[markvanH]

This post could also be called Where to Stop in the Search for Precision

I've been considering building a small machine to mill soft metal (steel would be a bonus but hey). Up to 0.5m x 0.5m cutting area. I could live with smaller. Not for anything particular, but parts for CNC and other machines could be the end game.

It appears that the only unpowered spindles that I can find that I can afford have a runout of 0.01mm at the spindle.

I presume that this runout fully determines how much effort and money I should be putting into rails and ballscrew accuracy (given suitable frame rigidity and accuracy of course).

For example, THK gives C7 accuracy as ±0.05mm per 100mm. I could accept that for a first machine, buy C7 ballscrews with double nuts and say that anything better than ±0.1mm accuracy on my machined parts is great. And live with that I'm not going to make any great CNC machine parts with that accuracy.

Or I could scan ebay for used C5 ballscrews and build a machine around them when I find them, in the hopes of attaining greater accuracy.

Similarly for the frame built from steel section: It's easy enough to go out and get flat and perpendicular surfaces milled (to what accuracy IDK) on steel section, or, at the cost of time, to do some hand scraping using the Withworth three plate method to get 'flat' surfaces for rail mounts.

But even then I assume I can only get ±0.02mm flatness using hand scraping, and that I may need to employ some degree of shimming, where the thinnest shims I would be able to find are one thou / 0.0254mm thick.

So yeah, I am very easily in an OCD-like fashion able to SPEND TOO MUCH TIME OBSESSING when for a first exercise I should just accept some limits, avoid spending too much cash, and be aware that the more time I spend building something, the less time I have with the other kind of fun, which is using it.

Any general thoughts on limits to accuracy sought, or indeed anything else raised by this post, please?

Thanks so much!

[markvanH]

I've passed by this thread before " Need Advice On CNC Vertical Mill Parts DIY Epoxy Granite Hard Milling Steel" https://www.cnczone.com/forums/cnc-d...cad-forum.html

User @joeaverage has this to say about ball screws in a comment there that is generally illuminating on ballscrews

"C7 screws are rolled and have a linear accuracy of 50um/300mm. That sounds good but they can and do have cyclic error, and is uncontrolled but can easily exceed 35um/rev. If you are targeting 0.1mm accuracy then these are ok. If you want 0.01mm accuracy they will not be up to it.

C5 and C3 are, and traditionally have been, made by grinding, and are mega expensive. A C5 screw is about 18um/300mm, which is better, but the real kicker is the cyclic which is 8um/rev.

C3's are better again, about 12um/300mm and 6um/rev.

So both C5's and C3's are desired and required if you want accuracies of 0.01mm."

https://www.cnczone.com/forums/cnc-d...ml#post2550404

[peteeng]

Hi Mark - Accuracy has two considerations (plus there's many more once you get into it). Repeatability and precision. If you require "precision" then point to point tolerances as you discuss are very important. However most parts require repeatability which will be comparably good with "low" tolerance screws. The search for precision is up to you and your application. Building a machine with 500x500 XY at better then 0.01mm is a commercial level machine and requires careful planning and resources. You mention steel and this is up another level again, especially for a 500x500mm machine. Look at commercial machines that do what you want to do and they will be huge, a 500x500mm mill is not a "small" machine. If you don't actually need 500x500 then 500x250mm is a more realistic size. Keeping machine parts commensurate quality for the machine is quite a juggle. If you use C5/C3 screws then you need other parts (and machined surfaces to match) to match those. The cost escalates exponentially once you work through those items. Keep at it. Peter

[markvanH]

Thanks Peter for your advice and for this

If you don't actually need 500x500 then 500x250mm is a more realistic size.

Maybe even 400x250.

I've been thinking as ever, and am tending to what I should do is build a machine with lower accuracy, use it, and in that process think about upgrades in two ways - if i need them for what I end up doing with the machine (and applications will emerge, of that I am sure) and, should I need to upgrade, how I might do that.

Of course, building a lower accuracy machine will have its challenges in itself but I look forward to them.

[peteeng]

Hi Mark - What do you want to do with the machine? A machine with a purpose is much easier to design then a general machine. I personally would not consider a possible "upgrade" approach. This rarely happens as the DNA of the machine is set at the point of design and the weak link in the design will knobble the "upgrade" plus the possibilities are too many to consider which makes too much thinking and not enough doing trying to cover so many bases. Think more along the terms of designing/building the best machine you can at the time (for its purpose) and in the future maybe build a better machine with the machine you have. This is how machine builders have progressed over the last couple of hundred years. If it's your first machine get the design roughed as quick as possible, have it reviewed on the forum (be prepared for tough talk) and then polish it a bit then get it built. You will learn more on your first machine so your No2 will be much, much better. You'll be making chips in about a year... Peter

[markvanH]

Hi Peter

Thanks for your ongoing advice.

What do you want to do with the machine?

Just a few thoughts

I can only give a few examples and characteristics: Right now I want to make a pantograph based engraving machine, nothing as sturdy as the machined cast iron ones I see from time to time, but for low occasional use. I can see a way to make parts for this on a CNC machine, but in a week or so I'm going to 3D print some bits and see how I get on with them.

Also plates and mounting hardware for other CNC machines (or for it itself).

And finally as examples, rather low accuracy bits and pieces for automobile projects, nothing challenging there.

But in general, and this may be challenging, because I think we may be talking about inference fits, I am interested in being able to embed round bearing shafts and ball bearings in thicker ally plate. Of course, there is always the possibility of heating up the plate, and putting the part to be embedded in the freezer and fitting tthem respec tively hot and cold, which lessens the accuracy requirements.

You'll be making chips in about a year... Peter

Yeah I can imagine that.

Think more along the terms of designing/building the best machine you can at the time (for its purpose) and in the future maybe build a better machine with the machine you have. This is how machine builders have progressed over the last couple of hundred years.

This is interesting on general principles, and I've been wondering about the general principle, can one use a machine (eg lathe, milling machine, even CNC machine) to manufacture parts to make a more accurate machine. I wonder where I read up on that, not that I have made any comprehensive search for material. Just musing here ...

[peteeng]

Hi Mark - A low tolerance machine can make high tolerance parts once you understand the machine, use careful procedures and have adequate reference tools & measuring equipment. But this is a slow process. This is how engineers have made better tools and machines over time. It is obviously impossible to make a more accurate machine to make that less accurate machine on less accurate machines (if that makes sense, chicken and egg stuff)... If you want production quantities at a certain tolerance then you need a machine that is say 10x more accurate at least then the tolerance you want to produce at. I'll find some info on this for you... This is interesting:

The Fascinating History and Development of Lathes | MachineMFG

I'm trying to find the story about a canon lathe I think. Maybe Wilkinson... He got a grant from the British govt to make the machine, hand made the first parts for the first one, took 3 years then used that machine to make a better machine then used that machine to make the final machine. 8 years I think from memory...

mechanical engineering - How are increasingly accurate machines made? - Engineering Stack Exchange

this discusses "Foundations of Mechanical Accuracy" by Moore. Its on the web and a very good read... Peter

Why make a pantograph? A cnc is the way to go? I used to play tennis with a machinist that was a pantograph specialist. He used to make rather large zipper parts, then use the pantograph to shrink it and make the zipper moulds... when they were metal and we made that sort of stuff in Oz.

There is one thread in the forum where a Maker builds a small plywood router, uses that to make a better bigger plywood and metal router then uses that to make a very good router or a mill I forget. I make plywood and aluminium laminate machines and I'm sure I could make a good mill from that combo. Easy to work with and stiff and damp...

[peteeng]

Hi Mark - I do recommend making a prototype machine first. Design or select your "tool train" ie CAD/CAM/controllers and whatever software / hardware you will use. This is a steep learning curve in itself. Then design your machine, use cheap rails etc so you don't break the bank. Heads up - do not buy anything until you have finalised the design and are sure that's it. I can guarantee that if you buy parts early things will change and they end up on the garage wall... Then make the machine in plywood or plastic etc This allows easy modification and identification of where wiring , cable guides, enclosures, extra holes, etc etc will go. Some of these are too difficult to do in CAD better done in the flesh. Then once the proto is sorted you can go into the real one with confidence. ie either same as P1 or P2 or with the many learnings. Some say move fast, make errors correct and go forward faster... I'm at machine 7 (designing 8 & 9) and I think I'm just getting to understand various subtle things about these types of machines...Peter

[joeavaerage]

Hi,
the real question here is do you need to cut steel or not? Steel is that damned tough that a machine for aluminum and brass is one thing but a steel capable machine is another.

If you want to cut steel then you need rigidity, everything else is secondary. For instance what matter is you paid mega dollars for C3 ballscrews but when the tool engages the steel workpiece the machine
flexes 0.1mm, and that is entirely possible, where then is you um accuracy C3 ballscrews????

The question is 'Do I want steel capacity?'....there is no such thing as an aluminum machine that can 'do a little bit of steel', it is either steel capable or not.

For the purposes of this post I will assume that you do want to be able to do steel, and in that case these would be my suggestions:

1) Reduce the size. Most of us want a big machine in which we can fit big parts....the problem is that big machines require immensely stiff components which in turn makes them too big and most
importantly too expensive to build. If you reduced your travels to 300mm X, 150mmY, 300mmZ then you are still going to make immensely stiff components but is a much smaller size and therefore cost.

2) Ballscrews and linear rails/cars are the heart of the machine and are a large part of the determination of accuracy and rigidity. They are not amenable to being swapped out at a later date. I would therefore decide
early what sort of accuracy you want, and if you determine C5's or C3's are required, get them right from the start. If you machine is small (see recommendation 1) the the probability of finding suitable C5's or C3's
as second hand or new old stock is very good. As the machine gets larger the choices (second hand market) diminish greatly.

3) Use solid sections. peteeng has described steel sections (RHS, SHS,I beams,C channels) as 'globally stiff' but 'locally compliant'. That is to say that they make great components for a house or a bridge. They achieve great strength
and stiffness despite their (comparatively) thin walls, whereas in small pieces that we would use in a CNC machine the walls are so thin that they flex. I would reccommend that you use steel billet. 50mm steel, 75mm steel and 100mm steel
are expensive to get cut certainly, and then you have to machine them...so not cheap....BUT.....still works out the cheapest for a component of a given (and required) stiffness. There is a good reason the steel is humanities
premier engineering material.

My first mini-mill (180mmX, 180mm Y, 180mmZ) was made out of cast iron....I got three cast iron elevator counterweights about 200kg each from the scrap yard and then cut and machined the axis beds etc. Cast iron is a great material, my new mill
I had three cast iron axis beds cast for me, 115kg each. The casting was expensive but the machining and stress relieving was much MUCH more, but the result is superb.

The bottom line is that if you want stiffness then solid steel and cast iron are the preferred and ultimately cheapest solutions.

4) Be realistic. I can only guess you are somewhat like me, a perfectionist. 'Perfection is the mortal enemy of good enough'. This is a lesson that I have had to learn again and again. I want N,M and P but can scarcely afford n,m and p.
The trick is to design and build the best you can, but you have to be able to build it. If for instance you decide that you want this particular toolmaker to machine your frame because of his undoubted expertise and equipment but is
going to cost $10,000....and that stops you dead in your tracks. If however you went to a local engineering shop and they could do the job for $4000, even if were not as good, but you could afford to do it ....just?????
A machine that you built, whatever its faults, is still a better machine than the one you can only dream about.

peteeng's recommendation about building a small machine first with a view to making a bigger and better machine down the track is spot on. CNC design is a learning curve and you'll never build a perfect machine...ever....and most
certainly not the first time. There are things that you will dismiss as unimportant, like flood cooling say, that is years to come that you'll change your mind and it goes to the top of the list or near to it. Conversely there are things
that you think are important like uber accurate spindle speed, only later to realise that is matters not a jot.

This sort of brings me back to the first recommendation, build small and cost effective. As your skill and appreciation grow THEN design a bigger machine.

Interestingly I used my first mini-mill for about seven/eight years before I built my new mill. I disassembled the mini-mill and robbed a few parts that were going to be reused. Recently (a few months ago) I gave it to Johnno,
my workmate, and he is busily rebuilding and upgrading it. The C5 ballscrews are still good, but the coolant got to and wrecked the linear rails.....so he's bought new rails and a new 1.5kW spindle. All-in-all it is very satisfying
that even my first machine has a useful second life.

Craig

[joeavaerage]

Hi,

But even then I assume I can only get ±0.02mm flatness using hand scraping,

Hand scraping is a useful technique to go from 0.01mm to 0.001mm, but any suggestion that you can remove as much as 0.1mm is a joke....you'll be there forever. Milling or grinding to within at least 0.01mm
is the answer, scrape if you think it will add anything but don't imagine you can remove any large volume of material.

Craig