A different perspective on ISO 2768 1

[fsincox]

I have seen some comments here concerning ISO 2768 and 8015 and it's disclaimer at the end. Personally, I think they are just being realistic. The standard basically says their intent at general tolerances is to not to specify extra large tolerances manufacturing theoretically doesn't need. Then this final statement is basically saying a part may not be bad if it is in one of these areas, the general toleranced areas, by that measure it may be true. We generally don't put extraordinary large tolerances on our prints either. I only see the good old (2) place +/-.03 and (3) place +/-.010 that is our general tolerance. I think that is lazier engineering. I believe both ANSI and ISO standard are works in progress and not final statements, they evolve.
Do you accept the concept that natural/default tolerances may need to grow as parts/features get bigger? I think it is more realistic after some 30 years of experience. Thanks, I look forward to discussing it with you and learning more about it.

 

[KENAT]

One important function of a drawing (or equivalent MBD) is to set a legal definition of what you will or wont accept.

By having that disclaimer they are basically nullifying that concept.

In practice I realize that under some circumstances it is desirable to accept parts that don't quite meet drawing requirements, typically due to cost or schedule pressure.

However, to have it as the default setting that any discrepancies have to be verified to not to be OK seems potentially problematic.

ISO 2768 may be a usefull guide for what tolerances can typically be held for a different size and grade of shop etc. but I'd have serious reservations over invoking it directly. Also, is the information up to date, or can modern CNC machines typically hold much tighter without significant cost increase?

I've posted in depth my own battles with a vendor using this standard and what they claimed it meant on tolerance stacks etc. over in Drafting Standards, GD&T & Tolerance Analysis.

KENAT,

Have you reminded yourself of faq731-376 recently, or taken a look at posting policies:

http://eng-tips.com/market.cfm?

What is Engineering anyway: faq1088-1484

 

[KENAT]

Forgot to say

"Do you accept the concept that natural/default tolerances may need to grow as parts/features get bigger?"

Generally yes, though there are other factors involved.

As to the typical ASME tol block varying based on decimals being lazy, it depends how it's used. Any 'default' tolerance system is prone to that.

KENAT,

Have you reminded yourself of faq731-376 recently, or taken a look at posting policies:

http://eng-tips.com/market.cfm?

What is Engineering anyway: faq1088-1484

 

[fsincox]

KENAT, I was hoping to talk with you. Iit seems you have been dealing with this subject for a while. I reread ISO 2768-1 & -2 last night and it specifically says: "general tolerances" so I am not sure it negates stated drawing tolerances at all, do you know that? Those are the ones (according to this philosophy) we supposedly care about.
As is obvious from my first post, we make mistakes all the time, I make them every day. To expect perfection is a very high bar.
I have here some data from the SFSA (Steel Founders Society of America?) a study that is adapted from ISO 8062-1994 (and it gives both metric and inches) basically supports the concept and does a statistical analysis of it, with some additional factors added in.
I do believe the ANSI standard is better (especially the new one) for comprehensiveness t I was involved in many arguments after the 1982 standard whether the second line of a composite location tolerance with a secondary datum restated was for orientation or location, thankfully they finally cleaned that up. A work in process.
I still get stares of unbelief from most shop and inspection people, when I say a position tolerance controls orientation or that parts must be perfect form at MMC. So if you are looking for trouble we already have it. The question becomes how do we best proceed. I happen to believe the metric system is a better engineering system, specifically; do to rounding issues with inch dimensions. I have had to fight with manufacturing about this since I started. This is also why I jumped to ANSI and the basic dimensioning system. But you need to apply tolerances somehow. I have worked with castings and weldments and had arguments about rounding dimensions to (2) places but don’t try to imply +/-.030 or +/-.060, they will just ignore it then. Is this a better way?

 

[fsincox]

KENAT, did you ever get an resolution on your issue?

 

[KENAT]

I am not saying that refering to the standard takes precidence over tolerances explicitly stated on the drawing, I'm fairly sure the opposite is true that explicit drawing tolerances take precedence, although I haven't looked at the standard for a while.

What I'm saying is that for any dimension that doesn't have a tolerance stated on the drawing, the relevant tolerance from the standard nominally applies. However, the disclaimer then says to the effect that just because the dimension doesn't meet this tolerance doesn't mean it should be rejected, instead look at function.

If function isn't defined/captured by the drawing then how/where? You end up in a loop, or having to analyze fit and function each time. To me a well prepared drawing generally draws the line in the sand as to what is or isnt' acceptable.

"Those are the ones (according to this philosophy) we supposedly care about. "

I think perhaps that's part of my problem with part of the approach this standard implicitly takes. I care about all the tolerances. If the tolerance needs to be tight based on function, I make it tight, if it can be loose I make it loose. For some tolerances there isn't an easy to define functional limit, at least in one direction, in these cases I'll try and base the tolerance on manufacturing capability, aesthetic concerns or the like. However, on most parts I've dealt with if these aren't the majority.

Any kind of 'default' tolerance has problems and is prone to people over relying on it and not looking at the individual tolerances based on function and manufacturability. In the other extreme to 2768 in the UK where I previously worked we had a single default value, that was selected for each drawing based on size, manufacturing capability, function etc. We then directly toleranced any dimensions that didn't use this. However, this is still open to abuse.

Not sure what you're talking about where you mention expecting perfection. Tolerancing essentially sets how imperfect something can be while still being acceptable.

I'm also not sure I see your point about metric being better for rounding etc. It seems you are touching on a lot of areas regarding tolerances. Some of these aren't inherently inch V mm or even ISO V ASME. US tends to put function first & is often based on worst case analysis using hard gaging etc., while I believe ISO leans more toward manufacturing capability, CMM & statistical etc.

I didn't get a satisfactory resolution to the problems I had with the German Vendor that used this 2768. They seemed to apply the standard in a way I didn't see supported by the text, and the US office of the German company actually agreed with me. This confusion makes me even more wary of the standard, I don't know if it was a one off or endemic, but it's certainly not good.

You may have got more of an audience over in GD&T, maybe I'll post a link to this thread there.

KENAT,

Have you reminded yourself of faq731-376 recently, or taken a look at posting policies:

http://eng-tips.com/market.cfm?

What is Engineering anyway: faq1088-1484

 

[fsincox]

The point I am trying to make with respect to perfection and my comments on problems I have had with the ANSI standard is that these documents are compromises by committees, not tablets from GOD so I have learned to expect some issues.
Do you now work with inch or metric drawings or both? You actually worked in the UK? Did they use ISO 2768 then? Thanks.

 

[KENAT]

I get what you mean now, the ASME 94 std isn't perfect, I doubt the new one is either but I haven't seen it. However, more often than not when I think it has a fault, I later come to realize it was me with the problem.

I currently work primarily in inches, although some optical stuff gets done in mm as that's "industry standard" and occasionally if we're interfacing with mm off the shelf parts we get the odd mm drawing.

In the UK we worked a mix, again depending on what we were interfacing with. New stuff was pretty much all mm but when interfacing with old or American inch stuff we'd use inches.

Just marks on a scale at the end of the day.

We more or less followed BS8888 in UK, but really were closer to the old BS308, though I wasn't as familiar with all the standards then. I'd never heard of 2768 until a couple of years ago, around about when I started posting on the subject. I only worked one place in the UK so can't say for sure we were typical.

KENAT,

Have you reminded yourself of faq731-376 recently, or taken a look at posting policies:

http://eng-tips.com/market.cfm?

What is Engineering anyway: faq1088-1484

 

[fsincox]

“However, more often than not when I think it has a fault, I later come to realize it was me with the problem.” I too have found that true many times, also.

I spent 25 years at a machine tool company that went out of business because we couldn’t compete with the Germans, we were told. The basis of the ISO standard was DIN, I believe. I contracted at a few places for a few years, and have now worked 7 years in defense/aerospace. All of my experience comes from the “rust belt” so maybe (I can always hope) my companies are more “old fashioned” than the norm (both companies started around the turn of the 20th century). I believe I have a good background in ASME Y14.5, (still reviewing 2009), and I am trying to be a “true believer” so to speak.

Next then lets tackle the metric/rounding issue. This is just my opinion, please bear with me and you may understand why I desire metric for engineering design work. If you took at all the things made or engineered in the world and averaged the size to get an average size engineered part. I tend to think there would be more small parts than big so the average would tend toward a smaller mean. I realize this may be my warped perspective from my experiences, but please, go along. Say the average is 30 in diagonal cube, I have no idea what it really would be, what do you think it would be? Since a millimeter is so much smaller than an inch it will be much less proportionally and it is much easier for a designer to round non-critical dimensions to the nearest whole number. Also, since a millimeter is only about .040” if you really need that fine of measurement you really do care, so, the precision of decimals seems more natural.
Years of arguing with manufacturing about .625 vs .62 so that we can have (2) place tolerances (which in many cases are not realistic, anyway) has made me desire a better way. Now if you desire to work with metric dimensions how will you to specify tolerances? We used to use (1) place and (2) place mostly. If that is your preference so be it, I think the natural expansion of tolerances, as in the ISO system is more realistic and flexible. Is it perfect? No, Will I be right every time? No.
I also realize that the grass is always greener on the other side, so I have come here to learn the potential pitfalls of the ISO system.
I was the GD&T guy at my company and we were using the 82 standard at the time. I mentioned the restatement of secondary datums in a composite position tolerance issue, earlier. I see here also others talking about the simultaneous vs separate issue, that was another issue i saw and avoided like the plague I knew it was a minefield. It was issues like this that killed it for us. People would say: “show me where it says that” and there was no good clear illustration of these issues then. The guys who don’t like change anyway grab that and say: "this is too complicated, we got by all these years bla, bla, bla," and the management who are clueless to the issues just goes along. The illustrations they show are simple, real world parts are not all so simple. This, by the way, is why I like the new standard best it has many more illustrations dealing with the issues I had to deal with 15-20 years ago. I hope you will find you like it too.

 

[fsincox]

Measuring land you might round to the nearest inch, not manufactured products.

 

[fcsuper]

Just to chime in, from my perspective, the disclaimer at the end of ISO 2768 is a self-invalidating statement that is very dangerous in the U.S., as it can (and prolly has been) abused by our legal system. ASME Y14.5 was born out of a multitude of lawsuits in the middle part of the 20th Century where ambiguities have been slowly purged from the industry. Once Europe starts experiencing this (and they will, especially if they continue to expand into the U.S. market), I can guarantee that disclaimer will go bye-bye.

The other reason why that disclaimer is out of place is that it is discussing manufacturing practices on a specification procedure, as it dabbles in both inspection and part acceptance (which are covered by completely different standards). As such, it is likely in conflict with those other standards. Of course, ISO is notorious for the creation of conflicting and redundant standards (which is another reason to stick with ASME for now...at least until ASME's current involvement in ISO can help clean ISO up).

Matt Lorono
CAD Engineer/ECN Analyst
Silicon Valley, CA
Lorono's SolidWorks Resources
Co-moderator of Solidworks Yahoo! Group
and Mechnical.Engineering Yahoo! Group

 

[KENAT]

fsincox, I'm really not convinced that the tolerancing problems are to blame on ASME or INCH etc. It comes down to the people specifying them not doing it properly. Blaming the tol block or equivalent is just lazy in my opinion.

Also, per the ASME 94 std, metric dimensions don't have trailing 0's so you can't really use the ".xx = +-.010" ".xxx = +-.005" etc. unless you have your own internal rules.

You have far more experience than I, especially in a machine shop environment so I'm trying to understand your point.

Customizing the block tol for each drawing addresses some of your concerns, but can introduce new problems, if people get used to it meaning one thing but then it changes mistakes get made - I've seen it.

There's at least one member over in GD&T forum who doesn't like default tols at all because people over rely on them. While I wouldnt' go so far as to say they should be eliminated I routinely see over reliance on them and frequently have to tell people to directly tolerance if necessary.

One thing about the tol varying by decimal and it's effect in CAD moddelling, people will model .375 but dimension .38. Well if a bunch of these start to accumulate you end up getting errors. If it's .38 model .3800000 if it needs to be .375 for geometry etc. then dimension .375+-.xxx as required, don't let the block tol drive you to fudge it.

The principal that tolerances tend to increas with size is I believe sound. A good designer/drafter should take account of this when tolerancing even if having a block tol. 2768 may have some use as a guide of what level of precision is possible compared to size of object etc. and how 'fine' the machining is but some claim it is out of date information as it's probably based on typical pre CNC capabilities etc. While maximum precision may not have improved much with CNC (some claim it seems to have decreased) I believe average precision probably has. I'm frequently told by people they can hold +-.001" or +-.002" whithout cost increase.

At the end of the day I suppose, what country's are more dominant in eingineering/machining etc., what standards do they use, and is there a clear affect of that on why they're successful. Based on this you might start to doubt the US system, but is it a consequence of the standard or other factors? You mention someone told you Germany beneffited from 2768 or the old DIN equivalent. I'll admit I'd be interested to find out more from someone that really uses it and deals with the ramifications of the disclaimer etc.

At the end of the day, most parts probably wont be near the limits statistically. And if limits are based more closely on manufacturing capablity then that will help.

Rambling now, gotta go.

KENAT,

Have you reminded yourself of faq731-376 recently, or taken a look at posting policies:

http://eng-tips.com/market.cfm?

What is Engineering anyway: faq1088-1484

 

[fsincox]

Thanks, Guys, I do appreciate your input.

"Also, per the ASME 94 std, metric dimensions don't have trailing 0's so you can't really use the ".xx = +-.010" ".xxx = +-.005" etc. unless you have your own internal rules." I absolutely do not want to do this; I feel it is the same thing, in my opinion. However, that is what was done in the past and I am only allowing that it seems some others may prefer it. I thought you might too.
How are you going to do it then? Specify all?
that may be practical for small parts, not so much for machine tools.
I will try to find that standard; I actually just had it, here, and was surprised to find it. We all have to clean out old files today and I ran into it. I do remember the date was very early 1980's because I specifically looked at that due to our discussions here.
The issue of CNC quality was brought up. I am constantly badgered by requests from manufacturing to revise tolerances on drawings 50 or 20 (whatever) years old. This has made me wonder about the quality issue. We did the best we could at our company to build good machines, but, my grandfather was a machinist and I suspect there was some art to it, not easily replaced. I look forward to rapid prototyping in real metals to become a reality to revolutionize manufacturing methods and default quality. .

 

[fsincox]

"I'll admit I'd be interested to find out more from someone that really uses it and deals with the ramifications of the disclaimer etc." I agree I was hoping to hear from someone that uses it.
The standard I have is DIN 7168 and it is in english, the first one I ever saw was not. This one is dated April 1991 and states "supersedes May 1981". It also states "This standard not to be used for new designs. General tolerances are now covered by ISO 2768....".
The other thing I thought I remember was the first ISO version I saw did not have a fine class and the German (DIN) I saw in German did.

 

[fsincox]

I have also noticed an earlier comment that the ISO by having many different standards to cover ANSI Y14.5 issues might be gouging people. I received a letter from Lowell Foster (form letter, not personal) former chairman of the ANSI GD&T committee, in which he says this: “In the ISO standards development agenda it is “one subject-one standard”. This philosophy and history alone could deserve volumes of coverage. Yet it can be said in brief, that this method evolves as the only reasonable way when being exposed to the lengthy travel, involved time consumed, cost outlays, national pride, language barriers, due process required etc. One subject at a time surfaces as the only practical and achievable approach. Obviously, related subjects must be discussed simultaneously (such as position tolerance and datums) but the detailed coverage on each ends up in separate documents of minimum size and very basic coverage” (November 1992).
The ANSI standard isn’t cheap either, I wish they were I would have had them all along time ago.

 

[KENAT]

I swear I just posted but I don’t see it, I think this is most of it minus some polishing:

Regarding using standard fits (eg 20 H13) for holes as you mention in thread1103-245261. Fine that gives the clearance/hole dia limits, what about positional tolerance?

Given that the equivalent positional tolerance using 2768-1 increases as the hole space increases then just relying on a single class for all holes nominally the same size regardless of spacing, with the ‘spacing’ tolerance from 2768 wouldn’t work.

Just because the space between holes increases, the allowable variation in position doesn't magically grow. If you want to use 2768 for your hole location you’d have to apply different fit classes to different holes of the same nominal size but with different spacing. I’d expect it to be better to have consistent Tolerancing of hole size for holes nominally the same size and choose an appropriate location tolerance that ensures fit while meeting manufacturing capabilities.

I’m not completely anti everything ISO. The idea of trying to standardize to limit the amount of tooling required etc. makes a lot of sense. However, on things like hole/shaft fits that’s fine for one shaft in one hole. When you have a pattern of holes & shafts then position comes into it.

Machineries gives recommended hole/c’bore/csk diameters/sizes for certain types of fastener. However, they don’t mention the position tolerance. I get people that blindly copy them and then rely on block tolerances for hole location and surprise surprise, you get interference issues.

Also there are tables of standard drill tolerances, can’t remember the spec they’re based on but the tol range increases as the size hole increases and they are typically from -.001 to +.00X. I’m not sure this is that much different from using the fits you mention for hole patterns.

Also my experience has led me to believe it’s generally best to have it explicitly detailed on the drawing, rather than cross referencing other standards where possible. However, this may indeed be in part due to less skilled manufacturing personnel. Then again, I thought mass production/assembly line and the like were all about simplifying/deskilling manufacturing to reduce costs.

I have ISO 2768-1 & 2 dated 1989, first edition. They were current as of about October 2007 when I got them. It has fine, medium, coarse & very coarse.

The ASME/ANSI system, and some of the associated conventions may not be perfect but seem a bit more coherent than ISO. They also seem to focus a bit more on function that reflects how I was taught and practiced in the US/UK. Although in fairness I’ve probably spent more time learning the ASME std’s in the US than I did the ISO’s in the UK.

However, maybe more attention should be given to manufacturability like ISO seems to, at least implicitly.

Sorry, long and rambling but you’ve brought up so many topics I’m having trouble deciding which to die in a ditch over;-).

KENAT,

Have you reminded yourself of faq731-376 recently, or taken a look at posting policies:

http://eng-tips.com/market.cfm?

What is Engineering anyway: faq1088-1484

 

[fsincox]

I am NOT advocating ISO default title block tolerances for hole inter-pattern location tolerances, I, myself, have never seen default tolerances used for hole inter-pattern location tolerances. You maybe are claiming you did, fine, as I said people make mistakes. Why do you think there is a whole standard on engineering revisions, they know that people will make mistakes. The drawings I have seen may have used the default tolerance to locate a pattern of holes but they always had a position tolerance, generally with the mounting face as a datum, for inter-pattern tolerances (precisely, because of the reason you stated above).
The other thread was about hole size callout methods and I stated another option I had seen. I also noticed there was a lot of default title block discussion so I mentioned that.
One of the things the ISO standard makes clear, in my opinion, is they don’t feel an engineer should waste time trying to find the absolute worst tolerance condition on less critical features if the process doesn’t need the tolerance anyway. All of these title block defaults are making the same assumptions except they are more fixed and in my opinion, less realistic in some (you)/many (me) cases.
I wondered: “what does your title block say?” I can’t help wonder if your parts may tend toward the simpler side, I may be wrong, I mean no offence.
The ISO standard states it’s intent is basically to focus on the critical features and not bury them. Have you ever seen an engine block drawing, all the datums? They have a chart on them just to keep them straight.
I happen to think it helps focus on the critical features. I wrote a treatise for my old company based on just the same concept.
I had noticed, essentially, that if you started from the “unimportant” edges like manufacturing always wanted us to (assuming the edges of the part were of the ISO kind, unimportant TO FUNCTION) to locate your important features, say, dowel pin holes. I would have to make them a datum; with of course the mounting face as primary, for the gear bearing bores. This is necessitated by the tight tolerances involved in a gear case assembly. I saw I needed to create more datums on the drawing and more, in my opinion, confusion. The drawing reflected process not function. If I would just start with the mounting face and the dowels and profile all of the unimportant features from them along with the gear bores I ended up with a more focused, simpler drawing and I didn’t have to try to guess were manufacturing wanted to start. At the time profile was scorned as a second-class tolerance no one wanted it on the drawings back then, I believe it still hasn’t seen its full potential 2009 helps even more, I really felt I could see where it was going. I discussed this issue with my mentor Dr. Don Shepherd, now a member of the ANSI committee, and I believe there currently is an illustration of a similar case in the standard. It was not in the 1982 version.
The next issue was what was the right profile tolerance for over all. Then I found ISO.
I feel the ISO standard intends to create this same effect. I may be wrong.

 

[fsincox]

For a picture look at ASME Y14.5M-1994 (pg. 71, fig 4-26 or ASME Y14.5M-2009 (pg. 79, fig 4-39). Again they simplify it a bit.
Per my other post sorry, Lowell Foster is not listed as the chairman he is only Vice-Chairman of subommittee 5 "dimensioning and tolerancing" 1973 & 1982, Chairman of that committee in 1994. Only listed as a member of the committee in 1966. Don is listed on the same committee in 1994. I don't know their actual start dates.

 

[fsincox]

Also, by the way, you need to invoke the more complex position tolerances you are starting to see pop up in the standards to accommodate the fact that you start somewhere you don't care about to get to the features you do. This assumes you want to give more tolerance from unimportant (functionally) features and not hold things unnecessarily tight for manufacturing to achieve.
Dr. Don used to say: "if you measure a football field does it matter which end you start with?" (in the proper framework, of course).

 

[fsincox]

More on the CNC issue, I am looking for the DIN standard, (found it DIN 7168 April 1991, supersedes May 1981) the quality engineer says to me: “we have another casting issue will you look at it with me”. We go to the shop floor and look at some housings, these housings have a cast electrical box at a slight angle from vertical. Think of a standard AC frame motor housing with an electrical box hanging off it, except the box had to be rotated about the centerline of shaft rotation slightly. The walls of the box are fairly thin and in each corner there is a rounded boss. For a picture of just the box, again reference ASME Y14.5M-1992, pg. 71, fig. 4-26 with 4 tapped holes instead of clearance (12.3-12.4, no dowels). These bosses only purpose in the life is to contain the four threaded holes for a box cover. The face the cover bolts to is machined perpendicular to the axis of the threads as in the fig (note: datum “A”).
OK, for reasons I won’t go into now, they start at the opposite end from that box to center on another cast boss that has a port in it. Because of the assumed +/-.030 they think the holes will be OK, the tapped holes break actually break out the wall, on many of them.
I stare in dis belief and proclaim: “this is not a casting problem it is our problem”, the quality guy then agrees, they have been having these problems since the line was brought here around 2 years ago., I was asked before about this problem and tried to get them to allow me change the hole locations position tolerance to be to centerlines derived from the box walls, it will probably happen now, the shop rejected it and an other engineer tried to get the casting house to “clean up their act”. Had the people who originally designed this been using ISO casting tolerances they may have foreseen these problems, with the security of +/-.03 they had no reason to think it wouldn’t work. The real world always interferes.
My grandfather was a machinist, I know machinist’s., because you push a button on a machine doesn’t make you a machinist any more than because you work on a train makes you an engineer (your kind).