View Full Version : Accuracy
daveyboyd
10th Jan 2008, 03:02 pm
While working on my current project I have been noticing that AutoCAD isn't as accurate as you might expext and I was wondering if one of you kind people could shed some light on the matter. For example, I want to draw a line 150mm (units) long I already have a line drawn so I offset the line by 150, neat. I then draw a line inbetween the two parrallel lines but overlap the ends. I use the trim command to trim the overlaps so that in theory the line is 150 dead.
I then check the length of the line by selecting it and looking at the properties box, my units/precision is set to max (.00000000) The line is 150.00000007.
Now I know this is no more than a nats knob as a distance but it isnt quite exactly 150 by any stretch. I have noticed this a few times on a number of projects but haven't worried as it is well within the stated tolerances but sometimes it's is greater than others and it's begining to annoy me now. Anyway, rant over but does anyone know what causes this?
Alan Cullen
10th Jan 2008, 03:13 pm
Dave, it would help if you posted the coords of one end of a line.
Back in the days of long ago, acad couldn't handle mapping coords too well. That's when they introduced things like Map. That may explain your errors. Acad uses coords to calculate lengths, so if you are in the realm of mapping coords, then that is where the problem lies.
Don't know of a fix for it.
Cad64
10th Jan 2008, 03:39 pm
I want to draw a line 150mm (units) long I already have a line drawn so I offset the line by 150, neat. I then draw a line inbetween the two parrallel lines but overlap the ends. I use the trim command to trim the overlaps so that in theory the line is 150 dead.
I then check the length of the line by selecting it and looking at the properties box, my units/precision is set to max (.00000000) The line is 150.00000007.
:huh: I just tried it on 2008 and my line is dead on 150.00000000.
Is everything in your drawing at Z0? Try running the FLATTEN command and then try again. A slight Z elevation on one end of your line will cause what you're experiencing.
daveyboyd
10th Jan 2008, 03:46 pm
Here is an example of what I mean . . .
Everything is drawn in world co-ords with no z values
NBC
10th Jan 2008, 04:01 pm
:lol: Call the X-Files man !
Alan Cullen
10th Jan 2008, 04:16 pm
Quite simple, mate. you are on mapping coords. That is why stock standard acad can't calculate it. The coord values are too high. Go check with your acad dealer for an explanation. It was about 2006 when this became a problem.
If you move that drawing back to coords 0,0 , it will work.
rkmcswain
10th Jan 2008, 07:44 pm
This has nothing to do with 2006, it's been a problem since AutoCAD was first created. AutoCAD is accurate to 15 significant digits. When you start using the majority of those digits on the left hand side of the decimal point, you have fewer on the right to be used in calculations.
Do your work on assumed coordinates (near 0,0) then move to real-world coordinates when you are done.
More details here: http://intelcad.com/pages/autocad/index.htm
daveyboyd
11th Jan 2008, 11:37 am
Cheers fellas, its funny but you would imagine that it would be able to cope with such stuff. A lot of my work needs to be coordinated as it is structural steel, I will have to make sure to create near 0,0 in future or I might have a few hairy steel errectors on my case!
Ta
jc4urcs
11th Jan 2008, 01:52 pm
it alway happen to me...specially when someone pass u the file to carry on....
what i do all the time is... set up a grid and put it at the centre of the drawing,and then check every single line b4 i do anything.... if not, then it all end up funny dimension and need spend more time to correct it; i also set the unit on 0.00 not 0.000000000
jc4urcs
11th Jan 2008, 01:57 pm
by the way i just spend 2 weeks to correct 1 ground floor plan because i havent check b4 i carry on at the beginning......and i still got 4 plan + sectin and elevation to do... poor me
rkmcswain
11th Jan 2008, 02:11 pm
It shouldn't be a problem unless your X or Y values get into the 10's of millions.
rkmcswain
11th Jan 2008, 02:12 pm
i also set the unit on 0.00 not 0.000000000
The setting in UNITS is purely for display purposes and has no effect on the internal calculations.
ReMark
11th Jan 2008, 06:17 pm
In structural steel, what kinds of tolerances are you normally dealing with?
Secretagdan
11th Jan 2008, 08:26 pm
I am the same way with the accuracy, but in rality, I do not belive that under most circumstances a person would ever be able to be that accurate, unless you are working with a plasma, or water jet machine in the end.
AVITWeb
9th Apr 2008, 03:27 am
I have begun to see this problem creep up again lately...But now its much worse. I offset a line 6'-8"...I then tell AutoCAD to move a block so that it's top most line is 6'-8" from the same line that I had offset, and, what do you know, the top of the block DOES NOT line up with the line I offset to 6'-8" above the first one. And It doesnt take a whole lot of zooming in to see that it is off. With my accuracy set to 1/256" AutoCAD measures this as being 0". But I will bet if I changed to inches, I would get something like 80.000000002.
Has anyone found out why this is happening or how to fix it...It is affecting our SNAPS and making quite a bit of a problem for us.
Thanks
Alan Cullen
9th Apr 2008, 04:28 am
Just a quick stab in the dark, but as soon as you said it is affecting your SNAPS, I immediately thought about VIEWRES. If you do a REGEN before snapping, it should work.
Check the value of your VIEWRES setting, and set it to something high. Mine is set to 10,000.
CarlB
9th Apr 2008, 08:22 am
Well, are you working with large coordinates, for which there was an explanation earlier in this thread?
Viewres and Regen should have nothing to do with snaps or acccuracy, just the display.
chrisdarmanin
9th Apr 2008, 10:29 am
i couldnt replicate the accuracy problem
take a look at this site; it says acad is EXTREMELY accurate!
http://www.intelcad.com/pages/autocad/index.htm
rkmcswain
9th Apr 2008, 05:24 pm
take a look at this site; it says acad is EXTREMELY accurate!
http://www.intelcad.com/pages/autocad/index.htm
It is, unless you are working in an area of the drawing where the coordinates are in the 10's of millions or larger - which is frequently done by civil folks working in SP coordinates.
AVITWeb
9th Apr 2008, 06:22 pm
Just a quick stab in the dark, but as soon as you said it is affecting your SNAPS, I immediately thought about VIEWRES. If you do a REGEN before snapping, it should work.
Check the value of your VIEWRES setting, and set it to something high. Mine is set to 10,000.
I did bump mine up to 20000 but afterwards there was no change...but low and behold, this morning i could not duplicate the problem for one of our support guys....
However, that being said, there is a drawing that contains many unit plans. When zoomed in all the way, you can see where they are not snapped together properly (1/256" off). We snap them to the right spot, save and exit. Next day, we open the drawing and they are all off again.
rkmcswain
10th Apr 2008, 12:23 am
I did bump mine up to 20000 but afterwards there was no change...but low and behold, this morning i could not duplicate the problem for one of our support guys.... As mentioned by CarlB, Viewres has nothing to do with drawing accuracy. Everything in an AutoCAD drawing is stored with the same accuracy (15 significant digits).
However, that being said, there is a drawing that contains many unit plans. When zoomed in all the way, you can see where they are not snapped together properly (1/256" off). We snap them to the right spot, save and exit. Next day, we open the drawing and they are all off again.
Just because something "LOOKS" like it's off doesn't mean that it is. If in question, measure the distance between two objects, don't trust the screen display. And remember the UNITS setting. If you are only showing 4 decimal places, and something is 0.00001 units apart, you won't see it.
Crank your units up to 8 places to see everything [well, almost everything... it's possible to use lisp/vba to see ALL the decimal places (15+).] --- but remember the units setting has no effect on placing or editing objects - it's only for display purposes.
Alan Cullen
10th Apr 2008, 12:57 am
Just to clarify, what I said was not that VIEWRES had anything to do with accuracy, but had a lot to do with display.
So, if VIEWRES is set to a low value, and you zoom in on an arc, the arc does not show up where it accually is. But, when you snap to the arc, the snap marker does snap to the correct location of the arc, and hence appears to be snapping to the wrong place, because it snaps off where the arc appears to be.
Please read my response with a little more care.
rkmcswain
11th Apr 2008, 01:20 pm
Please read my response with a little more care.
Ok, let's read it again.
...but as soon as you said it is affecting your SNAPS, I immediately thought about VIEWRES. If you do a REGEN before snapping, it should work
It appeared to me that you are saying that by altering the VIEWRES setting and performing a REGEN, this has some impact on SNAPS (which I presume you really mean Object Snap) - which it doesn't.
Alan Cullen
11th Apr 2008, 01:35 pm
Each to his own. You stick to your view, I will stick to mine. You obviously haven't come across this problem before. But then again, that is what I would have expected.
Alan Cullen
11th Apr 2008, 01:37 pm
If McS wants to keep going on about this, he can do so on his own.
rkmcswain
11th Apr 2008, 05:29 pm
Each to his own. You stick to your view, I will stick to mine. You obviously haven't come across this problem before.
See attached image.
This is the SAME arc and line in two tiled viewports. No matter which viewport you are in, if you SNAP to the INTersection of the line and arc, you will get the SAME POINT.
In summary, VIEWRES is not "affecting your SNAPS" (as you wrote on 8th Apr 2008 10:28 pm)
But then again, that is what I would have expected. Why would you expect that?
Powered by vBulletin™ Version 4.1.2 Copyright © 2012 vBulletin Solutions, Inc. All rights reserved.