“Oh, what a tangled web we weave, when first we practice to deceive!” attributed to Sir Walter Scott, 1808.
Geometric Dimensioning and Tolerancing, or GD&T, is a system of symbols, definitions, and practices designed to eliminate miscommunications between the description of an item and its real-world implementation. It removes unintentional deceptions.
This was supposed to be sharing my understanding, instead it highlights, to me, that I do not fully understand.
This image shows all the sides of the object as well as two projection images. This shows us what the object should look like, but it doesn’t tell us how to make it or the actual locations of each item.
Here I’ve added a center line and a dimension. The dimension says that this is 2 inches from the bottom to the top.
So where is the centerline? Knowing where the centerline is important as it tells us where the slot will go.
Worse, the computer has a concept of absolute sizes. In the model and on the paper, that part is 2.0000000000000000 inches wide. It is not realistic to make something that is exactly on size. In the physical world, we need things to be close.
How close is “tolerance”.
Consider a hole that is marked as 0.2500 and a pin that is marked as 0.2499. First, those measurements are difficult to reach. The tighter the tolerance, the harder it is to hit and the more likely you are to scrap a part.
If those two parts were made to those specifications, the pin should go into the hole. There are still reasons why it might not, but it should fit. There is a gap between the walls of the pin and the walls of the hole of 0.0001.
For perspective, a piece of printer paper is 0.004 inches thick. So when we are looking at hitting a tolerance of 0.005 we are looking for about the thickness of a single sheet of paper. If we are looking for a tolerance of 0.0005, then we are looking for about a tenth of the thickness of a piece of paper.
With my equipment and at my skill levels, I can hold 0.001 all day long. 0.0005 is a bit harder.
As a manual machinist, if we need to hit very precise tolerances, we will often times use abrasives to take off the last 0.0002 of a part.
So the engineers provide a tolerance for the theoretical values they provide so that they are achievable.
Which takes us back to my silly drawing. To know where something is, you need to first specify where to measure from.
These are called “datums”.
The A in a box with a filled triangle pointing to it tells us that this is a datum. The cool thing is that we can use that datum in many views. Since the dimension is based off the A datum plane, that means that is the measurement.
If the part was too thick, we would take it off the side that is NOT the datum.
(hey, these are just notes, don’t use them, look them up. The ASME Y14 spec is only a few hundred dollars.)
Here is the final top view. I have all the dimensions to machine this, but I don’t have the GD&T. I thought I understood, I don’t. I need to do some more reading. I know I need a feature control frame in there, I just don’t know where to put it and what it is supposed to say.
To be more clear, I know that I need to mark my datums. 3-2-1 rule applies, which I understand. I also understand the 6 Degrees of Freedom(DoF). I know how to mark the datums and to transfer those datum markings as required, to make it easier on the machinist.
I understand that I put the datums at the end of the feature control frame, so A|B would be a primary datum of A and a secondary datum of B. I understand how to make axis and centerlines datums. There is just so much going on here. Any way, another geek dump before I go look at Ezell
Comments
9 responses to “GD&T, oh my, I still don’t get it.”
Don’t feel bad, I got my drafting and design degree 19 years ago and I still don’t fully grasp GDT, because it’s not as common as they lead you to believe, and no school can afford ro supply you with the actual documents… have you looked in the bible yet? In the Inspection chapter they have a little on drafting and a decent amount of tolerancing.
“ the computer has a concept of absolute sizes”… actually it is not the computer, but you. I have not read your entire article (I will), but what is important to understand is what part interfaces with the slot and for what purpose. Decide how much slop in that interface allows the “purpose” to do what it is supposed to do and how best to relate that purpose to the rest of the system. GD&T is one of many techniques to describe that interface. It is a useful technique to be sure, but may (likely is) overkill for your Casinator project. The key to determining which technique to use is to use the simplest way that will get you *repeatable* results for produced individual parts/components with an acceptable scrap rate in complete Casinators and the ability to provide repair/replacement parts to service delivered units.
Stop thinking of describing individual parts and more about the entire system and the *relationships* among all the parts.
Thank you. What you are observing in real time is my obsession with learning. I’ve seen GD&T markings and, bluntly, ignored them. When Birdog357 pointed out some potential issues with my dimensioning, I went looking.
.
GD&T is what popped. Every part I make has a datum that I use for dimensioning. I spend the time making sure that I know what that datum is and how to reference it in the different setups.
.
I stand by the statement, “The computer has an absolute value”. There is no way to store 0.750+/- 0.005 in a value within the computer. The computer has absolute numbers. My intent is what tells us how to interpret those values. And I have to express that intent.
.
In the computer world, where I am a subject-matter expert, the value 0x30 is stored in memory. It is just a pattern of bits. We can express that as 0x30, 060, 48, or ‘0’. All of those are stored with the same bit pattern.
.
It is the intent behind those values that is important. You have stated it correctly.
.
I have NO disagreements with anything you have said here.
Not to be an ass, but … Strictly, the computer stores single values. The software you’re working with could have been written to have an entire set of values – records, arrays, etc. – associated with a single dimension: datum references, tolerances (of which there are many types, depending on what’s being dimensioned), etc. It could further have been written to display or suppress default or blank entries, etc. SolidWorks has provisions something like this, for instance.
.
Depending on the software, where you are in the process (e.g. 3d modeling, vs working up a drawing from that model) can determine where it’s easiest to enter tolerances etc.
.
Of course the software has to have provisions for recording and processing the full data set for each dimension, as well as properly processing the relationships between them. There’s a reason commerical CAD software is pricey…
This is where the experience comes in. Any idiot can model a part. But it takes an experienced drafter/designer to know when to add a bit here or there and to know how to properly format a drawing to convey the proper information quickly and concisely. Also, consult the bible.(Machinery’s Handbook) It has loads of charts so you know what to design.
Is there a rule, or a convention, that the number of digits in a dimension imply the tolerance if not otherwise stated? I’m used to that, perhaps from my father (a metrology professor) — so I would read that overall dimension 2.00 to mean “2.00 +/- .005”.
I still remember discovering how professionals can get caught when I reviewed a draft ANSI standard networking spec, and looked at the drawings of a fiber optic connector. It had cylindrical ferrules, with flat tops and 45 degree bevels. But the minor radius of that bevel (the radius of the circle where it intersects the front face) had a tolerance that allowed the max diameter to match the ferrule diameter. So I asked in the meeting if that bevel was intended to be optional. Got two responses: “no it wasn’t” and “hey, aren’t you a software guy?” 🙂
Yes, there are conventions on default tolerances based on number of digits past the decimal. My understanding is that it is 1/2 of the smallest value that can be expressed. 2.0 +/- 0.05, 2.00 +/- 0.005, 2.000 +/- 0.0005.
.
As a software dude, I’m required to understand what my software is doing in relationship to the real world. This means I’ve been known to ask questions relating to my clients’ businesses that they never expected to hear and often had not thought about.
Yes there is a standard interpretation based on the digits presented, fractional or decimal… the drawing will indicate how to interpret all dimensions that do not have explicit tolerances in the legend area of the drawing; it simplifies the drawing space.
Also, toleranced dimensions do not necessarily split it down the middle; there are instances where the tolerance is “minus 0/plus something “ and vice versa, as well as “minus one amount/plus another amount”.
As you are discovering, engineering (that is what you are doing) a device/product is a balancing act on many levels; it’s not a trivial exercise. Think of it this way: if you’re making onezee-twozee then you simply “make it work”… when mass producing, the tolerancing is intended to maximize acceptable end items while minimizing the cost of manufacture AND assembly, and minimizing scrap.
Choosing your datum(s) is (are) key.
I hope I’m not sounding like a broken record…
One more thing… you won’t get it right the first time… you will revise your design and your tolerances over time. Keep track of and document each revision; include the “why” for each change. Consider creating a Design Decision Record.
…
This is why I so enjoyed a career in engineering… always a huge, challenging puzzle.
Enjoy the journey.
Unequal tolerances: I’m not much on mechanical design and don’t know where I’d see it there, but it’s very familiar from electronics. Capacitors used for filtering (bypass) purposes often have -x/+y tolerances with y much larger than x (say, -20%/+80%). The reason is simple: for the intended purpose, a smaller than desired capacitance means it’s not as effective as it needs to be, but a larger than specified value, within reason, means it is more effective than needed and that is just fine.