> In this sense, the Apollonius perpendicular bisector construction appears to be sensitive to the errors of compass placement. It looks like errors are reason

> In this sense, the Apollonius perpendicular bisector construction appears to be sensitive to the errors of compass placement. It looks like errors are reasonably under control. It looks like the input error is a little over 5% of the distance AB and so it's not surprising that after a few steps, the output error in the position of C is around 15-20%. If you started with input error under 1%, the output error would be under 4%. There's no large error blowup here.