Rounding off dimensions 1

[EngJW]

Trying to find a setting to make SW round off a .125 dimension to .12 instead of .13. I checked in the GD&T forum and they said that rounding off is specified by an ANSI or ASME standard, but my options are already set to ANSI. We have this same problem with Autocad but have never been able to fix it.

Thanks

 

[3Dlvr]

We used to have that "problem" of .125 dims showing as .12 despite our settings or standards in SW. We always attributed it to a bug. As far as I know, we never discovered a setting to control it.

 

[felix7502]

We use ACAD R14 and you can round up or down by using the Dimension Style dialog box, Annotations, Round Off. It is stored in DIMRND. In SW you can always check Override Value.

 

[ShaggyPE]

If you want the part to display a dimension of .12 then model it at .12, not .125. Your model should always reflect what you want.

BTW solidworks rounds up to .13 because of the 5 in the thousandths digit. You may be able to find a setting where the dimensions are truncated. This essentially cuts off the remaining digits after the place that you indicate.

 

[EngJW]

Shag-

Most everything I do is drawn at the 2 place dimension unless I want 3 or 4. So much of our work is derived from old designs that use fractions, so I just round them off to 2 place decimals before putting them in. Also, it is convenient because our default tolerances are for 2 places. We have one guy who uses fractions but enters them in 4 places or more, and then tries to make the dimensions round off to 2 places. Kind of backwards I think.

I had one that I designed at .125 and then later decided not to hold it that tight, so I tried the lazy way of using a 2 place dimension but it came out to .13 instead of the .12 that I wanted. No big deal. I just went back and changed the model.

 

[Theophilus]

EngJW,

Be careful with that. You would be breaking a norm of rounding digits 0 through 4 down and 5 through 9 up. To round a digit of 5 down would be wrong according to scientific rules of rounding--so I don't recommend doing so or your dimensions will be misunderstood by all who read them. So your dimension of .13 is somewhat common for a two-digit representation of 1/8.


Jeff Mowry

http://www.industrialdesignhaus.com

Reality is no respecter of good intentions.

 

[EngJW]

Theo-

There is an old tradition that if the 5 is preceded by an even number you round it down and if an odd number you round it up (we're only talking about the case where the last digit is 5). Most people would round upward regardless. I checked at the GD&T forum and apparently there is an ANSI standard for this, although I have not seen it. Those guys called it "rounding toward the even number" which I never heard of but it means the same thing.

 

[Theophilus]

Really--I had no idea about that. Thanks for the info. Seems an odd practice except that it makes things more easily divisible by two.


Jeff Mowry

http://www.industrialdesignhaus.com

Reality is no respecter of good intentions.

 

[EEnd]

The justification that I have heard for rounding 5 down on even and up on odd is that if you always round 5 up, you have 1 2 3 and 4 going down and 5 6 7 8 9 going up. That’s 4 numbers going down and 5 going up. This could bias the mean / average of a group of numbers. Rounding 5 down on even is supposed to get it rounded down half the time and up the other half, keeping it from affecting the mean.

This seems more applicable to numbers from an outside source (recorded data) than numbers that you are making up (dimensional specifications).

For more info see:

http://mathforum.org/library/drmath/view/58972.html

 

[JonathanAnderson]

The practice of always rounding toward the even number is generally referred to as the even/odd or odd/even rule. As was mentioned it only applies when the last digit is a 5 and there are no other non-zero digits after the 5 (.125 -> .12, but .1251 -> .13). The reason for this is that statistically speaking, always rounding the 5 up results in a skew toward larger numbers (0.5 is the same distance numerically from both 0 and 1). By rounding it down half the time, this skew is corrected. So say we have a set of numbers to round:
1.3, 2.5, 3.6, 5.5 with the average of 3.225
Rounded using the simple method of .5 always rounds up:
1, 3, 4, 6 with average of 3.5
Rounded using even/odd rule:
1, 2, 4, 6 with average of 3.25
So you can see how that method of rounding helps prevent overall skewing of a set of numbers to a larger average.

Jonathan Anderson

 

[EngJW]

Interesting. Didn't know there was a good reason behind it.

Now, how do we make SW work this way?

 

[ctopher]

I agree with most.
I suggest do not change the settings. Make the models per whatever decimal places you want them, dimension on the dwg with tolerances how you need them. Let the software round the way it's suppose to. Whoever is working with fractions, learn decimal equiv per print.

Chris
Sr. Mechanical Designer, CAD
SolidWorks 05 SP3.1 / PDMWorks 05
ctopher's home site (updated 06-21-05)
FAQ559-1100
FAQ559-716