Official Luthiers Forum!

Attempt #1 failed. Trying to advance the flexspline caused the two gears to pinch the outgoing flexspline teeth and prevent them from getting out of the way. It was obvious that they needed something closer to the triangular profile from the original Harmonic Drive patent, so the two rigid gears pinching will tend to squeeze the flexspline tooth out rather than holding it in by friction.

On a whim while thinking about Trevor's post, I decided to try reducing the tooth height even though it doesn't reduce the necessary eccentricity. It had a surprising effect on the engagement pattern of the teeth for this style. While I had previously reduced the width of the flexspline teeth to mostly eliminate tooth collisions in the 3D model, the shortened teeth weren't making contact at all. Here's a screenshot of the grabby teeth, which does still have slight collisions with the black one, despite the flexspline teeth being much narrower than the spaces between rigid gear teeth.

And here's what happens if you make them too short.

After much fiddling around, I came up with a tooth profile that has relatively sloped sides, good contact with almost no collisions, and the flexspline can still be machined with a 1/32" mill to give 1.5mm face width on each of the rigid gears.

However I did have to increase the overall width to get the sloped sides, so I had to go from 44 flexspline teeth to 38. Pretty low for metal, but it turns out this style of flexspline is a lot more flexy than the cup type since it only has to bend in one dimension rather than two, so I think I can get away with it.

Here are the machined parts, still just using aluminum for the flexspline while I test tooth profiles and get a feel for the stiffness.

Minimum wall thickness is 0.18mm, equivalent stiffness to 0.125mm steel, and feels about right. The previous aluminum was 0.14mm, so I don't think I'll have trouble machining that.

Unfortunately the very last instruction of all the CNC programs for the post was bad and rammed the bit into it. New function for cutting the square shaft with rounded corners, and forgot to add the tool radius. Thankfully the post is all that was destroyed, no damage to the machine or end mill. I need to buy some 1/4" round stock to make posts. It's a pain cutting strips from 1/4" sheet and milling them round. I was too lazy to do it again so I just machined the remaining length of the wrecked post, making it unusable as a tuner but able to test the gears.

The knob is 3D printed again since it's a throwaway to match this aluminum flexspline. I made the wave generator area wider for stiffness. And made the amusing mistake of forgetting to leave depth for the little step below the body gear, resulting in machining through the ceiling, which gives a nice window to view the tooth engagement area.

I thought this one was a failure too at first. I had a heck of a time just getting it put together. It needs some chamfers to help things slide away from eachother as they're pressed together. After getting it together, the teeth wouldn't move much. But after abusing it for a while it loosened up, and now turns quite nicely (aside from PLA still being very high friction). Maybe just needs a few micrometers more clearance.

So whenever I find myself in the area of the local machinist's candy store Metal By The Foot, I'll buy that 1/4" rod to make a proper post for it and see how it handles string tension. And the steel tube to make a proper flexspline.

One minor issue is that it turns backward. Body gear is 36 teeth, post gear 34. So it should be the post gear that has more teeth. This means that cycloidal design is probably backward too.

I'm pretty sure pancake is the winner of all the tuner styles. All metal, turns the right way, and least total material removal. And since the rigid gears are only 1.5mm face width, I can do higher tooth count if I buy 1/64" end mills. But I think the 38 tooth flexspline will be fine, and personally I prefer lower reduction anyway.

I think I will use the banjo peg style mount, and make them from 5/8" round bar with a small plastic spacer to seal between the wave generator and body after all. Aside from saving material, that will also eliminate the need for grease there which can leak out a bit. I came up with a potential efficient way to make the spacers, by using a 3D printer to squirt just the right amount of plastic into an open-top mold.

This style of tuner will have an interesting appearance since the wave generator is part of the key, and is basically flush against the back of the headstock with no body of the tuner visible.

Does anyone know the proper name for these things and where I can buy them?

I've always called them grommets, but I'm not sure if that's correct. I could make them myself easily enough from from 3/8" hex bar, but seems like a waste of time for something that mass production equipment already exists for.

I'll also need washers, and standard M8 is rather loose. Any search terms for tighter tolerance 8x14x1mm or 1.5mm thick stainless steel washers?

Here's something fun I discovered. A while back I tried making a wave generator out of stainless steel to see if my machine would tolerate it (it does, but it's right on the edge of being too much vibration so I may avoid it until I have a lathe to do the bulk removal). It was before I found out the eccentricity was wrong, so it's not usable for anything, but I tried heating it to see if stainless turns the same colors as regular steel, and indeed it does. So this is an alternative way of getting pretty colors instead of anodized titanium. The part that would be visible turned a sort of blue green, and inside there are some areas of lovely violet blue which would be great if I could get that even all over.


For now I'll probably stick to aluminum and brass except for the flexspline, and maybe the post since it doesn't require much removal.

I'd also like to come up with a more attractive shape for the keys. They're machined from above, so nothing can extend horizontally beyond the 3/4" circular shape of the base. The tools I have are 45 degree chamfer, 60 degree chamfer, 12mm sphere with 6mm shank (can undercut a bit, which I use for the string wrapping area on the post), and the 1/4" ball end mill that I use for shaping the transition at the base of the key so it would be nice to use that since it would save a tool change. I can also do small 30 degree chamfers using a 6mm drill bit like a mill (which I use on the end of the post). And there's a cheap set of corner rounding bits on ebay I could buy.

Here's my favorite I've come up with so far. Circle with the 1/4" ball, and a small 60 degree chamfer.

There's also the option of machining followed by hand carving.

I’ve always known them as hex tuner bushings or threaded tuner bushings. I keep a small stockpile in the workshop as they frequently work loose and can be easily lost. There are ebay vendors for small volumes:

https://www.ebay.com.au/itm/224491449451?mkcid=16&mkevt=1&mkrid=705-154756-20017-0&ssspo=HtgJN-8AQaq&sssrc=2349624&ssuid=&var=&widget_ver=artemis&media=COPY

Also plenty of sellers on AliExpress:

https://m.aliexpress.com/item/4001108689093.html?spm=a2g0n.productlist.0.0.459516cdZE4LUQ&browser_id=231f74707edf4e938fe0f82ddeb20215&aff_trace_key=59c77fa98f2a433d9797f5a36fadf91a-1680615902771-08775-UneMJZVf&aff_platform=msite&m_page_id=vgohesjucabsheoa1881ce147be1be5a5b20fdf13b&gclid=&pdp_npi=3%40dis%21AUD%214.32%213.45%21%21%21%21%21%402102160416841117060492641d0745%2110000014432900212%21sea%21AU%210&algo_pvid=b111c246-b99a-4fb9-a205-cb40719c3cc7&_universallink=1&m_page_id=vgohesjucabsheoa1881ce147be1be5a5b20fdf13b&gatewayAdapt=Pc2Msite

Turns out that aluminum flexspline still wasn't really meshing properly, hence why it was difficult to get it started. I came up with a better technique to simulate the flexspline bending, where instead of just scaling it to an ellipse, I fit it into the wave generator ellipse by selecting and rotating one tooth at a time, with the pivot point at the thinnest area between teeth where most of the bending takes place. It sure would be nice if I could automate the process of copy-pasting the teeth and doing the bending simulation, so I could edit the tooth profiles and have the resulting mesh pattern change in real time. But probably not worth the time it would take to write a program for it.

After more trial-and-error, I came to this profile

which ironically is very similar to what I tried back at the beginning, making the eccentricity much less than the difference in tooth count. Stationary gear is 34 teeth, post gear is 36 teeth, and flexspline is 36 teeth, but the eccentricity is comparable to 54:56 teeth. I don't yet understand why it works, but I machined it, and it does in fact work. It looks like I could even use it with the cup type flexspline since the 34 tooth (black line) gear meshes nicely with the flexspline even without the other one involved. The other one alone would produce no relative motion since it has the same tooth count as the flexspline.

Thanks to the lower tooth count, I can make the main body of the tuner from 5/8" round stock without the need for a spacer between it and the inside of the wave generator, giving very efficient use of material and low part count. Flexspline is machined from 3/4" OD, 1/2" ID steel tube, so no full width slotting and not much total material removal necessary. One 1/32" end mill broke during the vertical nibbling between teeth, but after changing the program to take two bites instead of just one, the rest of the tooth cutting proceeded smoothly. I bought a proper 4 flute coated 1/8" end mill to cut the outside diameter, which performed well. Cutoff was done using a 1mm thick version of the undercutting sawblade type bit I used for the acrylic and delrin flexsplines. Then sand it on my extra-fine dia-sharp stone to remove the remaining burrs. The minimum wall thickness on this one is 0.18mm, same as the aluminum one I made last time. Due to the lower eccentricity, it is flexible enough.

I tried machining a stainless steel post, but it vibrated excessively and slipped in the workholding. Then tried brass and that worked fine. Gears and wave generator are just aluminum, but I could certainly do them in brass too. The threads in the body for mounting and on the post for the m3 nylock nut are done with a thread mill now, and I love it. I've wanted one for a long time and should have bought it sooner. Ebay user eversharpno1.

I decided to try machining the key from the side first and then the top, which gives a lot more freedom for the shape. This lower tooth count design only needs 18mm wave generator OD, and 19mm was already a bit narrow for fitting the key entirely inside the wave generator diameter (most guitar tuner keys are 22-24mm wide). I went with a clover leaf shape for this one, but I don't like it enough to make it standard. Haven't come up with anything better yet. It still needs a nylock nut (probably will arrive tomorrow), but is usable without it.


I dug out my first guitar to test it, since most of my others have 1/4" holes for open geared tuners, and I didn't want to put a barb divot in Galaxy's headstock. I had to fight a giant brown recluse, and the poor neglected guitar was covered in dust and webs. Thankfully I taped a piece of paper over the soundhole so the inside is clean. The soundboard is still full of cracks, and the bridge is still half way detached, but it's nice that it can serve a purpose. The tuner works well, so I think this will be the final design. Although I will also offer them with flange mount and 3-on-a-plate (the fixed gears will all be integral to the plate, which is pretty cool). Not sure if it will look too weird having the keys sticking out sideways from a slot head...

If anyone wants to be a beta tester, let me know what combo of materials, key shape, and mounting style you'd like and I'll make you a set

Some more traditional key shape candidates. Still no clear winner...

I do a little bit of turning on a micro lathe and use free-turning brass for most projects. It cuts better than any other material.

Nickel silver (copper, zinc, nickel alloy) works almost as easily as brass, tarnishes more slowly, and polishes to a nice shiny silver color. It might also be a good choice for small parts.

Clay, I haven't tried turning nickel silver, however, a little searching indicated it is about 5x the cost of brass.

Back working on these again! My mill was defeated by the steel flexsplines before. I tried making a set and got through 4 of them successfully, but the 5th started vibrating badly enough that the outer surface wasn't smooth enough to use.

After that I tried making flexsplines out of aluminum, which seems viable. It will eventually fatigue and crack, but probably not for many thousands of turns.

However, there was one major problem: Grease leakage. The headstock I had one mounted to for a long time has oil stains in the wood now. So I need to seal it well and/or remove the grease.

I tried a delrin flexspline, which works with little to no lubricant. Also added delrin rings between the wave generator and body, and between body and post, eliminating the need for lubrication there. It feels much nicer now. Aluminum-on-aluminum is sort of gritty feeling even with grease. The only remaining aluminum-on-aluminum is the faces of the gears, which aren't under any pressure so they may not need lubrication at all. But they do gall if rubbed together by hand, and a bit of graphite powder smooths them out. I tried graphite between the flexspline and wave generator too, but I think it's better without it.

I've also redesigned the knobs. The wave generator is now machined from a tube (much faster), and the wood key is epoxied into it by the side grain rather than end grain. The clover key broke off from a small impact with pliers when fiddling with the nylock nut, so epoxy on bare end grain is no good. The new style gives a nicer looking and more durable metal edge around the wood too. The knob is held on by a screw through the length of it rather than the nylock nut slipped in through the side, so impacts are transferred to the screw and post rather than to the glue joint. It needs thread lock to keep the screw from spinning with the knob, but seems to work ok. I'm using a sort of gummy low-strength thread lock, so there's no concern for impacts breaking it loose.


This keystone shape turns out to be excellent, because I can machine it from the top and then carve the side tapers by hand, rather than having to do a separate setup for machining from the side. Saves time and adds a small handmade touch Knobs naturally come in pairs. First I carve some flats on the stick for the vise to grip, and then flip it over and use some 3D printed cauls to grip that knob and machine another one from the end that was being held before.


It does occur to me that using a plastic flexspline defeats the purpose of developing the pancake style in the first place. I could just go back to the original cup type, especially once I get my lathe finished up to do the heavy work on the posts. But I prefer the look of this style, and it makes more efficient use of material. It also has the option of doing 3-on-a-plate. I tried making one in brass and it's gorgeous. It has a sort of scaly pattern made by repeatedly stabbing with the end mill.

I'm not sure if I'll be able to make these tuners affordably enough to sell, but I am quite happy with the design now. I'll keep trying to improve the machining processes.

I've installed this first set on The Robin's Nest

Amazing!
So great to see your progress on this. Thanks for posting.

Those look great Kevin (??!!) The 3 on a plate looks fancier than premium stuff from Rogers, or Rubner.
I'm happy to make wooden pegs, and my own tailpieces, and bridges from chunks of wood for violins and violas!

This is one a whole different level.

The knobs have a funky modern feel.

I have no clue where Kevin came from. The only Kevin I've ever heard of is the kid on Home Alone!

_________________
Why be normal?

Hmm, have you met my brother somewhere? His name is Kevin, but he's not a member here. He plays synthesizers and jaw harps.

Oddly enough, it would be more difficult for me to give the plate a plain flat surface, because the milling cutter always leaves swirl marks that would have to be polished out. This pattern is good to go straight off the machine. Here's a closer shot of it.

Hopefully it will remain shiny gold forever. I coated it with shellac when I made it 5 months ago, and it seems to be working. The uncoated areas (gears and flat underside surface) are somewhat tarnished now, and shellacked areas still look new.

It would be fun to do flat surfaces with engraving like Rodgers too. And patterns engraved around the outside of the wave generator. But I should refrain from unnecessary complications for now... except for one, because the desire for anodized titanium wave generators was my motivation for designing them to be machined from a tube in the first place. I can buy 18mmx14mm titanium tube so the outside doesn't need machined at all and inside only needs a small amount of material removal, so my machine might be able to manage it. I really want the 90-95V green anodization color with purpleheart knobs for an emerald starling themed guitar. And those colors would look great against the brass plate too.

Another fun variant will be 10:1 for my steel string harp ukuleles. 4:1 Pegheds work ok, but higher ratio would be better. It turns out my dumb-luck tooth profile still works even if you significantly reduce the tooth count, and fewer teeth means smaller overall size while still using the same tools. 10:1 should be around 12mm outside diameter, versus 18mm for the 17:1 guitar size. And plastic flexspline means it can deal with the tighter bending radius that comes with lower tooth count.