[Hirudin]

I think the title says most of my issue, but I'll explain further...

I'm trying to test the squareness of my machine. As of now I have not "messed with" the column; it's exactly as it was when I received it. As you can see in the video I set up a dial test indicator against a square positioned close to the table. I then raised the Z 12". ~3.5" up the square the indicator is reading ~6 thousandths off. To me this seems pretty bad. After that it stays within ~1 thousandth (within a thousandth of the reading at ~3.5" above the table which is between 6 and 7 thousandths of the reading just above the table). I hope this makes sense. If not, I think the video does a pretty good job showing what is happening.

Is 6 thousandths over less than 4" acceptable? Any advice on how to fix this?

I'm assuming it's not the square. It's brand new and is suppose to be accurate to within ~2 ten-thousandths at the top.

I'm a novice so if you see something in the video that I'm doing wrong please tell me.

[nomedia="http://www.youtube.com/watch?v=E3CNdoNdu2k"]YouTube - Novakon NM-200s2 XZ plane squareness test[/nomedia]

[mactec54]

Hirudin

I do not think that type of square will be that good, you really need a granite square, or a ground cylindrical square, to do that kind of testing

Try Traming the Head to the Table first & then do your test

[Hirudin]

The head is trammed in that direction, but as I understand it the tram shouldn't matter for this test.

Anyone know a place to get a good cylindrical square?

[cteufert]

Is it possible that what you saw was looseness at the head? maybe combined with lack of lube in one place cause the head to drag on one side. Perhaps once it started moving up it pulled to one side and stayed there. One quick check would be to repeat the test in the opposite direction (downward top to bottom) and see if the error is consistently in the same place. I'd also try twisting the head to check for too much slop in the Z way. If the way is bent it would explain what you are seeing. To check you could tram with Z close to the table then move up and use a spacer say 5" tall to check again. (use the same spacer and move it from side to side.)

Charlie.

[mactec54]

Hirudin

That is correct the test you were doing, the head does not need to be Tramed, I thought that machine had linear rails on the Z axes

[Hirudin]

Charlie
Thanks for the insight! I did actually do the test in both directions and the measurements appear to be the same while moving down as well. I also gave the machine several (too many) pumps on the single-shot oiling system shortly before doing this test I would be surprised if one area is less lubricated than the rest but I'll inspect it as well as I can to make sure.
Thanks for the suggestion about tramming at two different heights. It took me a couple rereads but I now understand what you're saying. Seems like a good idea and I will try it soon. I'm still unclear about what you mean when you say "use the same spacer and move it side to side" though.

I'll see if I can detect any movement in the Z by trying to twist the head.

Mactec54
This machine, the NM-200 series 2, does have linear ways for the Z axis.

About the square. I have thought about it and I can't figure out why this kind of square wouldn't work. It should be accurate to 2-tenths of a thousandth (0.0001 every 6"). As I said, I am a newbie - can you explain why a cylindrical square is preferable for this test?

[JoeBean]

OK, so I've done this myself and here's what I can say:
1. The type of ways on the Z don't matter when it comes to tramming for this. Tramming is unnecessary because the head can be at 90 it doesn't matter, you're not comparing left side to right side you're comparing travel in Z relative to the XY plane. (I know it seems counter-intuitive, at least it did to me at first. But think about it for a while and it starts to make sense, or your head explodes - either way, problem solved).

2. The type of square you're using is, indeed, not really suitable for this test. Here's a good link to explain the proper way of doing this (note, too, that you need to repeat this test at 90 degrees as well, in other words with the DTI facing forward - more about that below). The reasons why you need either a cylindrical square or (my preferred option) a granite square - note that both are a bit expensive - are probably numerous. These "straight edge" squares are often not truly straight for the whole length of their arm. Instead they bow, typically inwards. They may be spec'd for X variation over Y length, but that can be tip-to-tip square, not along it's length. Plus, of course, many Chinese manufacturers aren't honest about the specs. When I run this against a good granite square I now, after some adjustment, get .0003 one way, .0002 the other over 8" (see here). On my squares like yours I get a "dip" in the centre of several thou.

3. Anything, and I mean anything, under the arm of the square on the table will screw up readings, so be aware of that (though not the dip you're getting)

4. Take readings on different parts of the table as the table is not perfectly flat.

5. I needed in put a shim in the column and carefully move it around a little bit to get everything perfectly (or as near as req'd) square. The 2nd link I mentioned above talks about that. There are other ways of shimming, but this worked for me. One shim, in the right location, corrected both "lean" (left/right) and "nod"(front/back).

6. Make sure you're using a good DTI, make sure it's secured well in the spindle, and if possible lock the spindle somehow. I rigged up a clamp system at the drawbar with my trusty visegrips which had the advantage of being able to turn it in 90 degree increments.

7. Think about what's happening during travel to locate the cause of the error & correct it. If you think about it a bit your intuition will probably lead you in the right direction. Remember to measure lean & nod each time before a correction. Otherwise you could fix one and exacerbate the other.

8. Don't expect miracles. Set a goal of what you realistically need and try to hit that. I fluked upon the right setting on mine to get the accuracy that high. I probably would have been happy with .001 over 8"

Hope that helps!

[Hirudin]

Thanks for the reply as well JoeBean!

The square I'm using is a Starrett 50136 (American made). I bought it from Amazon so I highly doubt it's counterfeit or something like that. Also since it's brand new and was shipped in literally five layers of cardboard (including the Amazon box) I figure the chance that it was damaged in shipping is slim to none. If I can't trust that it's square I wonder if I can trust anything. The DTI is a Mitutoyo 513-504T (also new from Amazon) and is graduated in tenths-of-thousandths - an entire rotation of the dial is 0.010. I think I can trust the square and DTI to be accurate to a much greater degree than I can expect to achieve.

Reading the internet some it looks like people don't think Mitutoyo DTIs are very good. Maybe I'll throw my Shars DTI on there and see what it says.

I could be convinced that the shape of the square could be a slight concern, but at the moment I really can't see how. I cleaned off the table and the square (not that either was dirty in the first place) before doing the test. I'll test again (recording the readings every 0.5") and then turn the square around and test on the left side of the spindle. If I get different readings I'll take a hard look at the square and indicator. I'll also try different parts of the table as you suggested. Once I start shimming I'll surely alternate between the XZ plane and the YZ plane.

I've read that page on CNC Cookbook you linked to. Thanks for it anyway though; I'll read through it again. I can't stress enough that I am a beginner so maybe now that I have a tiny bit of experience the importance of one of the tips on that page will become more apparent. Although (and maybe because) I'm a beginner I still have to know *why* some things are the way they are. If a cylindrical square really is the only and best tool to use I'm perfectly willing to accept that fact, but I would really like to know why they're better suited for this job.

In case this might matter: the table is also pretty much perfectly level, or it's as close as I can get it. When I put my electronic level on the table it says it's less than 0.03º off.

At the moment the spindle isn't secure. I will take your advice and clamp it down somehow. I don't think it's moving, but I guess it certainly could be. Though I only recorded one, I moved the head up and down several times and got what appeared to be the same readings every time.

If I could get under a thousandth that would be great! I'll settle for anything less than 2 thousandths without complaining too much.

[cteufert]

My suggestion to use the same spacer was to eliminate any chance there was a height difference between to different spacers. You don't need anything super precision just easy to position when you tram from one side to the other. You could also mount your indicator to an extension to bring it closer to the table when checking the tram a few inches up.

Charlie.

Charlie
Thanks for the suggestion about tramming at two different heights. It took me a couple rereads but I now understand what you're saying. Seems like a good idea and I will try it soon. I'm still unclear about what you mean when you say "use the same spacer and move it side to side" though.

[Hirudin]

My suggestion to use the same spacer was to eliminate any chance there was a height difference between to different spacers. You don't need anything super precision just easy to position when you tram from one side to the other. You could also mount your indicator to an extension to bring it closer to the table when checking the tram a few inches up.

Charlie.

Oh, I see now. Thanks again.

[JoeBean]

The square I'm using is a Starrett 50136 (American made). I bought it from Amazon so I highly doubt it's counterfeit or something like that. Also since it's brand new and was shipped in literally five layers of cardboard (including the Amazon box) I figure the chance that it was damaged in shipping is slim to none. If I can't trust that it's square I wonder if I can trust anything.

Starrett's quality has been going downhill for some time (not just them, my beloved B&S has been heading the same way). Not that that proves there's a problem with this square, but just search for Starrett over on PM and see what problems have been encountered. As well, a lot of their product is being made in China, but still stamped with their US location (ie. not "Made in America"). Not that that proves it's off either - from some comments I've read over the last 2 years some Chinese made stuff from them is more accurate than the American made stuff.

So again, I'm not saying that yours is a POS, I'm just saying don't trust it automatically. If you have a good known surface plate and the equipment needed to check it that can give you some more confidence in the tool.

The DTI is a Mitutoyo 513-504T (also new from Amazon) and is graduated in tenths-of-thousandths - an entire rotation of the dial is 0.010. I think I can trust the square and DTI to be accurate to a much greater degree than I can expect to achieve.

Mitutoyo are trouble to rebuild (apparently - I have never taken one apart, just going from what I've read) but they're accurate and stay so until they're used up. That doesn't mean they're not accurate. FWIW check out Long Island Indicator for info on brands and models when shopping - they rebuild instruments (more than just indicators), sell parts, and have a HUGE wealth of info on the differences between brands and models. I use them all the time when looking to buy a new instrument and they've never let me down. And if you buy from them their prices are quite reasonable and their service fast.

As for trusting it, nothing's a sure bet.

Reading the internet some it looks like people don't think Mitutoyo DTIs are very good. Maybe I'll throw my Shars DTI on there and see what it says.

See above.

[/QUOTE]I could be convinced that the shape of the square could be a slight concern, but at the moment I really can't see how. I cleaned off the table and the square (not that either was dirty in the first place) before doing the test. I'll test again (recording the readings every 0.5") and then turn the square around and test on the left side of the spindle. If I get different readings I'll take a hard look at the square and indicator. I'll also try different parts of the table as you suggested. Once I start shimming I'll surely alternate between the XZ plane and the YZ plane.[/QUOTE]
There's an easier way to verify if it's the measuring tool's fault: As I mentioned in the earlier post, do one to test to check lean (as you're doing) then another to check nod. If they both show the same dip the problem is with your measuring equipment. If it only dips on the lean then it's your machine.

As for shimming, you should only need one appropriately sized shim. Loosen the 6 bolts on the sides of the column riser & remove the 2 bolts inside the column facing up (accessible by removing the bottom rear cover plate). Use some means to tilt the head slightly (this is why the bolts on the sides are only loosened slightly, a few turns at best: It keeps alignment and prevents you from pulling the whole head over). Push the head to tip it and insert the shim - either on the left or right, depending on lean and either forwards or rearwards depending on nod. Of course in the center would give you no adjustment of the nod or lean. You should need a very small shim - I used a .0005 feeler gauge. Once you've got it in place lightly tighten the 6 bolts and run your tests again. Once they're good tighten the 6 bolts all the way and retest, and again retest after installing the 2 lower bolts.

I've read that page on CNC Cookbook you linked to. Thanks for it anyway though; I'll read through it again. I can't stress enough that I am a beginner so maybe now that I have a tiny bit of experience the importance of one of the tips on that page will become more apparent. Although (and maybe because) I'm a beginner I still have to know *why* some things are the way they are. If a cylindrical square really is the only and best tool to use I'm perfectly willing to accept that fact, but I would really like to know why they're better suited for this job.

Cylindrical squares are not the only tool for the job. As I mentioned above granite squares work great as well.

The thing that both of them have going for themselves is that they're very stable. The likelihood of them going out of tolerance over time, barring any damage being inflicted upon them, is very low. On the other hand, with standard squares you have the joint that can be out (though that wouldn't account for your numbers,as I mentioned earlier), you have a thin metal arm that can "settle" over time after having been ground, and you have an item that most companies seem to be manufacturing sloppily. Yes, a proven good square would work, but they're not easy to come by that way to begin with and often require reworking to get perfect. Cylinder squares and granite squares are less likely to be off - not saying they can't be though! And, of course, their shear weight makes it easier to test - no clamping required, their weight will keep them from moving.

But you're correct, assuming the square is perfect in all dimensions there's no reason it can't be used for the test and give accurate results.

At the moment the spindle isn't secure. I will take your advice and clamp it down somehow. I don't think it's moving, but I guess it certainly could be. Though I only recorded one, I moved the head up and down several times and got what appeared to be the same readings every time.

This is mostly a convenience to prevent it from turning and popping off the edge. In your case, one benefit of using a thin edge is you'll know right away if that happens. On a cylindrical square it can be more subtle.

If I could get under a thousandth that would be great! I'll settle for anything less than 2 thousandths without complaining too much.

Try doing as I mentioned above - run the test again 90 degrees out, ie in the YZ plane (you already know what it will say in the XZ). If it gives the same, or comparable, dip then you know there's something wrong with your equipment, likely your square. If it gives no (or little) error you know there's something wrong with the mill. BLAMMO! Answer found!

On a related note, if you're not already familiar with the process, if you have a good surface plate & a surface gage you can check the square for arm straightness & squareness.

[JoeBean]

Also, how far down on the Z rails are you when you're starting this travel? I'm wondering if it's happening only at the very bottom of Z's travel, which would obviously be less serious.

Your actual error, btw, looks to be about 5 thou/6thou when compensating from cosine error (at least from the best measurement of the angle that I could make).

Finally, 2 other thoughts come to mind: What happens if you turn the square around and run indicator up the other side of the square (obviously you'd be starting with the indicator slightly higher, but you have the room with your setup)? And secondly, if you have decent 123 blocks that are spec'd for squareness have you tried running up one of their long sides to see if that also gives you the error? It should give you nearly the same over the 3".

[Hirudin]

Phew, thanks again for the advice! There's too much to address specifically, all is appreciated.

I've ordered a cylindrical square. It's true, every time I look at these machinists squares I think to myself "hmmm... I wonder if this joint has come loose".

I don't have a surface plate yet (not one I can trust) and I don't even know what a surface gage is yet (I'll research this... eventually).

Good idea to use 123 blocks (since most of my error is in the first 3.5 inches).

I still haven't made it out to my garage to do all the checks I should. It's just been too cold in there the last few days.

[kvom]

A quite accurate cylindrical square can be made easily on a decent lathe.

[JoeBean]

A quite accurate cylindrical square can be made easily on a decent lathe.

+1

And a lot of good articles/writeups on the process available online, with info on calculations of accuracy and everything.

[Hirudin]

OK I finally tested it again. This time with a cylindrical square.

Just to get this out of the way: the results weren't significantly different.

Sorry I didn't label the piece of paper (image 1) very well. The numbers are in thousandths so "2.0" is 0.0020". The top most measurement was at Z11.9. The one below that was at Z11.5. It continues down the Z axis at 0.5" steps until it reaches Z0, which is the bottom number on the list. The left column shows the measurements taken when the DTI was on the left side of the spindle. The "L"s and "R"s denote which direction the DTI needle was pointing (the "0" on the dial was at "6 o'clock" - image 3).

I had the spindle clamped down using a long bar-clamp which I think did a pretty good job because after finishing with the left side I returned the Z axis to its starting position and the measurement looked exactly the same as how it started (at least within the smallest fraction of a tenthousandth).

This time I measured with the DTI on both the left and right sides of the spindle. I'm pretty sure each individual measurement at every 0.5" increment was within 0.00015" of the equivalent step on the opposite side. Despite that, the left had a discrepancy of of 0.0055" top to bottom while the right was 0.0064". I now realize that I really should have started the measurements at the same point on the DTI. It's interesting to see that moving from Z0 to Z0.5 produced greater than a thousandth of error.

If it matters I did both sides without moving the table - I picked-up the cylindrical square and placed it on the other side of the table without turning the X or Y screws at all.

With the DTI at Z0 (its lowest measuring point - image 2) I was able to grab the head and move it 2 thousandths pulling it with my arm. I can't imagine this is a good sign. I'll go through and tighten every bolt I can find on the Z.

Thanks again all. I hope this new information will be helpful.

[JoeBean]

Ugh. Doesn't look good...

I did up a quick spreadsheet with a graph showing the data. I had to take a screencap to get the image to show so hopefully it's legible


First off, we had to correct for your right-facing test and left-facing test to make one direction of travel negative and one positive. That's why your numbers in the Left-Facing column have the sign reversed (positive is negative and vice versa).

I included your half tenth measurements even though technically the DTI is only accurate to 1 tenth. It won't make much difference though.

The difference shows you that your measurements from one side to the other are within about a thousandth of each other (ie. the line is nearly horizontal). So your measurements are fine for this purpose.

What's of primary import is the average between left-facing and right-facing measurements, which I left out of this to avoid crowding the image as it's fairly obvious where the average would be - right between the green and blue lines. The average gives us a better idea of the rate of change.

The easiest way of thinking about this is to turn chart 90 degrees and look at the average, with 0" on the bottom. What you are then seeing is how the head is shifting left to right during travel. I created a view like that here:


Since the line is fairly straight from 11.9-5" the rate of change is fairly even and the differences could be easily compensated for, if needed, by shimming. Essentially, as you can tell from the picture, if you added a small shim to the right side of the column (as viewed facing the mill from the front) you would create a nearly straight vertical line from 11.9-5", meaning the head would be travelling in a straight, vertical line. On the other hand, if you wanted to fix 5"-0 you'd need to put a much larger shim on the left side, and then, of course, 11.9-5" would be way off.

The obvious problem is at about the 5" mark. Something's causing the rate of change to go completely another way. I don't think tightening things down will change this, though with that much movement obviously something's amiss. You might want to check out my teardown thread as it shows all the bolts that could be loose, and I particularly remember the bolts holding the linear guides being loose.

But if these values were a matter of something shifting at the 5" point (for example, let's say your coolant hose snagged and the head was able to tilt slightly) we should see a jump at that point followed by a roughly straight line that would be parallel to the average from 11.9-5". In other words, all your values would simply jump to the left or right (looking at the second image), but the trend would remain the same. Just think of doing the test with the bolts for tram loosened, then hitting the head at 5".

That's not what we're seeing though. Instead, the rate of change is different and so the line representing the averages is on a completely different angle. The head is, in effect, following a curve with it's vertex at 5".

Unfortunately the only thing I can think of that would cause that would be if the rails were not straight, because anything else should cause an anomaly. You can take any part of the head unit and throw the value off, eg. by adjusting it out of tram or loosening the bolts connected to the guides, and if you redo all the measurements the values you measure shouldn't change, as the issue isn't in the shape of the head, as it were, but the geometry of it's travel. Again, if you can picture throwing the tram off and redoing the tests you can see that the travel of the whole head would still be the same.

The same should be true of binding on the rails. If you think about it, you could bind on one side and that should cause the whole head to tilt left or right a small amount. But then as you continue travel it would still follow the path of the rails. So you'd have an anomaly. At, say, 5" the head jams, and all the whole head twists, causing the values in your chart to shift to the right .001". It would then follow the same path, just offset to the right by .001".

Similarly, if it were binding on both sides off and on starting at the 5" mark you should see random anomalies as the head stutters back and forth. Instead we're seeing a fairly steady line.

So once you get the head off I'd check that the rails are straight and parallel to verify that this is the problem. Easier said than done! And fixing it's going to be a real mess because the mounting holes quite possibly won't be in a straight line. Hopefully there's enough room in the bolt holes of the linear rails to allow a little movement, so that even if you need to replace a rail/rails you could do it without repairing bolt holes.

[Hirudin]

Wow, that chart says a lot doesn't it?

Take off the head... this is turning into a nightmare. I also have ~3 thousandths of backlash on the X and Y (I haven't tried the Z yet).

Maybe a call to Khai is in order.

[JoeBean]

Wow, that chart says a lot doesn't it?

Take off the head... this is turning into a nightmare. I also have ~3 thousandths of backlash on the X and Y (I haven't tried the Z yet).

Maybe a call to Khai is in order.

It's no small job, but for some comfort removing the head isn't that hard. You just need something capable of holding the weight located above it. A shop crane works well, or I have eyebolts in beams above the mill that I use come-alongs on to lift it (and the table & large work) off with. Take it in pieces if you can.

The real issue is correcting the alignment.

[ihavenofish]

what the heck?

a bit much to read, so this might be already stated...

what direction is the 'bend" in travel? in the x or y?

if its in the X, take a straight edge and first check it on the table surface, then put it up against the side of one linear rail on the Z. you want to measure against the lower registration edge of the rail. if the rail isnt mounted straight, you should see that here. then check the other rail.

if this appears to be the issue, you may be able to loosen the bolts in the rails, then retighten one from the top down forcing it into alignment as you go. then repeat with the other side. do this with the saddle on, but it may be easier with head off for less weight so manipulate.

if the issue is in theY axis direction, you can check the top of the rails with a straight edge, however this may not be a valid measrment as the top is not a precision surface. you may need to take the rails off and check the column surface. if the column is an issue, you can likely shim it in spots to straighten the rail. again mount from the top and force the alignment on the way down.

the rails themselves have a tolerance of about .002" per foot for straightness, however they are "flexible" and are meant to conform to the mounting surface.

this is a really bizarre issue though. being out of alignment is common on all machines, but being effectively bent is very very strange.