It appears that I have posted some of the pictures included here elsewhere on the forum but not, I hope, in this thread.
The steering gear drawings are rather sketchy. The OD of the worm is given along with a lead of 1" and pitch of 1/2", ie, two start. The worm wheel is stated as having 22 teeth and an OD is given, but no other information. The worm wheel is clearly a helical gear rather than a worm wheel. Starting with the worm I assessed the options for machining. My lathe gearbox will cut a 2 tpi thread but not 1 tpi. I suspect that even 2 tpi would be pretty scary.
Another option was to use the 4th axis on the CNC mill. In theory this is fairly simple, the code:
G01 X100 A720 F200
should advance 100mm in X at the same time as rotating 720°, ie, two turns, around the A axis. The feedrate should be interpreted in mm/minute. Experiments with Mach3 showed that this was simply not the case for the feedrate. Then I discovered the 'radius compensation' option, which I though would compensate for changes of the diameter being machined. For the same numbers in the G-code the distance actually travelled increases as the diameter increases. Wrong again, experiments gave numbers that made no sense. Then I found some information on a forum implying that the radius compensation was for the axis of rotation being different to the axis of the machine. But in my case the axes are co-incident; if you set the value to zero it disables the function, so you need to set it to a small value, say 0.0001.
After more experimentation I thought I'd come up with some feedrate numbers that were sensible.
It was now pointed out to me that a two start worm is not prototypical, and it will backdrive, in other words the worm wheel will drive the worm if torque is applied. Back to the drawing board. A worm wheel will backdrive if the tangent of the helix angle is greater than the coefficient of friction between the parts. That's the theory anyway, but is highly dependent upon surface finish, lubrication and the amount of vibration present. I changed the design to a single start worm with a lead and pitch of 0.5". Flushed with success I re-wrote the G-code. I drew up the thread form 20 times size on graph paper and plotted the positions needed for roughing with a 6mm and then 4mm endmill, and finally a tapered endmill with an included angle of 30°. Of course normal Acme threads are 29° included, but no cutter was available for that value, and since I am making both gears, it doesn't matter. Good luck to any rivet counter who spots it.
I then machined a trial worm in Delrin. It looked great, except that it was the wrong hand thread.
Still, simple enough to change, just a minus sign in the G-code. Then I cut a couple of worms in steel. Oddly there was some chatter using the HSS tapered endmill and the finish wasn't as good as I expected. Slowly it dawned that the rotary axis hadn't slowed down for each successive tool as coded, it was just running flat out. Cue more experimentation.
After some hours of trials I contacted the manufacturer of my CNC mill. They said that Mach3 was
bleep-bleep in this area and I should use G93, inverse time feedrate. This is an odd one. The time taken to execute the line of code is the inverse of the value given, in minutes. So a value of 1 will execute in 1 minute, but a value of 0.2 will take 5 minutes. More experiments showed that this worked, with a small but consistent time offset, a few seconds. Having sorted all these issues out I could then machine the real worms. Here's a general shot of the setup with a 6mm cutter taking its roughing cuts:
And the finished worms with keyways cut:
Having made the worms I moved onto the worm wheel. I wanted this to be single enveloping, ie, the worm wheel wraps partly around the worm. Normally these are cut using a hob with the blank geared to the rotation of the hob to produce the correct number of teeth. I don't have facilities for this, so I resorted to free hobbing. This is where the worm wheel is gashed with embryo teeth using a standard involute gear cutter and the hob is then used to clean out and form the teeth. Since the worm wheel has been pre-gashed the hob can drive the worm wheel itself, at least that's the theory. Making the hob was simple, just a longer version of the worm. Here is the embryo hob before the teeth were cut. Material was silver steel:
The thread at the right is an odd one, 25mm OD and 20 tpi Whitworth to fit a Clarkson style milling chuck. As a precursor to machining the worm wheels from the supplied castings I made one in steel to test the process. This highlighted a number of areas where I needed to up my game. On the steel worm wheel I machined the curved throat on the OD by eye. This wasn't sufficient as there was some rubbing of the hob at maximum depth. For the proper worm wheels I filed a template from sheet steel and used this as a gauge to get a more accurate throat shape. To get the maximum depth I could gash the teeth to while allowing the hob to clean up I used CAD. I can't remember the value but it worked well. Except on the prototype steel worm wheel. I traced this to lack of care in centring the hob transversely, so it cut more on one side than the other. I took more care with the proper worm wheels. Of course I was keen to see if the technique worked or if the hob would disintegrate so I rushed the setup with the steel worm wheel. Starting the hobbing process is binary machining, you can't sneak up on using the hob, you just have to drop the clutch on the mill and stand back!
For gashing the worm wheel blank is set over by the helix angle of the worm, although it is not obvious in this picture:
For hobbing the worm wheel blank is set back to be perpendicular to the hob. Here is a picture of the hobbing of the worm wheel, notice the significant difference in tooth shape:
The material is cart iron so machining is done dry. After I had machined the steel worm wheel I noticed some faceting on the tooth faces. I eventually twigged that this is a function of the number of teeth per revolution of the hob. Essentially the hob cuts the tooth form as a series of flats. For the proper worm wheels I improved this faceting by moving the hob axially after full depth was achieved. The hob was moved one pitch of the worm, 0.5", in steps of 1/16" and at each step the worm wheel was allowed to make a complete revolution, driven by the hob. Although slight there was a definite change in the sound of the cut as the worm wheel completed a revolution.
Finally here are the worm and worm wheel installed on the engine:
Andrew