Need advice on gantry design

[jcarpenter]

First real post! Looking for advice on moving gantry design.

Proposed machine cutting area: 1220mm x 1220mm cutting area (1/2 plywood sheet)
Materials to be cut: Foam, Hardwood, Plywood, Aluminum to 6mm thick.

Proposed machine parameters:
Floor standing
Moving gantry with high wall frame
Gantry beam length about 1500mm
Twin X axis ballscrews, single Y and Z
Non-production environment
Need decent but not perfect finish on edges of aluminum plate.
Would like to hold plus minus 0.1mm on dimensions

Materials available for the router build:
80/20 extrusion 3030 (76mm x 76mm)
80/20 extrusion 1530 (38mm x 76mm)
6mm aluminum plate
15mm plate
20mm plate

Shop tools available:
20 inch bandsaw
Full size CNC knee mill
Small CNC lathe
300mm x 300mm CNC router
Small shear, small pan break, air compressor, die filer, etc.

Experience:
Amateur CNC machinist. Built a couple of small CNC routers and some custom automated machinery for pool cue
builders. Just enough knowledge to be really dangerous.


First picture is what my design looked like before searching around this forum for a while. Looks like I hit all the usual problems
such as:
Separated twin beam gantry
Ballscrew between gantry beams
Thin Z axis plate
Z axis rails not on Z axis plate
Single gusset plate to hold end extrusions to gantry carriage plate

Probably a lot more problems lurking in that design. So I am starting over. And the first thing I want to do is to settle what the cross section of the
gantry should look like using the materials available to me. The second picture is of some gantry and Z axis plate cross
section views with different configurations of extrusions. Trying to come up with some ideas that combine the recommended
box or L box beam construction with things I have on hand.

Do the drawings make sense? Any of my ideas decent? Got any suggestions?

Thanks,

John C

2321923220

[EddyCurrent]

Consider the information here; http://www.cncroutersource.com/do-it-yourself-CNC-router.html
That should be a guide to the best arrangement.

There was some discussion here about "stiffness", it's a pity more people did not contribute at the time. http://www.mycncuk.com/threads/7155-stiffness-measurements-cnc-mk3?highlight=stiffness+test
There may be some merit at looking at the numbers to see what each design offered.

[jcarpenter]

EddyCurrent, your first reference seems to recommend widely separated unsupported rails for the gantry. That might be better for a smaller CNC than I plan. The second reference concerns measuring the stiffness of the CNC machine after it is built. A great idea but I am far from even starting my build.

What I really was hoping that someone could advise me on what would be a good gantry design (cross section anyway) using square and 1/2 of square extrusions with using thin plate to box in the beams. Without advice from someone with experience I will probably need to do some simplistic FEA on configurations of hollow rectangular tubes to get a comparison of configurations. Even if I had an FEA program that analysis if done accurately will be somewhat of a task. How about a gut feel comparison of the cross section sketches in my original post?

Thanks,

John C

[EddyCurrent]

I think you missed the point in both cases.
The first link was not to show which components to use but rather the best geometry for success.
In the second instance, the stiffness, which you want loads of, was measured after the build, as you say, but that gives us the benefit of looking at the designs to see why those figures turned out as they were.
So now we can either use or avoid those ideas in our own designs.
Somewhere in this forum is a spreadsheet that calculates various gantry designs, you should find that and test your own ideas out with it.

[routercnc]

Here is the calculator I wrote some time ago if you want to compare sections
http://www.mycncuk.com/threads/2214-cnc-machine-stiffness-calculator

Of the designs you have drawn they all have the ballscrew next to the YZ axis which is compromising the gantry stiffness.

A popular design here is the L shaped gantry with ballscrew tucked behind. Away from pc so will post example later or someone might post it.
L

[jcarpenter]

Ok, thanks to both. Sounds like I need to go back to the math for a while.

John C

[jcarpenter]

Got some math working (I think) to compare different Gantry cross sections. Would appreciate any criticisms or comments on my methodology.

I have gotten my static FEA capabilities up and running and will attempt to analyze some different gantry configurations with it. My FEA refuses to deal with complicated aluminum extrusion CAD so I will simplify the extrusions as a hollow rectangular beams with internal cavity adjusted so that the hollow beam and the 80/20 extrusions have the same moments of inertia. Not looking for absolute displacement numbers, only comparisons between configurations.

First image is my starting CAD design for the gantry. Called Gantry_0
Second image is the FEA geometry used.
Third image is the loadcase used. I am fixing the beams ends in space and applying a line load of 500N at the bottom of the Z axis block.
Fourth image is the result. About 0.125mm. Means nothing except as a comparison to results from other configurations.

23233232322323123230

Next is my first configuration change where I box the beams with a plate underneath the rails. Gantry_1. Again the images are Geometry, Loadcase, and results. This time the displacement is about 0.086mm compared to the original of 0.125. So the plate helped some.

232362323523234

My next configuration (Gantry_2) is to box the other side of the gantry and see what I get.

232402324123242

Image progression is the same as above. Resulting displacement is 0.056mm for further improvement in the stiffness of the beam. Hard to see in the results image but the majority of the flex is in the lower beam at the Z axis attachment. See below.

23243

So I will add a half size beam along the lower beam to form an "L" shape to the gantry. Just like the sticky above recommends, haha. More to come.

John C

[EddyCurrent]

A circular tube or L shape are among the best ideas I think. In my own build I used an L shape with a flat plate attached to the front, that seemed to work well.

[jcarpenter]

I added an "L" shaped 1/2 size extrusion to the lower edge of the gantry and ran the deflection calculations again. Got another 20% reduction in the displacement metric. 0.056 to 0.040mm. Not as much reduction as I expected and the gantry is getting pretty heavy. I will try a full size extrusion soon and see how that works as an "L" beam. After that I might try 4 half size extrusions to form the frame of a 250mm square box tube.

23248232492325123250

John C

[EddyCurrent]

I don't think the weight will be an issue but are you also taking account of the spindle weight ?
Try a different L arrangement, http://www.mycncuk.com/threads/6565-Ready-Steady-Eddy?p=55076#post55076
Also try the Y rails on the top and bottom faces rather than the front face. This makes the rails further apart which is good and also allows the Z axis to be closer to the gantry.

[jcarpenter]

An easy change to the configuration of Gantry_3 was to increase the size of the "L" extension to the square 76mm tube used in other parts of the gantry. Called Gantry_4 and shown below.

23260232612326223263

With this larger "L" the displacement metric goes from 0.040mm to 0.034mm. My gains in stiffness are getting less and less now. The interesting thing here is that there is a more equal distortion level on the top and bottom of the gantry. A well designed mechanism will show a more distributed deformation than one with very high and low distortion areas. This suggests to me that I am approaching the limit of improvements in this configuration. This configuration is probably OK as far as it goes.

I will next try flipping the horizontal and vertical sections of the "L" so that the longer "L" section is positioned is resist bending in the lower part of the gantry.

[routercnc]

Interesting thread John, good work. Your simplified analysis is sound enough to draw conclusions and it is no surprise you are iterating towards the popular L shape !

Watch out if you reverse the L shape then you may gain for the current horizontal load case, but the spindle mass and plunging forces will cause more deflection vertically so I would say run vertical loading on it, and repeat on the initial design.

In general, as per Eddy's comments, putting the rails on the top and bottom, and moving the Z axis close to the gantry might get a little bit more stiffness, so that would make another good one to try out. Plus lots of builders put an additional strip of steel inside the box section where the rail mounts. This is to get good thread engagement but will also add some local stiffness and help spread the load into the gantry, again adding a bit more stiffness.

[jcarpenter]

I inverted the legs of the "L" gantry configuration and reran the loads. As before the images are respectively, Geometry, Loadcase, Results, and another view of results. Deflection decreased from configuration Gantry_4 of 0.034mm to 0.030mm for Gantry_5.

23266232672326823269

Better than the decrease of about 10% in deflection is the more uniform deflection distribution over the gantry. In addition the most distorted part is the Z Axis block which is fake anyway. Pretty happy with this configuration but as mentioned above I will need to do some comparisons with the base gantry before committing to the Gantry_5 design.

John C

[EddyCurrent]

1. Top and bottom rails need to be further apart, i.e. vertical part of the "L" needs to be taller.
2. Z axis is way too far away from the gantry, it needs to be as close as possible to it.
3. The Z axis plate should be wider, when looking at the front face of it. It needs 2 bearings on top and 2 on the bottom, these should be as far apart as practical to resist sideways deflection at the tool.

[routercnc]

1. Top and bottom rails need to be further apart, i.e. vertical part of the "L" needs to be taller.
2. Z axis is way too far away from the gantry, it needs to be as close as possible to it.
3. The Z axis plate should be wider, when looking at the front face of it. It needs 2 bearings on top and 2 on the bottom, these should be as far apart as practical to resist sideways deflection at the tool.

Agree, #2 and #3 will make any of the options looked at so far a little bit stiffer.

#1 Though is an interesting one. The results for torsional stiffness show that the horizontal L (0.030 mm deflection) is better than the vertical L (0.034 mm deflection). This is easy to understand as that load case has a combination of bending in the X direction (for which the section is now very deep), and twisting (for which the section is the same as before). The sum of these gives an overall benefit for the horizontal L.
But the vertical loading needs to be checked as this will now be less stiff for the horizontal L. This is important to resist sagging due to the spindle/Y axis mass and causing machined surfaces to be bowed.

From a practical point of view the vertical L is always going to be a good choice as it allows more fore/aft gantry travel for any given bed length. But from an analysis point of view it will be interesting to see which will win overall . . . . .

[jcarpenter]

Agree on the Z axis distance from the gantry. Z axis design is probably next. The Z axis plate will be as wide as I can make it with my already purchased ballscrews and the cutting area needed. I bought some components early when I mistakenly thought the design was done.



Did vertical loadcase on the Gantry_0 design (twin separated beams) and the Horizontal "L". Used 500N line loaded distribution in vertical direction.

Twin separated beams: 0.1mm vertical deflection which will be my vertical comparison metric

232742327523276

And the horizontal "L": 0.03mm vertical deflection.

232792327723278

Still happy with the horizontal "L" configuration. Also like that again the beam has a fairly nice distortion distribution with the horizontal and vertical deflections similar under similar loading. Not sure I need to increase the vertical height of the horizontal "L" configuration or not. Deflections are well matched in both directions and adding height will be a little complex.

This discussion is really helping me to think this out.

John C

[EddyCurrent]

I've just noticed you have the top and bottom rails right up to the front edge of the gantry, they would be better set back to the centre of the top edge I think.

[routercnc]

I would say they are a bit more supported on the front edge because they have the vertical face of the section they sit on directly underneath. If they are in the middle of the section they are in the middle of the unsupported membrane of the section. It also minimises the offset from the spindle to the rail (a little bit).

If they are steel they would of course need to be set back a bit as there will be a radius along the edge.

In terms of performance comparisons this confirms what I said earlier in that the vertical stiffness is easier to achieve than torsional stiffness. To get the same deflection in both directions you need a lot more material resisting the torsion/moment than the more straightforward vertical load, hence the horizontal L.

But I would also say that the vertical and horizontal L shapes would both give good performance in practice, and that in the end it will come down to whether the horizontal L leads to unacceptable loss of travel in the X axis for any given design.

[jcarpenter]

To round out this discussion I did a box gantry also. Deflection of 0.02mm is the lowest yet.

232822328323284

Dimensions of the boxed beam are the same as the widest dimension of the horizontal "L" gantry.

So all this playing around with FEA has basically validated the advice given in the sticky threads at the top of the sub forum. That is, use a box or an "L" beam. I think I will. Just gotta decide which one to chose.

Thanks to all.

John C

[routercnc]

Great thread John thanks for posting.

The box gantry will always be a clear winner from the deflection point of view, but then you have to put the ballscrew somewhere. You can put it on the front face but then you have to step the Z axis away from the gantry, loosing some stiffness benefit. Plenty of commerical CNC routers use this layout and they work just fine. Or you can go for the L shape, tuck the Z axis tight against the gantry and put the ballscrew just behind. Plenty of DIY builders on here have gone this route and that also works just fine.

In general section size is key for stiffness as it goes up with the power ^3 for bending and with the power ^2 for torsion. So a little bit deeper / wider goes along way.

In all cases the material needs to all be furthest away from the neutral axis, the bit somewhere in the middle where not a lot is happening. So tubes are good, squares are better. Avoid open sections such as [ or I beams. They are only good in one direction.

Once you have maxed out on the outer size, have a play with wall thickness as this can really boost the stiffness. The limit is then how much weight you are comfortable with.

[jcarpenter]

A Facebook CNC guy I respect asked for another run of the box configuration with the rails mounted top and bottom.

233262332723328

Deflection was 0.024mm. But the Z axis had to be extended downward to account for the increased separation between the rails. So the lever arm is longer and the displacement more. Depending on where the twist axis lies. Too complicated for the simplicity of the model. My personal opinion is the the rail placement is probably not much difference.

OK, now I am done. Gotta start a gantry layout now.

John C

[jcarpenter]

Time goes by so fast when you are having fun. 9 Months since I did the math for the gantry and just now have a new initial design for it.

240162401724018

But there is a problem. My original estimate for weight for the gantry was 40Kg. The weight estimate for this new much stiffer gantry configuration is 100Kg. My question is: Is a 100Kg gantry a problem? The math says I need a stepper that can deliver 1Nm up to 600 rpm to power the two X axis ball screws that move the gantry. That is with a factor of safety of 2. Stepper size looks doable. Does a gantry of this weight pass a reality check?

Thanks,

John C

[routercnc]

Hi John,

Usually up to 80 kg gantry is OK for typical stepper and drivers, above that may be entering servo territory (=money!).

See post #2 here:
http://www.mycncuk.com/threads/6457-Sturdy-and-Fast-all-Steel-CNC-my-first-build?p=48417&highlight=80kg#post48417

I'm intrigued how you ended up with the design in post #22. That thin wall, and very open box section looks like it will resonate in practice, and the intermediate extrusions (to space away for the ballscrew?) are going to flex locally. The ends look fairly tricky to build and maybe a bit complicated.

[jcarpenter]

To get to this design you gotta remember that as much as possible stock on hand is being used. Decided on a box beam as analyzed in post #19. Turned sides inside out to get the offset for the Y ballscrew to fit between the rails. Side plates don't cover the complete length of extrusions since 48 inch wide thin plate (5mm) stock is available but the rails (and extrusions) are 60 inch. Double plate gantry side supports have been recommended here so 16mm plate was used for those and the rest of the end plates.

As you can guess there is aluminum plate, sheet, and extrusions in abundance to use. All sunk costs. Below is what the structure inside the gantry looks like before adding the end plates.

24020

Ends are complicated but the CNC knee mill in the garage will sort that out if necessary. Could use some advice on how to simplfy them.

I am worried about resonance. If people here think that resonance will be a problem with a manufactured box beam like this I will probably abandon it and try an "L" configuration. Looks like a lot of people here are having success with less overkill than this. How do the chinese big box beam gantries avoid resonance?

Thanks,

John C

[routercnc]

If you have a mill then things are much easier. Are you going to weld this all together then skim flat?

If this is a fabrication then I would add a horizontal web inside the section joining front and rear faces. Easiest way would be to mill short occasional slots in the front and rear face plates and puddle weld down onto the edge of the web. Maybe even add 2 webs close in line with the unsupported edge of the extrusions. This makes several smaller boxes and will tie the extrusions into the section and reduce resonance. Some people fill open sections with all sorts of stuff but I would prefer adding a few webs (before considering filling).

Commercial machines I've seen using box section gantry are steel with thick walls. They sometimes weld 2 thick strips to the front face, skim them, and mount the rails to those. The drive is often on the top face, or sometimes they squeeze it on the front face as well.

For the end pieces you could go with RHS section trimmed into a triangle shape. Then crop the extrusions flush with the fabrication, butt up the triangular fabrication and insert short extrusion pieces inside the triangles so that when you bolt right through it will prevent them crushing and allow you to torque the bolts. See my mk3 build to see how I did it.

[jcarpenter]

No welding will be involved. Everything will be bolted together. Built a machine once that had terrible resonance. Don't want to do that again. I think I will back up again and layout what a "L" shaped gantry with three 3x3 inch extrusions will look like. And single plate gantry supports. More deflection but less chance of resonance.

[jcarpenter]

The box beam gantry configuration is the stiffest design. But having fought resonance before I do not want to do it again. Scared of those large thin sides of the box. And the 100kg weight of the model. So tried another iteration. This time its a simple "L" beam made of the three 3in x 3in 80/20 extrusions. Ran the math and while it is half the stiffness of the boxed beam it is still over twice as stiff as the original twin separated beam gantry design. Comparison deflection metric is 0.05mm. Going with this.

240532405424055

Be back for more comments after getting this style gantry modeled.

Thanks,

John C

[jcarpenter]

New gantry layout using three 76mm x 76mm T-Slot extruded aluminum beams in an "L" configuration. Based on the FEA above.


240762407724078

This is a concept layout not a design yet but please give me comments and criticisms. What have I missed? Gantry side plates are 25mm thick, Z axis plates are 19mm, and the extruded T-Slot beams are 1520mm long. Y axis rails are 25mm HiWin, Z axis are 20mm HiWin. Light preload.

Thanks,

John C

[routercnc]

Looks good John. Some thoughts are:
You could centralise the end plates on the horizontal plates so they are directly above the X bearing carriages and bolt from underneath and not include the angle supports. The bolts into the X bearing carriages may just about straddle this end plate but if the bolt access holes to the bearing carriages underneath is obscured then add a second horizontal plate to bolt to the carriages and then bolt these to each other.

One end plate could be extended rearward slightly to accommodate the stepper which then belt drives to the ballscrew.

Make sure the extrusions and joined to each other to get the section benefits.

Check how you are going to drive the gantry and connect the X ballscrew nut before finalising the exact layout otherwise you might need complicated joining brackets either on the gantry ends or the bed ends where the stepper and bearing blocks are fixed.

[jcarpenter]

routercnc,

Looks good John. Some thoughts are:
You could centralise the end plates on the horizontal plates so they are directly above the X bearing carriages and bolt from underneath and not include the angle supports. The bolts into the X bearing carriages may just about straddle this end plate but if the bolt access holes to the bearing carriages underneath is obscured then add a second horizontal plate to bolt to the carriages and then bolt these to each other.

Yes, I could use two 12mm horizontal plates instead of the one 25mm one. That might make for easier alignment. Could mill an alignment rabbet on the bottom of the carriage plate to square the two carriages to each other.

One end plate could be extended rearward slightly to accommodate the stepper which then belt drives to the ballscrew.

Ok. Was going to direct drive everything but belts and sprockets might let me move a little faster with the 5mm ball screw pitch. What are the pros and cons of direct drive vs belt?

Make sure the extrusions and joined to each other to get the section benefits.

I have ideas for both internal and external fastening methods. Will try them and see what works better.

Check how you are going to drive the gantry and connect the X ballscrew nut before finalising the exact layout otherwise you might need complicated joining brackets either on the gantry ends or the bed ends where the stepper and bearing blocks are fixed.

Yes this was just a layout, not enough detail yet for a design. Looks like it might be time to put in the details.

Thanks,

John C

[jcarpenter]

Ran some more math on my triple beam "L" gantry. With a little less simplification. Generally don't see any major problem. I think it is time to add the final details to the design and then start building the thing.

Using Fusion 360 now and its static FEA solver. Not sure I will stick with this although the integration with industrial design capabilities looks pretty cool.


241622416324164

John C

[hanermo2]

Fwiw..

My large cnc mill bed is 240 kg in mass, for the bed alone, and up to 100 kg of vice and workpiece, easily.
So 340 kg++ of mass.

Worked fine with one small nema 23 stepper of 3Nm.
48V, 2M542 drive.
1:3 HTD belt drives.

Heavy is better, not worse.

For non-commercial use the speed was never a concern.
I never needed higher speeds.
Always wanted more mass, more rigidity, more resolution ...
leading to better accuracy.

[jcarpenter]

hanermo2, I have some 2.8 Nm NEMA 23 low inductance steppers and some 60 volt PS and drivers so I will try that with this rig. Will mount the steppers off belts rather than the direct drive I previously planned. Sigh, back to the CAD.

John C

[routercnc]

hanermo2, I have some 2.8 Nm NEMA 23 low inductance steppers and some 60 volt PS and drivers so I will try that with this rig. Will mount the steppers off belts rather than the direct drive I previously planned. Sigh, back to the CAD.

John C
My previous machine ran off something similar to that spec with 5mm screws (you mentioned these in an earlier post) and it was direct drive. All worked fine.
But if you want better speeds for wood then belt drive allows 1:2, also reduces resonance effects in the stepper and most important of all(!) allows the motor to be tucked away.