[Whitw0rth]

Most of the DIY CNC machine builds that I've seen have the router/cutting spindle attached to the the Z axis drive which is attached to the top of the gantry. As has been covered before, this means that when the Z axis is at full travel there is deflection as it acts as a cantilever. In other words, the usual layout /attachment order is Spindle -> Z axis -> X axis -> Y axis

Now, what I have been thinking is to instead put the Z axis drives on the gantry sides, which means the attachment order will be Spindle -> X axis -> Z axis -> Y axis. You can see that the RepRap team did this for their extruder (picture attached). Now, other than the need for 2 rails for the z axis, rather than just one, as well as coordinating the two axis, is there any reason that this isn't done? This surely means the Z axis will be far more rigid than in the typical designs?

Attachment 233550

[wizard]

Realize that you can build basically a thing you want. How well it will work is entirely dependent upon how much engineering goes into it.

Most of the DIY CNC machine builds that I've seen have the router/cutting spindle attached to the the Z axis drive which is attached to the top of the gantry. As has been covered before, this means that when the Z axis is at full travel there is deflection as it acts as a cantilever. In other words, the usual layout /attachment order is Spindle -> Z axis -> X axis -> Y axis

Well maybe the lettering convention here for the axis is screwed up, there is actually a bit of a debate as to how the axis on these gantry mills should be dedicated. Usually it goes (top to bottom) Z, Y & X.

Now, what I have been thinking is to instead put the Z axis drives on the gantry sides, which means the attachment order will be Spindle -> X axis -> Z axis -> Y axis.

This is doable but I haven't seen it done on a small machine. Examples on large machines would be planner mills, vertical turret lathes (slightly different machine), way grinders and others I can't think of off the top of my head. As you can imagine some of these machines are rather huge. Also even though many of these machines have a Y gantry that moves up and down the individual heads often have quills to drive the spindles in and out. I say heads because one Y gantry might have multiple milling heads mounted. You need to consider one thing, if you go this route you may give up clearance that will come in handy from time to time because the spindle doesn't clear the Y axis beam as you work deeper and deeper into the subject material. This could be a problem if your intention is to profile a large part.

Obviously the clearance problem is dependent upon what you intend to use the machine for. Like all good machine design the final product needs to fit the intended usage. Of course it never hurts to experiment with new arraignments, just imagine how the work you intend to do will get done on the machine. A machine suitable for working on sheet goods might not work at all for profiling molds for a boat, car fender or simply working on an engine block.

You can see that the RepRap team did this for their extruder (picture attached).

Remember there are zero reaction forces on such machines.

Now, other than the need for 2 rails for the z axis, rather than just one,

You still need two rails! The location is different but you really can't rely upon a single profile rail (well the narrow ones) to support a machining axis.

as well as coordinating the two axis, is there any reason that this isn't done?

Well you can try it and tell us how it worked out.

I can guess though if that doesn't bother you. The problem is likely complexity, especially in smaller shops. You also need substantially stiffer up rights and in fact you may need a beam anyways from upright to upright. To keep everything from binding you will need to archive a very high level of precision in machining and assembly. Then you have to consider that the Z axis now has far more mass to deal with. This doesn't even get into what your uprights mount on and how you keep everything true.

This surely means the Z axis will be far more rigid than in the typical designs?

It doesn't mean anything really, this is a novel approach in a small machine so you don't get the benefit of all the engineering that has gone into other designs or the rules of thumbs that have developed. This is another way of saying there will be lots of engineering or considerable tinkering. Off the cuff I do believe that more material will be required to achieve the same results. Further the whole Y axis beam is now less rigid with respect to the rest of the machine.

In any event don't let me stop you, if you have access to the tools to do the parts right then go for it. Just don't ignore the negative due to the positives. I'd also put lots of work into design to make sure everything comes out right.

[Whitw0rth]

Hi Wizard,

Thanks for the reply, it definitely given me some things to think about. I did wonder about axis notation with these machines, I guess I got X and Y mixed up. I'll follow the (top to bottom) Z, Y & X notation from now on. The bed size I've envisioned is perhaps 600x400mm, with main materials cut being wood and aluminium.

Regarding the single rail, to clarify I mean two sets of rails. The X axis would have a set of rails on either side of the machine, with the a centrally driven gantry, as is common in a lot of machines. The gantry columns will still be connected under the machine at the top of the machine. Where the design will depart is that on each of the vertical gantry columns there will be a set of rails and a lead screw which will provide the Z axis movement. To these rails will be attached the Y axis, which will behave as a normal Y axis does, and will have the cutting spindle mounted directly to it.

With this design to stiffen the z axis means to stiffen the gantry columns, which I believe is far easier than trying to stiffen the traditionally placed Z axis. Now, whether the Y axis will be as stiff running on the Z axis as opposed to being in direct connection with the gantry columns remains to be seen.

Part of the reason for moving the Z axis to the gantry columns is because I want to make this machine a true 5 axis -adding the weight of 2 more servos, pulleys and bearings to the Y axis is considerable. I know this is a challenge, but I plan to spend the next 3 months playing with CAD and FEA to test these configurations.

[wizard]

Hi Wizard,

Thanks for the reply, it definitely given me some things to think about. I did wonder about axis notation with these machines, I guess I got X and Y mixed up. I'll follow the (top to bottom) Z, Y & X notation from now on. The bed size I've envisioned is perhaps 600x400mm, with main materials cut being wood and aluminium.

Realize that the axis designations aren't written in stone here and there are some discussions about which is the right way. I generally default to naming the longest axis the X. The Z is kinda agreed upon as it is accepted that that axis moves the spindle up and down.

As to your machine size, it is nice in that it leaves you with lots of options that are fairly easy to execute build wise.

Regarding the single rail, to clarify I mean two sets of rails. The X axis would have a set of rails on either side of the machine, with the a centrally driven gantry, as is common in a lot of machines.

This is what I'm commenting about you would only need one rail oer vertical column. When you say sets of rails I see two rails per column. I really don't think this is needed but further mechanical analysis probably will help.

The problem I see here and frankly it doesn't matter how many rails you use, is getting both columns to line up correctly and this hold position on your linear slides such that they don't bind or jam during operation. Your machine size is small enough that you might be able to weld up a box frame, weld some mounting plates to the face of the frame, heat treat and then machine everything to keep the rails in the same plane and parallel to each other. A decent machine shop would have a mill big enough to do a 600 x 400mm frame.

The gantry columns will still be connected under the machine at the top of the machine. Where the design will depart is that on each of the vertical gantry columns there will be a set of rails and a lead screw which will provide the Z axis movement. To these rails will be attached the Y axis, which will behave as a normal Y axis does, and will have the cutting spindle mounted directly to it.

I more or less follow this. I don't think it is impossible to build at all. I do wonder if there is a real advantage in these smaller machine sizes. For a small machine, of the more conventional design often found in these forums, it is pretty easy to build a stiff Z and gantry. After all 400 mm is about a 16" span, a piece of 5" tubing over that distance would be very stiff. At that point you would be focused on making the Z axis slide stiff, stiff enough to meet your needs. I'm just not sure that there will be a huge pay off considering the design complexity.

With this design to stiffen the z axis means to stiffen the gantry columns, which I believe is far easier than trying to stiffen the traditionally placed Z axis. Now, whether the Y axis will be as stiff running on the Z axis as opposed to being in direct connection with the gantry columns remains to be seen.

Given enough time and materials I suspect that you would have a stiffer machine. They problem is that you will need more resources to do the required machining and your other expenses increase substantially.

Part of the reason for moving the Z axis to the gantry columns is because I want to make this machine a true 5 axis -adding the weight of 2 more servos, pulleys and bearings to the Y axis is considerable. I know this is a challenge, but I plan to spend the next 3 months playing with CAD and FEA to test these configurations.

OK now you have lost me. If by 5 axis you mean a machine that can rotate the spindle via one of this axis, how would this work if the spindle was fixed to the Y axis saddle? In other words the minute you tried to rotate the spindle you would start to loose clearance with the Y axis beam. If you cantilever the rotation mechanism off the side of the Y axis saddle they aren't we back to the same problem with extending the Z spindle and supposedly loosing stiffness? If you want a 5 the axis I'm not sure how this moving Y axis beam design would work at all or at least be a significant advantage considering the additional complexity.

Obviously a design and some mechanical modeling will address these seat of the pants concerns. I'm sure many here will be very interested in what you come up with. When you throw in the 5 the axis though, I'm left wondering if this is the right approach at all. This is mainly due to figuring out how you would get useful motion and clearance. The minute you overhang the spindle rotation mechanism you loose much of the advantage you think you are getting here with this design.

In any event I will wait for drawings to see what you have in mind.

[Whitw0rth]

Looking around the forums there doesn't seem to be a good deal on 5 axis builds, but I see some have been done. So, perhaps what I'll do is use this thread as build thread (if that's allowed?). It will most probably just be me talking to myself, but hopefully it might be useful for others.

The material for construction will be aluminium extrusion, either the Bosh Rexroth system or the 8020. I'm based in the UK, so the 8020 may be harder to get hold of (although there does seem to be a UK distributor - I'm awaiting their price list).

To begin with, I want to make a comparison between stiffness's on 3 axis using FEA. To save on processing power, I'll be modelling the aluminium extrusion simply with the sections being simplified to squares. To begin with, I'll also ignore the baseframe forces and restrain the gantry at the bottom. I expect to monitor the forces on the bottom of the gantry and see how equal they are, firstly with only cutting force and then after with component mass + cutting force.

This will give me the first useful metric, which I will call the gantry force ratio. This will be defined as gantry forces cutting / gantry forces cutting + gantry forces mass. This ratio will always be less than 1, however, the closer to 1 the more efficient the design for the gantry is in terms of mass.

Also, another measure of efficiency that I'll define here is bed efficiency. This will be Available Cutting Volume / Total Bed Volume. The available cutting volume will be the dX, dY and dZ extents that the spindle can move, the total bed volume is self explanatory, other than the Z axis, which will be defined as the distance between the bed and the lowest Z point of the gantry.

A few other specification points:
1) The design will allow the 5 axis to be changed to 3 axis easily, so the extra 2 axis can be removed and the spindle transferred. This may mean some mass redistribution to equalize gantry support forces, which will be dependent on the weight of the extra 2 axis.
2) The design will consider either a planetary gearbox or HTD pulley set. The ratio will probably be 5:1, which for a standard 1.8 degree stepper motor will be a 0.36 degree movement increment per step. The decision as to which will come down mostly to cost & backlash. The planetary gearbox would be far simpler to implement. Figures that I can find for backlash of a typical planetary box is 0.2 degrees, which I believe is measured at the output shaft. Not sure on the typical backlash on the HTD set with a proper belt engagement, so I'll have to do some research into that.

[wizard]

Looking around the forums there doesn't seem to be a good deal on 5 axis builds, but I see some have been done. So, perhaps what I'll do is use this thread as build thread (if that's allowed?). It will most probably just be me talking to myself, but hopefully it might be useful for others.

I'm not the owner of the site but this is your thread. I think people will be interested.

The material for construction will be aluminium extrusion, either the Bosh Rexroth system or the 8020. I'm based in the UK, so the 8020 may be harder to get hold of (although there does seem to be a UK distributor - I'm awaiting their price list).

This will likely leave people up in arms but it really doesn't matter whose extrusion you use. Well except in one way, some makers have much heavier webbing and are willing to machine an extrusion flat for you. Most extrusions while pretty good are not guaranteed to be flat in the context of machine tools.

To begin with, I want to make a comparison between stiffness's on 3 axis using FEA. To save on processing power, I'll be modelling the aluminium extrusion simply with the sections being simplified to squares. To begin with, I'll also ignore the baseframe forces and restrain the gantry at the bottom. I expect to monitor the forces on the bottom of the gantry and see how equal they are, firstly with only cutting force and then after with component mass + cutting force.

Take this with a grain of salt as I'm not a FEA expert at all, but wouldn't that lead to significant errors? Of course this depends upon the extrusion but there is no contiguous wall around the perimeter of the section.

This will give me the first useful metric, which I will call the gantry force ratio. This will be defined as gantry forces cutting / gantry forces cutting + gantry forces mass. This ratio will always be less than 1, however, the closer to 1 the more efficient the design for the gantry is in terms of mass.

Interesting concept. Often mass gets thrown into a discussion as a way to avoid good design, i would love to see what this ratio says about the various designs you come up with.

Also, another measure of efficiency that I'll define here is bed efficiency. This will be Available Cutting Volume / Total Bed Volume. The available cutting volume will be the dX, dY and dZ extents that the spindle can move, the total bed volume is self explanatory, other than the Z axis, which will be defined as the distance between the bed and the lowest Z point of the gantry.

Have you decided upon moving gantry or fixed gantry yet? I would thing that with a moving gantry you would design for 100% bed converage or even a bit more.

A few other specification points:
1) The design will allow the 5 axis to be changed to 3 axis easily, so the extra 2 axis can be removed and the spindle transferred. This may mean some mass redistribution to equalize gantry support forces, which will be dependent on the weight of the extra 2 axis.
2) The design will consider either a planetary gearbox or HTD pulley set. The ratio will probably be 5:1, which for a standard 1.8 degree stepper motor will be a 0.36 degree movement increment per step. The decision as to which will come down mostly to cost & backlash. The planetary gearbox would be far simpler to implement. Figures that I can find for backlash of a typical planetary box is 0.2 degrees, which I believe is measured at the output shaft. Not sure on the typical backlash on the HTD set with a proper belt engagement, so I'll have to do some research into that.

I'm under the impression that belt drives are easier to implement. The problem I have to wonder about though is ultimate speed on your axis with stepper motors. A 5:1 reduction is significant and may make the machine pretty slow.

[Whitw0rth]

You're 100% correct about the FEA, the change in section will alter the moment of area which completely changes the deflection results. However, the purpose of it is comparative. So, if version 1 of the design with 40x40 square is 50% stiffer than version 2 with 40x40 square, then it is safe to assume that ratio will exist between the versions when constructed out of the extrusion (the extrusions have 2 planes of symmetry which makes this assumption valid - if they didn't then you'd have to look at which orientation they are in). I will at some stage run a simulation with the 40x40 extrusion, but it will take considerably longer to compute, so I don't fancy running too many of these simulations.

I've decided upon a moving gantry as these machines seem to be more efficient space wise. On balance though, moving the bed would be easier than moving the gantry as the bed + workpiece + fixturing will most likely be lighter than the gantry, as I'm not likely to be machining huge chunks of metal.

I'm not sure what speed A and B will be running at, but I think these axis will be least affected by the reduction, since movements on these axis are more gradual. With 0.36 degrees per step, it means a 1kHz stepping frequency is needed to get one full revolution in a second.

Interestingly I found this photo (below) earlier of a gantry z axis. You can see that because it's for a 3D printer they've just used lead screws instead of rails on the gantry. In a CNC application though, this does pose the question, if you have rails here instead of a lead screw, then where do you locate the z axis lead screw/s

Attachment 234014

[wizard]

You're 100% correct about the FEA, the change in section will alter the moment of area which completely changes the deflection results. However, the purpose of it is comparative. So, if version 1 of the design with 40x40 square is 50% stiffer than version 2 with 40x40 square, then it is safe to assume that ratio will exist between the versions when constructed out of the extrusion (the extrusions have 2 planes of symmetry which makes this assumption valid - if they didn't then you'd have to look at which orientation they are in). I will at some stage run a simulation with the 40x40 extrusion, but it will take considerably longer to compute, so I don't fancy running too many of these simulations.

Sounds like you need a computer upgrade or a little GPU processing.

I've decided upon a moving gantry as these machines seem to be more efficient space wise. On balance though, moving the bed would be easier than moving the gantry as the bed + workpiece + fixturing will most likely be lighter than the gantry, as I'm not likely to be machining huge chunks of metal.

It is certainly a trade off that one needs to address, for most people space is far more important. On the flip side I see moving table machines as far more accurate, easier to assemble and easier to maintain in good working order.

I'm not sure what speed A and B will be running at, but I think these axis will be least affected by the reduction, since movements on these axis are more gradual. With 0.36 degrees per step, it means a 1kHz stepping frequency is needed to get one full revolution in a second.

It really depends upon your needs but generally you don't see many stepper based systems running large reductions.

Interestingly I found this photo (below) earlier of a gantry z axis. You can see that because it's for a 3D printer they've just used lead screws instead of rails on the gantry. In a CNC application though, this does pose the question, if you have rails here instead of a lead screw, then where do you locate the z axis lead screw/s

Well if I was using steel tubing I'd run the leadscrews right through the tubing in front of the linear bearings. The other option is off to either side. Of course running through the tubing implies boring the tube and making up some sort of insert to get good results.

[Whitw0rth]

Just a quick update on progress so far. I've started from a bottom up approach and have found a very useful book called Manufacturing Automation by Yusuf Altintas. It's good a lot of content, but at the moment I will be spending my time with the cutting forces section to calculate cutting forces and feed forces at the cutter. There's a lot of coefficients needed in the typical formulas here so I'll have to contact a major tool manufacturer or two and request some information.

[wizard]

Just a quick update on progress so far. I've started from a bottom up approach and have found a very useful book called Manufacturing Automation by Yusuf Altintas. It's good a lot of content, but at the moment I will be spending my time with the cutting forces section to calculate cutting forces and feed forces at the cutter. There's a lot of coefficients needed in the typical formulas here so I'll have to contact a major tool manufacturer or two and request some information.

Tool forces can vary dramatically especially in these smaller machines that may find extensive use of engraving and other fine cutters. Obviously you don't want to design for non existent forces so you need to zero in on what you would expect to be the highest load. Then go beyond that because on a general purpose machine it is very likely that you will find new uses for the machine "if only it where a bit more rigid".