Surprisingly, I've never done this experiment myself.

Distortions are common, and will be more detectable at large scales. Even very small initial errors with home point position can cause significant distortions out at the edges.

A slightly trapezoidal shape can indicate a home point that is slightly too high or two low. It is possible that there is also some built-in distortions that come from the geometry that is created as the cords change the angle that they drape.

The internal mathematical model of the machine assumes that the cord always hangs from the point on the sprocket that is closest to the other sprocket - so the inside rim. Of course, this is rarely true in practice, and so the drawing will rarely be entirely true.

I'd be interested in hearing other people's experiments in high-accuracy Polargraphs.

sn

I just did an experiment like this, to see how accurately I could place a drawing onto my paper. I beleive it shrunk on the y axis about 1cm during printing, but remained true elsewhere, appropriately aligning with the paper.