.0625 Rounding up to .063 in generic model, but down to .062 in Family Table

The standard for rounding engineering drawing dimensions is to always round to an even number, so the Creo drawing is wrong and does not follow standards. This is defined in an ANSI standard. I found this explanation in another forum.

Maybe we need an enhancement request to PTC for Creo to follow industry standards!

Rules for Rounding Off
--------------------------------------------------------------------------------
Ever since the calculator replaced the slide rule, people have been able to get results to six or more places, therefore it's critical that we know how to round the answers off correctly. The typical rule taught back in elementary school was that you round UP with five or more and round DOWN with four or less.

SORRY, BUT THIS RULE IS WRONG!

However, please don't rush off to your elementary school teacher and read 'em the riot act!

The problem lies in rounding "up" (increasing) the number that is followed by a 5. For example, numbers like 3.65 or 3.75, where you are to round off to the nearest tenth.

OK, let's see if we can explain this. When you round off, you change the value of the number, except if you round off a zero. Following the old rules, you can round a number down in value four times (rounding with one, two, three, four) compared to rounding it upwards five times (five, six, seven, eight, nine). Remember that "rounding off" a zero does not change the value of the number being rounded off.

Suppose you had a very large sample of numbers to round off. On average you would be changing values in the sample downwards 4/9ths of the time, compared to changing values in the sample upward 5/9ths of the time.

This means the average of the values AFTER rounding off would be GREATER than the average of the values BEFORE rounding.

THIS IS NOT ACCEPTABLE.

We can correct for this problem by rounding "off" (keeping the number the same) in fifty percent of the roundings-even numbers followed by a 5. Then, on average, the roundings "off" will cancel out the roundings "up."

The following rules dictate the manner in which numbers are to be rounded to the number of figures indicated. The first two rules are more-or-less the old ones. Rule three is the change in the old way.

When rounding, examine the figure following (i.e., to the right of) the figure that is to be last. This figure you are examining is the first figure to be dropped.

If it is less than 5, drop it and all the figures to the right of it.
If it is more than 5, increase by 1 the number to be rounded, that is, the preceeding figure.
If it is 5, round the number so that it will be EVEN.

Keep in mind that a zero is always considered to be EVEN when rounding off.

The even rounding rule is stupid for a number of reason.

• If you're averaging numbers, you shouldn't be rounding them before they are average, you only round after they are averaged.
• With computers doing all of our calculations, measurements, etc. the phantom "0.01" isn't an issue.
• A large part of the design/manufacturing industry works in fractions, and a large number of those people are going to automatically enter something stated as 0.63" to 0.625" w/o giving it a second thought. You have to know your customer (whether that is a sales guy, fabricator, or end customer).

There are 2 huge issues here that no one has touched on.

1. PTC doesn't release ANY information that I am aware of on what standards they do/don't follow, or on how their system works. If they would simply say what they are doing, then we could accurately work with it regardless of how it is designed.
2. You cannot have a system that works with numbers in multiple ways.

A great example of #2 is that we built relations into some of our start parts to automatically right descriptions. One of those is for pipes. Based on the OD and schedule we assign to the part, the relations set the wall thickness and then write the description. After writing these relations we spent hours trying to work through what it is rounding, what it is truncating, and how to accurately get the results we wanted. It was the same issue point out above... in some cases numbers were rounded, in some places they weren't at all, and in others they might have been truncated but we never could tell for sure. We finally got what we needed, but that was because we had a limited amount of data that we had to troubleshoot. If you're using family tables and/or relations there could be a unlimited number of results and we should be able to know what the system is doing to take full advantage of this. That isn't currently possible without knowing how PTC is handling every number it deals with, and it is extremely difficult since numbers are handled differently depending on where they are being used.

And I work for a company where generally 1/32" of an inch is extremely tight tolerancing. I cannot imagine what it would be like where thousandths matter.