tan tan radius circles

[STACEROGERS1301]

I've struggled with this problem for years...

When creating circles using tan/tan/raduis method the circle often does not intersect with the object selected. I know this only happens when my drawing coordinates are big (around X=400,000 Y=6,400,000 from the origin) an is due to Autocads methods of handling big numbers.

What I wat to know is if there is a way of overcoming this without moving my objects closer to the origin. My coordinates are important as they relate to a horizontal datum.

I tried the UCS command and set an origin close to me objects. This had no effect.

[hwalker]

Are you sure it's not working properly.

As a quick test. I drew a line from o,o to 400000,6400000. I then drew another line at an arbitary angle from that.

I then did a tan tan radius circle of 500 and it sits at the junction of those two lines just fine.

[nestly2]

I remember this being an issue with AutoCAD. If I recall correctly, FILLET will place the arc correctly when CIRCLE>TTR fails... which seems odd since you'd think the arc center point would be calculated the same way.

[STACEROGERS1301]

Thanks for everyone's comments. I know filleting is an option but when filleting between lines and arcs or arcs and arc, sometimes acad seems to chose the wrong end of the arc, in which case TTR circles are my only option.

[STACEROGERS1301]

I've attached a drawing I'm working on which is a large roundabout. Lot's of big arcs etc. the entire dataset is situation around X=400,000m Y=6,400,000m from the origin.

When working with data this far from the origin is when TTR circles become unreliable. If I move all my drawing objects close to 0,0 then everything works fine. This has been an issue for many years across all versions of acad.

[rkmcswain]

STACEROGERS1301 wrote:
What I wat to know is if there is a way of overcoming this without moving my objects closer to the origin. My coordinates are important as they relate to a horizontal datum.

Not that I am aware of.

We draw curb geometry in state plane coordinates and occasionally run into this too (I was able to recreate it in your drawing). Quite frankly, other than the annoyance of not being able to find the INT point, we don't worry too much about it - it's 0.000001' in a drawing where the contractor can't even construct something ±2 or 3 inches sometimes...

[LeafRiders]

Ensure that your VIEWRES setting is at 2000 or more. Also, if you set your UCS to a point that is close to where you're working, I use UCS, enter, 3, enter, click left, and a point right on ortho... to ensure I'm still properly setting the x axis left to right. REGENALL a few times, save / close / reopen, and I often succeed. It's not 100%, but changing your ucs essentially is bringing your objects closer to "Origin" (in theory). It's about time they figure this one out.

[rkmcswain]

It's about time they figure this one out.

Nothing really to figure out. That's just the nature of floating point precision.Here are a couple of links that help explain.

http://autodesk.blogs.com/between_the_lines/2004/01/more_on_autocad.html

https://web.archive.org/web/20070710163032/http://www.intelcad.com/pages/autocad/index.htm

[nestly2]

I kinda disagree that it can't be figured out/fixed.

I believe the example where Fillet may yield proper results when Circle>TTR fails demonstrates that there's more than one way to calculate a result.  Additionally, AutoCAD is already "aware" of the pickpoint(s) on selected objects, so IMO it "could" also calculate based on the picked coords, not just the absolute origin, not dis-similar to the way UCSDETECT temporarily changes the Coordinate system based on the selection.