Profile Cutting Gear Wheels

Profile cutting of gear wheels does not seem to have many advocates on the Internet. In my opinion it has several advantages and is the way that nearly all gear wheels should be cut.

Mark Winder
October 2005. Updated 2007
The source code for the cog.pm module is available. It can be downloaded from cpan.
Comments and suggestions to
mail: Mark.Winder4 at btinternet.com

What Is Profile Cutting?

Profile cutting gear wheels is any method that involves a cutter moving in x and y plane. A good example is a traditional vertical cnc milling machine using a cylindrical cutter. These days there are many examples of other types of cutting machine, but all can work on a similar principle. This cannot be done with a non-cnc machine (well maybe with a special fitting!) because x and y slides need to be moved simultaniously.

Advantages of Form Cutting

The advantages of profile cutting are as follows:

Within very wide limits the wheel and tooth parameters are not dependent on the cutter but can be software generated.
The same software can be used to generate wheels in a variety of different materials, eg wood, and metal: historically different methods have been required for wood.
Unusual shapes are no problem, eg eleptical wheels.
Use of general purpose machinery means that very large wheels can be cut by those in possession of appropriate general purpose machines.

The disadvantages are as follows:

Thick wheels can pose a problem particularly when fine teeth are required. This is particularly the case when using milling machines.
Bosses on wheels may be time consuming or expensive if machined out of a solid block.
Unusual shapes can be a problem: eg worm, or bevels.

cog.pm

This document presents cog.pm, a Perl module that generates g code suitable for a CNC milling machine for form cutting small gear wheels out of sheet metal. The software is designed so that you may call the functions with what ever parameters you desire and G-code for the wheel will be generated.

The software currently covers horological tooth shapes, spoked wheels, and there is also a graham style escarpment and yoke.

Cog.pm uses one of two modules for generating output: G-code.pm generates G-code in a file, gdcode.pm generates a jpeg image of the outline that would be generated.

Parameters are provided for feed and multiple passes all of which can be controlled from a high level.

A function for stacking objects one on top of the other, eg a pinion on top of a wheel, is also provided. The stack routine automatically machines out the space around the smaller top object so that the lower object can be machined.

Some Examples

Below are some examples, with code, photographs of the finished objects, and comments. Naturally these examples have been made over time as the code evolves, so that I cannot promise absolutly that all examples are currently error free, although I believe they are and would like to hear from anyone one who thinks otherwise!

Example 1
Example 2
Example 3
Example 4

Example 1

Code

#!/usr/bin/perl

                                 
# This contains the cog cutting functions. 

use
cog;                         
# This contains the jpeg output functions

use
gdcode;                      
# and this the G-code output functions

use
G-code;                       
# Actually you only need one of these. 



$c=newcogpair cog(3.0,7,18);      # make a
pinion and a wheel module 3, 

                                 
# with 7 and 18 teeth.



$c->cutset({cuttersize=>0.125,passes=>3,passdepth=> -0.025
}); 



$c->{wheel}->hole(0.125); 

$c->{pinion}->hole(0.25); 



my $feed=3; 

#my $g=new
G-code("test.ngc",$feed,5);        #
for G-code out use this line

                 
# output file, feed rate, tool number for offset commands 

my $g=new gdcode("test.png",4.50 ,1300,1300);# for jpg use this line 

                 
# file, size in inches horizontally, x and y size in pixels

$g->ginit();                              




$c->{wheel}->cut($g,0,-0.75,0); # x,y,z. 

$c->{pinion}->cut($g,0,1.25,0);



$g->gend();                
# finalise graphics operations, write files etc.

This is about as simple as it gets. The code generates a large jpg image, on your screen you may need to zoom in as the lines are just a single pixel wide and some viewers do not show this at full size. You can change the resolution to suit. Below is a 200 x200 image produced by the code above.

Images

About The Images

The following points can be noted about the images:

There is a fairly obvious mistake in the lower wheel on east-most tooth. This while unfortunate, is nothing to do with any code!
On first appearances the holes in the center of the wheels do noyt appear to match those on the jpeg. In fact it seems to be missing on the lower wheel on the jpeg. This is because the jpeg takes no account of the size of the tool. The hole on the larger wheel, which is 1/8" is the same size as the cutter, so it is a simple plunge and would be represented by one pixel on the Jpeg. The hole on the pinion would show as a 1/8" hole on the Jpeg, since that how it has to move to generate a 1/4" hole with a 1/8" cutter.
The hole in the pinion is probably too big for this wheel!
So if thats the case, why then do the teeth clearly not show similar characteristics? The reason is that for holes, the cog.pm calculates its own tool path, and the inbuilt G-code cutter compensation is not used, while for the teeth it is. So that what is plotted (after turning on compensation) is exactly what you get. Sorry, you just have to know this!
Note in particular that these were the first two piieces I made, and are nearly as small as you can go with a cutter as large as this.
I do not advocate mettal gears this size (2.5" accross) !!!
The green lines in the jpeg are movements made with z > 0 , assumed to be non-cutting movements.
Note that the order of operations is considered significant. Normally, the exterior, at least, will be clamped. It is therefor better to drill holes first and then cut the teeth since otherwise you would be drilling in an otherwise unsecured piece of work. With other clamping methods (double sided tape, vacuum tables etc) this may not be an issue.

Gcode

The following GCODE is produced:. This is of course rather long, and in subsequesnt examples I do not intend to quote the code (It will be longer!) but it gives you an idea of the output.
Click to see the G-Code

Example 2

Code

#!/usr/bin/perl



use cog;

use gdcode; 

use G-code; 

$VERSION=0.02; 



# process options. Use getopts if you like. 

my %opt; 

$opt{allowed}='p'; 

@ARGV=grep { s/^-([$opt{allowed}])//?(($opt{$1}=1)&&0):1 }
@ARGV; # set options allowed in opt

grep { m/^-/ } @ARGV and die "Illegal option in @ARGV"; 



my ($g);

if ($opt{p})

{

  $g=new gdcode("test.png",1.5,300,300);   # this uses
the GD module to generate a .png file size 1950 pixels square. 

}

else

{

   my
$feed=3.0;                          
# feed rate for cut. 

   $g=new
G-code("test.ngc",$feed);        
# and this produces G-code in the file test.ngc. 

}



$c=newcogpair cog(1.6,7,16); # Creates a pair of meshing wheels, with
teeth module 1.6 (mm) one with 7 teeth, one with 16. 

$c->{wheel}->hole(0.16); 

$c->{wheel}->cutset(0.0625,4,-0.015); # cuttersize, passes,
passdepth 



$c->{wheel}->trepan(5,0.2,0.175,0.35,0.05,1.0,0.05);

#        $spoken, # number of spokes

#        $wos,     #
total width of spokes

#        $bsf,    # boss
radius in inches

#        $rsf,    # rim
size factor as proportion of pitch radius. 

#        $roe,    # radius
of window edge for trepan

#        $wobf,   # width at
base factor, > 1 for narrower at spoke rim 

#       $rot,    #
rotation factor for inside of spokes relative to outside 1= 1
revolution 0 to 0.2 are good. 



$c->{wheel}->bossindent(0.25,-0.01,1); # diameter of indent,
depth of indent, how many passes, feedrate



$g->ginit(); 

$c->{wheel}->cut($g,0.0,0.0,0.0); 

$g->gend();

Images

The caliper shown in the second picture has an upper scale in inches, lower in cm.

Notes

A few new features are introduced in this example. First, the use of a command line option to choose between graphical output and G-code. Pretty essential in practice, but absolutly nothing to do with the example in fact.

Second, I introduce the trepan function. This allows a wheel to be "lightened" by cutting out sections from the center of the wheel . The result is that the rim is supported by spokes. This function allows the number of spokes to be varied, and also allows the center of the wheel to be "roted" with respect to the rim. (Think spokes made of rubber.) This is really for artistic effect. But you could argue that it increases the strength of the wheel. You can also choose to have tapered spokes. Naturally, since you are going to cut this out with a circular tool, all window corners must be radiused at a radius greater than the tool radius.

Lastly there is a feature called bossindent that is implemented on this wheel. This creates a small circular indentaion which was to be used to aid in registration if fixing an external boss to the wheel. Not sure if this is useful or not.

Example 3

Code

#!/usr/bin/perl



use cog;

use gdcode; 

use G-code; 

$VERSION=0.02; 



my %opt; 

$opt{allowed}='p'; 

@ARGV=grep { s/^-([$opt{allowed}])//?(($opt{$1}=1)&&0):1 }
@ARGV;

grep { m/^-/ } @ARGV and die "Illegal option in @ARGV"; 



my ($g);

if ($opt{p})

{

  $g=new gdcode("test.png",1.25,800,800);   # this uses
the GD module to generate a .png file

}

else

{

   my
$feed=5.0;                         
# feed rate for cut. 

   $g=new
G-code("test.ngc",$feed);        
# and this produces G-code in the file test.ngc. 

}



$c=newcogpair cog(1.6,9,16); # Creates a pair of meshing wheels, with
teeth module 1.6 (mm) one with 9, one with 16 teeth. 

$c->{wheel}->cutset(0.0625,4,-0.0125); # cuttersize, passes,
passdepth 



$c->{pinion}->{passes}=8;             
# need more depth on pinion. 

$c->{pinion}->{fillet}=1; 



$s=new stack(0.125,4,-0.025,-0.005);   #
$cuttersize,$passes,$passdepth,$facedepth

$boss=new boss(0.125,8,-0.0125,0.13);

$boss->{name}='Jim';                  
# used in comments in G-code



$c->{pinion}->{name}='pinion';        
# used in comments 

$c->{wheel}->{name
}='wheel';          # used
in comments 



$s->add($boss,$c->{pinion},$c->{wheel});



               


$g->ginit(); 

$s->cut($g,0,0,0); 

$g->gend();

Images

Notes

This example introduces the concept of stacked objects. The idea is that objects can be defined and then stacked one on top of the other. The cut function for the stack object calls the individual cut functions for all objects in the stack, and also inserts extra objects that are necessary for the machining to go ahead. These objects extra objects are of two types. Firstly, if two or more objects are machined on top of each other, it is assumed (at the very least) that the top object is smaller than the lower one (other wise you couldnt machine it could you!) An anulus is machined outside the top object so that the lower object can be machined. Secondly this anulus is widened so that the top anulus is at least as wide as the anulus or all objects below it. This (nearly) enables wheels to be stacked on top of each other and machiened from a soliid piece of metal. In the exaple 3 items are stacked. The wheel from example 2, a pinion to fit it with as many leaves as I could fit in, and lastly a boss with the same diameter as the dedendum of the pinion. All this out of a solid piece of brass 1/4" thick. I use two cutters. A big one for roughing out the anulus's 1/8"" and 1/16"" for cutting the teeth. Note that except for the outermost cut, cuts cutting an annular space are at slightly different radii and appear double in the diagram above.

The second type of cut that is made here, I call a fillet. This is to get round the problem that when you cut a wheel "one sided" as it were, and leave it atached to a back plate, as is the case with the pinion, its no good just cutting out the profile, as this leaves some pieces attached to the wheel. These are removed by the lines that go into the gaps between the teeth. The algorithmn is very ad-hoc but seems to work ok-ish in the few cases I've tried. To activate the fillet function you need to set the fillet flag as is done in the pinion above. The remains of the fillets can be seen well in the second picture: it seems like I didnt go quite deep enough when I made this piece.

There are many objects all generating similar code, but crucially with different numbers. In order to debug this I have allowed objects to have names that can be used in the comments printed out in the G-code and that you can see as the metal is being cut. In the above, the wheel boss is called Jim.

Its not shown at any stage here, but a manual tool change at the right point is needed here. How you do this will depend on your hardware.

An attempt has been made to make cuts at different depths come out in different colours. If you stare very hard at the image you will see that some cuts are black and some brown. The blacker lines are supposed to be nearer the viewer. You'll notice that the two innermost circles in the image above are black for example. This doesnt work well, not sure why.

There is no particular reason for there being no hole in the center of this piece.

Example 4 - A Graham Escapement

Its worth pointing out that far more able people than me have written whole books on the graham escapement. I have concentrated on making it possible to make. This work is not completed yet and although I have made one escapement, so far it has not been possible to test this.

Code

Images

Notes

The wheel on the left was a first attempt. After making it (and yes there is one small mistake that probably makes the wheel unusable in any case!) I decided that it was built like a tank and that as angular inertia is very important to the graham escapement a second attempt was warranted. This had the following changes:

Much thinner rim.
Many more spokes to give more even support to the thinner rim
Tapered spokes to further decrease the angular momentum.
A slightly deeper boss indent, 0.015 instead of 0.01 inches.
Bigger boss at the wheel center. (Weight at the wheel center is not an issue so far as momentum is concerned. )
The Angle subtended at the tooth top was changed from 6 to 8 % of the tooth gap, making the tooth tops wider and less sharp. I doubt you can see that in the picture though.

The new wheel weighs about half the old one, and has a far lower moment of inertia.