[Adoug]

I'm just so tired of trying to find a way to get a relatively accurate solution worked out. I know it mostly comes down to getting a machine with good repeatability and then dialing it in with dial indicators, test geometries and so on. I am willing to do that. But I am always so frustrated in so many departments on how what I need is not on the trodden path, that I always seem to have to be a pioneer, when this is pretty basic stuff. I mean not entirely, there is linuxcnc, which is excellent work, and it has very basic capacity for software compensation for squareness error, apparently, although I have yet to determine if it is working/completed.

What I would like is to buy a pre calibrated machine at a reasonably price point. They sell highly accurate machines for making PCBs, but they have z travels that are too small, and they are ridiculously expensive. Like $15,000 for a 3040 sized machine, if you want accuracy of about 10 microns.

The thing that bothers me is also the gap between the absolute requirements of mechanical engineering and what the machines can deliver. I mean, if you have a post that needs to fit in a hole, and they are both off by 50 microns, that's just not going to do at all, is it? It's likely not going to fit to the bare minimum. So many other examples of basic features and aspects of machines and systems that just need far higher tolerances than that. And you can get that from expensive machines. I have read that a nice HAAS machine will give you in the range of 10 microns of accuracy in it's motion, right out of the box.

I think maybe we need to develop a calibration technique. I am baffled that for PCB machines etc. that this approach is not used - make a highly repeatable machine, then attach the spindle to a measurement device, which can be very expensive, a CMM machine, basically, and run a program that moves the machine around, and builds a table or set of polynomials or whatever, relating commanded position to actual position. Such machines, instead, appear to be made from extremely accurate and therefore expensive components that are laboriously assembled in ways to reduce error stack up.

I know this won't help with some issues, the spindle axis still has to be square with the z axis motion, the runout has to be ok, things like that. But it would help a lot. And yes in reality tool deflection and wear and so on all matter ( I also aim to get a tool change spindle, that's another case where I am muchly frustrated by the lack of foundation, however you can buy tool changers for 3040 and thereabouts sized machines through alibaba, they are just pretty expensive and hard to source).

Anyone else with me, wanting a more accurate machine? How do you hope to go about it/ have you gone about it?

I think we have a real problem in society where people see they can't really do things, like make accurate parts, then they don't demand it, then the market doesn't offer the tools to do so, then it just gets pushed out of imagination. We need to move forward. Demand higher accuracy machines, to open the doors that we can see will open when mechanical engineering can be really acted upon, the concepts made real, have their potential realized, becomes more accessible and widespread.

[awerby]

There are very accurate machines, and there are very cheap machines. What there aren't is highly accurate machines that are also very cheap. Precision costs money; you can demand all you want, but if you're not willing to pay for it, you won't get it. If you've got ideas about how to make a high-tolerance machine for less money, go ahead and build one, put it into production and offer it for sale at the price of a cheap Chinese router. The world will beat a path to your door (if it really works).

[Jim Dawson]

What Andrew said ^^^^^

First you start with a massive high quality machine, then add to that. I do have a machine that will hold +/- 1 micron, at least over distance of 4 inches, verified on a high end CMM. It's mechanically very tight and has no detectable backlash in any axis. It also weighs over 5000 lbs. But it was not cheap to it get to that condition. Even in this case the resolution at 1 micron is really too low for extreme accuracy, should be 0.1 micron resolution. It is really not important anyway, I couldn't measure that close in my shop. Without thermal control, that kind of accuracy is pretty much impossible. I'm perfectly happy if I can hold +/- 0.01mm consistently on the rare occasions that I actually need to do so.

[deadlykitten]

hy i work mostly on okuma lathes ... turret and spindle are no longer aligned, yet i can drill 6d

with the tailstock, i have conicity of 0.02/1000 ....

precise as they are, actually, more important is how you use them; for accuracy, i use custom toolpaths, like, for example, always approaching from same direction, just like on a mill, if you would need to deliver 3 holes, one should go 123, not 132, since there will appear backlash errors

backlash compensation technique (parameter declaration), indications and trials, simply shows that accuracy depends on travel and speed, and in it's setting, one has to average rapid and feed scenarios ( unless an abso scale is there )

with all these, i really imagine the hard time that someone has with a machine that needs to go at home position from time to time, so to reset it's zero, or not to handle thermal deformations ...

i really have no clue about building machines, especially accurate ones, but if i should, i would contact guys that implement automation ( so they know what their screwballs can do ?! ) and also guys that build fancy routers, or whatever, or maybe someone that does both ... i don't know, good luck with whatever you wish to do

if you know what to do, i agree that you can retrofit a solid machine, and really get something in return, like a solid accurate cnc, cheaper than from a dealer; i can't do that ... but jim dawson can, and also there a few others that i know that can do such things; speak with them

[joeavaerage]

Hi,
just as an example of why precision costs as much as it does....

Just about all CNC machines rely on ballscrews to translate rotary motion from a servo or stepper to linear motion.
A C7 grade rolled ballscrew, the type which abound on Ebay and the likes, can be had for say $150 for a 20mm diameter screw, ballnut and support blocks, and C7 is 50um/300mm, 35um cyclic.
The same size screw in C5 grade is about $1500 for screw, ballnut and support bearings.....that's ten times the price!. C5's are 18um/300 and 8um cyclic, and could still only be called entry level
for accurate machines. C3's are better again but about $3000. You can see where this is going, despite a C7 ballscrew looking nearly indentical in size and construction to the C3 along side it the
C3 is twenty times the price.

Highly accurate components like ballscrews and linear rails/cars are essential components in an accurate machine, but if the machine flexes then its potential accuracy is lost. Thus an accurate machines
may weigh ten times what a similar sized hobby (flexy) machine might weigh.

I wanted to build an as accurate as I could make mill. I bought near new but secondhand 32mm C5 BNFN double nut THK ballscrews and supports, new old stock 20mm HSR rails/cars by THK
and 750W Delta B2 series servos. I had the three axis beds cast in iron for $3000 and had them machined/heat treated for another $5000. You can see where this is going.....it costs.....it costs....it costs.
The build is in service, it has travels of 350mm x 350mm x 350mm, g0's of 25m/min, accels of 2.5m.s^-2. I was hoping for a resolution/accuracy of 1um but have in fact something like 8-10um
resolution accuracy....IF....I don't overload and start flexing it! Getting a lot of use and much enjoyment and satisfaction but I have paid dearly for that satisfaction.

Craig

[Algirdas]

The new branch of industry is up to start with totaly new philosophy of accuracy. You can get high accuracy without rigid frame, without casted body if you use high accuracy vision or whatever other method to know where your work point is.

[deadlykitten]

you are right mr bunny for example, last year, there was an automation&digitize workshop : there where only 1-2 technology providers for the automotive automation sector, and many many new players, with only a few months old, playing into big data management fields ... something is changing

[joeavaerage]

Hi,

The new branch of industry is up to start with totaly new philosophy of accuracy

What will also be new is the truly eye-watering prices such systems will command. If you think this is the way to affordable precision CNC then
have at it...and good luck to you. The rest of us will have to make do with the best components we can afford in the most rigid frames we can make.....
ho-hum....at least it works.

Craig

[cncmakers001]

There are very accurate machines, and there are very cheap machines. What there aren't is highly accurate machines that are also very cheap. Precision costs money; you can demand all you want, but if you're not willing to pay for it, you won't get it. If you've got ideas about how to make a high-tolerance machine for less money, go ahead and build one, put it into production and offer it for sale at the price of a cheap Chinese router. The world will beat a path to your door (if it really works).

Couldn't agree more.

[CitizenOfDreams]

The new branch of industry is up to start with totaly new philosophy of accuracy. You can get high accuracy without rigid frame, without casted body if you use high accuracy vision or whatever other method to know where your work point is.

Not exactly a new concept. One common example is a CD/DVD player: the laser head is positioned by a flimsy and sloppy leadscrew driven by a crude stepper motor. Then the optical lens makes sub-micron fine adjustments to keep the laser beam on track.

Unfortunately, it's not so easy to do when you need to apply some cutting forces to the tool. Try carving something with a knife that has a piece of rubber hose for the handle.

[Algirdas]

Mr.CitizenOfDreams That's exactly what I mean. The human can carve manualy quite precise parts because of reaction to variation of conditions. Traditional machine tools need rigidity for precission, which is not right. Rigidity must ensure the potential force applied to the workpiece. Accuracy should be achieved my means of observation and keeping the alowance. That's 21st age, people, there are technologies alowing us to do that.

[joeavaerage]

Hi,
skilled trades and craft people have often surprised and delighted the rest of us that they can manually achieve precision they do.
Humans learn the ability to apply force to a tool sufficient to cut, say, but in addition supply the extra movement necessary to overcome
the inevitable flexure of the tool, the work holding and even the human operators arm. Human ability to learn and do things like tis
is amazing.

Conceivably its possible for a machine to do similar things. Its called kinematics.

In a regular CNC machine if an axis is required to move 50mm the servo must rotate an exact amount, lets say 720 degrees, with a 25mm pitch screw that is 50mm linear.
But what happens if there is an opposing force....the servo rotates its 720 degrees but only gets to 49.95mm. If the controller knows or can otherwise measure the opposing
force it might apply a correction, this would be called feed forward. If the machine can measure the position then it can and issue further steps to the servo to make up for the motion
lost to flex, this is called feedback.

The most advanced CNC machines use both methods.

The mathematical complexity and consequent computational power not to mention the high precision high bandwidth feedback measuring devices are very expensive,
way more, I argue, than just using precision components and high rigidity machines.

May I suggest that this 'pie in the sky' stuff is all very interesting but if you want a machine to make parts....then come down to earth and see how parts are made and on what machines
to get a desired level of accuracy. You and your budget will just have to slip in there somewhere.

Craig

[stevehuckss396]

You must be because i'm happy with the machines I have. On my lathe, once the spindle warms up, i can make parts for hours at +-0.0005. That's good enough for what I am doing and in some cases over kill..

[deadlykitten]

i remembered that a while ago i saw a video about a fast router/minimill, that was not fixed to the ground, but somehow it was mounted inside another chasis, having some dumpners or something

during cutting, it was shaking, while the outer chasis was fix; something like you have mini cnc machine that cutts, but you start to shake it, thus it will continue it's work, only that it will be shaking

i think that such a design was there to handle vibrations associated with high speed repositioning ? i don't know, maybe i will find a video / kindly

[CitizenOfDreams]

Are you talking about the Shaper Origin?

[pippin88]

What I would like is to buy a pre calibrated machine at a reasonably price point. They sell highly accurate machines for making PCBs, but they have z travels that are too small, and they are ridiculously expensive. Like $15,000 for a 3040 sized machine, if you want accuracy of about 10 microns.

I think maybe we need to develop a calibration technique. I am baffled that for PCB machines etc. that this approach is not used - make a highly repeatable machine, then attach the spindle to a measurement device, which can be very expensive, a CMM machine, basically, and run a program that moves the machine around, and builds a table or set of polynomials or whatever, relating commanded position to actual position. Such machines, instead, appear to be made from extremely accurate and therefore expensive components that are laboriously assembled in ways to reduce error stack up.

Once off calibration of a repeatable but imprecise machine will not be the ultimate solution.
What happens with wear?
Something gets knocked slightly out alignment and has to be adjusted? Your whole map will be off

Screw mapping etc are already used in commercial machines.

Sent from my SM-G970F using Tapatalk

[deadlykitten]

Are you talking about the Shaper Origin?

hy citizen ( of dreams ) that's an interesting machine ... never heard of it, and it seems to be pretty cool for someone that really needs it, to make random stuff

i just made a sketch : the inner is shaking, while the outer is fixed, mounted to the ground; i don't know how is called, or why it is like that, but i remember that it also had a vice, of maybe 100mm ?! it was not big, but also not tiny; it was labeled something like fastest router in the world, also precise ?! fully enclosed, with doors, and a control panel / kindly

[peteeng]

Hi All - Some clues to making high accuracy, precise and repeatable machines. Unfortunately rigidity will always be king, you can't make a wobbly machine that cuts material and expect the electronics, lidar, scales, mapping and all the other strategies correct for the soft structure. When laser subtractive matures this maybe not the case as there is no contact forces. Peter

[mactec54]

I'm just so tired of trying to find a way to get a relatively accurate solution worked out. I know it mostly comes down to getting a machine with good repeatability and then dialing it in with dial indicators, test geometries and so on. I am willing to do that. But I am always so frustrated in so many departments on how what I need is not on the trodden path, that I always seem to have to be a pioneer, when this is pretty basic stuff. I mean not entirely, there is linuxcnc, which is excellent work, and it has very basic capacity for software compensation for squareness error, apparently, although I have yet to determine if it is working/completed.

What I would like is to buy a pre calibrated machine at a reasonably price point. They sell highly accurate machines for making PCBs, but they have z travels that are too small, and they are ridiculously expensive. Like $15,000 for a 3040 sized machine, if you want accuracy of about 10 microns.

The thing that bothers me is also the gap between the absolute requirements of mechanical engineering and what the machines can deliver. I mean, if you have a post that needs to fit in a hole, and they are both off by 50 microns, that's just not going to do at all, is it? It's likely not going to fit to the bare minimum. So many other examples of basic features and aspects of machines and systems that just need far higher tolerances than that. And you can get that from expensive machines. I have read that a nice HAAS machine will give you in the range of 10 microns of accuracy in it's motion, right out of the box.

I think maybe we need to develop a calibration technique. I am baffled that for PCB machines etc. that this approach is not used - make a highly repeatable machine, then attach the spindle to a measurement device, which can be very expensive, a CMM machine, basically, and run a program that moves the machine around, and builds a table or set of polynomials or whatever, relating commanded position to actual position. Such machines, instead, appear to be made from extremely accurate and therefore expensive components that are laboriously assembled in ways to reduce error stack up.

I know this won't help with some issues, the spindle axis still has to be square with the z axis motion, the runout has to be ok, things like that. But it would help a lot. And yes in reality tool deflection and wear and so on all matter ( I also aim to get a tool change spindle, that's another case where I am muchly frustrated by the lack of foundation, however you can buy tool changers for 3040 and thereabouts sized machines through alibaba, they are just pretty expensive and hard to source).

Anyone else with me, wanting a more accurate machine? How do you hope to go about it/ have you gone about it?

I think we have a real problem in society where people see they can't really do things, like make accurate parts, then they don't demand it, then the market doesn't offer the tools to do so, then it just gets pushed out of imagination. We need to move forward. Demand higher accuracy machines, to open the doors that we can see will open when mechanical engineering can be really acted upon, the concepts made real, have their potential realized, becomes more accessible and widespread.

You can make any machine with just ( 2 ) main requirements, C3 Ballscrews and AC servos with 20Bit Encoders, the next requirement you need is a rigid frame, and you will have your 10 micron machine or better, even an old CNC Bridgeport can do better than 10 micron with these parts. the control does not have much to do with it, any control with the right parts in the machine build will do what you want

[mactec54]

Hi,
skilled trades and craft people have often surprised and delighted the rest of us that they can manually achieve precision they do.
Humans learn the ability to apply force to a tool sufficient to cut, say, but in addition supply the extra movement necessary to overcome
the inevitable flexure of the tool, the work holding and even the human operators arm. Human ability to learn and do things like tis
is amazing.

Conceivably its possible for a machine to do similar things. Its called kinematics.

In a regular CNC machine if an axis is required to move 50mm the servo must rotate an exact amount, lets say 720 degrees, with a 25mm pitch screw that is 50mm linear.
But what happens if there is an opposing force....the servo rotates its 720 degrees but only gets to 49.95mm. If the controller knows or can otherwise measure the opposing
force it might apply a correction, this would be called feed forward. If the machine can measure the position then it can and issue further steps to the servo to make up for the motion
lost to flex, this is called feedback.

The most advanced CNC machines use both methods.

The mathematical complexity and consequent computational power not to mention the high precision high bandwidth feedback measuring devices are very expensive,
way more, I argue, than just using precision components and high rigidity machines.

May I suggest that this 'pie in the sky' stuff is all very interesting but if you want a machine to make parts....then come down to earth and see how parts are made and on what machines
to get a desired level of accuracy. You and your budget will just have to slip in there somewhere.

Craig

Pie in the sky, " no " this is easy to do, even on a manual machine, 10 micron is a big number, 0.01mm=.000397"

Any basic machine shop with modern machines can do 10 micron work, or skilled machinist's can do even better than this on manual machines, so that is a big tolerance for some shops machining parts, this is easy to achieve, for some of us.