Holes have a taper

This seems very unlikely to be a LinuxCNC issue, but I'm stuck, and there are lots of sharp people in this forum. I'm trying to cut 20mm diameter holes 3/4" deep in MDF. The bottom of the hole is around 0.008" smaller than the top.

I have a 30x30" cnc router that I designed and built 10 years ago. It uses ball screws x and y. I'm changing my spoil board so it has spots for precision dogs. Cutting the dog holes has turned into a project. I'm getting decent accuracy at the top of the hole, but around 8 thousandths less at the bottom. The holes are about 20mm x 3/4" deep in MDF. I'm not saying this is a new problem, I haven't done precision holes before, so have no idea if this is new or old.

Hole cutting: With a 1/4" straight Whiteside carbide bit (2 flute), each level is cut with 50% stepover. The 3/4" depth is done in 4 layers, 0.2" each, leaving a 0.001" skin. Each layer is cut conventional, with a single additional pass on each layer (finish pass). After the hole is cut, there are two additional finish passes done at full depth, one conventional, one climb. The inside of the holes looks pretty clean.

My CAM software (DeskProto) does not use circular arcs, it uses a lot of x-y positions to cut a circle. They are very nice circles to the eye, no segmentation. I searched the gcode to find max/min x and y, they are correct for the size hole I want. I did a hole from gcode produced by Kiri Moto, same results

My first thought was bit deflection due to loading, but the final two finish passes should have very little load.
Another possibility is if my cutter was ground with a small taper. I don't measure anything like that, the flutes are parallel the full length (1" long flutes).

I've tried a 1/2" bit and a brand new 1/4" Whiteside, still the same sort of taper.

I'm at a loss, please advise.
Steve

i would check you slides on your z axis, sound like something loosened up on the lower part of the z axis. if the machine is 10 years old and shop built that would be the most like cause. 0.008 could just be a loose screw on your slide

My "rails" are 4" wide cold rolled steel, with skate ball bearings (12 per truck). No way would it be accurate enough for metal, but all I do is wood on this machine.

I checked all the bearings on Z, they all are well clamped to the cold rolled, could not budge them with finger pressure, so I think they are good.
I put a dial indicator on a pin in the router and moved the assembly by hand. I don't know how much pressure, two fingers maybe 10 pounds? This results in +/-10 thousandths motion at the bit. I'm thinking I have that much flex. The moving parts of the Z axis are the 1/4"x4" cold rolled plus a piece of 80/20 1.5 x 3" extrusion.

So maybe I'm stuck with the taper, unless I rebuild the Z axis.

not necessary try a finish pass see if it cleans us

I'm doing two full depth finish passes, one CW one CCW now. Also, I've moved the router, measured the hole, and then tried the same gcode on the same hole. At most it changes 0.001 or 0.002, it's pretty repetitive.

Couple things that may be worth trying:
1- cut another hole in a new piece on top of the target part. This will point out if there is a Z height thing going on. If the bottom of the hole is correct but the top is .008 too big, something is happening with Z.
2- cut a plug the same size as the hole. It is much easier to measure the plug than the hole. If it is parallel and correct, see if it fits in the hole. Could be simply a measurement thing if you are measuring the hole from the top.

aside form tool spring it is hard to imagine what could be wrong. Seems like the only other answer is something loose - spindle bearings/slide (but must be loose in x and y direction the exact same amount)

The cutter is not tapered I hope
Maybe your cutter is not as sharp at the tip as it is on the higher part.
The tip is seeing more cutting action and gets dull earlier.

I would try with another cutter again, to see if the cutter is involved in the problem.

I have cut holes in my old spoilboard, and also on a piece of 3/4" attached on top of the old spoilboard. Same result, so I don't think it's a taper in the Z axis. I'll do another check with an indicator, raising and lowering z while measuring side position.

I'm using a small telescoping gauge and micrometer to measure the holes, also I have a parallel plug gauge (i.e. the precision round dogs). The holes are clearly tapered.

Good ideas, I'll recheck Z for taper in the axis motion.

Timo, I did look at my cutter edge. Nothing I could see (no glints), but when I felt the edge there was a small difference in texture at the end. I had a new bit, still in it's tube and tried it. Same result.

I've improved the taper from about 8 thou to 5 thou by reducing acceleration and feed rate. I was at 50 ips^2 and changed to 20, feed from 100 ipm to 30 ipm. Further reductions did not help.
I also made 15 holes, all the same gcode. I didn't measure every hole, but they vary by a few thou. I've remeasured holes and get within one thou.
I think I'm as close as my machine will get. The 15 holes vary from too tight (which got better with multiple insertions) to OK, I just wish it was better. I'm going with what I've got.
Thanks for the ideas people!