Homebuilt CNC in Tokyo

[Milland]

Mithril's no use for home applications, too many orcs about. The glow would keep you up all night.

 

[drcoelho]

Are you suggesting I have a frame custom cast? Because I'd love that, but the idea is to build this out of reasonably accessible components, and even a small (2-4 kilo) casting here in Japan would be thousands of dollars before the machining that would be required to make it even useful. Maybe I misunderstand what you are suggesting, it isn't really clear.

As far as bolting things to big hunks of metal, my base plate that this will be built onto, and form the basic bed is a 72KG cast Iron Surface Plate that had been abused over what I can only assume is the last 50-60 years. It took days of hard work getting the layers of congealed dirt and corrosion off of it, I'm pretty sure it had been buried in the dirt for a while at some point, but the ancient layers of grease and quality iron kept it in recoverable shape. It was primed and then painted with several coats of 2-part epoxy machine paint.

It is bolted to a stiff aluminum undercarriage and machine casters to help move it around, the total weight will be close to 300 kilograms when complete. It will ultimately be tethered to reinforced mounting points on the floor. Neat fact! I have heard recently that some people with an unfortunate overabundance of fecal production require specialized toilets that actually exceed the total weight of this machine!

The main structural members of my design are going to be the much derided Aluminum extrusions. They are however much larger and stronger than typical, made with JIS A6N01S-T5 alloy and 100x200mm M10 profile. Are there stronger materials I could be using? Yes, and they are not secrets, even a section of mild steel square tubing would likely be stiffer. But there are tradeoffs to be considered, and one of the biggest is the material processing requirements. There are also weight, cost and recyclability factors I outlined above. The extrusions are turn-key, drop shipped to my door perfectly sized, and have a multitude of mounting opportunities without any post processing required.

The fact is that there is a certain amount of 'bootstrapping' that is made available by building it this way. Whatever the capabilities of the Al turn out to be, it is highly probable that it will be enough to build its own reinforcements. Incrementally targeting weaknesses that are identified. Potentially being able to properly machine the surfaces and attachment points onto steel plates or even, if it turns out to be necessary, a replacement steel gantry tube.

So I respect the thinking, and I understand the wisdom of copying what's come before, but please realize that I have a very different set of constraints (not all of which are easy to get, although I'll continue to do my best to explain them) and have zero intention of doing any kind of radical design change at this point. Short of being presented with some persuasive evidence that it will not work, the layout you see is the layout it will be. These parts are sitting next to me, I'm just sharing what's happening.

Yes, I am suggesting starting with a solid cast iron frame. Just having a solid base as you have proposed doesn't do it, IMHO. What you want is a super stiff frame. Every hobby type project of this type I've seen always compromises the design right up front by using aluminum for the frame and then the whole project is compromised. I do understand that you have constraints, money and otherwise, BUT maybe if you were to make just one investment in the project it would be to get the frame super stiff. If I were doing such a project, I would look at all the professional CNC machines out there and understand how they designed their frames and then scale down in size. As mentioned, my brother speedio CNC has a LOT of metal, all one piece, that makes up the frame of the machine. Anything that moves leverages off that big chunk of metal. It is not built out of aluminum struts that flex.

 

[EmGo]

This is built at home....so is home built....
Epoxy granite cnc mill walk around - YouTube

aka Harcrete®. That may have been mentioned earlier ...

 

[Bakafish]

Yes, I am suggesting starting with a solid cast iron frame. Just having a solid base as you have proposed doesn't do it, IMHO. What you want is a super stiff frame.

This is obvious. No one thinks that rigidity isn't a priority.

Every hobby type project of this type I've seen always compromises the design right up front by using aluminum for the frame and then the whole project is compromised.

Please show me actual examples, or even a single example of a 'failure', that is using anything close to the dimensions to what I'm using. This isn't 80/20 it is specialty 100/200, and the unsupported span is a mere ~500mm. It is easy to say all aluminum is bad, but in reality aluminum just needs to be used with its properties in mind. You can design a part that is just as strong as steel. It will be bigger no doubt, it will also be more expensive, but it will be every bit as strong. If you want to provide a structural analysis showing I'm wrong I can point you to the actual materials, otherwise you are stating an opinion backed by anecdotal evidence that you are not even providing.

I do understand that you have constraints, money and otherwise, BUT maybe if you were to make just one investment in the project it would be to get the frame super stiff.

Ignoring the fact that I'm in Japan, what are you actually proposing here? Do you want to provide a link to this 'thing' you are vaguely describing? Pretend I have $10k to drop on this mystery lump, where's the link to it? I'm fascinated that you think this is a thing that exists at any price. You realize that custom castings like this are the main reason that professional machines are so expensive right? It makes zero sense from a homebuilt perspective, and the closest you could come is gutting some old machine, which I think I've made clear is highly cost prohibitive here as you would pay many times the value of the machine (were you able to miraculously find a suitable candidate) to just dispose of the unneeded parts.

In Japan scrapping is ridiculously expensive, and governmental waste management will not accept "industrial" waste which is a far broader category than you would expect. Most building and construction materials falls under that, so you have to arrange for private companies to pick that stuff up at even higher rates than the already expensive government facilities.

If I were doing such a project, I would look at all the professional CNC machines out there and understand how they designed their frames and then scale down in size. As mentioned, my brother speedio CNC has a LOT of metal, all one piece, that makes up the frame of the machine. Anything that moves leverages off that big chunk of metal. It is not built out of aluminum struts that flex.

Well, I know you are trying to help and I'm coming off as a jerk, but this project isn't something that I came up with on a piece of toilet paper this morning. If you are going to offer unsolicited advice on a project, it is going to be welcomed in proportion to the effort you've made to prove your case and provide actionable information. I've been doing research and obtaining parts for very close to a year now. The key point here is that until you start actually trying to make one of these things, you naturally underestimate the difficulties involved. And once you have an actual design, you will often find that the materials (like this BCM (big chunk of metal)) are not things that exist outside of your design and that manufacturing them is not only costly, but ruins the "self-built" aspect when all of the needed processing takes place somewhere else.

That's not even taking into consideration the risks involved... you've spent thousands of dollars on the BCM, and thousands of dollars having it machined, and then you find out that something is spec'ed wrong, or a crash causes a crack at a critical location because of a hidden defect or because you are not really a structural engineer. And what if, despite being a really BCM and a giant investment of time and money, it actually performs like crap anyway? What then? No part of my design should be so integral that it can't be easily changed or replaced, no part should eclipse the value of the project as a whole.

And just because it seems not to be getting through somehow I'll keep saying it, if the Aluminum doesn't work then it will go. The total cost of the 3 spars delivered to my door is less than $700, and they are intrinsically useful for other projects should they not serve this application satisfactorily. My pessimistic goal for them is to be just strong enough to machine their replacements out of steel tubing, which due to the design are virtually drop in if that is necessary. If they can't even pull that off, epoxy granite or another composite is the next most viable material. But I feel, with enough effort, there is a very good chance that they will serve their purpose despite the rampant scepticism. We will all know soon enough, I'm here to document it, and am happy to admit when I'm proven wrong.

 

[Bakafish]

This is built at home....so is home built....
Epoxy granite cnc mill walk around - YouTube

I was going to add this one to my original list, it is a great build. There are a few more out there I can think of as well. But I was just trying to make a point that homebuilt need not mean garbage. There is also a lot of different ways to build a machine, and the ability to build it to your specific needs is one of the most important aspects. Mine still might turn out to be crap, but people dismissing my efforts just because I'm not a professional machinist are clearly close minded.

As for the materials, those machines are really heavy, which is great, mass is great, but not for my application. I have to keep it as light as I can. If the business works out then a cement floored facility with 3-phase will make sense, and if this lightweight design is too compromised then it will be cannibalized and re-built with heavy composite members. But I'm still at phaze 0, and getting this working is the priority.

 

[SVFeingold]

Ignoring the predictable and pointless drama, here's my 5-minute take as a fellow engineer:

I agree with the suggestion to center the X-axis ballscrew between the rails. It's not so much a question of whether the rails can "handle it" - they can - it's more a question of machine deflection. More is more, and less is always better. The force on the tool + the counter-acting force of the ballscrew create a torque couple which will be greater with the ballscrew positioned as it is in your OP.

On the topic of servos vs. steppers, I don't necessarily see any issue with the stepper you've chosen. Oriental Motors are solid products. Is a servo better? Yes, but tend to be much more expensive. If ever you do want to move to servos, I have three suggestions that might work for your situation:

1. Check auctions. Chances are it will take quite a while for something suitable to come up, so I wouldn't hold up the machine construction on that account. Along with servos you'd need a single-phase servo drive. Which do exist but are relatively rare on the auction circuit. I've picked up about 10 servo motors this way. Including 3x brand new Kollmorgen servos, for $100/ea, which retail for about $1500~$2000. Getting matching single-phase drives will be difficult, but possible. And they don't necessarily have to match. It's a pain in the ass option, especially including the shipping to Tokyo and the extra cost to have someone pack and ship them...but it's possible!

2. That aside, check out Teknik's Clearpath integrated servos. They are real servos and run off of 75VDC. They won't be as high performing or reliable as "true" 480V 3-phase industrial servos, but they're solid devices. Assume $300~$600 per servo (which includes an integrated drive). They also have models that are drop-in replacements for steppers, and will accept the same step/direction signals. In other words you can keep the exact same control system.

3. I've not used them, but a few months ago I came across an affordable servo system that was higher performance than the Clearpaths. They are real servos and drives, not weird homebrew/DIY stuff. I thought I had it bookmarked but apparently not, and the name escapes me. I'll keep looking and post here if I track them down. The gist is that you can kit out a 3-axis CNC system for a few thousand bucks. That includes all servos, drives, and IIRC a motion controller. Add the cost of one motor since you're dual-driving an axis.

That aside, I will whole-heartedly endorse whidbey's note on stiffness. A 2.5" aluminum plate with 90% of the material removed will be stiffer than a 1" thick solid steel plate. It will also weigh a fraction as much. This really only matters if you have the freedom to choose either larger extrusions, or have a custom part machined. No idea what that's like in Tokyo but it sounds like it's not easy. Of course you could always have someone here mill it for you and ship it as "scrap metal."

Focusing in on the rails themselves, a few things to note. Maybe you already know this, but lots of engineers don't:

1. Pay close attention to the preload of the bearings. There are varying levels of preload including "clearance" or "no preload." You don't want either of those. The OEM's datasheet or catalog will tell you which is which and what options are available. Any particular bearing size from an OEM is usually offered in a number of preload configurations. Don't assume on that front...confirm with the OEM. Higher preload = stiffer bearings. A big bearing with no/low preload may be less stiff than a smaller bearing with high preload. Yes, the bigger one will be stiffer after you take up all the slop and initial deflection, but by that time things have already moved quite a bit. The downside of higher preload is increased rolling resistance. Another downside is that as bearing preload goes up, the tolerance to system misalignment goes down. Which brings me to...

2. Have a plan for aligning bearings. Not only aligning the axes relative to one another, but aligning the rails to one another. The "move it back and forth while tightening and checking for resistance" kinda-sorta works, but doing it with an accurate indicator is even better. It's common to have a clean reference surface to precisely locate one rail, and then indicate the other one in. There are all kinds of methods you can Google if you've not done it before. The key takeaway is that even very tiny misalignments (i.e. rails not parallel) can generate enormous forces within the bearing blocks. Which you might not notice, until they literally explode. I've seen it happen. The stiffer your bearings and system, the more attention you need to pay to this. Which brings me to...

3. Think about ensuring you are not overconstraining the system. If you have two very stiff bearings, and a very stiff carriage between them, you risk overconstraining the system. Whether due to operating forces, or mismatched amounts of thermal expansion (e.g. your spindle carriage vs. the machine base), or whatever - the result will be much higher loads within the bearings than intended which can cause binding and/or premature failure. To prevent this, usually there is some compliance required somewhere in the system. This can mean a number of things. It could mean running one bearing with a high preload and the opposing bearing with light preload. It could mean compliant structures that are stiff in one direction and compliant in the other (e.g. for your Y axis, it would be stiff in the Y axis and compliant in the X-axis). Servo couplings are a great example of this.

It could also mean just having a sufficiently compliant carriage between the bearings that it's not a huge concern because it isn't capable of generating high enough forces to cause issues for the bearings. The key point is one that is often lost on engineers and laypeople alike: "stiffness" isn't a material property, it's a structural property. Choosing stiff bearings in isolation won't mean much, nor will choosing a stiff machine structure with sloppy bearings. Gotta look at the whole thing as a system. My gut feeling is that the extrusion will be sufficiently "soft" that they provide the necessary compliance. Something to think about.

4. Roller bearings are better than ball bearings. But pricier.

5. If you can, I would strongly consider spreading apart bearing blocks (which share a rail) as much as practical. I definitely see some opportunity there in your Y-axis. It will reduce the impact of any "give" in the bearings, and may lower the forces that the bearings see in use. It will also help prevent binding, but that's not a top concern here IMHO. It may increase the machine footprint slightly (or reduce the work envelope) but if you can do it, it'd be a plus.

That's all I got for now. Good luck! Definitely interested to see how this goes. Here's another home-brew epoxy-granite mill that definitely goes above and beyond the usual maker-fare.

Also LOL at the linear motor suggestion. That will double the cost of this project, if OP is lucky, and he'd get to take advantage of approximately zero of the potential benefits in this application.

 

[Milland]

Also LOL at the linear motor suggestion. That will double the cost of this project, if OP is lucky, and he'd get to take advantage of approximately zero of the potential benefits in this application.

Sometimes an idea becomes a "signature", rather than a practicality. EG signs his name "Linear Motor".

 

[EmGo]

Also LOL at the linear motor suggestion. That will double the cost of this project, if OP is lucky, and he'd get to take advantage of approximately zero of the potential benefits in this application.

Also LOL at you. I know people selling them and making them. They are widely used in chip automation and pick-and-place robots already and they even built a little demo machine for testing switches, which was fun to watch. They have some nice rotary tables, very competitive with conventional systems (except they perform better).

So carry on, my wayward son, LOL about some more incorrect stuff.

Wait a minute, maybe I am wrong. Compared to the useless stepper motor shit he was talking about, perhaps you are correct. I've never priced stepper motors because I would barely use those to rewind excess toilet paper that was overpulled.

 

[drcoelho]

This is built at home....so is home built....
Epoxy granite cnc mill walk around - YouTube

This is exactly the kind of machine I was suggesting, great video, and smart guy. ALL hobby CNC prospective builders should review this video....

 

[EmGo]

This is exactly the kind of machine I was suggesting, great video, and smart guy. ALL hobby CNC prospective builders should review this video....

Does look nice. Table looked a little skimpy under the t-slots but that probably won't matter for his use. Maybe he'll even need a chip conveyor

 

[Bakafish]

Wow, this is packed full of good stuff. I really appreciate your insight here.

Ignoring the predictable and pointless drama, here's my 5-minute take as a fellow engineer:

I agree with the suggestion to center the X-axis ballscrew between the rails. It's not so much a question of whether the rails can "handle it" - they can - it's more a question of machine deflection. More is more, and less is always better. The force on the tool + the counter-acting force of the ballscrew create a torque couple which will be greater with the ballscrew positioned as it is in your OP.

I get this, I really do. The question is, what are the tradeoffs of the 4 possible ball screw locations?

In location A as the model I uploaded here is shown, the obvious drawback is that the force of the ballscrew is actually adding to the forces exerted on the cutting tool. The lever arm is much shorter, I calculated roughly an increase of 30-40% over the actual load seen by the cutting tool, but clearly not helping. The advantage is that I gain over 80mm of Z height clarence over location C and reduce the nod inducing (force b) ~50mm of additional stick out for ball screw clearance that your preferred B location would have.

Location B as you rightfully point out, acts on the center of rotation, and contributes no twisting torque (force a), but we need to provide space for it to fit between the rails, and that means pushing the spindle out from the gantry tube's axis of rotation, and increasing nod forces (force b). I suspect that the loss of precision in this direction will actually be larger than what I would lose in the rotational (force a) direction due to the geometry, but I could be wrong.

Location C actually has the benefit of directly counteracting the forces on the cutting tool (and transferred to the rails) and so would seem to me the ideal location. But it means giving up a ton of Z axis clearance (not height per se, just the gantrys ability to traverse an object.) And I suspect that sacrifice will become painful at some point. Raising the entire gantry to compensate has effects on the overall rigidity, and the C location has difficult packaging issues as well.

The unmarked D would be locating the ball screw on the backside of the gantry and transferring the torque via arms in the A and C locations. I think it puts a lot more strain on the ball screw itself, and adds a lot of fiddly elements, I put it here for completeness.

On the topic of servos vs. steppers, I don't necessarily see any issue with the stepper you've chosen. Oriental Motors are solid products. Is a servo better? Yes, but tend to be much more expensive. If ever you do want to move to servos, I have three suggestions that might work for your situation:

1. Check auctions. Chances are it will take quite a while for something suitable to come up, so I wouldn't hold up the machine construction on that account. Along with servos you'd need a single-phase servo drive. Which do exist but are relatively rare on the auction circuit. I've picked up about 10 servo motors this way. Including 3x brand new Kollmorgen servos, for $100/ea, which retail for about $1500~$2000. Getting matching single-phase drives will be difficult, but possible. And they don't necessarily have to match. It's a pain in the ass option, especially including the shipping to Tokyo and the extra cost to have someone pack and ship them...but it's possible!

2. That aside, check out Teknik's Clearpath integrated servos. They are real servos and run off of 75VDC. They won't be as high performing or reliable as "true" 480V 3-phase industrial servos, but they're solid devices. Assume $300~$600 per servo (which includes an integrated drive). They also have models that are drop-in replacements for steppers, and will accept the same step/direction signals. In other words you can keep the exact same control system.

3. I've not used them, but a few months ago I came across an affordable servo system that was higher performance than the Clearpaths. They are real servos and drives, not weird homebrew/DIY stuff. I thought I had it bookmarked but apparently not, and the name escapes me. I'll keep looking and post here if I track them down. The gist is that you can kit out a 3-axis CNC system for a few thousand bucks. That includes all servos, drives, and IIRC a motion controller. Add the cost of one motor since you're dual-driving an axis.

I really did try to go with Servos. They wouldn't have cost a lot more than my closed loop steppers. The issue was that the used market is exclusively for 3-phase drivers. Since I need this to run off of a 2-phase 60A service, this isn't an option. My OM steppers all use 200v single phase drivers, which are ideal for my purpose. Yes if I bought 'new' servos like the highly regarded Clearpath units, I could likely meet that constraint, but the cost would have ballooned. You will have to trust me when I say that this was the best I could do with the constraints (some real, some maybe more questionable) I have. I promise if I build another CNC after this one I will use Servos.

That aside, I will whole-heartedly endorse whidbey's note on stiffness. A 2.5" aluminum plate with 90% of the material removed will be stiffer than a 1" thick solid steel plate. It will also weigh a fraction as much. This really only matters if you have the freedom to choose either larger extrusions, or have a custom part machined. No idea what that's like in Tokyo but it sounds like it's not easy. Of course you could always have someone here mill it for you and ship it as "scrap metal."

Focusing in on the rails themselves, a few things to note. Maybe you already know this, but lots of engineers don't:

1. Pay close attention to the preload of the bearings. There are varying levels of preload including "clearance" or "no preload." You don't want either of those. The OEM's datasheet or catalog will tell you which is which and what options are available. Any particular bearing size from an OEM is usually offered in a number of preload configurations. Don't assume on that front...confirm with the OEM. Higher preload = stiffer bearings. A big bearing with no/low preload may be less stiff than a smaller bearing with high preload. Yes, the bigger one will be stiffer after you take up all the slop and initial deflection, but by that time things have already moved quite a bit. The downside of higher preload is increased rolling resistance. Another downside is that as bearing preload goes up, the tolerance to system misalignment goes down. Which brings me to...

I said I'd be honest when I made mistakes. This was something I absolutely didn't know when I bought my rails and only discovered when researching them more deeply. The units I am using are Z0 "light preload" rails, and had I been paying full price I would have really been hurting over this oversight. I know that this will hurt my precision a bit, but let's be real, this machine isn't going to be chasing microns, so I try to not think about it. But it is a great point that I wish I had known earlier, just so I was better informed.

2. Have a plan for aligning bearings. Not only aligning the axes relative to one another, but aligning the rails to one another. The "move it back and forth while tightening and checking for resistance" kinda-sorta works, but doing it with an accurate indicator is even better. It's common to have a clean reference surface to precisely locate one rail, and then indicate the other one in. There are all kinds of methods you can Google if you've not done it before. The key takeaway is that even very tiny misalignments (i.e. rails not parallel) can generate enormous forces within the bearing blocks. Which you might not notice, until they literally explode. I've seen it happen. The stiffer your bearings and system, the more attention you need to pay to this. Which brings me to...

I invested heavily on good metrological gear, class A squares, straight edges and high end dial indicator stands. My ability may be lacking, but I'm tooled up to make sure everything ends up properly oriented.

3. Think about ensuring you are not overconstraining the system. If you have two very stiff bearings, and a very stiff carriage between them, you risk overconstraining the system. Whether due to operating forces, or mismatched amounts of thermal expansion (e.g. your spindle carriage vs. the machine base), or whatever - the result will be much higher loads within the bearings than intended which can cause binding and/or premature failure. To prevent this, usually there is some compliance required somewhere in the system. This can mean a number of things. It could mean running one bearing with a high preload and the opposing bearing with light preload. It could mean compliant structures that are stiff in one direction and compliant in the other (e.g. for your Y axis, it would be stiff in the Y axis and compliant in the X-axis). Servo couplings are a great example of this.

It could also mean just having a sufficiently compliant carriage between the bearings that it's not a huge concern because it isn't capable of generating high enough forces to cause issues for the bearings. The key point is one that is often lost on engineers and laypeople alike: "stiffness" isn't a material property, it's a structural property. Choosing stiff bearings in isolation won't mean much, nor will choosing a stiff machine structure with sloppy bearings. Gotta look at the whole thing as a system. My gut feeling is that the extrusion will be sufficiently "soft" that they provide the necessary compliance. Something to think about.

I agree that there is likely going to be enough give everywhere that this isn't an issue, but I will keep this in mind.

4. Roller bearings are better than ball bearings. But pricier.

Would love some, but they seem to be rare where I shop. I honestly got my 6 new in box rails/guides for I think ~$300 total, they were pretty cheap for factory Hiwin heavy duty 25mm rails. My only regret was over the preload as mentioned above,, but it could have been worse I guess.

5. If you can, I would strongly consider spreading apart bearing blocks (which share a rail) as much as practical. I definitely see some opportunity there in your Y-axis. It will reduce the impact of any "give" in the bearings, and may lower the forces that the bearings see in use. It will also help prevent binding, but that's not a top concern here IMHO. It may increase the machine footprint slightly (or reduce the work envelope) but if you can do it, it'd be a plus.

I've targeted a 500x500mm working area, and so the spacing is exactly what's required to support that. It isn't obvious, but part of that work envelope actually extends past the front edge. This is because I want to be able to work on taller items that can be affixed to the front of the machine. It will also help when I add a 4th axis as that can sit below the main base plate in front.

That's all I got for now. Good luck! Definitely interested to see how this goes. Here's another home-brew epoxy-granite mill that definitely goes above and beyond the usual maker-fare.

Also LOL at the linear motor suggestion. That will double the cost of this project, if OP is lucky, and he'd get to take advantage of approximately zero of the potential benefits in this application.

Yeah, I expect we'll be making "It just needs linear motors" jokes for some time... :-)

 

[SVFeingold]

Also LOL at you. I know people selling them and making them. They are widely used in chip automation and pick-and-place robots already and they even built a little demo machine for testing switches, which was fun to watch. They have some nice rotary tables, very competitive with conventional systems (except they perform better).

Pretty sure I never questioned their utility or existence. They are way more expensive, tend to be considerably more bespoke than bolting on a rotary servo, and require scales and more sophisticated drives. Meaning more $$$. A lot more. And speaking of those rotary tables (I assume you meant direct drive, not linear motor) they're fast and accurate, but they also can't handle the same torque/cutting forces as a "traditional" rotary table for a given size/power envelope. It's a tradeoff.

Pick-and-place and other electronics assembly equipment all have thing in common, which is that the loads essentially constant - you're always moving the same mass, give or take a few %. Makes it much easier to really dial the system in and ensure that the performance is repeatable. CNC machines not so much. Plus the electronics equipment can tangibly benefit from the greatly improved acceleration and accuracy. Most PnP machines are still running "regular" servos and ballscrews. You see linear motors when you get into the really high throughput high-volume equipment. Up to that point it doesn't make a ton of sense, and even there they usually only have certain axes driven by LMs, the rest are servo + ballscrew. None of these potential benefits would be realized for OP with that machine. Most CNC mills period would not benefit from such a retrofit.

Part of this whole "engineering" thing is choosing the right thing for the right task. There's rarely just a thing that's "the best," only the best for a specific application, and even then only when the constraints and application actually require it.

Is OP's machine going to benefit from 5G accelerations vs. 1G? Nope. Is OP's machine going to benefit from sub-micron accuracy? Nope. Is OP's machine going to benefit from increased motor stiffness? Nope.

I'm happy to be wrong if you can think of a tangible benefit for OP from switching to linear motors, and/or if he can outfit 3 axes with linear motors for less than $5k - which will buy you a very nice servo system. Otherwise sure they're neat but they're a pointless expense and a whole lot of additional hassle to boot.

There I go again.

 

[Bakafish]

Wait, should I cancel that big order of linear motors I just placed?

 

[EmGo]

Pretty sure I never questioned their utility or existence. They are way more expensive, tend to be considerably more bespoke than bolting on a rotary servo, and require scales and more sophisticated drives. Meaning more $$$. A lot more.

I'm happy to be wrong if you can think of a tangible benefit for OP from switching to linear motors, and/or if he can outfit 3 axes with linear motors for less than $5k - which will buy you a very nice servo system. Otherwise sure they're neat but they're a pointless expense and a whole lot of additional hassle to boot.

There I go again.

One again, you do not have an arfing clue. About a year ago we spent several full days visiting a company making these things. Not just the motors, the drives and the controls AND complete machines of several varieties. Also direct drive rotary tables, and a bunch of other stuff. Mitsubishi thought enough of them to drop a 10% stake into the company. If I were younger I'd be peddling their machine tools, they are pretty nice, but they actually have bigger plans.

You remind me of some other people here who have pooh-poohed various things, but just because you cannot do something does not mean that no one can.

About the op, I'm far beyond caring what the hell he does, it's just another C N C Zone thing. Not interesting.

 

[SVFeingold]

Sometimes an idea becomes a "signature", rather than a practicality. EG signs his name "Linear Motor".

The general vibe I've gotten is "Everything was better in 1980 and everything new sucks. Except for linear motors. Computers haven't improved since 1975."

About the op, I'm far beyond caring what the hell he does, it's just another C N C Zone thing. Not interesting.

Though interesting enough to waste your time on. From the reply you'd think I said "linear motors are fake and nobody uses them." Do you have an answer to any of the actual substance, or are you going to keep trying to convince me that linear motors exist for some reason? I can think of few things less relevant to OP than the investment strategy of a $25B+ global corporation. It's nice you know a guy and saw a demo. Few whole days? Wow. Translate that into something meaningful in the context of this thread.

They exist and they're great if the application calls for them. If it doesn't they're likely a waste. Prove me wrong by showing us the linear motors on your garage door. Drop an estimate for what it'll cost to outfit OP's 3-axis machine with linear motors, drives, and scales. Assuming you know. Which you should, as a linear motor expert. Point out the realizable benefits for OP's application. If you can't do that then what are we even talking about? It's not like linear motors are new anyway, I mean we had linear motors back in 1915! Nothing new to learn, move along, the world stopped changing when I was young, blah blah blah.

Bottom line: if your requirements don't outpace the capabilities of servos + ballscrews, linear motors provide no benefit and many potential downsides. Enough on this baffling tangent unless you can show a picture of your garage door kitted with linear motors. Because they're more better for everything, duh.

 

[SVFeingold]

@Baka, sounds like you've thought about most of this and you're on a good path. Like most such machines I wouldn't expect miracles in the rigidity department, but hell...people cut steel with Shapeokos and you should be well ahead of that barring any major screwups.

As far as the bearing preloads, this usually comes from the bearing block, so if ever needed (or you just have nothing better to do) you can always order new ones with increased preload. Would be an interesting A/B test running the same part/tool/program. You won't be taking heavy cuts with this machine so the benefit is questionable. At some point either have to simulate it, or trial it, to get an idea of the benefit (if any). Guidelines only go so far.

Steppers vs. servos, again depends if you'll notice/care about the difference. Especially closed-loop steppers with encoders. Servos have a higher power density and better dynamics. Does that matter for you? Probably not - and by the time it does you can easily upgrade.

For the diagram, not sure I agree with some of your assumptions.

- The ballscrew is tasked with resisting the cutting forces regardless of where it is. I wouldn't expect to see any difference in loading on the screw (and resultant motor torque) between any of these positions. The whole point of the rails, after all, is to carry all of the loads except those parallel to the rail.

- Even with a ballscrew centered between the rails you'd still have some torque created, just a bit less. In all cases, regardless of ballscrew position, the bearings will barely notice this torque. It only matters in that it will technically reduce the deflection. Will that reduction even be noticeable? Impossible to say without a deeper dive and testing/sims. This will be easiest to test with a dial indicator once it's assembled, but of course that's the hardest time to make changes.

- In consideration of the above I agree it's not ideal to push the spindle away from the X-axis rails for the sake of fitting a ballscrew. I was thinking more to make it fit "inside" the extrusion (cut a trench or some such). But eh. You can accomplish more by just popping in some stiffer bearings someday.

- The weakest point of builds like this, IMHO, is the aluminum extrusion. Unfortunately the hardest to replace. All of the motion components are off-the-shelf, but for the actual frame you'd need to either get lucky/DIY or get something custom. Maybe for another day.

- On the metrology gear, as long as you have what you need to attach a dial indicator to one bearing block and use it to sweep the entire length of the opposite rail, you should be good to go. With a system like this and light preloads...you honestly might be fine just doing the "roll it back and forth and tighten the rails incrementally" method. But actual measurements make this easier so use it if you got it.

- Do your bearing blocks include scrapers, or just rubber seals? Rubber seals work great in general but they're not really intended to remove large chunks and bits of metal. They may get chewed up in a hurry if your rails aren't covered. Eventually, provided you maintain them, they will start pissing grease everywhere as it accumulates, and things like to stick to that grease. This is easy to solve with waycovers and also a quick wipe with a rag now and then. Just don't expect the bearing seals to last toooo long if they're tasked with scraping chips off the rails.

The lever arm is much shorter, I calculated roughly an increase of 30-40% over the actual load seen by the cutting tool, but clearly not helping.

Can you elaborate on this a bit? Not entirely sure what you mean.

 

[Milland]

If you can find a bearing ball supplier who stocks micron +/- on either side of common diameter (3mm, 3/32", etc.), you may be able to reball your linear trucks, as well as the ball screw. But that's a bit of a challenge if you can't measure the balls correctly or repeatedly. And of course, the right size is always the one not stocked...

I think it's pretty clear that you need a properly supported known-straight reference to get the rails installed correctly. Just sweeping one rail to the other can give you a parallel squiggle, not ideal for precision machining.

The two Y Al extrusions will have to be packed to be any use, even with just a lightly bonded sand (mixed with a weak glue, like Elmer's?). Otherwise they're just two flexures in X. A water-based glue might change with evaporation, but would be easiest to use and dispose of afterwards (just break up the clumps, nothing hazardous in the mix). Plan on installing endcaps at the least.

 

[Bakafish]

@Baka, sounds like you've thought about most of this and you're on a good path. Like most such machines I wouldn't expect miracles in the rigidity department, but hell...people cut steel with Shapeokos and you should be well ahead of that barring any major screwups.

I looked at the Shapeoko (before the Pro came out) and said, "I can do better than that..." so I guess that was my admittedly low initial target :-)

As far as the bearing preloads, this usually comes from the bearing block, so if ever needed (or you just have nothing better to do) you can always order new ones with increased preload. Would be an interesting A/B test running the same part/tool/program. You won't be taking heavy cuts with this machine so the benefit is questionable. At some point either have to simulate it, or trial it, to get an idea of the benefit (if any). Guidelines only go so far.

Yes, in my initial panic I was strategizing how to get them to sell me replacement bearings to increase the preload. Then I decided to relax and try them. :-) As much as I like the Hiwin's, if I were going to dump more money in the rails it would likely be to replace them with something even bigger and more domestic market :-D

Steppers vs. servos, again depends if you'll notice/care about the difference. Especially closed-loop steppers with encoders. Servos have a higher power density and better dynamics. Does that matter for you? Probably not - and by the time it does you can easily upgrade.

For the diagram, not sure I agree with some of your assumptions.

- The ballscrew is tasked with resisting the cutting forces regardless of where it is. I wouldn't expect to see any difference in loading on the screw (and resultant motor torque) between any of these positions. The whole point of the rails, after all, is to carry all of the loads except those parallel to the rail.

I likely didn't explain it well or I have a poor understanding of the forces. Feel free to smack down my poor understanding of the mechanics here. There would seem to be several forces acting on the ball screw, but I was focusing on the one in opposition to the force created by the cutting tools movement through the material. That force is transferred to the spindle backing plate through the ball nut and that plate is located in space by the 4 bearing blocks. I mentally picture the force coming from the cutting tool as a rigid arm that is applying torque around the backing plate's center of rotation (indicated by the small blue circle and arrows marked 'a') in that drawing. Obviously they resist that torque, but the structure and all its elements will be deflected by that force, and it would seem to me that force is the one most influenced by the ball screw location.

What I was trying to say is that locating the ball screw above that center of rotation, when we apply force through it to move the cutting tool, we are contributing additional torque in the same direction of rotation that the cutting tool is, so the rotational torque is increased by the difference in the ratio of the two lever arms. (If that makes any sense.)

The amount of leverage created by the cutting tool would vary (distance from cutter to rotational center from ~350 down to 80mm at max Z height), as the spindle moves up the lever arm would become shorter. The lever arm length of the ball screw would be fixed (at ~140mm.)

Conversely, being located in C the rotational torque would be reduced as the two lever arms are in opposition, so the ball screw would be canceling out some of the force being exerted. And in location B it would be applying force at the center of rotation and have no contribution to that specific force one way or another.

The one thing I may change is that currently all the 28mm ball screws are only rigidly retained at one end, the other sides are using a floating deep groove bearing support block. I know that pre-loading the X axis ball screw along its length by using another pair of angular bearings at the end will add rigidity and help it resist this rotational torque (if it isn't a figment of my imagination) by way of reducing the vertical deflection of the screw that has to happen if the plate rotates. (again, I hope I'm not speaking gibberish.)

- Even with a ballscrew centered between the rails you'd still have some torque created, just a bit less. In all cases, regardless of ballscrew position, the bearings will barely notice this torque. It only matters in that it will technically reduce the deflection. Will that reduction even be noticeable? Impossible to say without a deeper dive and testing/sims. This will be easiest to test with a dial indicator once it's assembled, but of course that's the hardest time to make changes.

Yes, this is my feeling as well. I looked into what I'd have to do to simulate it, and decided that a few minutes on the finished unit with a sub micron Mitutoyo would be far more illuminating. Something I will try to test on a sub assembly as soon in the process as possible.

- In consideration of the above I agree it's not ideal to push the spindle away from the X-axis rails for the sake of fitting a ballscrew. I was thinking more to make it fit "inside" the extrusion (cut a trench or some such). But eh. You can accomplish more by just popping in some stiffer bearings someday.

The model I uploaded isn't very detailed and I wouldn't expect you to have easily seen that. I can create a similarly dimensioned C frame out of smaller extrusions that would have space for a central screw, but I suspect the loss of strength isn't a good tradeoff.

- The weakest point of builds like this, IMHO, is the aluminum extrusion. Unfortunately the hardest to replace. All of the motion components are off-the-shelf, but for the actual frame you'd need to either get lucky/DIY or get something custom. Maybe for another day.

My feeling is that it is hard to replace on the initial build, but then I could use the machine to at least lightly face and mark drill points on a steel square section of tubing. Or add steel plates bolted to the large surfaces of the extrusion like an exoskeleton. Or fill the extrusion with epoxy granite. There are things to try if this is too floppy.

- On the metrology gear, as long as you have what you need to attach a dial indicator to one bearing block and use it to sweep the entire length of the opposite rail, you should be good to go. With a system like this and light preloads...you honestly might be fine just doing the "roll it back and forth and tighten the rails incrementally" method. But actual measurements make this easier so use it if you got it.

Pass up a chance to bring out all my little treasures? Are you mad? I'm going to indicate this poor machine to death... it's going to visibly shudder at the sight of a Mitutoyo box.

- Do your bearing blocks include scrapers, or just rubber seals? Rubber seals work great in general but they're not really intended to remove large chunks and bits of metal. They may get chewed up in a hurry if your rails aren't covered. Eventually, provided you maintain them, they will start pissing grease everywhere as it accumulates, and things like to stick to that grease. This is easy to solve with waycovers and also a quick wipe with a rag now and then. Just don't expect the bearing seals to last toooo long if they're tasked with scraping chips off the rails.

They came with scrapers and seals, but I want to cover everything up anyway. Exactly how is yet to be determined, I've seen a lot of solutions, but few that are really impressive. It's a hard problem though, worth pondering.

Can you elaborate on this a bit? Not entirely sure what you mean.

Hopefully I explained better above. Thanks again for your adult and constructive engagement!

 

[Bakafish]

If you can find a bearing ball supplier who stocks micron +/- on either side of common diameter (3mm, 3/32", etc.), you may be able to reball your linear trucks, as well as the ball screw. But that's a bit of a challenge if you can't measure the balls correctly or repeatedly. And of course, the right size is always the one not stocked...

I think it's pretty clear that you need a properly supported known-straight reference to get the rails installed correctly. Just sweeping one rail to the other can give you a parallel squiggle, not ideal for precision machining.

The two Y Al extrusions will have to be packed to be any use, even with just a lightly bonded sand (mixed with a weak glue, like Elmer's?). Otherwise they're just two flexures in X. A water-based glue might change with evaporation, but would be easiest to use and dispose of afterwards (just break up the clumps, nothing hazardous in the mix). Plan on installing endcaps at the least.

Good thoughts. I was just responding about the ball bearings :-)

I have multiple class A straight edges of various lengths... some might think I have a hoarding problem... they wouldn't be wrong.

I was thinking that a water soluble plaster and sand mix might be good, adding some water soluble glue wouldn't hurt though. Was also thinking about a heat releasable adhesive that I could coat the insides with and then something like epoxy granite could be freed by applying heat. I want to start with the least permanent solutions though.

I will have all the end caps covered as there will be structural plates on all of them.

 

[Milland]

some might think I have a hoarding problem... they wouldn't be wrong.

When it comes to metrology and machine tools there's no such thing as hording.

Just inadequate storage.