[BrandonBe]

I'm a Mechatronics Engineering student at the University of Waterloo and I'm planning on building a CNC Mill for aluminum and wood. The XYZ extents will be 2' X 2' X 1'. I will model it after the attached image. I plan on using ball screws and profile linear rails from the manufacturers in china. I have a whole bunch of questions so I'm just going to list them and hope some get answered.

1. Can I expect to get near 0.001" to 0.003" accuracy under $1500?
2. What grade of ball screw would be suitable? Is C5 accurate enough?
3. Are 12 mm diameter ball screws sturdy enough?
4. Is 1" X 2" X 0.065" Steel tube suitable for profile rails to be mounted to or do I have to use aluminum extrusion?
5. What size stepper motors would be suitable for such a machine, assuming ball screws have 4 mm pitch.
6. What kind of router would be suitable for wood and aluminum?
7. What do industrial cnc mills use to get to a 0.0001" accuracy range?

I'd appreciate any advice!

[trailerparkboys]

My best advice as an engineer would be to go and find the answers to these questions yourself. I don't mean to sound like a dick, but in doing so you will know how you came to the answers and will be able to be flexible when things change. Doing a project like this in your spare time will give you a wealth of knowledge that you will be able to apply to many mechatronics problems. Also, I landed my first engineering job because I was able to present a bunch of "engineering" projects I had done outside of university.

A great place to find information on ball screws, linear guides, etc. is catalogues and datasheets from the major manufacturers (Bosch-Rexroth, Hiwin, etc.). All of these will have an engineering section that lays out all of the required calcs you need to do to arrive at a particular component choice.

Good luck!

[louieatienza]

1. You likely won't find a set a ballscrews with .001" accuracy over 2' for under $1500. Rolled ballscrews on eBay are listed at about .002" accuracy over 1', but a few have found their backlash exceeds their accuracy rating, and resort to double nuts to reduce lash. A C5 screw is almost always ground, and has an accuracy of about .001" over 1'; C3 is half of that. I would say realistically rolled ballscrews with double nuts will give you about .002"-.003' over 1'.

2. Read 1.

3. Depends on what you're doing.... I'd say no, given the length you want.

4. I would say no, as the strength of the profile rail exceeds that of the tube you want to use. And I would recommend against relying on your profile rail as structural material, though it could add to it.

5. Depends on the weight of your machine, but you likely could use 300in-oz steppers. Finding C5, 4mm pitch ballscrews are highly expensive, even surplus on eBay.

6. Aluminum is pretty free-machining. However, their bearings are not made to take the stresses of cutting metal, and certainly not for the side loads the machine can exert. That said I've had success with 1-1/4HP trim routers as well as 2HP routers. The structure you posted would absolutely not be suited on anything but the lightest of cuts.

7. Aside from C3 double nut ground ballscrews with matching end bearings and though-coolant, glass scale linear encoders and high-end servos, heavy Meehanite castings that are precision ground, expert assembly and tuning... Even then they use automated probing and laser measuring of tooling, with wear compensation. And the machine would be in a climate-controlled environment.

[ger21]

I will model it after the attached image.

That's a terrible design, and makes the answers to the rest of your questions mute.

A few things to add to Louie's comments.


1) .001 over 2ft? Not likely. You can get repeatability withing .002-.003, but not accuracy.

4) Aluminum extrusions are not as straight as the tolerance you're seeking. For steel, you'd probably want 2 x 2 x .1875

5) Depends on how fast you want to go. With 4mm pitch screws, you don't need very large motors.

6) A chinese spindle would have bigger bearings than a handheld router, and those bearings should last longer.

7) About a half million dollars of what Louie said.

[BrandonBe]

Could you please give me an idea of what is wrong with that design? I know it isn't a great design obviously but why is it terrible? And is steel tubing typically straight enough to attach profile rails directly to it? And what kind of accuracy would you say I could reasonably aim for? I found C5 ball screws for a reasonable price and they should have a travel distance error of no more than 0.00118" over 630 mm and the profile rails have an accuracy within 0.001" per 500 - 1000mm. I wouldn't think that 0.002-0.003" is not a reasonable accuracy. What am I missing here?

[trailerparkboys]

Could you please give me an idea of what is wrong with that design? I know it isn't a great design obviously but why is it terrible? And is steel tubing typically straight enough to attach profile rails directly to it? And what kind of accuracy would you say I could reasonably aim for? I found C5 ball screws for a reasonable price and they should have a travel distance error of no more than 0.00118" over 630 mm and the profile rails have an accuracy within 0.001" per 500 - 1000mm. I wouldn't think that 0.002-0.003" is not a reasonable accuracy. What am I missing here?

What accuracy are you talking about? Absolute positional accuracy, repeatability or final part geometry tolerance?

[BrandonBe]

What accuracy are you talking about? Absolute positional accuracy, repeatability or final part geometry tolerance?

I'm interested in finalnpart tolerance.

[louieatienza]

Could you please give me an idea of what is wrong with that design? I know it isn't a great design obviously but why is it terrible? And is steel tubing typically straight enough to attach profile rails directly to it? And what kind of accuracy would you say I could reasonably aim for? I found C5 ball screws for a reasonable price and they should have a travel distance error of no more than 0.00118" over 630 mm and the profile rails have an accuracy within 0.001" per 500 - 1000mm. I wouldn't think that 0.002-0.003" is not a reasonable accuracy. What am I missing here?

Gantry is a bit too high for the bearing spread. Expanding it to 2' will cause racking. I think the Z axis is OK< but for 1' travel like you propose I believe it is far better to put the rails on the moving carriage. The carriages holding the gantry are not at their most advantageous configuration. You'll have lots of unsupported area under the table, so you'll lose quite a bit of Z travel if you want a table that won't flex. And you'll want that if you have ANY aspirations of cutting aluminum. No gusseting at the welds likely will lead to lots of flexing under load, and you'd be surprised how much force these steppers and screws can produce. Extruded tubing will always have stresses built into them and may or may not be straight, and can and will warp somewhat after welding. Weldments are usually stress relieved, and mounting surfaces machined.

I'm also not a fan of using standoffs for the stepper motors as shown. There should be a proper plate that the stepper mounts to, and that plate can be mounted to the standoffs. For one, a stepper plate acts as a heatsink, and these motors do get hot. Second and more importantly, as I mentioned these steppers multiplied by the screw pitch can produce a lot of force. If you run into a hard stop, that force is transferred to the motor casing, of which the base is a casting and prone to break.

Finally this design leaves not much room for adjustment. You end up with what you got after welding. And it does not look the mounting surfaces were milled or ground. Granted, the aluminum standoffs for the round rail do take up some irregularities, but that leads to accuracy issues. But the machine only needs to be as accurate as it needs to be.

As to profile rails - they are not necessarily straight believe it or not. The accuracy is largely due in part to how well the're mounted, or more specifically how well the mounting surface and datum ledge are prepared. They can telegraph any irregularities of what they're mounted to. As to accuracy of parts - all errors are cumulative. Being within .001" here, .002" there... it all adds up. Part of what makes parts accurate are the machines they're made on, and another part is the experience of the operator. I've gotten within .002"-.003" tolerance on parts using a machine that probably has .004"-.005" total accuracy. But I can't just slap any piece of material down, press "Play" and expect it.

Even end mills are given at nominal size, and some of them can have a tolerance of over .002". They even make them in plus and minus tolerance for specialized use.

There was a book published in the late 90's or early 2000's so about CNC Machining, and I believe this machine is an offspring of that early design; when folks were just getting started in this as a hobby. Today there's so much more information out there (and here), the price of higher quality components has gone down, and electronics are readily available. Every design, good or bad, has its trade-offs, and I think this particular one trades off too many structural issues in exchange for being cheap; the components could better perform with some design changes.

I think the take-away here is to temper your expectations. Many folks here are using C7 rolled ballscrews, even ACME screws, and far less rigid materials and getting very useable parts. eBay has a plethora of used ballscrws but there's no guarantee of the accuracy of them when they're used, and screws that were originally anti-backlash may develop lash due to wear. And the accuracy of the screw is only as good as the fixed and free end bearing blocks they're mounted to, and what those blocks are mounted to. On the picture the screws are mounted directly to the stepper via a coupling, which is fine for shorter lengths but to be avoided with longer spans.

[trailerparkboys]

Could you please give me an idea of what is wrong with that design? I know it isn't a great design obviously but why is it terrible? And is steel tubing typically straight enough to attach profile rails directly to it? And what kind of accuracy would you say I could reasonably aim for? I found C5 ball screws for a reasonable price and they should have a travel distance error of no more than 0.00118" over 630 mm and the profile rails have an accuracy within 0.001" per 500 - 1000mm. I wouldn't think that 0.002-0.003" is not a reasonable accuracy. What am I missing here?

You have to think about all of the variables that make up the accuracy of your machine, not just the specs of your linear components. The largest contributing factors to the accuracy of your machine will be how accurately you can build structure of the machine and the stiffness of your machine as a whole. You will need to ensure your machine is straight, coplanar and/or perpendicular to a tolerance much finer than 0.002" across it's whole travel. You will then need to ensure that no element of the machine can flex more than a tolerance much finer than 0.002" under cutting load. Then you have to ensure that your chosen spindle motor has run out much less than 0.002". The list goes on. At the end of all of this, you should be able sum all of the tolerances across the machine and reach a final machine accuracy much better than 0.002" because, at the end of the day, you will still need headroom for tool deflection.

I'm not trying to be discouraging here; building precise machines is a very interesting pass time! At the end of the day, you should just build your machine with the best knowledge you have at the time and fix any mistakes in versions 2, 3 and beyond.

[G59]

At the end of the day, you should just build your machine with the best knowledge you have at the time and fix any mistakes in versions 2, 3 and beyond.

I agree 100% with that statement. After all the first version of anything is nothing more than a rough concept with high expectations.
The first of anything, has gone through many design revisions before reaching any appreciable accuracy. Think how the first lathe was made. More than likely out of wood.
The first machines to have reached 0.0001" accuracy was more than likely machined with another less accurate machine, which was shimmed in every concievable way and had levelling tweaks made to it, so it could make the parts for such a machine. Not counting the manual massaging that would of been needed.
In the end however, they succeeded.

[wizard]

I'm a Mechatronics Engineering student at the University of Waterloo and I'm planning on building a CNC Mill for aluminum and wood. The XYZ extents will be 2' X 2' X 1'. I will model it after the attached image. I plan on using ball screws and profile linear rails from the manufacturers in china. I have a whole bunch of questions so I'm just going to list them and hope some get answered.

The attached image leaves a lot to be desired.

1. Can I expect to get near 0.001" to 0.003" accuracy under $1500?

This is a good question, do you have access to a machine shop at your university? I ask because the only way you will hit your price goal is with free access to machine shop tooling.

2. What grade of ball screw would be suitable? Is C5 accurate enough?

This depends upon your control software to some extent. If it can map the ball screw and apply corrections then you might do good. However if you model your machine on the one in the picture I don't think you have a chance in hell of holding tolerances thus the ball screw isn't the biggest factor in failure.

3. Are 12 mm diameter ball screws sturdy enough?

For wood they would be fine in a machine this size.

4. Is 1" X 2" X 0.065" Steel tube suitable for profile rails to be mounted to or do I have to use aluminum extrusion?

No & no! For one you never have to use aluminum extrusions and in some cases extrusions are less than ideal.

As for the tubing it is pretty light for a machine that wants to do aluminum. This machine isn't that large at 2 foot square, however your Z is pretty high so I'd opt for larger tubing, probably 3" square at the minimal. The gantry beam itself would need to be bigger.

Wall thickness is an issue too, you need material thick enough to drill and tap for the rail mounting screws. If not that thick then you need different approaches. One approach could be to weld mounting pads for the rails and then machine them afterwards. Another option is to buy the rails drilled and taped, fastened in place with through holes in the beam.

5. What size stepper motors would be suitable for such a machine, assuming ball screws have 4 mm pitch.

It depends upon the performance you want.

6. What kind of router would be suitable for wood and aluminum?

One that can take half inch shanks. Further you will likely be better off with a single speed model that you can apply an external speed controller to. However expenses explode rapidly so you should consider a Chinese spindle.

7. What do industrial cnc mills use to get to a 0.0001" accuracy range?

Very few do

I'd appreciate any advice!

You need to become more realistic on your design if you want to come anywhere near the accuracy you have talked about here. That means lots of beefing up of the design.

[hanermo]

Use 150 mm tube, of 8 mm wall thickness. In steel.
Have them surface ground two sides.

Use 30 mm rails.
Use 25 mm screws.
Use servos.

This will give you local accuracy of 0.01 mm, to about 100-200 mm cube.
And repeatability upto 3-5 microns, depending.

[ger21]

Use 150 mm tube, of 8 mm wall thickness. In steel.
Have them surface ground two sides.

Use 30 mm rails.
Use 25 mm screws.
Use servos.

This will give you local accuracy of 0.01 mm, to about 100-200 mm cube.
And repeatability upto 3-5 microns, depending.

Can you tell him how to do that with his $1500 budget?

[hanermo]

Industrial machines have a typical basic accuracy of 0.0001", or 2.5 microns, plus or minus.
== 5 microns max error.

Haas, leadwell, all of them..
This is *easily* achieved by
1. making the machine rigid enough,
2. using ground screws and servos

+/- 2.5 microns is easy.
+/- 1 micron is easy.

Manual machine have done this 40 years ago. Jig grinders. Whatchmakers jig bores. Sip. Secondary offsets adjust for screw error (from early 1900s). etc
The screw accuracy is critical.

But..
its expensive, in terms of the screws.

Today, secondary feedback via glass scales makes 1 micron accuracy easy, and relatively cheap.
Its very rare, and I dont know anyone who has it running.
About 2k€ for controller, 2-3k for scales.

I will do it, on lathe, but my lathe is very expensive, by most users standards, about 8 k€ wholesale just in parts (10k € retail).
My industrial csmio-ip-s controller supports it (mach4), but most dont.

2.5 micron machine accuracy+resolution is NOT machining accuracy of 2.5 microns.
It means potential accuracy to 2.5 micron (or better).
Methods and technique matter quite a lot, after 0.01 mm or 10 microns.

Best-in-class VMCs, like Mori Seiki DCG machines, can machine a round hole to 0.7 micron TIR or roundness, about 7x better than 2.5+/-.
Best lathes, like Moore Nanotech, do a round nickel ball to 0.3 microns, with 0.01 micron glass scales and linear drives.

My lathe retrofit has about 1 micron or better incremental resolution, using 0.2 micron step sizes (w ac servos).

This means I can makes features with 1 micron features.
Its not the same as accurate to 1 micron.

Ie I can make a graduated pin with 1 micron step sizes.
Say 50.000 mm D with 5 decreasing steps.
I wont know in advance, if step 1 or 2 is 50.000 mm.
But I DO know the steps themselves are 0.001 mm, and one of them will be 50.000.

Using tests, trial and error, I could make any nr of pieces with the exact same 50.000 size.
Via probing and references to known-good standards, like gage blocks, or gage pins, or setting ring gages, etc.
This is not a machine "accurate" to 1 micron.
(It will be, if the scales work.)

Its using manual techniques and leveraging automation to deliver highly accurate results.

Commonly stated temperature etc stuff is mostly (or partly) hookum.
Mostly, its very hard to make large parts very accurately.
After about 200-300 mm or 14" or so, it gets very hard, very fast.

Hand lapping, or adjusted/applied to homebrew cnc, can get there, and better, with the right technique, but its nothing to do with Moore Nanotech.
Hand work (or cnc-hand work) can do 0.1 microns easily.

Most accuracy stuff to 1 micron is FUD.

I expect to make ultra precision leadscrews on my lathe, for sale ie commercial, at some point soon.
So far, 100k€ in current stuff and 10 years/15.000 work hours, and I am getting there.
Spindle rebuild or c axis was 200+ hours.

So far, z axis rebuild is == 300 hours, and maybe 3-10 hr to finish (version 2) to working state.
My ground screw (used) was crap. 100 hours wasted.
No pics atm- sorry.
Commercial reasons.

For example, I use a 2 cm thick tool steel plate at the back of lathe, to connect to saddle.
200x300x20 mm == 16 kg.
Z axis coupler mount plate = 250x200x50 mm = 20 kg.
Ie its 2 inches thick tool steel.

Also, for example on the mill, "capacity" or "strength" is 50 tons metric, for 500 kgf force or 2 kW spindle power.
50:1.
Accurate machine = rigid = much stronger than stress used.
Ie mill is 50x stronger than stress, as that => accuracy.
Mill uses 35 mm rails, and overconstrained, 12 blocks, of 4000 kgf each.

You need to vastly upgrade your expectations re: strength/rigidity.
By about 20x.
Smile, grin - So did I, way back when.

[wizard]

What is interesting about this project, is what can be obtained within the $1500 budget. 1500 dollars is pretty tight for a machine that user wants to maintain tight absolute accuracy upon. This is why I ask about the colleges and hopefully a student accessible machine shop. In the end I don't think the poster will come even close to the results he wants without a machine shop facility that i.e. Can run with little or no cost to him. Being a relatively small machine he should be able to handle the parts (gantry beam) even on a smallish Bridgeport type machine. Effectively he would need a machine with two feet of travel plus the saddle widths plus a bit of cushion. Say 34" of travel.

The next issue is the 12" of Z space. While I can see some usages where such a Z clearance can be useful it does add challenges. For ne this forces spreading out the bearings for the gantry support saddles. On a small machine this is where it makes sense to consider a fixed gantry moving table design. With such a design one can make a much stiffer gantry beam. This is important because the twisting effort on the gantry will be pretty high while machining aluminum not to mention what happens in a crash. What is more important is that a moving table won't explode machine length greatly over what would be required for a properly supported moving gantry.

In a nut shell I think we can get this guy closer to what he wants, within his budget, but that really depends upon being able to access the tools to do it. That means welding up a frame, ideally stress relieving it and then machining everything to dimensions. The ultimate accuracy won't be as good as he wants but you can't work miricals in this business.

The big problem with the machine he presented as a model is that it simply isn't stiff enough to give him the numbers he wants. A lot can be done about that, that won't completely sink the budget. In the end the positional accuracy just won't be there at this budget level.

[wizard]

This is all well and good but you do realize that this is a college student with a tight budget - right? I really believe he can do better within his budget (basically scrap this design) by starting over with an eye towards rigidity. Even then $1500 dollars doesn't go that far. So the goal should be to get the best possible out of the budget at hand.

Having spent considerable time working on high precision lathes I have a good feeling for the costs involved, glass scales would effectively sink the budget before a single beam was fabricated.

Industrial machines have a typical basic accuracy of 0.0001", or 2.5 microns, plus or minus.
== 5 microns max error.

Haas, leadwell, all of them..
This is *easily* achieved by
1. making the machine rigid enough,
2. using ground screws and servos

+/- 2.5 microns is easy.
+/- 1 micron is easy.

Manual machine have done this 40 years ago. Jig grinders. Whatchmakers jig bores. Sip. Secondary offsets adjust for screw error (from early 1900s). etc
The screw accuracy is critical.

But..
its expensive, in terms of the screws.

Today, secondary feedback via glass scales makes 1 micron accuracy easy, and relatively cheap.
Its very rare, and I dont know anyone who has it running.
About 2k€ for controller, 2-3k for scales.

I will do it, on lathe, but my lathe is very expensive, by most users standards, about 8 k€ wholesale just in parts (10k € retail).
My industrial csmio-ip-s controller supports it (mach4), but most dont.

2.5 micron machine accuracy+resolution is NOT machining accuracy of 2.5 microns.
It means potential accuracy to 2.5 micron (or better).
Methods and technique matter quite a lot, after 0.01 mm or 10 microns.

Best-in-class VMCs, like Mori Seiki DCG machines, can machine a round hole to 0.7 micron TIR or roundness, about 7x better than 2.5+/-.
Best lathes, like Moore Nanotech, do a round nickel ball to 0.3 microns, with 0.01 micron glass scales and linear drives.

My lathe retrofit has about 1 micron or better incremental resolution, using 0.2 micron step sizes (w ac servos).

This means I can makes features with 1 micron features.
Its not the same as accurate to 1 micron.

Ie I can make a graduated pin with 1 micron step sizes.
Say 50.000 mm D with 5 decreasing steps.
I wont know in advance, if step 1 or 2 is 50.000 mm.
But I DO know the steps themselves are 0.001 mm, and one of them will be 50.000.

Using tests, trial and error, I could make any nr of pieces with the exact same 50.000 size.
Via probing and references to known-good standards, like gage blocks, or gage pins, or setting ring gages, etc.
This is not a machine "accurate" to 1 micron.
(It will be, if the scales work.)

Its using manual techniques and leveraging automation to deliver highly accurate results.

Commonly stated temperature etc stuff is mostly (or partly) hookum.
Mostly, its very hard to make large parts very accurately.
After about 200-300 mm or 14" or so, it gets very hard, very fast.

Hand lapping, or adjusted/applied to homebrew cnc, can get there, and better, with the right technique, but its nothing to do with Moore Nanotech.
Hand work (or cnc-hand work) can do 0.1 microns easily.

Most accuracy stuff to 1 micron is FUD.

I expect to make ultra precision leadscrews on my lathe, for sale ie commercial, at some point soon.
So far, 100k€ in current stuff and 10 years/15.000 work hours, and I am getting there.
Spindle rebuild or c axis was 200+ hours.

So far, z axis rebuild is == 300 hours, and maybe 3-10 hr to finish (version 2) to working state.
My ground screw (used) was crap. 100 hours wasted.
No pics atm- sorry.
Commercial reasons.

For example, I use a 2 cm thick tool steel plate at the back of lathe, to connect to saddle.
200x300x20 mm == 16 kg.
Z axis coupler mount plate = 250x200x50 mm = 20 kg.
Ie its 2 inches thick tool steel.

Also, for example on the mill, "capacity" or "strength" is 50 tons metric, for 500 kgf force or 2 kW spindle power.
50:1.
Accurate machine = rigid = much stronger than stress used.
Ie mill is 50x stronger than stress, as that => accuracy.
Mill uses 35 mm rails, and overconstrained, 12 blocks, of 4000 kgf each.

You need to vastly upgrade your expectations re: strength/rigidity.
By about 20x.
Smile, grin - So did I, way back when.

[hanermo]

I did not advocate glass scales for the OP.
I explained, as he asked, how commercial stuff gets much better.

I advocated proper frames, of very little extra cost.
The rest will depend on what he can get & afford.

Budget is unrealistic, yes, ime.

[BrandonBe]

Thanks, I appreciate all the great advice! I understand the frame will need to be much much sturdier. And I see that 0.001" is a bit hopeful. Realistically I would be pleased with anything below 0.01". Would a fixed gantry made of 3"x 3" x 3/16" steel be sturdy enough with proper gussets and cross members? And will 15 mm profile rails be strong enough? They are rated way beyond what I will be using them for, somewhere around 700 kg force vertically, but they do look kind of small. And I originally thought 12mm ball screws would suffice but i am thinking 16mm will be required now. Or will I need to go to 20 mm screws? I also have access to a machine shop so machining parts is no problemo. And I will be using C7 ball screws.

[louieatienza]

Thanks, I appreciate all the great advice! I understand the frame will need to be much much sturdier. And I see that 0.001" is a bit hopeful. Realistically I would be pleased with anything below 0.01". Would a fixed gantry made of 3"x 3" x 3/16" steel be sturdy enough with proper gussets and cross members? And will 15 mm profile rails be strong enough? They are rated way beyond what I will be using them for, somewhere around 700 kg force vertically, but they do look kind of small. And I originally thought 12mm ball screws would suffice but i am thinking 16mm will be required now. Or will I need to go to 20 mm screws? I also have access to a machine shop so machining parts is no problemo. And I will be using C7 ball screws.

I think that's more in the right direction... Since the gantry's fixed, there's less a concern about having to move that weight. I think 15mm rails should be more than fine. A THK HSR15 bearing block has a dynamic rating of about 1800lbs force. You'll have 4 per axis. I think 16mm would be good. 12mm may work if the lead is coarse enough, over 10mm, but at the cost of some resolution. I think at 4mm pitch a 12mm screw could potentially whip during rapids speeds. C7 is fine, though remember anti-backlash does not mean zero-backlash. I would order them with double nuts, if accuracy is a concern.

[hanermo]

3x3 - no.
15 mm - no.
I use 50:1 for strength.
Ie my blocks are good fo 50.000 kgf, when used in 500-1000 kgf use.

Dont mix things.
Top industrial stuff will work fine with 15 mm.
But, you wont afford the UP (ultra precision) blocks and high precision ledges, surfaces they are mounted on. (At 4-6 k).

Top manual built stuff could do it.
If you need to ask, then its not yours.
Long granite standards, scraping, etc etc.

Plan for 50:1 in "strength" for milling.

Thanks, I appreciate all the great advice! I understand the frame will need to be much much sturdier. And I see that 0.001" is a bit hopeful. Realistically I would be pleased with anything below 0.01". Would a fixed gantry made of 3"x 3" x 3/16" steel be sturdy enough with proper gussets and cross members? And will 15 mm profile rails be strong enough? They are rated way beyond what I will be using them for, somewhere around 700 kg force vertically, but they do look kind of small. And I originally thought 12mm ball screws would suffice but i am thinking 16mm will be required now. Or will I need to go to 20 mm screws? I also have access to a machine shop so machining parts is no problemo. And I will be using C7 ball screws.