to the moving mass, and because it will not restrict the sideways (Y) spindle/tool movement.

That's my opinion, but I think there are now 6000 members on this forum, so you still have 5999 opinions to go.

[bgolash]

"So I would go for one screw, longer rails and closer to a square pattern of 4 bearings."
I was thinking the same thing. I did a test and with a bearing spacing of 12 inches, it
seems to provide good tracking.
Thanks for the advice.

[Cold Fusion]

My X axis ballscrew is offset to one side and it has caused a small amount of problems. I think that if I make a few changes to the gantry bridge it would be fine though. Bottom line is that if you don't use thk and a preloaded ballscrew it is a bad idea.

[bgolash]

Hi Cold Fusion

I was thinking about a spacing between my 1.25 dia.thomson linear bearings of
12 inches. Do you think the spacing would provide good tracking. I`m wondering
what changes you`d make to your design to enhance its function.

Thanks

[bgolash]

This is a direct quote from Thomson`s application engineer. It would
seem my design as is, should be modified.

If I interpret your design correctly, you have shafts spaced 36" apart that are approximately 67" long and you have bearings 12" apart on each rail. That puts the shaft spacing to bearing spacing at 3:1 and that is not safe if you plan on driving from one side only. The stiffness of the frame will have a lot to do with the final performance, but at best, it is a marginal design.
It would be much safer to put a ball screw on each side and link them together or put one screw in the center of the shafts.
You concern of the carriage "racking' as it tries to move under the existing design is well taken.
Our suggestion is to alter the design to either of the above mentioned options.
I have requested that a new set of catalogs be sent to you.
Regards,

[uscra112]

Barry, et.al.

Please forgive me if I horn in here, but I can speak as one whose career has largely been in the machine tool design/build field for the last 25 years or so. In 200+ years of machine tool building, every 3-axis structure imagineable has been tried, and I've probably seen or at least read about all of 'em.

Agreeing with ESjaavik, there has long been the dictum that, if rapid accel-decel forces need controlling, the drive should run thru the center of mass. If cutting thrust forces dominate, then thru the center of the cutting forces. The less rigid the structures, the more critical this becomes.

For a gantry, there almost always has to be a major compromise, because it's hard to design a gantry with a central drive, (it CAN be done, but IMHO the "under-the-table" design isn't it!), and the cutting forces move left to right across the crossrail.

Here's how we do it today. To get good tracking of a gantry, the way bearings on ONE side of the gantry are laid out to have as high an aspect ratio as possible, i.e. the bearing pads that contact the guideway should be as far apart as possible. At least as far apart as your gantry is wide. This keeps all the yaw and pitch motions of the gantry under firm (we say stiff) control as it reverses direction. It does necessitate that the control-side guideway be longer than the bed, but it's a good tradeoff.

With the pitch and yaw under control, we can and do mount the gantry drive right next to the control-side way.

The company I now work for builds some of the most accurate 3-axis machines in the world, and that is exactly how they are laid out. The opposite guideway (we call it the "outboard way") should NOT try to control the gantry in the horizontal plane at all. It basically just holds the outboard end up. We only use ONE bearing on the outboard side.

To get good results when following this principle, you will need to have a good rigid gantry, of course. But that's relatively easy to do.

The spacing of the two ways makes no difference to the tracking. It's the aspect ratio of the control-side way bearing alone that does that.

Many years ago I worked on really big gantries used for aerospace machining. These had crossrails 20-30 feet wide, and could machine a badminton court in one setup if necessary. Their gantries were driven from both ends to keep the yaw under control. Expensive! And a minor nightmare for the controls engineers. But most of the time it worked. On small machines like these, screws on both sides driven from a common motor would be possible, and even adviseable. Although it adds a little to the cost and mechanical complication, it can cut the overall length of the machine bed down a bit.

Cheers!

Phil

[ESjaavik]

@Phil: You're very welcome "horning in" here!
It is a great asset having someone with that kind of background onboard.

Could you elaborate on "aspect ratio"? Would that be the ratio of distance between the bearings on the control-side and the distance from the control side rail to the worst case center of attack for the accelerating and cutting forces? Or to the support bearing?

And it should not matter much if it is an equal sided triangle or a square cornered one? The last one would be a lot easier to design.

This delta shaped gantry is the one I keep coming back to when thinking about how to make a stiff, yet simple design.

Having one screw each side of course adds to the cost, but also to the inertia. This means the size of servo has to be increased. So what if using smaller diameter ballscrews to keep down inertia? Well if the spindle/tool is all the way over to one side, the screw on that side will have to take most of the forces it seems. So it will bear more than half of the total. Is this reasoning correct, if so to what extent?

Also the resonance frequency of the smaller screw will be different, so whip will occur at a lower speed. Which sends us back to 2 screws of almost the same diameter as if we have one. And a motor that needs to be bigger than with one screw. A pro will be that this bigger motor will be less affected by cutting forces than the smaller one.

Please shoot down my reasoning where appropriate. I'll be "building on paper" for a while still before I do a gantry mill. But when I do, I only want to do it one time over.
BTW: My desire is for a high speed small mill for Aluminum and plastics.

[uscra112]

The aspect ratio of a guideway is related to only one way. It is the length of the bearing area divided by the width (or diameter) of the way bar itself.

Thus, if you had a Thomson rail one inch diameter, and two bushings spaced 10 inches apart (outside ends) the aspect ratio would be 10:1. Greater aspect ratio always improves yaw and pitch control.

If your gantry bridge is going to be 10 inches wide, then 10 inches spacing of the bushings on the control side would be good. What you do on the other side doesn't matter an awful lot, so long as your bridge structure is nice and stiff.

We are usually hanging the Z-axis (if there is one) off the front face of the crossrail, so stiffness of the rail in torsion is important. Flat plates are poor in torsion. Tubes, especially round ones, (and the larger the better), are great in torsion. Many full-scale machine tool designs make use of large tubes as the main structural form for this reason. They're also good in bending. For us, the problem is that they're hard to attach things to, unless you've got a welder. So square tubes get more play at our scale.

You're sharp about the screw size and whip. That's been a bugaboo since the dawn of CNC. Small screws can't go fast because of whip, and they're springy to boot. A screw large enough to not whip has so much inertia you can't accelerate and decelerate it. For really large machines, the screw alone is the largest single inertia component, even though the slide may weigh several tons! Keeping the screw SHORT reduces whip a lot, so smaller screws can be used. I've seen some machines that had screw supports spaced out along the length, to control whip. Another thing you ALWAYS do in full-scale machines it to put the screw under tension. That increases the critical speed quite a bit. But of course you need lots of stiffness between the screw bearings or your bed will be a banana. And the bearings have to be ball-bearings, or they'll heat up like crazy.

Time to go home - work has had me for today . . . .

Phil

[jeffs555]

Phil,
Thanks a lot for the excellent insight into machine design. I had read before about only constraining the horizontal movement on one side, but never so authoritatively, and with such detail. I have been designing a machine with screw drive and horizontal way bearings on only one side of the gantry. I was surprised to see that you use only one vertical bearing on the outboard end. I had decided on 2 spaced vertical way bearings on each end of the gantry, because with only one bearing on the outboard end, I was worried about torsion on the gantry. It seems that with only one bearing on the outboard end, it would be hard to build the gantry with enough torsional rigidity to keep it from twisting due to cutting forces when the tool is at the outboard end.

Jeff

[uscra112]

You guys are thinking well.

Actually MACHINE TOOL gantries do use two bearings on the outboard side, for exactly the reason you cite - to control crossrail torsional deflection.

We don't because our machine is actually a coordinate measuring machine. We have no cutting forces to resist, and we want the lightest possible structure so we can accelerate it really fast.

Mirror-imaging the bearing layout from the control side to the outboard side is the right thing for you to do, but the outboard bearings should still not try to constrain the bridge in the horizontal plane, only the vertical. Doing that with Thomson rails is a bit tricky, since the bushings want to constrain in all directions. There is a way. We use it in the probe heads of our machines. Consider a thin, flat spring. It can bend, but it can't be compressed or stretched in it's own plane. If such an element is oriented vertically and parallel to the way, and connects the bridge to the way bushings, you have constraint in the vertical but none in the lateral direction of the horizontal plane. JUST what you wanted! So if the outboard vertical leg of your gantry is a thinnish flat plate, rather than a stiff box structure, (as it should be on the control side), you will have the problem licked. The leg can flex laterally, but will be stiff vertically and also along the torsion axis of the crossrail. 2mm to 4mm aluminum sheet might be just the ticket.

Occurs to me that a good crossrail plan-form would be a right triangle, with one leg parallel to the control-side way, the right-angle leg forming the crossrail, and the hypotenuse behind the crossrail. That, plus a control-side leg that is stiff in torsion around the vertical axis, and you'll be golden.

To return to the original question - now you can drive close to the control-side leg, and nothing will twist or bend very much.

[DR-Motion]

Sorry for not replying sooner but I must have missed this thread.

The off-centered ballscrew drive mentioned by Bgolash is the one I implemented in the Durham Robotics design. The design is based on the adage " ask the experts... then do what they say is impossible." Not to offend anybody, but often, that which works in practice is not necessarily what is predicted by theory.

At this time there are over ten machines in circulation... not one has been reported as having any racking problems. The small machine displayed on the website has a working envelope of 16" x 8" x 4" and is capable of machining aluminum. (there is now a movie of this ).

The secret of success here is the use of massively overspecified linear bearings... and yes, as the experts have pointed out, the use of staggered carriages on the in-board side is used on machine models with 16" or greater Y axis travel.

This is now, IMHO, a proven design and I have no hesitation in recommending its use in simplifying the gantry configuration .

There is at least one major drawback ... the ballscrew is very exposed to cutting debris and swarf. Solution is to use ballscrew covers.

Keep pushing the envelope of the possible... beat the impossible into submission.

regards Gary

[ESjaavik]

The secret of success here is the use of massively overspecified linear bearings... and yes, as the experts have pointed out, the use of staggered carriages on the in-board side is used on machine models with 16" or greater Y axis travel.

This is a bad use of resources IMO. Linear bearings are expensive, machine frame is relatively cheap. So don't use those expensive components to make the machine stiff. Make a stiff frame, and sound distribution of forces, and you can use less massive (cheaper) linear bearings. The fact that no amateur can make linear bearings at home, but most people can make a more rigid frame, makes it even more beneficial to do this.

Try to place such a rail supported only at both ends, and a measuring clock at the center, then push lightly on it with your finger. Then you will realize that the stiffness of such a rail is very low! It needs support along the full length to make a stiff machine. When you write that many are satisfied with what you propose, what did they compare it with? I dare suggest they have no reference.

The problem is that many do not know how best to bring this stiffness into the construction, and that is where uscra112 contributes valueable information.

Your point that the ballscrew should be protected from debris is a very good point to ensure that the inital performance is not degraded over time. And on my current project incorporating such a cover have caused me much headscratching. But there is no way without it.

[dansutula]

Dear Phil & Folks,
It appears that we are all chewing on the same problem with respect to racking. I’m collecting parts and trying to finalize plans for a heavy duty 4x8x1’ router. My latest idea is for a single center screw drive (under the table). I’m thinking about adding a third linear rail (centrally located) on which the ball nut would be guided. The nut would be located significant forward or aft of the normal guide bearings (in this case 4) located on the side rails. The structural members connecting the nut to the gantry would be stiff in tension or compression (spar elements). Thus, the third rail effectively reduces the adverse aspect ratio of the wide gantry. The diagonal elements of the connecting frame provide a "load-leveling" effect. A racking load applied to one side of the gantry will be resolved largely between that side's guide rail and the center rail. Attached is a preliminary illustration of the concept looking from the underside of a gantry table. I have not done the math yet, but conceptually I believe that the concept is valid. Has any body tried this? Will it work? What do you think?
Please note -the attached image is a view of the underside of the router table.
Sincerely,
Dan

[jcc3inc]

Gentlemen,

Interesting group of comments!

As to solving the "racking" or skewing of a gantry, use rack and pinion on both sides of the gantry, either connect the two pinions with a common, stiff shaft, or synchronize the two drive motors.

It sounds like youall are doing a lot of speculating about how stiff a gantry is with a certain design. Why not build your idea of what should be a good design and test it. Lock the drive motor and apply +50# then -50# to one side of your gantry; do the same for the other and measure the deflection. Post your results for all to see so that others may benefit from your efforts.

Of course the TOTAL deflection is measured at the cutter bit; it includes the sum of all spring deflections plus the lost motion. You might say That Is The Bottom Line!

Regards,
Jack C.