4th Axis with Harmonic Drive – Build Log

The shop has a full size CNC knee mill and currently under design is a CNC wood router for carving gunstocks.  Both of these machines need 4th axis capability.  And of course there is no money for any of that.  So this is my effort to build a low cost 4th axis with harmonic drive.


4th AxisAbove is the end result of the build.  Basically it is a standard commercial harmonic transmission with the input shaft replaced with a stepper motor and a four jaw chuck attached to the output shaft.


Harmonic TransmissionThe build started with a commercial Harmonic Transmission purchased off ebay.  This one cost about $130 but I have seen them as low as $60.


Harmonic DriveThe harmonic drive works by nutating a inner flexible spline connected to the output shaft around a outer rigid spline fixed to the housing.  An oval shaped eccentric (wave generator) is driven by the input shaft and the longer ends of the oval walk the flexible splines around the rigid ones.  The inner and outer splines have unequal number of spline teeth and that results in rotation of the flexible spline shaft with respect to fixed spline.  If the fixed spline has 200 teeth and the flexible shaft has 199 then every 200 revolutions of the oval eccentric result in 1 revolution of the output shaft.


Input Assy OffThe transmission needed disassembly so that the input shaft could be replaced with a stepper motor. The input shaft and endcap are on the right.


Inside the caseThis transmission was filled with old – very old, grease.  Some of them are filled with oil probably for high speed operation.  There are oil level windows in the side of the case .  I had not originally planned to take the whole thing apart but the old grease really needed replacing. Grease will be fine, in fact better than oil, for this slow speed application.


Clean HousingTook the out put shaft and guts out of the housing and cleaned it up. Old grease stinks.


Harmonic Cup and BearingsThis is the output shaft assembly.  The shaft itself is welded to the flexible spline cup on the right. Two bearings with a spacer are held onto the shaft with a clip ring.  The spacer has a groove machined into the OD.  This ring is used to preload one of the bearings.  A setscrew with a cone end fits into a threaded hole in the transmission case.  The cone of the setscrew pushes against the sloped side of the groove in the spacer and forces the bearing OD toward the end of the output shaft.  The clip rings rides against the housing and keeps the ID of the bearing from moving and thus the bearing is preloaded. Never seen that before.


Clean Input AssyThe input assembly has two bearings. One in the housing and one on the end of the shaft.  The one on the end of the shaft fits into a cavity inside the flexible spline cup.  Two bearings are needed as the input shaft usually has a pulley and is belt driven.  This bearing will not be needed with the stepper drive as no significant side forces exists with the direct coupled stepper motor.


Input Assy Exploded Input disassembled.  I planned on machining the endcap housing to mount the stepper motor. The two lobed eccentric wave generator is at top center.


Wave Generator EccentricEnlarged view of the wave generator.


Wave Generator ExplodedThe hole in the wave generator hub (far right) is larger than the OD of my stepper shaft.  I have to reduce the diameter of that hole to fit my stepper.

Cast part Endcap where stepper will be mounted.  At least this was the original plan.


OopsAs I was machining the endcap to accept the stepper motor I cranked down too hard on the vise that held it on the mill.  And it cracked.  Plan B is now to machine a new endcap.


Cover3CAD design of the new endcap.  Making a new one made for a better motor adapter anyway.  So I told myself.


New_cap_botttomMachined the new endcap on the CNC knee mill.


Modified Oddman HubThis is the wave generator hub,  as shown above.  The hole was too big for my stepper motor shaft so the ID needed to be reduced.  This was done by press fitting an aluminum rod into the hole and reboring the hole to the correct diameter.  This eliminated the keyway so a couple of holes for setscrews were drilled and threaded and set screws installed. The stepper motor shaft is not flatted so the original intent of the manufacturer was set screws with no shaft modifications.  Torque of the stepper is only 1.25 foot pounds so I think I am OK.


Harmonic Trans ModsPresto! All back together.  Sorry, in the heat of getting the transmission reassembled pictures were forgotten.  The sequence of reassembly was: Grease the bearings and slide the output shaft back into the housing.  Mount the stepper motor onto the new endcap. mount the wave generator hub onto the stepper shaft and assemble the wave generator components onto the hub.  Mount the endcap onto the housing.  This entails pushing the wave generator into the flexible spline cup that is attached to the output shaft.  The flexible spline is deformed into an oval during this push.  And you need to do this straight so that the splines engage correctly of either side of the cup.  I took it off and put it on several times to get things lined up.  I can’t think of any way to check the alignment so I just took it off and on until it seemed to be settled in.  If it is mis-aligned supposedly if turns rough. Well, a greased 200 to 1 gear reduction doesn’t turn with your fingers on the stepper shaft.   I put a knob on the other end of my dual output stepper motor (important that it be dual output!) and it seemed to turn ok but stiff.  Best guess is that it was OK.


Chuck Backplate DesignThe intent for the 4th axis is to have a 4 jaw chuck on the output end.  At the start there was no intention to pull the whole harmonic transmission apart.  And thus no intention to do any machining on the output shaft.  So a clamping backplate was designed for the chuck.  The hole in the backplate is a 0.001 inch interference with the shaft. The plan was to heat the backplate up until it slip on the shaft.  After it cooled that slit would be pulled tight with bolts across the slit.


Backplate SectionHow the bolts are mounted to clamp the plate at the slit.  Movement of the chuck mounting threaded holes in the backplate was calculated to be moot.


Chuck BackplateThe back plate was machined out of aluminum on the lathe and mill.  It is as thick as it is since the transmission shaft  is larger diameter than the through hole in the chuck. So the back of the chuck can’t get any closer to the transmission than the end of the shaft.  And I thought the more massive the better.


Modified Trans with Backing Plate Chuck backplate on shaft.  Good thing I had the slit.  After heating it the first time I slid it on the shaft only to discover I put it on backwards. The slit had to be wedged open to remove the backplate. Second time was a charm. 350 deg F opened up the interference fit so that it slid on easily.


Runout checkEndplay and radial runout was checked. Endplay is zero with the setscrew preloaded but also appears to be a linear function of how tight the setscrew is torqued.  Makes sense I guess. TIR was about 0.0015 inch on the backplate and 0.0025 inch on the chuck. Close enough and this is a 4 jaw chuck anyway.


Bearing PreloadThe setscrew in the center of the photo is the preload setscrew.  If I build another one of these I will make three more setscrew holes all at 90 deg to each other.  That will load the ring equally  around the perimeter.


4th AxisBuild is complete!  Now for some testing.


za_BackplotThe Excello CNC mill runs LinuxCNC.  A wraparound toolpath was manually written in G Code and simulated. Plot is  LinuxCNC backplot of first metal part cut.


Milling Test

4th axis mounted in milling vise.  Turns out this was a bad idea.

Here is a link to video of the 4th axis milling.     Milling Test


zb_First MetalFirst metal part turned out pretty good.  But I noticed some chatter is a couple of places in the groove.  So it was decided to do some tests to see how rigid the 4th axis chuck mount was.


zc_Deflection TestThe stiffness tests consisted of applying side loads to the chuck in question while measuring the load applied and the deflection result. A luggage weight checker was used to determine the side load and a test indicator the side deflection.


zd_DeflectionThe test showed that the 4th axis when mounted in the vise was very flexible.  This was how the part was machined and it is a wonder it turned out as good as it did.  There is not a straight side on the commercial housing and when you pushed on the side of the axis you could hear it it clunk.  This was also shown by the large amount of compliance measured and shown in the chart.  Mounting the 4th axis to the table made it almost twice as stiff.  Then running the preload on the setscrew up as high as an allen wrench could be pushed resulted in a stiffness three times the original vise mounted stiffness.  So mounting and preload are important to good results with this 4th axis.


Project complete.   Ahhhhhhhhhhh now some beer.



Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>