Reasons for circular interpolation to cut out of round?

We bought a used CNC router from another shop to get us by on a particular plastic fab job for a few months until we can buy a nice new one. It's an old Larken System 50, 4'x5', with stepper motors. Larken Automation The System 50 is not shown on their current site. The axis guides are kinda lame. For the long axis (Y) the guide system is a round linear guide on the "left" and a flat plate on the "right" with ball bearing cam followers on the top and bottom. The stepper driving the Y axis is on the right, with a drive shaft going over to the left; rack and pinion on both sides. The X-axis gantry has the same stepper driving rack/pinion, and the guide system is a flat plate with a 90º vee on each edge, and cam followers on each side of the vees, top and bottom.

I'm working on cleaning the machine up and diagnosing/fixing issues. It needs a new spindle/router, as the current one has bad bearings, resulting in terrible surface finish. And with finger pressure, the cutter moves back and forth .005" in each direction... Also, I took a couple circular interpolation test cuts and they measure .01" larger in the Y-axis direction. Both CW and CCW, around a boss (pic attached.) 11,000 rpm, 55 ipm, 2-flute HSS, 1/2" thick polycarbonate, roughed with .01" left for finish pass, cuts started/ended at 45º, 1.03"dia and 1.13"dia. X-axis was right on size (as close as the bad surface finish would allow measuring.)

What are the possible causes for cutting out of round, and how would you suggest diagnosing?

(BTW, I hope nobody has a problem with this post because it's about a router. Some "precision" machine shops look down on "fab" work as if it's below them, but we focus on filling all of our customers' needs, and sometimes that's large pieces of plastic. I consider this no less CNC machining than a smaller part on a VMC, the machine is American made, not Chinese, and it's certainly not a hobby.)

I would check the X and Y leadscrew pitch and backlash. If the pitch is off, it is possible that the machine has lost a leadscrew compensation file. By going both ways around the boss, you have reduced the effect of backlash - but maybe not eliminated it.

try going very slowly and just draw the circle with a pencil on paper. See if the "interpolation" is off, or if it's hardware.

If you go slowly enough, You will see the lost motion.

Does the control have a lost motion compensation? (As BG mentioned above) Steppers often need the comp to deal with mechanical slop elsewhere.

CalG's point about lost motion compensation is valid, but not what I was talking about. I should have call it a leadscrew map rather than a compensation file. This is often used on machines that have less than top precision lead screws - particularly rolled screws. The screw pitch is not precise and can vary over the length. The manufacturer maps this error and puts it into the control. As long as there is no wear, the control can compensate for pitch error. Either pitch error or lost motion (backlash) can come into play after the lead screw has some wear.

FYI with that kinda guide setup, that might be it realisticly, Takes a lot of unworn slop free parts to have a accurate gantry set-up.

Above all replace the spindle, then do a similar cut test, with a nice sharp cutter! but at each 4 corners of the table. If there all out equally then its the Y axis. But You may well find that one side of the table is accurate far side is not. Due to slop in the Y axis drive from one side to the other.

Don't take this comment the wrong way, but that machine even when new would not split thou's, 3-4 thou is probaly a realistic accuracy expectation on small parts and your going to have to set-up some accurate comp tables in the control if you want good longer distance accuracy across the bed.