Odd sized results - different from design

Hey guys, wondering if anyone can see what I’m missing here, I’m sure it’s something simple.

I’m making boards like the image below out of 19mm hardwood, and getting different sized results from the CAD/CAM setup. I know being hardwood there may be deflection, but it seems like a lot, and uneven throughout the job.

The grain of the board is running sideways to the image above if that makes a difference.

I’m using the DeWalt router, with the aluminium flat bar top and bottom siffening the X-axis, and get close to zero chatter most of the time. I’m using a 3.175mm Amana Tool downcut end mill with a 1/4"shank. Using Fusion 360 for the CAD/CAM setup

The dimensions should be
Full Depth Contour/Profile Cuts:
Height: 170mm
Width: 250mm
Pocket Operation (2.5mm deep)
Circular Pockets: 62.5mm diameter

I’m running 1mm DOC @ 2200mm/m feed rate.

The measurements I’m getting of the finished piece are
Full Depth Contour/Profile Cuts:
Height: 172mm
Width: 251mm (pretty close!)
Pocket Operation (2.5mm deep)
Circular Pockets: 59.2mm diameter

Does anyone know why it’s off, especially the circles? I went back into the CAM in Fusion 360 and set the pockets to “Stock To Leave” radial stock to -0.5mm to make the results a little bigger, but technically that should be producing pockets with a diameter of 63.5mm and I’m only getting 61.25mm so I’m still losing over 2mm somewhere…

Could be a few things.

  1. Loss of steps, possibly from a loose/tight belt, loose pulley screw, loose/tight V-wheels, or pots that need to be adjusted.
  2. Improper calibration of steps/mm.
  3. Improper bit measurement, such as assuming a 3.175mm bit is actually 3.175mm instead of measuring and learning it’s actually 2.9mm.

I’d start by taking the bit out and measuring it as accurately as possible, since the bits are typically smaller than the stated size due to the manufacturing process. Then give the machine a once-over to check the items in #1. If that’s fine, put a bit with a point in and check its movement against the most accurate scale you have, to verify the calibration is accurate.