PDA

View Full Version : Lathe test bars



loose nut
03-05-2017, 09:20 AM
I need to "level" (yes I know it doesn't actually need to be level) my lathe again. The preferred method is to put a bar in the chuck and turn the outboard end and then a bit near the chuck and adjust the feet until it turns parallel. The size of bar usually quoted is 1" in diameter but it is also suppose to be extended about 1' from the chuck without any tailstock support. With that amount of overhang it seems like there will be a lot of chatter and the bar will spring which makes turning a bitch.

Is this right or wrong and if so what is a better way and don't say get a "master precision level" and do it that way, my pockets don't run that deep.

Sun God
03-05-2017, 10:04 AM
Supposed to be done with tailstock support. You gotta dial in your tailstock first (with the tailstock center against the headstock center, or as short as you can make it). This gives you the case where the tailstock is coaxial with the headstock at that point only. Then with the bar between centers (and the tailstock some significant percentage of the length of the bed away from the headstock), you turn your gauge rings. This gives you the general case to test - the difference in axial location between 'nil' gauge length between headstock and tailstock and 'large' gauge length is the twist of the bed.

The difference in diameter of the gauge rings should be caused by the offset of the tailstock from the headstock, caused by the twist of the bed; the 'handedness' of the difference tells you which direction the bed is twisted.

You then adjust the 'level' of the bed, recut your gauge rings, and measure. Repeat ad nauseum until the gauge rings turn the same diameter.

When the rings turn the same diameter, the 'twist' has been taken out of the bed, and the tailstock center should be coaxial with the headstock.

Recheck the initial case (tailstock center against headstock center); if it still holds, you're done. The tailstock can then be moved to any point along the bed and will remain in the same axial relationship to the headstock. The lathe is 'levelled'.

Edit: The first test is done between centers because it mostly eliminates errors in headstock and tailstock axis with the bed. You can then do that test again with a shorter bar, unsupported, to test how axial the spindle axis is with the new levelled bed. And around and around in circles you go until you've reached a level of axial parallelism you're happy with. I normally go for a thou per foot or better.

QSIMDO
03-05-2017, 10:37 AM
Keith Rucker just did his lathe; https://www.youtube.com/watch?v=5bWpgElMm8c

lakeside53
03-05-2017, 11:12 AM
The test needs to be done without the tail stock support; the tail stock comes in AFTER the head is aligned to the bed. Use an aluminum tube as big as you can measure and grab well, (weight sags), super light cuts with sharp tooling. Remember in the real world your chuck jaws can cause misalignment (ok, so once the part turned it goes away, but...); using the spindle MT as a cross check or even your initial check is a good idea. You can heat shrink an AL tube onto an MT stub.

Tons of long arguments over on PM about the best way to do this. If all you are concerned about is 1 foot from the chuck, you can fake it, but for that go for a few 10ths, not thou. If you want 60 inch... yikes.. not really practical with bar and mics. IMO... you need to level the lathe or you have no real reference. Level the lathe, then test cuts to confirm whether other items need adjustment. Adjusting the base to compensate for a misaligned tail or head stock is backwards. I have real levels, and it's amazing how much a 5000lb lathe bed follows the floor, and how the 8 jack screws can change things with the smallest adjustment.

I have an $80 level, and a $zillion level, the latter I got from an auction for $300. I use the $80 level most of the time to stay sane. The other stays pristine in its beautiful wood box, just as it did with the prior owner ;)

J Tiers
03-05-2017, 11:22 AM
... IMO... you need to level the lathe or you have no real reference. ....

+10

If the machine is not at least roughly aligned, you have no idea where you are starting. If you CAN align the head, and on most common US machines this can only be done by scraping, then you must level first to be sure the bed is straight, in order to have some consistent reference for alignment.

There are endless arguments about "RDM" as a substitute for a level.... but he level is needed to give an outside reference. It can be a simple medium level, like a Starrett 98, no need for a "master precision" as a prelude to removing twist with "two collars" test. In fact a sensitive level will drive you crazy if there is much error.

David Powell
03-05-2017, 02:31 PM
My " precision " level is a ww 2 inclinometer, presumably from a large gun. I was taught that once the bed is levelled as best as possible, to take a piece about 1" diameter, stick it out about 4 " and without tailstock support, light cuts and sharp, very slightly radiused hss tools see what the results are. If satisfactory then to move on to the tailstock alignment with longer pieces. I got within two tenths in 4" on my Colchester Bantam, and now my Southbend 9" gets within half a thous in 4" I seldom have to turn all along pieces longer than 3" or so nowadays, and I am my own inspector just building steam engines . Regards David Powell.

partsproduction
03-05-2017, 10:32 PM
The master precision level I bought 30 years ago on ebay (?) is graduated .0002" per 10", I think I paid $80 or something like that for it. It's been a good level, but looking at cheap master levels in ebay recently I didn't see any of them that were that sensitive, more like .0005" per 12". If I were setting up a lathe I'd find a way to get access to one. Beg borrow or rent.

cmantunes
03-05-2017, 10:59 PM
I leveled my lathe using this machinist level from Shars: http://www.shars.com/8-x-1-9-16-x-8-precision-frame-spirit-levels. After leveling, I machined a piece of steel and measured 8 microns per 300mm (roughly half a thou per 18"). I am now a fan of that machinist level, particularly considering the 80 bucks price tag.

BCRider
03-06-2017, 02:22 AM
I don't know where you heard that the bar needs to be only 1". It is better if it's up around 2" diameter for all the reasons you are worried about. The ONLY downside to this is the need for a 1-2" micrometer instead of a 0-1".

It's well worth hogging off .04 to 06 through the middle and leave just a collar at the end so you can traverse quickly to the collar by the chuck without touching the part between.

For just getting into the ballpark even a builder's level will be close enough that the lathe doesn't rock corner to corner.

loose nut
03-06-2017, 10:04 AM
Most of the old books and newer videos use a 1" bar. I can use 2" no problem.

pinstripe
03-06-2017, 10:34 AM
The master precision level I bought 30 years ago on ebay (?)...

Nope, eBay didn't exist 30 years ago. The World Wide Web wasn't even around. The technology is so pervasive that it feels like it's been around forever. I was discussing this with my wife today. I've got a computer science degree, but I never had an email address at university.

lakeside53
03-06-2017, 12:30 PM
Most of the old books and newer videos use a 1" bar. I can use 2" no problem.


At 12 inches there will be sag.. all depends on what "accuracy" you are aiming for. Try a 2 inch aluminum tube - less sag, and lighter tooling pressures "pushing" at the end point. As a test case, mount up something with 12 inches of overhand and measure deflection when pushing/pulling by hand. Tooling pressures are way higher.

J Tiers
03-06-2017, 12:54 PM
Sag will have nearly NO effect on a larger diameter workpiece, due to the geometry. So go large.

It may also have less sag to start with, of course.

BCRider
03-06-2017, 02:46 PM
Sag will have nearly NO effect on a larger diameter workpiece, due to the geometry. So go large.

It may also have less sag to start with, of course.

That's true. But a 1" bar sticking out a long way will certainly flex enough to make a big difference. That's why I used a large diameter round bar when I did my own test cuts. As pointed out above even with a large bar we are able to pull on a 12" long piece enough to easily show on a dial gauge. The trick is to reduce the flex that occurs to something below an amount that can be a source of significant error. And the easy way to do that is to make the diameter larger.

If using pipe for this you'll want to use a piece with a thicker than normal wall. Otherwise the jaws will distort the pipe. And even with a rounding cut the change in wall thickness can cause a release of the pressure and reverse clamping distortion. So if you have your druthers it should be solid round bar. And the larger you can go the better. Whateevr you have micrometers that can fit.

loose nut
03-06-2017, 06:56 PM
My mics go to 9" so 2 or 3" bar isn't a problem.

I was always told that unsupported projection of a bar was 4 x the diameter. so 4" for 1", 8" for 2" etc. I suppose 8" can be stretched to 12" for this.

J Tiers
03-06-2017, 07:11 PM
That's true. But a 1" bar sticking out a long way will certainly flex enough to make a big difference. That's why I used a large diameter round bar when I did my own test cuts. As pointed out above even with a large bar we are able to pull on a 12" long piece enough to easily show on a dial gauge. The trick is to reduce the flex that occurs to something below an amount that can be a source of significant error. And the easy way to do that is to make the diameter larger.

...

A 1" diameter bar or tube, deflected down (or up) by 0.02", which is a good bit, will account for a diameter change of about 4 tenths on the cut. You can do the math and see if that is correct. Most of the machines in question, especially Atlas and similar, will have a basic accuracy of between 5 tenths and 1 thou, and an error of a few tenths is not reliably measurable with a common 0.001" calibrated mic, so it is not much of an error.

I would not suppose most would stick a 1" bar out far enough for it to HAVE 0.02" of sag, so even that tiny error is not likely.

BCRider
03-06-2017, 07:32 PM
My mics go to 9" so 2 or 3" bar isn't a problem.

I was always told that unsupported projection of a bar was 4 x the diameter. so 4" for 1", 8" for 2" etc. I suppose 8" can be stretched to 12" for this.

For the sort of light skim cuts using a nice sharp HSS tool to remove about a thou or two extending a 2" round bar out to 12 " isn't going to be an issue.

The test for proper cutting is that after you do the pass if you reverse over it the cutter does not take any more material off. At at least not more than some light "dust" that sits on the end of the cutter's tip. If it does take off any significant material on the way back then it flexed the test bar away from the cut and the results are not valid. You want the cutter to remove the metal without any flexing of the bar away from the cut.

When I did my own setup last year I was able to achieve that without any issue. Just the usual shaping off an HSS cutter and a light skim cut.

BCRider
03-06-2017, 07:52 PM
A 1" diameter bar or tube, deflected down (or up) by 0.02", which is a good bit, will account for a diameter change of about 4 tenths on the cut. You can do the math and see if that is correct. Most of the machines in question, especially Atlas and similar, will have a basic accuracy of between 5 tenths and 1 thou, and an error of a few tenths is not reliably measurable with a common 0.001" calibrated mic, so it is not much of an error.

I would not suppose most would stick a 1" bar out far enough for it to HAVE 0.02" of sag, so even that tiny error is not likely.

Totally... A .02 flex in any direction on a 1" bar with even 8" of extension would indicate that some major forces were at work. And far more force then a decently sharp tool taking a skim cut of a couple of thou would ever generate.

Where does this "play"come from on the Atlas and similar lathes? Is it related to the models with plain headstock bearings? I'm asking because with my Asian lathe it was no big deal to get results that matched on the test bar easily and consistently to a tenth or two. I could make a cut, adjust the tension on the support and hold down nuts on the mounting studs and see a direct and consistent correlation during the next cut. It was easy to zero in and even a second cut to double check showed the same results to within my ability to read the 10's marks on the mic barrel.

I went into the whole procedure not really knowing what to expect. And needless to say I was highly impressed when I saw how things worked out. I guess I figured that if I was able to work to that sort of tolerance first time around without really going out of my way that it would be pretty much standard on most other lathes. But perhaps that isn't the case. I guess I can see a plain bearing headstock which needs a little bearing clearance being harder to get this level of consistency. But I still like the way the test bar method can be used to directly tune the machine to as good a degree as the machine is capable of reaching. And the fact that it can also be a nice direct way to test for not only removing any residual twist from the bed but also for correcting any arching of the bed and even indicating how far out the tail stock is from being centered to the head stock axis.

In fact while typing this out I realize now how I could also use a DTI or good plunger style indicator to check the tail stock to see if the ram axis is coaxial to the spindle axis or not.

J Tiers
03-06-2017, 09:09 PM
Any individual machine may do better than a thou, sure.

But I see no evidence that Atlas, or Southbend, for instance, were designed to hold consistent tolerances below 0.001" The spec, if you can find one, is usually not better than a half thou or so for spindle TIR, for instance. While that does not preclude holding tighter tolerances by itself, it is an indicator of the design intent.

I am certainly not suggesting that an Atlas etc WILL show 0,02" of slop, just pointing out that even THAT much SAG (not slop), which is relatively huge, will not change the diameter more than the basic capabilities of the lathe account for.

And, with all due respect, if we are going to get into holding to microns on an Atlas again, I want no part of the discussion. Resolution and accuracy to small numbers of microns is just not what they were made for (A micron is 0.001 mm, which would be about 0.4 tenth, so holding to a tenth is essentially holding to 2 microns). Start with the calibration marks on the dials, and the accuracy of the feedscrews..........then the uncontrolled temperatures in the shop, measuring tools and calibration of them............ Let's just not GO there, OK?*


* You don't even need the calibrations on the dials, I agree with that.... but you are getting into an area of advanced work that is not typical for folks here. And there is still the temperature issue.

loose nut
03-07-2017, 09:08 AM
A thou over 10" or there about would be good enough for me.

softtail
03-07-2017, 10:05 AM
Chuck up your bar and do the rough turning/creation of the rings by whatever means necessary.. use your tailstock if you like. Once done, remove tailstock, and take light passes on the rings.. if you use Al and take a light cuts with a sharp tool, deflection should be minimal.

BCRider
03-07-2017, 01:22 PM
A thou over 10" or there about would be good enough for me.

OK, this is a bit of blue sky thinking as I've only used the method the one time on the one lathe. I'm thinking that on a plain bearing head stock that there may be some allowance but I wonder if the more predominant effect of the bearing clearance might show up more as a slight chatter in the cut of the farthest out ring. Or maybe the moving oil film is strong enough to center the spindle with a fair degree of force much like it does in an engine? Or would the oil film deflect by some amount?

If there is a little deflection of this sort but without any actual chatter perhaps once the test bar is matched up the bed might actually have a small amount of twist held into it so that the cut itself is parallel? Now that might be interesting.

When I started out with the process I was thinking the same way. But as I made cuts, measured and tweaked it soon became apparent that I was going to be able to adjust it to a lot tighter difference than a thou/10". And it wasn't a fussy sort of work either. Now granted I'd set up the lathe so it sat on studs with nuts as the seating points below the bed and chip tray and then hold down nuts above. But by playing the tightening torque between the jacking and hold down nuts I was rapidly at around a thou over the 10 or so inch bar I was using. And it was only a tweak or two more to get it to where it was a small part of a thou. Working with shims might prove to be less ideal but even there I suspect that it would not be too hard to get the twist set to something under a thou over 10".

The ease that I was able to make the cuts and see this level of accuracy is why I'm such a fan of this method. It didn't take a lot of old world esoteric skill or even a pointed hat with stars and moons. Just a freshly sharpened free cutting tool bit, a good micrometer and a DTI with a magnetic stand. As I said above it was my first time truing any lathe up so I was all cautious and hesitant. But once I got into it and saw how predictable and responsive it was to adjustments that all went away.

partsproduction
03-07-2017, 06:43 PM
Nope, eBay didn't exist 30 years ago.

I was wondering about that actually. I know I got into it early, but later than many. I had that MPL back when I worked in the machine shop at a large lumber mill in 1985, the head millwright asked me to borrow it to set up some machine in the mill. I said fine, a long as I'm there (To make sure it wasn't abused) and he got all upset at me.
I must have bought it through MSC or perhaps Enco, it was not expensive. Back then MSC had a lot of real bargains, things changed when they bought Enco, I think. I bought a 16" Bison 3 jaw chuck for a pittance because it had a bit of rust, otherwise new, I think like $280!


Still use it often when restoring machine tools and occasionally when setting up work that needs to be done to close tolerances.

loose nut
03-08-2017, 07:31 PM
The net went commercial in1994 or there about's. I don't think that it was open to the public 30 years ago, just gov, universities and the US military.