3/4" diameter cutter in MDF with Shapeoko 3

2-flute carbide router bit. Wanting to make some some 1/4" deep scallops in a project.

Survey of the Carbide3D 3-flute feeds and speeds chart, the Shapeoko wiki materials page, and a CNC Cookbook page has surface speeds for MDF of ~1100 SFM, 1036 SFM, and 1500 SFM.

With a 3/4" bit at 10,000 rpm (slowest my Makita will go), that’s ~1964 SFM. Since I’m only going 1/4" deep, I won’t hit the full diameter, but instead about 0.707". Which works out to a surface speed of ~1851 SFM. I saw Winston using a 3/4" bowl bit in his walnut catch-all tray video, so I’m assuming the extra surface speed is not a big deal (unless the differences between walnut and MDF are somehow going to screw things up, such as the resin in the MDF).

This page from GDP Tooling recommends a 0.025" chipload for MDF with cutters 1/2" and larger:


Dumping 10k rpm, 2 flutes, and 0.025" chipload into my chipload spreadsheet yeilds a feedrate of 500 IPM (over 8 IPS)! From what I read / calculated, the max speed of the Shapeoko is just under 200 IPM (over 3 IPS) based on Grbl settings. Don’t think I’ll be trying that kind of feed either.

Can someone recommend a less ridonculous chipload for this size cutter on the Shapeoko? 1/4" cutters have chiploads in the 0.0016 to 0.0017" range for MDF on the Shapeoko. If I plug in that chipload for the 3/4" cutter, feed comes out to a much more reasonable-sounding 33 IPM. Doubling the chipload to 0.0033" doubles the feedrate to 66 IPM.

I’m tempted to start with the 33 IPM. Can anyone chime in on practical experience with 3/4" cutters in MDF on the Shapeoko?


I get the feeling that the chipload table from GDP was intended for large industrial machines like this Haas router:


15 hp spindle and max cutting feed of 800 IPM. I got to see one of its cousins (a GR-512) at a semi-local vendor years ago. It was quite impressive. We were having some large aluminum tooling plates machined there.


Yep, second that, the Shapeoko will make nasty noises if you run those feedrates.

I use a 19.1mm flat cutter for surfacing stock, when running this cutter I send it fast through the stock with a shallow DoC and the machine sounds happy, these feeds are MDF with about 1mm of DoC

These feeds are Maple with 3+mm DoC

You don’t want the cutter hanging around recutting and getting things hot. I’d suggest starting with whatever max RPM the cutter is rated for, 1500mm/min (60 imperial inches) feed rate and 2mm Doc (0.05" or 0.1"), listen to the cut and then progressively increase the Depth of Cut until the machine doesn’t sound happy any more or the cut quality starts to deteriorate.

Beware, these bit cutters at high speed fling the chips a long way, straight through the dust boot is not uncommon, so expect a messy workshop :wink:



Thanks for the response.

Your 19.1 mm cutter is very close to 3/4" at ~0.752". Your surface speed is about 3,937 SFM at 20,000 rpm, which is significantly higher than recommendations that I found for MDF.

A chipload of 0.05 mm (~0.002") sounds reasonable, as does 2 mm (~0.079) DOC. Winston’s video on MDF feeds and speeds recommends 0.060" [~1.5 mm] DOC for 1/8" and 1/4" tools. I was planning to do the 0.250" deep troughs in my project in 4 Z levels of 0.0625" each.

I’ve worked with MDF with traditional woodworking tools before and definitely agree on the dust/chip issue. The circular saw is especially notorious for being a mess-maker. With a speed of 6,200 rpm and a 7.25" blade diameter [184.15 mm], it’s running at a surface speed of ~11,768 SFM [~3,587 m/min]. Which makes your 3,937 SFM [1,200 m/min] surface speed seem fairly tame now. :slight_smile:

I need to get my dust collection / containment up to snuff before I start cutting MDF, but today I ran a couple test cuts in blue extruded polystyrene foam with the 3/4" ball bit. The short cut was done in 4 Z levels and the longer one at full 1/4" depth. I am pretty tickled with the way the polystyrene foam cuts on the machine and will surely be using it to do some lost-foam aluminum casting in the future.



This may help as you were referencing MDF. I used to get many issues cutting it but these days with better work-holding and tramming as well as more accurate belt tensioning, I now get a completely smooth and flat floor to my cuts. I have just cut a 150 x 150mm (6 x 6") pocket using a 2 flute straight cutter of 6.35mm (1/4"). The feed rate was 1500mm per min with a plunge speed of 500mm per minute. The cut depth was 3mm and the stepdown was .5mm with a stepover of 3.175mm. The penultimate image may be instructive.


Wow - that pocket in the fixture looks great! Like a factory surface.

Can you share what you’ve done in the way of belt tensioning?

Your feed and plunge rates work out to ~60 and ~20 ipm for those of us on the other side of the pond. 0.5 mm is ~0.020". What spindle speed is this running?


In addition to @jepho’s (future) input, here’s a thread describing a process to set belt tension in a “scientific” way

The “belt tone” method boils down to downloading the Gates app on your phone, move the gantry to one side, insert a 1/4" dowel under the middle of the belt, pluck it, adjust tension until reading ~100 to 120Hz

What’s important beyond absolute belt tension value is to ensure equal tension on the left and right Y belts.

I also started with slower RPMs in wood but certain people here kept urging me to turn up the RPM and send it…

It’s mostly about limiting cutting forces to be suitable for the rigidity of the machine so taking more cuts per second works out getting more material removed before hitting the machine limits, even if it might be a bit faster than recommended for the cutter.

Now that sounds interesting.

1 Like

I was never a great fan of the pull on the belts being applied via an ‘L’ shaped bracket. I could never apply enough tension because getting the bracket to meet the screw thread while under tension was not the easiest task. Given that belt tension and equality of tension between the belts is fundamental to cutting performance, I chose to modify my belt tensioning system. You can read about it at the following link.

I use the Gates app to tune the belts and since July 9th I have checked tensions each week. There has been some decay of the initial 135Hz values set but the belts are now tripping the Gates app as follows:-
Y1 = 124Hz - Y2 = 125Hz and X = 124Hz. I think that is a very creditable performance and all done by turning a single screw, which is of course much finer grained adjustment that just securing the screw to an ‘L’ shaped bracket. I also tend to think the straight belt path provides a more sensible application of mechanical force rather than twisting the bracket through 90° and then applying the force for tensioning. My selected values for tension of the original GT2 Gates belts supplied with my standard sized SO3 machine are anywhere between 100 ~ 150Hz.

I dislike the Gates app for its apparent sensitivity/inaccuracy but I take an average of 10 readings to get the reading I believe is the accurate one. Possibly an expensive frequency counter may be better but the resonant frequency of the belts when tensioned is in a low register and any extraneous sound can produce false readings. Potentially, a high quality microphone attached to the phone may help this issue.

I was running the spindle at 18,000 RPM and the cutter used was an Amana carbide 2 flute straight cut #43824. linked:

1 Like

I don’t use this method, @Julien. My extrusions are 600mm in length and I subdivided them thus: 160 - 280 - 160mm. This means that I hold the belt away from the extrusion by 1/4" at the 160 and the 440mm mark from one end of the extrusion. I use square cut pieces of hardwood that are 1/4" high from any face.

I also use a wooden plectrum I made for the purpose of consistently plucking the belt without adding or removing any resonances. I had started out by using 1/4" metal dowels but there were too many harmonic frequencies created so that the Gates app was often confused about the precise frequency that it was detecting.

explanatory pix:

This image shows the index marks using permanent sharpie. They are highlighted in red at 160 and 440mm from one end of the extrusion.

This image shows the highlighted wooden 1/4" blocks to raise the belt away from the extrusion and the highlighted plectrum so that the sounds are as pure as can be managed by this primitive method of tuning the belts and tensioning them correctly.

1 Like

Go on, you know you want one of these to really tune the belts properly


Have you tried a guitar tuner app to see if it’s any less temperamental?

1 Like

:grin: Thanks for the laugh, Liam. Much appreciated.

/ Jeff turns his SO3 into a CNC based telecaster, with the fingerboard being stolen from a Gibson SG. It would of course have to be fretless like the Fender Jazz bass guitar. Jaco fans watch out! :wink:

No, that is a great idea. I will look into it. With @Gerry using bass guitar machine heads and headstock for belt tuning, we could possibly get a rock band together.

I wonder if you have noticed how my home-made plectrum and belt tensioning supports look as if they may be carved from ebony. The wood is Canadian sharpie maple. This variety cannot be grown because the permanent sharpie colour has to be added during the manufacturing process.

It reminds me to ask whether you had considered buying genuine ebony sticks of say 1.5 inch square by 12 inch from Exotic Hardwoods, to fix your Yamaha Music Stand problem. It is likely to be cheap enough at this level and it could save a whole lot of work.

1 Like

I have had a look now, @LiamN. There are three basic types of tuner which are easily available and not too expensive. I’ve discounted those that plug into the guitar and those that work by sensing the vibrations. This leaves the handheld range and my specification was that it should be capable of detecting low frequencies (below my preferred range of 100 ~ 150Hz) via the sound made.

I picked A0 as the lowest frequency detectable (27.5000 Hz) so that I could be certain that my preferred belt frequency could be detected and C8 (4186.01Hz) to ensure that the tuner would cope. I have placed an order for this tuner (13 beer tokens). I think this idea is great and should be enabling insofar as being able to carry out the tuning/tensioning of Gates drive belts more accurately.

I do not know where the best tension lies… between not bending the pulley shafts of the stepper motors and not wearing the belts out through excessive stretching. Do you know?

The frequency range which I have selected as my tensioning range looks like this on a musical note tuner: I will let the community know once I have used the tuner as to whether it represents an improvement over the Gates app.

103.826 = G ♯ / A♭(2)
110Hz = A (2)
116.541 = A ♯ / B♭(2)
123.471 = B (2)
130.813 = C (2)
138.591 = C ♯/ D (3)
146.832 = D (3)

I used the vibration one in my experiment here:

I have three clip-on vibration tuners, one on a baroque lute, one on an acoustic guitar and a third on an electric. All are pretty much spot on.

(edit: this tuner is using A=415 because it’s used for baroque tuning which is a semitone lower than current A440)

(edit: the point was that if you clip a tuner on to the endplates it will almost certainly give you a reading if you pluck the belt)


I had wondered about that and suspending the belt on blocks. Having that beautiful headstock and machine head adjuster ensures that the clip-on type would be a good option.

Never having used a tuner before, I had thought that the extrusions would not necessarily transmit enough vibration to the tuner. I may yet be proven wrong but I went for a sound frequency detector in the hope that it would be a substantial improvement on the Gates app, which appears to lose the lowest frequencies.

Yes, noted. I really like the baroque period music; with Bach and Scarlatti among my favourite composers.

I think it’s likely the hollow aluminium will resonate properly, just like an aluminium bass.

If you clipped it to your Canadian sharpie-maple “bridge” then that would certainly pick up the vibrations.

1 Like


I had no idea ebony was available at that sort of price, I assumed you needed to be Pablo Escobar to afford any reasonable amount.

As for belt tension, there’s an updated table in this post

The simple summary is that I did not find any reason to believe that there was any reduction in backlash or vibration with higher belt tensions once the minimum working tension is achieved. I was not able to get the rating data for the Shapeoko steppers but pretty much all of the steppers in this size range have a load limit which is pretty close to the minimum tension, after that the shafts will eventually work harden and fracture.

The minimum (and maximum) tension is set by the stepper max torque, both sides of the belt need to stay in tension, that’s about 18lbs which comes out around 80N.


No worries. Looking at the rest of the Ebony stock, even Pablo Escobar may have winced a bit. Ebony is fearfully expensive but the amounts you need are probably going to be cheap enough. On the other hand, it is such a beautiful wood that I may yet purchase a small block for a project I have in mind.
sotto voce: I hope my wife is not reading this over my shoulder.

Great! Thank you very much for this gargantuan work, Liam. I have just read through it again. I would buy the book and give it a prominent place in my CNC library.
:+1: :+1:

That’s a very interesting finding. The arithmetic is a bit beyond me and it is a rather counter-intuitive conclusion. I would have expected that as the tension increases, the tendency to vibrate or deflect or highlight the inherent backlash would be reduced.

I understand nothing of metallurgy. Do you know anything about the number of cycles the shafts will tolerate before fracture? is it correct to state that the work hardening would be negatively impacted by increased belt tension? This would make the argument for having the belts in as relaxed a state as possible, commensurate with accurate movement.

The updated Range of Belt Tension table was very welcome. Keeping between 110 and 150Hz looks to be ideal for my standard sized SO3 machine. I also suspect that when combined with the Z plus, the fixture tooling plate and the modular vices, I possibly have obtained (by serendipity) the very best combination of features and rigidity set against its price, for the SO3 series.

Oh, OK. We may yet get to find out if my choice of belt tuner was bad.

No no, it’s fine. The CA-50 you chose is one of a long line of very good Korg tuners.

1 Like