dimensioning error

[Anonymous]

Occasionally, I get an error in dimension strings. For some reason dimension strings will read just a bit under or over what I've modeled.

I've included a screen print of the problem. The wall was modeled as a 40'-1" wall. I put in windows and doors and then adjusted the window and door locations. The problem is that the interval dimensions do not add up to the overall length. If you look at the second interval dimension from the left you'll see that the interval dimensions are short by 1/32". THis is pretty typical of this problem.

This is a problem for us that I want to get solved. What is causing this and how can I prevent it?

We are aware that we could customize the dimension text. We would like to really fix the problem not just use a workaround.

Thank you,
John

[Karl Ottenstein]

Hi John,

You won't like the answer: there's no real way to avoid these things at all times.

There was a big (and I mean big and long) discussion of this issue on the old escribe list in April 2002. Here is just one message from that - my response which talks about numerical representation and error:
http://www.escribe.com/software/archicadtalk/m15278.html

The context of that message (thread index):
http://www.escribe.com/software/archicadtalk/index.html?by=OneThread&t=THIS%20is%20the%20dumb%20dim%....

The other threads discussing this issue (at length!!) are:

http://www.escribe.com/software/archicadtalk/index.html?by=OneThread&t=%5Bwishlist%5DMAJOR%20Dimensi...

http://www.escribe.com/software/archicadtalk/index.html?by=OneThread&t=MAJOR%20Dimensioning%20Bug-%2...

http://www.escribe.com/software/archicadtalk/index.html?by=OneThread&t=MAJOR%20Dimensioning%20Bug%20...

http://www.escribe.com/software/archicadtalk/index.html?by=OneThread&t=MAJOR%20Dimensioning%20Bug


The best thing that the user can do is to always enter amounts in the coordinate box. (Certainly NEVER drag a wall, look at the number in the coordinate box, and click when the number looks good. That is a formula for disaster.)

But, because of the numerical representation of decimal numbers (and fractions) in a binary computer (and in your and my case, numbers converted from Imperial to the internal AC metric values), you are more or less guaranteed to see these things at times.

(It could happen less if different representations and math were used inside of ArchiCAD. But, that is not something GS can easily change.)

The workaround is to manually 'fix' any incorrect dimensions. Not a happy solution. Alex's tip was to use red for the text of any such dimensions that are manually adjusted, so that if you change the model, you know to revisit those dimensions.

Karl

[Aussie John]

I have found this is as a result of rounding. (I will work in metric for my convenience)
Strictly this is not a computer error but an operator one.

two elements may be 10.4mm apart but appear as 10 on the dimension. Another 10.2mm apart appears as 10 on the dimension also)
The combined string will show dimension 21mm (20.6 rounded up) when you expect it to be 20

The solution as Karl says is be scrupulously accurate in setting out the elements being dimensioned.

As a temp measure increase the accuracy of your dimension string (& working units) and you will find the errors. You can then adjust as necessary

[Jefferson]

Hi John,

Until reading Karl's post I didn't fully understand that this was in part the fault of my computer+ ArchiCAD, I assumed as always, pilot error, but it's nice to share the blame . My experience is my errors come to life when dimensioning, and having my tolerances set to 64's, the highest possible in my effort to build the "perfect model" they show up, pronto.

The workaround I use, thank you Mr. Ellis , is to nudge the object, i.e. alt+x 1/128". This almost always places the object being dimensioned where I thought it was in the first place and corrects the dimension string all at once and I can move forward. Zooming to AC's full extent doesn't reveal anything being out of place either. Occasionally I will have to repeat this once or twice, not usually. I am very reluctant to insert false dimension numbers as this nullifies the point of being accurate, but in the rare instance the RED text trick is VERY valuable. I am coming to believe that both the computers and software we drive require more adaptation on our parts to them than we'd like to think. Like elections, grin and bare it..........

[Anonymous]

Karl,
Thanks for the links. The wall dimensions were set by snapping the walls to layout lines. Each of these objects were placed randomly on the wall and then shifted to their final locations using the coordinate box. Maybe this system has something to do with the errors. I'm trying to think of another way to size the walls and locate the doors and windows on the walls.
I've also had this problem when dimensioning interior walls intersecting with exterior walls. Most of the time I use the coordinate box to nudge the interior walls to a even (hopefully) dimension.

Jefferson,
In my case nudging the objects didn't seem to work as the dimension string on the other side would reflect the change. I have had luck with completely deleting the object and then reinserting them into the wall.

Thank you,
John

[Jefferson]

Hey John,

I was hasty and didn't do the math on your dimensions,

It sounds like your setting yourself "right" with layout lines which makes this all the more frustrating. I also set a single overall dimension string and then multiply it, and dimension the other "sets", as in windows and doors from one of the copies this way I know for SURE that the overall string started out correct. For what ever that might be worth..........

[Anonymous]

Jefferson,
I have often done the same thing with dimension lines. It seems like it should be a bulletproof solution to dimensioning errors.
When I layout the exterior walls I pull one overall dimension from outside corner to outside corner. Then I replicate that dimension line three times for the intermediate dimension strings. Next I select the dimension string and snap to the windows, doors, jogs in the walls and interior walls that I want dimensioned.
Doing dimensioning this way I am assured that the overall outside dimensions are snapped to the same points for all of the dimension lines. Unfortunately, I sometimes still get the error I wrote about above.
Thanks,
John

[Aussie John]

I have never found the snap method to suitable for the accuracy reason and to move an element using the coordinate box is best