DoN. Nichols wrote:
On 2008-04-06, Bill Schwab wrote:
DoN. Nichols wrote:
Well ... the sensors on the mill are not being confused by
backlash in the leadscrew and nut, which the dials progressively are.
I respectfully disagree: the dials are tricked by uneven wear in the
leadscrew (and the DRO is not). Backlash confuses the _operator_ if
same is not paying attention.

The sensors in the mill are part of the DRO, not part of the
leadscrew/nut/dials assembly.

I know that.


Backlash results in positional errors both depending on the
direction of approach,

But you cannot call mis-reading the dials relative to a good screw an
error of the machine; it is operator error. I know Harold gets a lot of
grief, and I am starting to feel it too, but I read, I thought, I
practiced, and I came to understand it. Backlash does not cause a
positional error in reverse, period; improper interpretation of the
dials causes the error. All one has to do is retreat and go back
forward, or work against the dials based on measurements, or re-zero in
the direction (though I have yet to do that).


and in absolute errors when the leadscrew is not
able to move the table to the position which the dial numbers would
suggest.

Agreed. That is part of what I said above.


The cross-slide leadscrew on my Clausing when I first received
it was badly worn -- enough so that it looked like this in the middle
(use a fixed pitch font like Courier to avoid distortion of the ASCII
drawings.):

I freely admit that a DRO can compensate for this, and said so above.
Personally, I would rebuild the machine to have fresh new screws, but
you are correct.




/\__/\__/\__/\__/\__/\__/\__

Instead of like this as it should
_ _ _ _ _ _ _
/ \_/ \_/ \_/ \_/ \_/ \_/ \_

The total backlash was 0.070" (out of a total of 0.100" expected motion
for a full rotation of the leadscrew. Since the leadscrew was only this
worn in the middle, you could expect an error of about 0.035" in the
middle if it was zeroed at the unworn sections near the end. And in
reality, since the majority of the wear was during power crossfeed to
part off workpieces -- most other work was done by a bed turret, the
error on feed was probably significantly greater than on the return as
would be used when boring the ID of a workpiece.

But a DRO would be measuring the actual position of the
cross-slide, not what the leadscrew dial thought it was commanding, so
it would be free of the error and the lathe would be better used with
the DRO readings and with a blank dial until the leadscrew was replaced.
(It has long ago been replaced, of course.)

We are saying the same thing, except that I do not consider the
dead-space in the dial readings to be a failure of the machine; being
tricked by it is a failure of the operator.



CAD drawings meet my needs, but they could certainly be simpler if they
needed to show dimensions only for the first of such holes. As it is, I
end up creating extra dimension layers for my more complicated parts.
Two layers are fairly common, usually split among milling and drilling
operations. I think three is my record to date.

O.K. And these depend on the accuracy of the leadscrews in the
absence of a DRO.

Very true.


Note that backlash in a milling machine's worn leadscrew can
also result in serious damage to the workpiece and/or the cutter. If
you are climb milling, the cutter can pull out the backlash resulting in
too strong a chip load on the cutter.

True again, which is why I think the manual process is a good idea: one
ends up with the slack removed, limiting the opportunity for what you
describe. Any pilots among us might think of it being analogous to a
procedure turn. However, the amount of backlash is not terribly
relevant to the DRO/manual debate, as long as it is consistent. When
the screw wear varies greatly over the length, the DRO wins - I do not
deny that.

Bill