Shimming an Atlas lathe bed

[JPMacG]

Has anyone else gone down the rabbit hole of shimming their Atlas bed for a constant diameter cut? I shimmed my bed to be as level as I could get it in X and Y. Then I took a test cut (not using the tailstock) and found that I was cutting a taper of about .005 diameter change over a 4.5 inch length. So I ignored level and experimentally shimmed the bed for a constant diameter. Now the diameter variation is about .001 over 4.5 inches but the lathe bed is far from level.

 

[graham-xrf]

Hi Jon
The thing about leveling took a while before I fully unpicked why we do it, how good it is, and when is it appropriate. There is a dividing line between lathes that have to be leveled simply because they have parts which are so massive that gravity alone will distort the lay of their ways, introducing twist, and bend, and other mis-alignments, and those smaller lathes that can keep their shape when fixed down to a solid surface or frame. Use a level with both kinds. A big lathe may have six or eight support points, and getting everything right is a major pain!

In both cases, the planet gives us a free, and easy to access, insanely accurate reference line to the horizon. We use the sensitive precision levels as a way of measuring the surfaces on our machine, to find out everything about the ways and surfaces, which is a different reason than ensuring the self-mass will not droop into distortion. The lighter machines could, in theory, be bolted onto any slope, even something not level! Lathes installed on ships have to use a ship's reference line when being installed. In effect, the poor guy doing it has to deliberately always use an offset bubble, or more conveniently, put his level on a ship reference surface, and adjust it to tell the right kind of lies.

The trick still works, even if the local gravity field is pulled aside a bit by nearby terrain. e.g. All buildings in Cape Town are built with their verticals not pointing to the Earth's centre, because the Table Mountain mass pulls all plumb bobs a little to the side. Lathes set up there with levels do not suffer at all !

Distortion over time?
I am told it happens. However many support points are under anything, only three count, and the rest are kinematic redundant. The only way the redundant supports share any load is if they are adjusted up to take up the gap, and start to share force without causing any other support point to develop a gap. The only way this happens is if the entire machine structure is compliant enough to let it do that, in effect making two supports act as one as if placed in a line somewhere between them.

Cast iron, semi-steel, and generally the stuff our machines are made of, if subject to a "twisting" or "bending" force, will slowly distort to that shape, and if the force is removed, can "relax" back, though often not all the way. This is why a lathe with twist in the ways, fixed by "leveling", can be taken back to the state it was in when made in the factory, though one may have to keep checking it in stages, over weeks, until it gets there. Smaller lathes can have ways so twisted, they may always have to be forced by bolting to a stiffer frame or floor, but it is very uncommon to have one so messed up it needs to be re-ground. That happens if the whole machine is dropped, or suffers a major crash.

Non-constant diameters?
The classic reason for this is the tailstock is off to the side a bit. Less commonly, the ways have a twist "wind". Use the level to discover it, or use other tricks, like "winding sticks parallels". Non-constant diameters is the feature that one exploits to get the machine straight. The "dumb-bell" test bar you make yourself, and measure with a micrometer, making tiny cuts between each adjustment. There are procedures that allow one to get the machine figured out with such a thing, like one known as "Rollie's Dad's Method", but you don't have to go that way. I simply got a reasonably affordable (made in India) test bar of known accuracy as a fast way to bypass all that.

All the tricks, nuances, explanations, and methods are fully searchable in this forum. The dedicated scrapers here have posted widely what they have done to their machines, and also what they had to do over to recover what they messed up. The combination of accurate level and test bar, including the turned dumb-bell is unbeatable. Keep in mind that an over-sensitive level can be a pain. Easiest are those that sense about 4 tenths inch.

 

[graham-xrf]

@JPMacG

Hi Jon. Just getting back to your original dilemma, you adjusted under supports to get the lathe "level", and found it cutting a taper.
At this point, the taper might have been from tailstock misalignment alignment.

Regardless that possibility , you then ignored level setups, and forced the supports to to "improve" the taper, by so distorting the rest of the machine as to mostly, but not entirely, compensate for what may have been simple sideways displacement of the tailstock, or a quill side play under force, or other stuff. Do not bend on your machine "until the symptom goes away". Be very sure what the real cause is!

1. I suggest start again. No matter whether you use a level, or some other means to establish the ways alignment, be sure you have that right. By far the easiest way is to use a level, and measure across the ways all along X at about 10cm (4") intervals. Even before that, walk the level along the ways X-axis, and establish whether the top surface of the ways is still unworn (it should be), and so get to know if the ways are are bowed or not. Keep notes, or write on the machine with a marker.

2. Next is to be reasonably sure the headstock is pointing along the axis correctly, though you can assume this, and more easily establish that afterwards, if problems persist. By now, you will have discovered twist, if any, and maybe even set it up "level", and untwisted it.

3. Next is to know absolutely whether the tailstock is set up right. Yes, you can do this with a dumb-bell test bar you make, but first put a dead centre in the tailstock, and also in the spindle taper (not a chuck). Chucks have their own separate alignment and run-out issues. Push the tailstock up to the headstock, and hold a small, thin thing, like a 6" rule between the points. You will quickly discover if something is too high, or low, or sideways. Shim the tailstock to be the right height, and move it's sideways adjusts until its dead on, and be sure to close up and lock the remaining gaps in the adjust screws.

4. Now get critical of the quill. Set up a dial test, first against the topside, and then the nearside of the quill, and wind it in and out, Try the tailstock at two or three places along the bed. Discover what the tailstock is doing. I do not expect any silly stuff would happen here. You should be OK. It goes without saying that when you do make test cuts, everything that can be tightened to lock up movement is done OK.

5. Test cuts. By now, you should have the bed level, as the way of discovering it is not twisted. You should have walked the level along X-axis. You might also have put an indicator on the saddle, and tested down onto an unworn reference, or onto a block placed across the ways, and tested along the length as best can be traveled. You can also do one at a time, testing onto a gauge block placed against a way's running surface. Get this far, and you will either have fixed the problem, or you can clearly see what it is.

6. Finally, unlikely, but I mention for completeness. Is the headstock aligned? Putting a known test bar, not too long, but at least 30cm, into the cleaned spindle taper, and testing along it's top and side with a DTI will tell you. It must at least point along the same direction as the saddle is headed along the ways. My South Bends do not even have any aspect of this adjustable. The headstock sits on the ways, and is never much disturbed from then on.

I don't really know any more on this. I am sure there are HM folk who have taken lots more pain than me in setting up. I think I just got lucky.

 

[JPMacG]

Thank you for the very helpful, comprehensive and well written response.

I should have been more detailed with my original post. I did not use the tailstock for my test cuts. I am using an 8" long very stiff test bar in the headstock chuck with no tailstock support. Turning between centers is not a problem.

The motor of my Atlas is hung (cantilevered?) from the bed casting. The bed casting is relatively light and not very stiff. The motor imparts a significant twisting force on the bed. The bed has three mounting points (holes) - two on the headstock end and one on the right end. The mounting points are not adequate to resist the twisting force of the cantilevered motor. In order to remove the twist in the bed from the motor I have to shim at multiple locations under the bed near the motor and use the mounting bolts and differential shimming to instill a twist force opposing the motor.

I learned this over several frustrating days using a Starrett 98-6 level. After a lot of creative shimming I did get the lathe very close to "level" in X and Y, testing at many points along and across the bed. I then tested my work, using only the headstock, and found that it was turning a taper. So then I experimentally reworked the shims and succeeded in reducing the taper to what I consider acceptable. But now the bed is not level - in fact it is so far off that the bubble on the Starrett rails to one side. So I don't know if there is twist or not.

I wonder if other Atlas owners have faced the same challenge and if so, how the dealt with it. I suspect the problem is unique to Atlas lathes - the method of mounting the motor and the relatively flexible bed are the driving features.

 

[great white]

Thank you for the very helpful, comprehensive and well written response.

I should have been more detailed with my original post. I did not use the tailstock for my test cuts. I am using an 8" long very stiff test bar in the headstock chuck with no tailstock support. Turning between centers is not a problem.

The motor of my Atlas is hung (cantilevered?) from the bed casting. The bed casting is relatively light and not very stiff. The motor imparts a significant twisting force on the bed. The bed has three mounting points (holes) - two on the headstock end and one on the right end. The mounting points are not adequate to resist the twisting force of the cantilevered motor. In order to remove the twist in the bed from the motor I have to shim at multiple locations under the bed near the motor and use the mounting bolts and differential shimming to instill a twist force opposing the motor.

I learned this over several frustrating days using a Starrett 98-6 level. After a lot of creative shimming I did get the lathe very close to "level" in X and Y, testing at many points along and across the bed. I then tested my work, using only the headstock, and found that it was turning a taper. So then I experimentally reworked the shims and succeeded in reducing the taper to what I consider acceptable. But now the bed is not level - in fact it is so far off that the bubble on the Starrett rails to one side. So I don't know if there is twist or not.

I wonder if other Atlas owners have faced the same challenge and if so, how the dealt with it. I suspect the problem is unique to Atlas lathes - the method of mounting the motor and the relatively flexible bed are the driving features.

Set my TH42 on the cabinet, leveled with a a starret 98-6 and no issues from that point forward. I use a tailstock when turning between centers/taper checks.

 

[JPMacG]

Here is the motor arrangement. Sorry about the poor lighting.

 

[JPMacG]

Yes, I get good results when turning between centers too. Interesting.

I was getting a slightly convex shape when facing a large diameter workpiece. That is how I fell down the rabbit hole.

 

[graham-xrf]

Here is the motor arrangement. Sorry about the poor lighting.

Please forgive that get sidetracked here, but what I get from your picture is..
.. is that a bookcase, and a Hi-Fi speaker, and carpets??

For a minute there, I thought you might be the kind of hero who sets up his machine in his lounge living room, or maybe has turned his shop into a multi-purpose living space, but then I see the wall blocks behind the motor, and the surface plate gauges near what looks like a W2AXZ QSL card!


The motor drive assembly may be hanging out in such a way that it does look like it would apply a twisting force to the bed, but here I run out of knowledge about what is OK for that lathe. Surely the Atlas was designed to be like that. Here is where we should maybe get the attention of Atlas owners, and ask for their input.

 

[pdentrem]

You are chasing partly the wear on the ways. When I had mine, I placed the level at the far end where the tail stock sits and has the least wear, and right up to the head stock. Those are the best places to get a reading to even out the bed. It does not have to be level just the same reading at both ends. Twist in the bed gives the varying sizes of course. Do a two collar test after doing the bed leveling.
My lathe sat on thick metal pads and those were on adjustable threads rods to allow movement. The table itself was just a heavy wooden table made of 2x4s and covered in aluminum sheet. Adjusting was required as the seasons passed during the year.
Pierre

 

[FOMOGO]

Much of what you're seeing is probably in the chuck it self. Mike