Lathe Levelling - how level?

May seem a daft question, but how much discrepancy in levelling the bed is too much?
I'm aware it's bed twist that really matters, rather than end-to-end or
exact side-to-side levelling, but how much twist?
I've found that my grandad's WW1 clinometer (used for setting up the sights on Vickers .303 machine guns - he was the final inspector in one of the factories!) is accurate to 1 minute of angle , which I think is an inch in about 95 yards, and can be set +/- 15 degrees consistently- accurate enough, too accurate?
Thanks all,
Dave H.
--
(The engineer formerly known as Homeless)

"Rules are for the obedience of fools, and the guidance of wise men" -
  Click to see the full signature.
Add pictures here
✖
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
On Nov 5, 4:53 pm, "Dave H."

There were articles about making a machine level in Model Engineers Workshop a few years ago, but if I remember right a gun clinometer was considered the thing to get.
John H
Add pictures here
✖
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
On Fri, 5 Nov 2010 16:53:28 -0000, "Dave H."

Hello Dave, You want it as good as you can sensibly get. 1 minute is certainly not too accurate, it's almost exactly 1:3600 or an inch in 100 yds as you say, but more usefully a thou in 3.6".
A decent lathe should have twist of less than that, say your tool height is 4.5", then one minute of twist would move the tool postiion by a bit about 00125 or if you'd rather, put 0025 taper on the work. Assuming no other errors etc. obviously.
Richard
Add pictures here
✖
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
"Richard Shute" wrote...

OK, thanks Richard - so it'd probably be best to level it as accurately as poss' with the clinometer, then once running take a few test cuts (assuming I can get the tailstock in line...) to check for any taper and adjust the feet to suit?
The clino's *calibrated* in m.o.a., but I can probably estimate down to about 10-15 seconds in a pinch, so that should get me within a thou" of taper (centre height's approx' 6-3/8" above the ways, slightly less than half the swing) if I follow the calculation correctly? I hope the 2 Imperial Tons of Old Iron should be pretty rigid to start with, particularly with an (almost) integral cast-iron box cabinet to sit on...
Dave H.
--
(The engineer formerly known as Homeless)

"Rules are for the obedience of fools, and the guidance of wise men" -
  Click to see the full signature.
Add pictures here
✖
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Dave H. wrote:

My understanding is that test cuts for bed twist are done without the use of the tailstock. Once the degree of taper is acceptable, a test bar is fitted between centres and the tailstock adjusted so that the bar can be clocked with minimal deviation along the length of the bed.
Both these methods do need the saddle and bed to be in a good state of adjustment.
Bob
Add pictures here
✖
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
On Mon, 8 Nov 2010 02:24:14 -0000, "Dave H."

. The use of a precision level simply a method of measuring bed twist. It really doesn't matter whether the lathe is level or not provided the bed isn't twisted.
Most texts describe placing the level directly on the bed ways and moving it along the ways to measure the level at all points. More convenient and, slightly more accurate is to mount the level on the crossslide and simply observe the change in bubble position as the carriage is traversed end to end.
For aligning the tailstock I prefer to use a lever type dial gauge mounted by a 3 or 4 jaw chuck with the measurement tip bearing on the INSIDE of the tailstock bore.
Because the dial gauge is rotated by the chuck it uniquely defines the headstock rotation axis.
Because the tip bears directly on the tailstock bore it eliminates errors relating to the use of tailstock centres.
It also shows up directly the shifts resulting from clamping the headstock to the bed or the barrel to the body. It is also easy to check the change in alignment at different extensions of the tailstock barrel .
Jim
Add pictures here
✖
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

Add another tool to the Xmas list then - a lever DTI! Although I have a decent (half-thou") plunger indicator or three, and I've seen a few turn-it-to-a-lever adaptors in the DIY tools mag's / web pages, may have the urge to make one!
Without adjusting / levelling, I ran the carriage from head to tail and back a few times: once I had the Clino' set I couldn't see much bubble movement if any at all, so I assume I'm free of twist, although it slopes about 30 m.o.a transversely and about a degree and a half head-down at the moment...
Re the "ground test bar" that's been mentioned, I have a number of print-head rails (about 18" long. 3/8" diameter), I suppose these would be adequate (they're uniform to 0.0001" on diameter and at least as straight as my surface plate's flat) if a little short?
Thanks again, Dave H.
--
(The engineer formerly known as Homeless)

"Rules are for the obedience of fools, and the guidance of wise men" -
  Click to see the full signature.
Add pictures here
✖
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
On Nov 8, 1:37 pm, "Dave H."

The problem is chucking them so they run true enough, though Rollie's Dad's Method works if the runout is within the short range of an 0.0001" indicator.
I prefer a test bar (or cylinder square) with a center hole running in the dead center, since it represents working conditions more closely and you can use it to recenter the tailstock after turning a taper.
jsw
Add pictures here
✖
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
On Mon, 8 Nov 2010 18:37:51 -0000, "Dave H."

Used properly, although it may be convenient, a test bar does not have to be ground, straight and constant diameter.
The essential requirement is just for the bar surface to be truly circular at the near and far measurement locations. At these locations the headstock rotation axis is uniquely defined by the average {half (max reading minus min reading)} clock reading and is independent of both local bar diameter and eccentricity.
A few nicely finished turned locations on a mild steel or light alloy bar is all that is needed
Jim
Add pictures here
✖
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
snipped-for-privacy@yahoo.com writes

But surely Jim, though that may be a good test for concentricity of spindle rotation with spindle nose fitting, it would do nothing to test whether the spindle axis was parallel with the bed.
What the non-ground bar you describe *would* be useful for is to make a test cut; if the "near and far" positions are given a light cut, and come out at exactly the same diameter, that does show that spindle and bed are parallel. In fact, you could say that is the most revealing test of all, as that is mostly what you want from a lathe.
David
--
David Littlewood

Add pictures here
✖
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
On Tue, 9 Nov 2010 12:13:31 +0000, David Littlewood

A bar of the type described may be secured in any convenient 3 or 4 jaw chuck.
The chucking errors simply add to any errors in diameter and eccentricity which the bar already possesses at the truly circular machined measurement locations. The sum of the errors is precisely cancelled out by the average clock reading indication method.
It's important to understand that the clock must be mounted horizontal and rigidly fixed to the lathe carriage so that there can be no chance of the clock zero shifting as the carriage is traversed between the measurement locations.
With a perfect bar slowly rotating in a perfect lathe the clock will remain at its abitrary reading (conveniently zero) as the carriage is traversed between measurement locations and this defines the spindle axis to be parallel to the bed in the horizontal plane. Vertical clock mounting would establish parallelism in the vertical plane.
In our imperfect bar there will be some resultant eccentricity at each of the measurement locations but. on truly circular measurement locations the average of the minimum and maximum clock readings still precisely defines the centre about which the bar is rotating.
This measurement result is the axis of spindle rotation and is not affected by errors in the spindle nose fitting.
Jim
Add pictures here
✖
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
On Nov 9, 5:50 pm, snipped-for-privacy@yahoo.com wrote:

Your description is better than the one I wrote and discarded.
If you mechanically average the readings by zeroing the indicator at the minimum reading, turning the spindle to obtain the max reading and then backing out the slide to halve that number the indicator should read zero plus or minus equal amounts when you check eccentricity at any position along the ways. This is also a quick way to center round work in a 4-jaw.
The low-reading zero may have to be one or two pointer revolutions in from its stop.
The plunger spring causes an offset error if the rod is too flexible.
jsw
Add pictures here
✖
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
snipped-for-privacy@yahoo.com writes

That's fine, if the bar has a constant diameter (or at least the two test regions are the same diameter). The post I replied to said it didn't need to be.
David
--
David Littlewood

Add pictures here
✖
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
On Wed, 10 Nov 2010 00:43:39 +0000, David Littlewood

The bar does NOT need to be the same diameter at each of the test regions.
In our imperfect bar there will be some eccentricity and diameter difference at each of the measurement locations. On truly circular measurement locations the average of the minimum and maximum clock readings still precisely defines the centre about which the bar is rotating.
Jim
Add pictures here
✖
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
snipped-for-privacy@yahoo.com writes

Then how does your "test" distinguish between (a) a perfect lathe with a bar of identical test areas, and (b) a lathe with say 10 thou misalignment of the bed between the test areas and a bar with a 10 thou difference in diameter between the two test areas?
David
--
David Littlewood

Add pictures here
✖
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
On Wed, 10 Nov 2010 10:40:28 +0000, David Littlewood

.The average reading always relates to the true axis of rotation - the actual local diameter and eccentricity do NOT change this result.
The basic premise of this method is that, if any truly circular object rotates about some arbitrary centre line, this centre line line is necessarily located equi distant from the points of minimum and maximum eccentricity and is independent of both the diameter of the object and the degree of eccentricity
I really don't know how to explain this more clearly - perhaps interested onlookers can suggest a more easily understood version
Jim
Add pictures here
✖
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
On Wed, 10 Nov 2010 21:48:26 +0000, snipped-for-privacy@yahoo.com wrote:

Sorry to reply to my own post but I need to don a hair shirt!
The method truly corrects for eccentricity but, because the measurement point is located at a point on average half the diameter away from the axis of rotation, it is NOT independent of diameter.
The basic premise ........... statement is correct but, because the measurement is made at a point (on average) half a diameter away from the axis, half the diameter is added to the clock reading. This means that, if the diameters differ, the clock reading should be corrected by half the difference in diameter.
Sorry about the confusion
Jim
Add pictures here
✖
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
snipped-for-privacy@yahoo.com wrote:

You also have to take the diameter of the bar at the point of measurement into account. Suppose a bar 1 inch across but perfectly round and perfectly centered, the center is 1/2" from the measurements (both are the same in this case, and equal the average). If the bar is 2 inches across at another point, but still perfectly round and perfectly centered, the center is 1" from the measurements.
For Rollie's Dad's method add the two measurements and the diameter[8] of the bar - this total is twice the distance to the center.
[8] measured across the maximum/minimum plane at the point where the other measurements are taken - you should measure the diameter for each set of measurements, unless the bar is known to be parallel enough that you don't have to.
-- Peter Fairbrother
Add pictures here
✖
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
On Nov 10, 5:02 am, snipped-for-privacy@yahoo.com wrote:

Think about a conical test bar running nearly true. To find the axis of rotation you have to add the diameter to the low reading, turning it into the high reading on the far side. Then the average of the two is the central axis. If the diameters are the same at both ends they cancel out since you really looking for a constant offset (the bar's radius) from the axis.
jsw
Add pictures here
✖
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

Oops, my antecedent is ambiguous. <slaps wrist>
The average of the near and far side high readings is the distance to the central axis.
jsw
Add pictures here
✖
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

Polytechforum.com is a website by engineers for engineers. It is not affiliated with any of manufacturers or vendors discussed here. All logos and trade names are the property of their respective owners.