【转帖】positional tolerance and datum precedence

yang686526

positional tolerance and datum precedence
suppose we have a cylinder, with an off-center slot extending right across one end face. suppose the surface of that end face is defined as datum a, and that the axis of the cylinder is defined as datum b by using a datum callout on the diameter. also, suppose one side of the off-center slot is defined as datum c, and that its radial location  from the centerline of the cylinder is specified with a toleranced dimension of some sort. now suppose that, using a positional tolerance, it is required to locate a circular set of six holes in the said end face, such that the hole pattern is centered on the axis of the cylinder (datum b) and is angularly aligned using the full length of the slot (ie datum c) in some specified manner.
if in the position tolerance frame we use the three datums a, b, c in that order, it would contradict the required 3 point/2 point/1 point rule used to define the datums in the ansi standard, since c requires two points because of the fact that we desire to use it to define a line for angular orientation. (this does not lead to ambiguity in most of the examples shown in the standard and other texts, because they always show very short slots which in conjunction with datum b can then provide a localized single point angular locator without ambiguity). if on the other hand we use the sequence a, c, b, as the standard seems to require, would the position tolerance then be referenced relative to a point at the intersection of datum c and a perpendicular lying on datum a and passing through the cylinder centerline (defined by datum b)? if so, that would of course not meet the original intent. or is it the case, as i have been told by certain others, that the position tolerance would be implied to be referenced to the b datum with c used to define only the orientation? i realize i am probably displaying my poor understanding of g d & t here. however, any comments ?
actually, i've just noticed on this recent
englishmuffin,
   your datum system is described explicitly in asme y14.5m-1994.  see figure 4-6.  datum a locates one face, z, at whichever three points stick down the furthest.  datum b is a feature of size which locates in x and y.  you should consider calling up b at mmc in your tolerance boxes.  datum c is a face that controls rotation.  
   the abc datum system immobilizes the part.
   your acb specification is different.  a locates the z face.  c locates the x face, and b becomes a feature of size that locates the y face.  again, you must consider an mmc specification when you call up datum b.  i think this is a weird way of doing things, but perfectly legal.  if the holes must be located accurately with respect to the slot, it may even be a good idea.
   acb immobilizes the part too, but in a different location.
                 jhg
thanks drawoh.
this question came up because of an argument between two people, and one of them casually asked my opinion. person #1 is the best expert we have in the plant and is gd & t certified at technician level (he operates our large cmm and consequently gets heavily into the subject on a day to day basis). person #2 is a representative of a large corporation for whom we are making the part in question. the actual part actually does not have an off center slot - it has two in-line short slots symmetrically disposed either side of the b axis, and which form a common datum c. but to frame the argument more clearly i am specifying the slot to be a single long one offset from the b axis. as i see it, the reasoning for both situations should be basically the same.
person #2 says it should be a, b, c. on the other hand, person #1 says (as i understand him):
1. datum a is located by 3 a minimum of 3 points
2. datum c is located by a minimum of 2 points
3. datum b defines two planes regardless of whether it is placed in the second or third box on the callout. i think he also is implying that he would then locate b with a single point (minimum) at the center of the part.
he therefore argues that c should precede b. if he is actually right about statement "3" then that would conflict with your statements about the acb sequence being different - the holes would aways be referenced to the b axis in two orthogonal directions.
i'm afraid i don't have a standard at my fingertips so i am unclear how much importance should be attached to how many points are needed and their exact relevence to datum precedence. but it would seem to me that once a is defined with 3 points, one can then go on to define b with 1 point, and one has now fully defined the location of 3 planes in translation, with two of them yet to be defined in rotation (although their relative perpendicularity to each other has already been covered). you could stop there if the hole pattern was not required to be oriented to anything, or if one of the holes was required to be a datum you could then orient the planes using one more point. so one could set everything up in that case with (3,1) points or (3,1,1) points, and in the original case with the c slot it would be(3,1,2) points. in this latter case, none of the planes go through "c". in the case of an a,c,b callout, one of the planes would go through c and the remaining plane would go through b and be perpendicular to c. it would appear to me that the 3,2,1 convention is only a guide, and does not always apply religiously, especially with round parts.
that's how i see it anyway, and i think it agrees pretty much with what you are saying. so with person #2, that makes three of us. but i guess it all comes down to whether person #1 is really correct in claiming that the "b" callout defines two planes intersecting the axis of the circular part, regardless of whether b appears in the second or third box, and whether the 3,2,1 point convention has to be slavishly adhered to. note that on the thread i referenced on my second post, the statement is made that  section 4.4.2 says the datum planes always intersect on the axis, which if correct would partially support person #1.
not being really familiar with the standard, i would be interested in your further comments and/or anyone elses.
i believe that a problem that you are encountering is at least in part due to using one side of the slot, as a datum feature, c.  the situation would seem to work better if only a point on the surface were designated as a datum feature.
it is difficult even on a computer to 'orient' a line to a surface.  by definition orientation requires a point.
that would serve to satisfy the orientation of the part more adequately and fit with your 3,2,1 point precedence considerations.
my opinion: a primary, b secondary, and c tertiary.
ringman:
i don't see how you can angularly orient the two planes through b, using the orientation of c, with only one point, regardless of whether c is defined as only one side of the slot or the slot feature itself. with the a,b,c designation, the third plane does not necessarily pass through c, (whether it is actually offset from the center or not), so you you would have to use a minimum of two points lying on c in order to define its angle, at least as far as i can see.
but at least so far we all seem to agree that it should be a,b,c.
orientation is accomplished by the theroetical line thru axis b and point c.  there of course would have to be an angular relationship specified for the mutually perpendicular planes that are associated with axis b.
voila!
ringman - c is derived from an off-center feature (the slot)- so as far as i can see, you cannot angularly orient the two axes through b using c, with only one point on c - you would have to use two points - both of them on c. drawoh seems to be in agreement with this, unless i misunderstand him. even if c were a short slot at the periphery, as is so often shown in examples, it would seem to me that this still implies that strictly speaking one should use two points on c, although that would not lead to very satisfactory results in practice. i would have thought that if, on the other hand, the intent were to orient the axes through b the way you describe (which would work for a short radial slot) then the callout should strictly be a,b,b-c. but i could well be wrong - i don't understand g d & t very well - hence my confusion.
orientation is accomplished by the theoretical line thru axis b and point c.  there of course would have to be an angular relationship specified for the mutually perpendicular planes that are associated with axis b.
(copied from above with theoretical corrected.)
you are sorta right.  one point will not orient.  you have to have the b axis and the association of c as a point with a specifiied angle to orient.
think of it if you will: equator= datum a, north star and earth = axis b,  grenwich uk, prime meridian, = datum point c.  does that work at all for you?
with those you can navigate to wherever.
you do know how to incicate a single point as a datum feature rather than the entire surface?
for this particular part, couldn't you use the center of one of your holes as datum c?  then you could control the hole pattern and slot location to your satisfaction, and there would no longer be an arguement about how to define c or how many points it should take.
englishmuffin,
   look at the drawing.  if the drafter shows the datums in order as a,b and c, then b is the secondary datum, and c the tertiary datum.  if the drawing shows a, then c, then b, c is the secondary datum and b the tertiary.  this has nothing whatsoever to do with the geometry.
   both schemes are valid for the drafter, and they are different.  if you are reading the drawing, the precedence is whatever the drafter indicates.
                   jhg