Keyway Cutter design and project


@John_Bump posted a video of a keyway cutter mounted to a lathe toolpost that got me thinking. And modeling. I really should have one in my shop, and I enjoy making tools anyway.

I had just accidentally purchased (don’t ask…) an extra #3 MT quick change tool holder for the toolpost on my lathe, and thought that maybe I could use it as the basis of my tool. Here’s an animation of what I came up with:


The basic idea is a shaft with interchangeable bits at the end. The bits can have different sizes and shapes for different keyways and materials; very little rake for brass, somewhat aggressive rake for aluminum, moderate rake for steel. A simple linkage to push the shaft through a sleeve in the toolholder, and the cutting edge of the bit lined up with the center of the shaft. After learning a lot of FreeCAD, particularly how to use Assembly4 for animation so that I could test that the range of motion would work, I made some drawings and got to work.


I made my first part, the sleeve, out of 1117 cold-rolled steel. (1117 is a sulfur-bearing steel that machines fairly well; I’d rather use sulfur-bearing than lead-bearing steel like the ubiquitous 12L14 because lead.)

This took me two attempts.

The first try was frustrating. I started by cutting a very good #3 morse taper on my first ever attempt copying the tailstock taper, and was really proud of myself. Then I tried using a 7/64 aircraft drill as a pilot drill, and managed to bury the tip and break off the first inch or so of it right in the middle of the part. I spent way too much time trying to rescue that part before deciding that if I’d cut a good MT#3 taper once, I could do it again, and that I should start over again with another chunk of stock.

For my second attempt, I used a 3/16" x 6" aircraft drill and tried carriage drilling. I drilled 1/10" per pass between cleaning out the flutes. I used power feed for the carriage at the slowest feed (1117 doesn’t work-harden as far as I know).

Sadly, the bit wandered, and the exit hole was slightly off center after traveling nearly 5" through the stock. I didn’t measure, but maybe 1mm?

I decided that was likely to work well enough, and invested the time to cut another #3 MT taper. Again it was a close fit, so I guess I figured this out. In retrospect, it was a mistake to cut it early as it limited where I could hold the part. I should have skimmed the surface to make a good concentric surface, and cut the taper after I had a straight hole. Live and learn!

To fix the wandering hole, I used a two-flute ⅜" end mill in each end to make a centered hole about 1" deep. Then I used a ⅜" drill to follow the hole from one direction, and I could see a much more accurate transition where the drill met the milled hole from the other direction. Maybe 0.1-0.2mm? Then I drilled 7/16", reamed 7/16, and finally had a straight enough hole. I drilled with a 39/64" (1/64" undersize for reaming ⅝") Silver & Deming bit, which is stiff and was also long enough to go straight through. I drilled 1cm between cleaning chips from the bit to avoid them packing.

Then I pulled out my H7 reamer, forgot to put a chamfer on the hole, and started to ream. After about 1cm, I pulled the reamer and saw and felt bad chatter marks. Yes, they really are as bad as they look:

I turned the work around, pulled out a chamfer bit, and tried to chamfer with the bit. That chattered too, so I switched up and used a good carbide boring bar to cut a 45° chamfer. I had to work out all the chatter with the boring bar, so I ended up with a comically oversized chamfer that you can see in the picture. Reaming from the chamfered end was smooth, and while I can still see and feel the chatter marks in the other end, the ⅝ drill rod is a good fit.

I had intended to put an oil groove inside that end, but discovered that I didn’t have a tool capable of putting a groove inside a ⅝" hole. The fit for the drill rod is loose enough that the oil grove probably isn’t necessary. I can spend a few hours to grind a new tool if I need to, but I won’t do that unless it’s a problem.

I put the sleeve in its tool block and put it in the mill to drill out the oil hole. In retrospect I should have done that before final reaming, but I did manage to clean up the hole anyway.

Sadly, the bearings on my lathe are going, so I’ll probably have to put completing this project on hold until I change the bearings. That’s likely to be its own substantial undertaking; no clear idea how long it will take.

The PDFs of my drawings, such as they are, and the realthunder FreeCAD files, are all available at:


I finally found a thread on hobby-machinist that explained why it wandered.


This rabbit hole is deep. With the spindle out, I’m starting to wonder about tearing out everything in the headstock for cleaning and possible complete set bearing replacement… This Grizzly was full of grit that I just couldn’t see clearly until I took the spindle out.

At least I’ve confirmed substantial wear to both sets of bearings, so I’m not tearing it apart for no reason.


This project is back in business!

I got some real help on hobby-machinist on replacing my spindle bearings.

Then to get back in business here, I started work on the shaft for the keyway cutter.

I grabbed my length of ⅝" W-1 drill rod, and realized that I needed to use the spider. But I remembered that during this process I had the idea that I could instead make a coupler for an ER40 collet and use it instead of a spider, so I made that part first.

It worked; I used a ⅝" collet in the chuck and a 16mm collet as a spider and pocketed the ⅝" rod nicely.


After looking at the animation full screen and zooming in a lot to get details right in small corners, I keep being surprised by how small this is. Somehow it feels much larger in my mind. I start to wonder whether the 40mm of stroke I designed for is enough.

I tried using Brownell Oxpho-Blue on the sleeve, and even though I cleaned it with acetone, it still went on looking blotchy. I can probably keep doing it more. I just wanted some contrast. Well, I guess blotchy is contrast! Here’s where I am so far.

Here’s my process and what I learned.

I found two bits of necessary information missing from the drawing for the shaft while I was making it, so I added those. It ended up looking like this.

Fitting the ⅝" shaft through an 8.5mm tall slot took some work. I thought it was a good idea to make a rounded edge, as shown in this detail:

This meant that I carefully side-milled it, 0.5mm per pass, conventional milling the whole way. The final conventional pass was 0.25mm to dimension, then I did a spring pass climb milling to improve the finish. I keep forgetting that light side-milling steel produces billions of twisty little razer-sharp steel splinters. After inserting one of those into my thumb, I got wise and put on thick nitrile gloves. It took a long time running my X feed as slow as it would go, drenching the mill with oil to carry away the slivers. (And cleaning up. Don’t get me started on cleaning up all those slivers.)

After I had milled it to dimension, it was a tight fit and hung up in places going through the 8.5mm tall tang slot at the back of the #3MT QCTP block I’m using. I used a grinding stone until the shaft was a smooth fit. I used non-drying prussian blue to identify high spots and ground those down. This was a nice fit, and is one of the good aspects of this build.

However, once I had done that, I discovered that the rounded approach jams in the block on the backstroke. I guess I’ll set up in a collet block and mill out that nice-looking feature that I won’t see anyway, and make it stop jamming.

That would have been a lot easier to mill if I’d just milled down the top, alternating sides, with a machinist’s jack under it.

The new design is simpler.

I used the QCTP block as part of the fixture for drilling, reaming, and tapping, to make sure the holes were precisely perpendicular to the flats.

I finally bought the zero-flute chamfer tool that Blondihacks recommends, and it sure did a nice job on the hole. This also shows the random scratches from honing these flats to get them to run smoothly.

I milled a 6mm flat for the set screw hole. After the pilot hole, I drilled it out #19 and tapped M5 with a spiral flute tap. I love spiral flute taps! They eject the chips so beautifully! Then a short M5 set screw to engage the bit.


To mill the fillets square, I had to come up with a fixture that would hold the existing flat faces to line up exactly with the faces in my square collet. I came up with this fixture:

Then after just a few minutes on each side, matching the existing height on the mill, followed by some filing and stoning, it now slides smoothly to the stop. It still leaves the set screw outside the opening to the sleeve, so that won’t bind and damage the sleeve. Lubricated with way oil, it runs very smooth. Success!

I think the next things I’ll work on are the handle assembly pieces. This is about half the piece count of the entire assembly, but it’s all straightforward lathe and mill work. Two brass bushings, two spacers that should be a press fit on one of the pins or glued in place to avoid parts falling all over the floor when I disassemble it, two mild steel bars, and the actual handle part that I hold, which I can freehand from pretty much whatever I find handy in my shop; I don’t feel like spending a 1" aluminum bar on this, and I have enough various scrap that I’ll just scrounge.

This section view shows more of the internal structure:

The washers bear on the bushings, which will run in .251 reamed holes. The bushings are threaded internally to hold the screws, one in each direction, which are intended to provide most of the strength of each pivot.

If I were smart, I’d buy some ¼" bronze for my bushings, but I have plenty of ¼" brass, and if I wear them out they will be easy to re-make.

Here’s the drawing I’ll be working from:

The “Tension Bar” in that drawing goes between the handle and the collar, so I might as well do it as the same time. It’s just two reamed holes in a length of the same mild steel stock as the handle bars.

I plan to redesign the collar a bit to make it easier to manufacture before I start working on it, though. 3D printing has given me some habits that are bad in the context of design for machining!


I finished the bars and made the handle bushings last night.

I have realized that my design of having two screws meet in the middle of the threaded bushings is a bad idea. I did it to avoid a screw sticking out of the bottom, but it’s going to put strain on the bushing and break. Especially the one connecting the tension bar to the top handle bars. So what I really want is a screw through to a nyloc nut, since I’ll want to be able to disassemble this, and to do that at least the screw through the shaft should be easy to remove. (I may choose to instead loctite the tension/handle joint; that I don’t intend to disassemble.)

On the other hand, threading the holes in the handle proper and using two short screws meet in the middle through it would make lot of sense. That’s where my hand is, and where low profile and no sharp thread edges makes a lot of sense.

The stackup now looks like this:

Showing the screws more clearly with hidden line view and removing the other parts:

I cut a chunk of scrapyard aluminum to make the handle. The only important dimensions are that the tang be at least as thick as the shaft tang but not much thicker, the tang be long enough for the handle bars to fit, and that the two tapped holes be 20mm apart to match the holes in the handle. Beyond that I’ll just freehand it until it feels good.


The hidden line view helped me be sure that the screws wouldn’t touch each other.

Note that for the two bushings, the washers bear on the bushings, so the screws are tight but the bars move freely.

I freestyled the handle. I used a SEAN/SEKN/SEKR insert face mill to square the cuts, then put it into the 4-jaw on the lathe. Indicated it in, center drilled for tailstock support, then used a HSS aluminum cutting tool I ground earlier (better than carbide for interrupted cuts) to round off the corners until it felt comfortably like a handle but still had flats left.

I chamfered the end (“because chamfers are what separate us from the animals”) and moved from the lathe to the mill. I used the same face mill to cut the ends down to 8.5mm across for the handle bars, then drilled two holes 20mm apart #19, tapped M5, to mount the handle bars to the handle. Broke all the hard edges, and it’s looking pretty good to me!

I think I’ll try making the collar from aluminum and see if that’s strong enough. If not, I do have some ½" steel that I can use, but aluminum will be quicker at least to start.

This is starting to resemble the final product.


It’s got to be so satisfying to not only CAD this up but then machine all the parts and have it work like the CAD drawing shows it to work. Great work!


I’m glad I didn’t start by trying to make the collar out of steel. It’s my first time using the new rotary table, and trying to fixture it and work out the moves in X, Y, and A (at least Z was fixed!) was tricky.

But the first part was more conventional. I laid out a few reference lines and drilled an initial hole.

Then a larger drlil bit to make room to finish with the boring head.

I forgot to take any pictures of enlarging the hole, but that’s boring anyway, so you probably don’t care. :crazy_face:

Then I superglued it to a piece of sacrificial aluminum plate from the scrapyard and managed to lift the rotary table onto the mill table without throwing my back out.

After using my center indicator to align the rotary table in X and Y on the mill table, I cut a quick mandrel on the lathe to center the part on the rotary table by referencing the spindle. I forgot to take a picture of that setup, but — spoiler alert — here’s what it looked like, but inserted into the part after cutting the outline.

This was where I was glad that the rotary table has the same slots as my my table. I started with two strap clamps, but had to reposition them without moving the part, and ended up alternating between several strap clamp arrangements as I cut.

I tried to work from the DRO but honestly found that it was easier to just work to the scribe lines. I decided to make the collar thicker because of the switch from steel to aluminum, but of course did that after cutting, so the end is a little narrower. That’s OK, it needs to bend.

When I released the last strap clamp, the superglue gave way, so I think I got pretty lucky that it held through the entire milling operation. That wasn’t a given!

Notice the flat at the right side; that’s from me changing the design after cutting the stock. Oops.

I forgot to take any pictures of the slotting operation. I used a 2.5mm saw with two cuts to make the 3.5mm slot up to the line you can see scribed halfway between the circle and the outboard end, and then I put it on the bandsaw to cut a smaller slit the rest of the way through.

Also lost track of pictures of drilling the hole, but I spotted it with a stubby 5mm, drilled through #19, end-milled to depth with a ¼" end mill, and tapped the remaining hole M5.

The rough oversize dimensions on the hot-rolled tension bar interfere with the 3.5mm slot, so I’ll need to finish that end of the tension bar to allow the collar to close properly.

But even with the collar wobbling slightly, it runs. I can tune it later.


Just adding a separate clamping screw helped but it wasn’t enough. It still wobbled.

I widened the clamping slot to about 3.5mm and the slot around the bars to about 4.5mm (freestyling) and ground the end of the tension bar down a bit for space, and it just barely fits. So I have updated the design to be a 4mm clamping slot and a 5mm slot for the bar.

It doesn’t wobble much now, but I can see it flex, so I’m likely to re-make this from proper 6061 instead of scrapyard mystery metal, using my larger design.

The oil hole does work to oil the shaft in use.


I removed the sleeve from the QCTP block, installed the collar on the sleeve, reattached the tension arm, and tightened everything up aligned, and the collar wobble is mostly gone. If I break this one, I’ll use the new design, but this will work for now.

The last bit is, well, the bit. I turned about an inch of O1 ⅝" rod down to 10mm (yay, mixed systems, my favorite), and faced it to length, then cut it off.

Then I went to cut the first milled features. I’m really enjoying the ER40 system, but less so my new ER40 collet blocks.

I took a break and watched Blondihacks’ latest video that had just dropped, Indicator Holder for your Mill - Let's make one! - YouTube which would probably have helped me do a better job of cutting the collar if she had released it earlier and I had watched it before cutting the collar!

I ended up using my 5C square collet block to hold the bit to cut the set screw flats on the 10mm section, and the reference flats for cutting working ends on the bit with the set screw flats hidden in a collet. See here the witness mark from testing-fitting:

The flats are cut on both sides so that the bit can be used either way:

Here, the blank bit is installed on the shaft:

This is running really nice.

I updated the drawings for the bit to make the tolerances clearer.

Now I have to take a breather and remember what my purpose was in making this in the first place, so that I can remember what shape to cut the bit, then harden it.

This rabbit hole was deeper than I intended.


I finally made a keyway cutter bit, for 5mm keys. I used a different geometry than I had originally intended, and I’m not sure that was a good idea. I used it to cut POM, and it didn’t cut flat through. I ended up setting the toolpost to aim in at an angle. I definitely had rigidity problems with the compound and cross slide somewhere; I haven’t figured out exactly where, as the movement is subtle but present.


I’m having real trouble cutting a 5mm keyway in aluminum; no surprise that it’s harder to cut than plastic. I can’t see what’s shifting, but something is; I’ll start cutting, and after a bit I can feel something shift, and then the bit doesn’t engage and I have to reset the toolpost and cross slide to engage the cut again.

So it looks like it is going to take a long time to actually cut that slot.

I’m not sure this design was a good one. I wish I could see what is moving so I could know what to change to fix it…

This project is a bust. I think it’s a combination of small amounts of bad implementation and large amounts of bad design. I did cut that 5mm keyway, but there were complications.

It turns out that the reamer I used for the sleeve wasn’t really a tight enough tolerance for the tool rod, and it wiggles. Not very much, but when you are trying to take off about .001" at a time, it really doesn’t take much.

There’s a bit of play in my compound. Not much, but trying to use this, every little bit matters. I should really make a block that can replace the compound for when I’m not using the compound, for additional stiffness.

Sometimes the QCTP shifted. I had put wrench flats on the custom top nut I made, so I locked the QCTP down harder using a wrench on the flats, which did improve things a little.

I was cutting a blind keyway, which means there is a hard stop at the end. This means that every cut ended with a shock to the handle. It broke at the attachment to the rod, at the weakest point, exactly where you would expect it to break if you look at the design. I didn’t expect these kinds of forces on it; I thought of .001" passes as something that should put light pressure on.

I locked that in place with shims and a clamp and used the carriage to finish making the cut with the spindle still locked (with aluminum cutoffs clamped to the outboard end of the spindle, since I don’t have a spindle break).

In the end, I discovered that I had locked the quick change tool post top nut down so hard that when I tried to unscrew it, the draw bar screwed out of the t-nut, and even clamping the draw bar in a vice, I can’t get the nut off the draw bar. So lots of failures to learn from. Fortunately I had an old QCTP, t-nut, and top nut that came with the lathe; I don’t like them but they are better than nothing.

So, I really think that before I cut more keyways, I need to go back to the drawing board.