[Roguish]

A very basic question for the machine builder. How square does a machine structure need to be? And how does one build it square?
A standard 3 axes machine (cnc or manual) needs 3 orthogonal planes, within its working envelope. How does one actual fabricate that 'squareness'? And how does one measure the 'as fabricated' squareness or orthogonality?
Do we measure in degrees (of tilt) or inches per inch, or ???
Also, how it the flatness of each plane established and measured?
There have been many threads dealing with machine accuracies and tolerances and backlash compensations and lead screw/ball screw compensations, but I have not seen one dealing with planarity and orthogonality.
So, here it is.
What do you think? How square is square and how do ya get there from here?

[VWSatOz]

I believe your machine is only as good as it is square & straight, & these are the two most elusive qualitys to achieve. It's not much use having .01mm accuracy of achievable tool postion if your X&Y slide motions are out of square by .1mm over a rectangle area travel of 300 by 600, as then if two components are milled or bored in mirror image they will have a mating error of DOUBLE the out of squareness ie a mismatch of .2mm!
To check for STRAIGHTNESS of a machines slideway place two thin long as possible plates on top of each other, clamp down without distorting the jobs & m/c along one edge (or both edges if you can) to the best finish you can get. Then remove the pair & roll/flip the top one over to hold/mate against the lower one. Any errors of straightness, waviness etc will be DOUBLED and easy to see... simple!
If you have machined both sides to get parallel strips then you measure accross the assembly with a micrometer, you should get twice the original width all the way along, or you can use feeler gauges between the centre joint or just look for light between the edges.
To check for SQUARENESS mc a thin plate say 300mm square along two edges X&Y ie one corner, then keeping the Xedge as the Xface flip the job over, clamp it down again clocking the same Xedge to zero with a dial indicator. Now when you clock up the prev. mcd Y-edge it should read zero again! If not then it shows DOUBLE of your machines out of squareness error.
If you have built your machine so these axes can be adjusted in relation to each other then it is a simple thing to improve your accuracies & get straighter or squarer! Just nudge & repeat the tests! No fancy tools needed! You can even "create" your own pair of square edged plates using the same method & use these held at 90degrees to clock along to set up your vertical z axis.
I have read where the most accurate surface grinders made in the world have been checked using this same simple method, mirror-mating of a ground pair.

[mrvinton]

Get this book it explains everything. It is the best book I have ever see on the subject. Published by Moore Too Co. It starts out explaining what flatness,straightness ect. is and how it is obtained and measured and then goes through the process of showing how a Moore Jig Mill is scraped and aligned and what tolerance everything is held to. Expensive book $150 but well worth it.

http://mooretool.thomasnet.com/keywo...pe=2&keyType=P

[Rick Linnabary]

I have used the same sort of method as VWSatOz said except with testing the accurace of a level. How can you tell the accuracey of a precsion level? You put it on a flat surface and shim or adjust the surface is level within .0005 or better per foot, then you turn the level around 180 degree's and see if it is still level and/or off the same amount and same direction as it was before. Then try turning the level over (if it has 2 sides) and repeat the previous test. You can tell alot using an test indicator and a precision steel angle block to check the Z axis along with using the test indicator and swing a large circle on the machine table to test and see if the spindle is square to the machine table. You may also use a precision level that has gradulations of .0005 per foot to make sure that your machine table is flat and level. Make sure to test the front, back, left and right sides of the machine table. You can also use a hieght gauge that is square to the surface that it is on and a test indicator to check the Z axis in both directions to see how square it is to the machine table. You can do alot with a test indicator if you use your imagination and a bit of thought.... I would get the machine table flat and level to start and go from there.

[lerman]

Taking a somewhat contrarian approach, I'd like to interject a somewhat different opinion.

The axes need not be square to one another. It is sufficient that the actual angles be known. A cnc machine can easily correct for known errors in orthogonality.

It is, however, necessary that the table surface be square to the Z axis. Otherwise, it wouldn't be possible to drill a hole perpendicular to the surface. But squaring the surface is pretty easy. Just surface it with a fly cutter.

On the otherhand, this approach does require that you be able to accurately measure the errors.

Ken

[otisba]

According to MACHINE TOOL RECONDITIONING, by Ed Connelly 1955. page 436-7, With a dial indicator in the spindle on a 6" radius bar, squareness should be within .001 error, low at the back, if at all.
Odie

[Geof]

Are you asking the question for general interest or are you seeking methods you can use for amateur construction?

[VWSatOz]

Thanks for all the comments, I was just 1st in to start the ball rolling &getting the dust out of the old brain box.
It all reminds of a new Bridgeport mill we had of Singapore build, it had an error of squareness between the knee vertical Z axis and the Ycross slide ways (we did not know about at first). You would clock up off the top of the table with an indicator flying about 5" radius and think nothing can go wrong. Then after picking up datum with the Digital Readout's X&Y, you might c/drill with a little short tool, drop the table heaps maybe 4" to drill a pilot hole, then drop the table to ream, or drop it still more to bore the hole with a boring head, and whammo! the accurate diameter hole would be out of postion in the Y-direction (&maybe the X too I can't remember) from where the DRO said it should be. We had to send out to a machine rebuilder for scraping & he made it worse, squarer but all tight & rolly &stick-slip, so we had a go ourselves, scraping filing lapping & found out how hard it is to correct a "non adjustable" axis alignment, the column laying on it's back & the knee dangling off a crane hook it's almost impossible to get a blue reading off the dovetail fit.
It shows that clocking up the table method to get the Zaxis square only works if you have a sliding quill design of Zaxis, and won't work if you have a spindle mounted on a slide, as most of these hobby CNC mills use. You must 1st ensure the table top clocks dead true under the motions of X&Y axis travels, then adjust the Z slide to traverse dead square to this table top, then finally adjust the spindle mounts until the fly-clock is true to the table top
If you draw a sketch with any type of error exagerated you can the easily see what occurs! Think BIG errors, and this will show you the way to design/build your machine so they can be easily ADJUSTED out! Think first & build last! Don't build it like a Bridgeport where to ONLY adjustment is of the angle of the quill, and requires the old scraping methods and expensive checking equipment.

[j.t.d]

I was shown a smashing way to check a T square by a craftsman. Draw your line off a straight edge then turn the square over and draw another from the same base point. If the lines diverge, you know your square is out.
This is an example of fascinating questions that deserve a section of their own. Proofs of accuracy and precision can make great reading. As to creating your own square, you could use a 3,4,5 triangle, or use a compass to draw intersections above and below a base line.
I wonder how many ways there are?

[Roguish]

Thanks for the interest. The question is a serious one. I, like many others, would like to build a benchtop mill. The design portion is fairly straight forward. The controls are also fairly easy, as well as the motion mechanicals. I'm trying to 'design in' and 'fabricate in' accuracy.
So, where does one start? The X-Y plane? Blanchard ground flat? Then what? Machine the other components in a cnc, like a Haas or Fadal? won't they be only as good as the machine they are made on? or actually a little less good?
Measure against angle plates? They are not perfect.
I like the idea of the 'mirror-mating' even though it sounds like a lot of adjustment. Better design/build in the adjustability!!!

[1stcnc]

Good thread, The very thing I have been investigating on my new x2 and vice. And I'm finding out there's alot of truing to do. I hopefully will get access to a surface grinder so so I can go thru the hole machine from the base up. I've managed to true up the z column within .0003 and moved the head a couple inches with it repeating the same. What a difference it made in how the mill cut or cross hatch cut that is. It didn't seam that bad before though no cross hatch pattern, till I put an indicator on it and found out it was .030 out front to back can't have that. Even with the head now within .0003 I still suspect there are some errors in the table and slide. When I'm done I hope to have it within .001 in movement and repeatability. That is of course with no load? LOL Being it doesn't take much to move it out 2 or 3.... or 4 thou LOL. Not used to these small machines. I'm also looking at removing the z pivot and making it a solid 90* with only a few thou for tramming adjustment. It looks as if I might be able to get another inch out of y with my idea I hope I hope Let you all see when it gets done, though it might be awhile, being I might have a casting made to do it.
Keepen it square, Larry

[Mcgyver]

A very basic question for the machine builder. How square does a machine structure need to be?

the slightly smartass but germane answer is how square does what it cuts have to be?

as to the rest of your questions, they're kind of all encompassing of the metal working craft, ie how do you make any parts flat, square and parallel? Each part of whatever machine is just a part, you have know how to accurately make all of the surfaces and holes relative to each other, same as any other part....difference is there is a need for higher accuracy than usual as everything you make thereafter will inherit all the errors AND the parts might be a bit bigger than usual....but they're still parts. so to ask how do i make a bunch of parts with surfaces flat, parallel and square to a very high degree of accuracy is asking how do i become a master machinist

Fortunately, with out having massive grinders and planing mills, there are some techniques to work this larger parts into a high degree of accuracy in the home shop - such as scraping. for the insanely enthusiastic you could even generate from scratch your our reference flats and squares to use in getting parts flat parallel and square. deciding to scrape in a mill is not for the squeamish, but its doable. Thing is, if building from scratch and you're not willing to go to this degree what sort of machine do you end up with...yeah i know linear rails, but they still have to be bolted to something and the something needs to be flat, square and parallel, right?

Taking a large structure and keeping surfaces flat parallel and square is not trivial - to paraphrase the book title, its the foundation of accuracy...now add in that it has to remain more or less so under load and can see why manual mills cost 20k and weigh 2000 lbs. I think these challenges are grossly under estimated by a lot builders....imo that is because its one of the most difficult to deal with. if you're serious about addressing this important part of machine building, the Connelly book would be a good starting point.

[Roguish]

No offense taken. I myself am either a smartass or a cynic...or both.
Anyway, the referenced book sounds interesting. I may try to budget it in.
In the mean time, I am not into becoming a 'scraper' though it's probably an admirable skilled trade.
And yes, even linear rails must be fastened to something. Having been in the design side of linear motion, that's called 'as installed' straightness. No rail company guarantees any straightness once installed.
And yes, size and mass are good things. I have a BP series II in the shop and it's pushing 5000#. The little cuts I make with it don't affect it much at all. We like mass and stiffness!!!
But how are commercial machines made? How the heck does Haas (or any other maker) get their accuracy? Do they have a room full of 'scrapers?' Optical flats tracable to NIST?
I would be ecstatic if I can eventually get my 12"x12"x6" machine envelop to a true one mil accuracy, both statically and dynamically!!!! without spending a fortune.

[VWSatOz]

The linear rails do not have to be bolted to somrthing dead flat IF they can be finally potted in with something like WickIn Loctite. As long as any gap is not too big to allow cure between the fully supported rail-base and the rigid surface base plate or weldment etc. Once you get everthing set up accurate on shims or jacking screws, & proved all straight & square & accurate, THEN you set it all solid with the loctite (or injected epoxy if the gaps are huge), & when this potting is hardened you can tighten up all the bolts & it stays just as accurate! I have built big machinery this way & it works! You can then take heavy cuts now as all the mated parts are strongly bedded, ie held apart &screwed tightly together "potted in situ". There is no need to accurately machine two faces that are to be bolted together. It is similar to the methods used with polymer concrete machine bases & elements.
If you havn't got access to huge great surface grinders etc, just make parts a flat as you can can but strong & rigid, adjust your rails or slides until accuracy is achieved using light cuts & checking with the mirror method, & "Pot"!
The important feature of any machine is what I call "the specific rigidty"... ie how much does the tip of the tool deflect in relation to the job when a force is applied between them, using a spring balance, lever & a dial indicator, and in various conditions & directions of XYZ etc. The deflection is the sum of all the deflections/springyness of every part of the machine. All things deflect when placed under a stress load, it's just a matter of how great or little the deflection will be, and what the effect of this distortion will have on the job we are trying to do, the cuts we are trying to make etc
It's a waste of time building a machine that is accurate if it not also rigid enough to do what you wanted it to! I see lots of people in these posts say "now Im going to build another machine with fully supported rails, instead of just round bars supported by their ends". Enough said.

[MonoNeuron]

If you want a VERY accurate square that is really cheap then head on over to your friendly mechanic shop or industrial earthmoving equipment shop and ask for a gudgeon pin (Piston pin) from one of their old motors. Some have very little wear on the bearing part that can't be measured with a micrometer but if they have a new one then it's better. they have holes through them that can be used for a clamp bolt to slide through to clamp them to the table. It really should have the bottom face recessed a few thou so that it only sits on the outside edges but we aren't looking for the micron sizes here, just a 90 deg surface.
2 clamped side by side and clocked up on the table can be useful for clamping jobs to to machine the top face square with the side. I have been using them for as long as I can remember and mine came from a 3406 Caterpillar engine. They beat the hell out of an angle plate for squareness and the price was about right. The engine we rebuilt had very little use so I swapped out 2 old gudgeons for the new ones.
I know this isn't strictly on topic but it has a foundation that may be useful.
Rich.