Does radial runout change as the router heats up?

I designed some small boxes in Fusion 360. There is a top and bottom that fit together with the top overlapping a lip on the bottom. With a lot of testing, I was able to work out the gap between the inside of the top lip and the outside of the bottom lip so they fit snugly.

All was going well until I tried to step it up and make 9 boxes on one board (3x3). At just under 20 minutes per box, it takes almost 3 hours to do all 9. In order that they were milled out, the fit gets tighter and tighter. The last three are extremely tight and sometimes don’t fit at all.

It is like the bit is getting thinner over time. I checked the diameter before and after a run. At most, it was down by 0.0005”. This is within the tolerances of my calipers and not enough to explain the differences in fit. Is it possible that the radial runout improves as the router heats up on a long run, decreasing the effective radius of the bit over time?

I have been working on this one for a few months. Any insights would be appreciated.

That is a very interesting issue and something I’ve yet to encounter. Excuse my thought experiment here is that is not what you looking for.

If you pause carving and leave the router running until the bit completely cools you may be able to eliminate bit temp build-up it as the source. Maybe some compressed air blowing on the bit during that time to speed the process up.

Conversely, the router itself can introduce a temperature delta into the bit as it runs over time. That could be tested by running the router for an extended period prior to beginning the carve project. Seemingly, that should result in a tighter fit for all of the boxes.

The router, being air cooled, should reach its quiescent temperature fairly quickly but, as you seem to suspect, that cooling may not extend into the bearings themselves. I’m not certain how that could be effectively field tested and would probably only relate to your individual router anyway.

OK stop, now my old brain is hurting.

1 Like

I have two versions of the G-gode; a version for 1 box and a version for 9 boxes. I did make a single box first thing this morning with the same bit. It was OK.

I am toying with the idea of getting a 2.2kw spindle.

Just a thought - is the router body heating up, and creating a tighter fit in the mount?
Reducing vibration and play, as you progress and generate the most heat (eventual reaching that thermal equilibrium point), towards the last few.

Nothing would reduce in size, with heat… right? haha

1 Like

I would think the bearing parts would expand fairly consistently and the bearing itself wouldn’t increase the runout noticeably with heat. With a 100 degree temp change you’re looking at about .0006 change in a 1" diameter bearing. As @pwpacp suggested you could let the router run for a period of time to reach it’s operating temp before cutting the first part.

I have noticed different pieces of the same species of wood and direction of the grain can affect the precision of the cut, it’s the nature of working with dead tree carcass as a medium :wink: I run into this with fitting truss rods in guitar necks and end up test fitting and adjusting tool paths (in Fusion changing the stock to leave value) until I get a proper fit. FWIW I run a spindle.

1 Like

Thanks everyone for your input. The current tool path works on one box at a time. I am going to try again with ‘order by operation’. That way the temperature (or whatever it is) will be consistent when doing the finishing at the end. The gap will have to be adjusted, but hopefully it will be more consistent.

1 Like

consider a good cooldown period, and re-run the finish pass. See if any more material is removed, or the fitment is better.

Unfortunately, the parts have to be removed from the stock to check the fit. I guess I can try and measure them and then run a second finishing pass if needed.

Hey Martin,

by wanting to reproduce your precisely determined gap, you place a demand on precision from the machine. You should consider

  1. that a hand trim router is the cheapest type of milling motor you can ever get (and it’s not even allowed for use in a stationary machine).

  2. Due to the design of the Onefinity and the position of the trim router, the cutting edge of the cutter is unnecessarily far away from the attachment point on the Z-axis (and therefore also from the X-axis), resulting in unnecessarily high leverage force exerted on router/Z gantry assembly, and subsequently in more freedom to be diverted by the counterforces during milling. One would like to clamp the router closer to the bearing to at least reduce this distance a bit, but this is not possible because the router would then hit the top of the z-stepper’s cage:

If you think of switching to a serious milling motor aka spindle, be aware that with 65 mm stock motor mount, the problem with the motor not clearing the stepper will persist:

Image: A 65 mm spindle does not clear the stepper and therefore has to be slid much downwards inside the 65 mm stock motor mount in order to not bump the stepper cage with its back :frowning:


In terms of cutting dynamics at play, consider that with an 80mm spindle, you are also adding an additional 15mm of outward leverage to the equation. So while an 80mm spindle may unquestionably have substantially more power than anything that fits into the 65mm holder, the added 15mm of leverage is the down-side which will theoretically work against you. I use the term “theoretically”, because in actual use it may be of very little consequence.

Image courtesy of @mchoonmaker (Mike)


:smiley: :smiley: :smiley: The run with ‘order by operation’ selected just finished. The finishing passes that are critical to the gap are near the end of the process. Expecting the boxes to all be tight the code was adjusted to increase the gap by 0.002". All the boxes fit snugly as designed. None were too loose or to tight.

I think this means the run-out does “tighten up” as the router runs, decreasing the effective radius of the bit.


Glad to hear you found a work around for this. Probably a good tip for us all.

@Marty - In reading your initial post I was expecting the same and was going to suggest the order of operation route to put your most critical tolerances at the end after everything has warmed up and settled in.

In a past life I worked as a process and maintenance technician in a CNC machining environment for a company that machined primarily copper electrodes and nozzles with pretty high tolerances. There was one brand (Okuma) of CNC turning centers which required the operators to run production for 30 minutes straight in order to get the thing to warm up and stabilize in order to run consistent parts without having to do a tool offset every other part to maintain the -0.0/+0.0005 orifice ID. The Mori Seiki machines would just start up and make good parts right from the first piece.

@Machinist - I agree with the added leverage on the spindle mount with the extra weight of the spindle itself and wires/hoses coming off of the top of the longer spindle motor. This would all cause higher deflection on the gantry rails compared to the trim router on top of the higher cutting forces you could apply with the ability to increase feed rates and use larger bits. TONS of variables…

Lots of good thought processes in this thread.


Hey Kyle,

yes but with a 80 mm spindle (which clears the stepper) you don’t have to slide the spindle that much downwards like you have with the trim router or the 65 mm spindle shown above. And this way you can have the cutter much, much nearer to the attachment point on Z Gantry (and subsequently on X Axis) thus decreasing the leverage force and the freedom of deflection:

Image: 80 mm spindle: Clears the stepper

I don’t think the weight of the spindle and the wires/hoses is that relevant in comparison to the counterforces exerted by the milling itself.


I’m going to approach this from a different direction than most in the thread. What are the chances that The issue lies in geometry of the machine/workholding/setup such that the tighter fit is caused by the location of the part on the wasteboard?

This would be easy enough to test with the single part g-code that you have already proven but place your stock in the location of the difficult parts from the array g-code.

1 Like