Bed warp

[rockyrude]

I have heard discussion regarding bed warp but I don't see any way to determine if a bed is warped

 

[Ulma Doctor]

a simple test is to mount a test bar in the spindle and sweep the bar longitudinally with a DTI mounted on the saddle
if there is large needle swings, you may just have some warpage

you could perform a 2 collar test on a piece of scrap, also indicates warpage

 

[Richard King 2]

This is a Clausing Lathe, but it is the same concept. Set a precision level on the bed and compare the head end to the tail end.

http://vintagemachinery.org/pubs/182/25340.pdf

look at minute 6

 

[benmychree]

Keith is using a surface for reference that was not ground originally and is not necessarily in alignment with the ground surfaces of the ways, this surface was planed, not ground. Also, he uses a Starrett #98 level, which is not sensitive enough for final leveling. A appropriate level would have sensitivity of .0005" per foot, not .005" per foot of the #98.

 

[sdelivery]

Keith is not looking to determine the accuracy of the ways but merely twist in the bed.
Comparative analysis vs measuring error.
He is not trying to put a number on the error but compare one end to the other.
Placing a straight edge diagonally across the bed ways will also show twist.

 

[benmychree]

Keith is not looking to determine the accuracy of the ways but merely twist in the bed.
Comparative analysis vs measuring error.
He is not trying to put a number on the error but compare one end to the other.
Placing a straight edge diagonally across the bed ways will also show twist.

I stand by my statement; Keith is using a reference surface that is not necessarily in close alignment with the way surfaces, so that in spite of the fact that he has comparable readings of transverse leveling at both ends of the bed, the way surfaces themselves are not necessarily in alignment with each other. A level with .005" per foot sensitivity is not sensitive for machine tool final leveling.

 

[jwmelvin]

Keith is using a surface for reference that was not ground originally and is not necessarily in alignment with the ground surfaces of the ways, this surface was planed, not ground.

So, better to use the two flat ways (one for the carriage and one for the tailstock)? Or better to put the level on the carriage and move it back and forth?

 

[Richard King 2]

I was there when Keith checked the machine. Those levels are .005" per graduation in 12", so if the bubble stays on the line it is within .001" or better. The device he made set on the flats that were ground and he was comparing one end to the other end. Heck yesterday someone showed me a picture of a carpenters level to test twist. I have checked the clearance surface between the ways and I would guess 1 out of 100 were not co=planer to the ways anyway. A Starrett 198 would better then a Starrett 98 - but it will get you close enough and then take test cuts and sweek the bed to cut straight.

 

[wa5cab]

On all beds for Atlas metal working lathes. whether the swing is 6", 9", 10" or 12", the entire top surface of the front and the rear ways is, before use, finish ground to all be in the same plane. However, neitherthe headstock, saddle or tailstock are supported by the entire width of either way. This is usually quite obvious on a used bed as usually you can clearly see where the saddle and the tailstock have been running. And a narrow strip running from headstock to near the right end of the bed shows where nothing has been running. So setting a precision level of adequate sensitivity across the bed and comparing the readings at both ends will give you the amount of twist in the bed for the unworn portion only. To indicate twist in the regions where the saddle actually runs, you really need to shim the level up more than the way wear front and rear (which is unlikely to be the same). However, as a practical matter, unless the bed is really badly worn, you are probably just trying to split hairs if you worry about that. The original accuracy of the lathe was only claimed to be +/- 0.001". Moving the decimal point one place to the right could be expected to move the decimal point in the dollar cost of the machine a like amount. To quote a well-known author of yester-year, TANSTAAFL.

 

[benmychree]

Yes, the #98 is good enough to get it close, then use the #198 for the final leveling, that is what I do.