Circle does not fit

[TheEnglishPatient]

What am I doing wrong? I am trying to fit a three point circle to [10,10], [30,15], [40,30]. QCAD does happily so with centre [14,36.5] and r = 26,800187. This has been independently verified (r = 26,800186566515) However, what appears on the display is a circle that seems slightly larger, and does not go through the points exactly when one zooms in. I can clamp a radial line to the circumference, which shows the same gap, although the numeric length is again given as exactly r. What am I doing wrong, or is there a numerical/scaling/display issue with circles? Could send you a screenshot if it helps. [win7 64 QCAD 3.0.5]
Thanks.

[J-J]

You picked my curiosity! I agree with your findings for ver 3.05. However it seems correct on ver 3.0!
JJ

[TheEnglishPatient]

You are (almost) right - in 3.0 the circle is much closer to the points, although again not exact. In my configuration it seems to be just a tad inside where it should be, but that's only visible with large zoom factors. (It's all a bit annoying as I am trying to illustrate a paper on circle fitting to survey data, where typically distances are large and angles are really small, so this effect somewhat ruins the whole point of the paper)

[clive]

Hello there
just tested this on win 7-63, with latest snapshot 3.0.7 and couldn't find any problems with this.

First circle placed through 3 points exactly, when zoomed in - no issues.

circle test 1.png (2.21 KiB) Viewed 741 times

New layer set up and test circle placed using radius and coordinates only.

circle test 2.png (13.64 KiB) Viewed 741 times

Fits perfectly

circle test 3.png (5.25 KiB) Viewed 741 times

[clive]

Amendment:

After more closer inspection,testing and 'extreamly close' zooming I can now reproduce this problem.

@TheEnglishPatient, further to your recent correspondence I have forwarded the details on to Andrew for him to have a closer look at this. Your observation regarding the graphical rendering of the file could be correct.

(Note -I have enlarged the point for visual purposes).

Test circle.png (35.5 KiB) Viewed 677 times

[clive]

Update:

I have re-tested this with an additional option - 'Render circular arcs precise' now available in the snapshot release 3.0.8.
in 3.0.8
Go to > Application preferences > Graphics view > Appearance - and tick Render circular arcs precise.
The result using the same above coordinates fitted perfectly !

Circle render test.png (110.22 KiB) Viewed 423 times

[TheEnglishPatient]

Clive,
thanks for the answer and improvement. I don't want to quibble with your use of 'perfectly' but if you look really close at point 40,30 the circle was previously +0.063255 to the SE from the point, it is now -0.00210853 to the NW. So, it's an improvement of an order of magnitude, but still not quite there. But if it's a rendering issue, it might be of course just my graphic card/monitor...? Again, the reason I am harping on about this is that I am working on some rail track surveying, where you have globally large distances and radii, but locally only very small angles and deviations from straight lines - so these things unfortunately do become important and visible. In any case, we are down from the 'unworkable' to the 'a little annoying' level, so thanks for that again.
Best Regards

[clive]

Snapshot release 3.0.9 (from customer download area) may fix this ?

In my test it looks good.

circle fit test.png (2.22 KiB) Viewed 275 times

[TheEnglishPatient]

Maybe - unfortunately I cannot test this any longer as my download has just expired. (Can I suggest that you put a crosshair on the centre of the circle instead of the yellow fill - it might it clearer)

[andrew]

TheEnglishPatient: I've sent you some direct download links via e-mail.

[clive]

Maybe - unfortunately I cannot test this any longer as my download has just expired. (Can I suggest that you put a crosshair on the centre of the circle instead of the yellow fill - it might it clearer)

Hows this, still looks good to me.

Circle fit 2.png (2.85 KiB) Viewed 229 times

[TheEnglishPatient]

Clive and Andrew,

Agree, I just had a quick look into my test files under 3.0.9 and it seems that this version has cured the problem now completely! Well done and thanks again for making a great program even better.