Re: new harmonic drive v2.xxxxxx
On Thursday 06 May 2021 19:51:13 Michael Möller wrote:
resend, I credited the wrong contributor.
Hi,
I'm no math wizard, but the code you show seems confused. (but
probably due to your extracting the vital part from bigger code). A
few general points, which have more to do with "good Programming
Practice/Style" rather than the problem you have.
Firstly, and I'm probably exposing my ignorance here, why put a "$" in
front of variables? Totally not needed. The "special" variables like
$fn, $t and so on have the $ to show they are special, in the sense
they are part of the OpenSCAD core.
Several times your comment says variables do not work - I can not see
what didn't work. Variables work fine for me, especially the type that
just is declared at the top.
Line 27: "0.000000" : It does not become more accurate, with many
trailing zeros.
Thts is a carry over from writing my own gcode for the last 20 years. And
I'm fond of loops. I have a saw blade sharpener routine of 80 lines.
But sharpening a carbide tipped table saw blade is slow work, so it
takes 3 days to run, which takes around .003" off the face of every
tooth of an 80 tooth ATBF blade. Sharpest blade I ever used, it was
nearly 3 years old before it made the first burn mark on some cherry I
was buulding an entertainment center cabinet from.
LinuxCNC's IF/WHILE/REPEAT etc code compares for equality out to about 8
digits right of the decimal when determining equality.
That also is only about the 2nd or third bit of openscad I ever wrote and
there are w/o a doubt, some unused defines in it.
rotate [0,0, ... is just as accurate as
rotate[0.000000, 0.000000 .... I have not had rounding errors after
several years of use, so $fx = ([ 360.00000000 / 60.00000000 ] or
$fx = [ 360/60 ] gives the same result. BTW, you do not use this $fx
elsewhere. OpenSCAD will make everything float so you do not get
truncation of dividing "integers". But tastes differ, you might feel
that it looks cleaner (lines up easier, used to reading engineering
diagrams...) with "0.00000"
Your mail states you "design" in 10-times scale and then scale down.
It makes absolutely no difference to the final accuracy. It might make
things more readable, if your "design" bit was in nice simple
integers. I've done it that way myself.
I do not know if you have a high vacuum metal deposition 3D printer. I
have the ordinary plastic kind and accuracy is 0.1mm on a good day, so
at my maximum bed size 3 significant digits is enough for any
variable.
So I have observed, and changing printers is a no-no as I'm finding parts
made on printer D don't fit parts made on printer A, B, or C often by
more than 2 percent.
Line 10: Why did you make that an array? You overwrite it later. AFAIK
there is no need to worry about variable memory allocation that your
comment alludes to. (Yes, that variables are not "variables" in
OpenSCAD but more like "defines" is a troublesome concept if you are
used to procedural languages.
It has taken some getting used to.
OpenSCAD is a declarative language, it
describes a static object,not a run time execution)
Line 20: citing various odd values... if it is the result of a
calculator, put the calculation there, like 360/62 instead of 5.860...
Internally it will be even more precise than what you write.
I actually tried that, but the syntax diffs, whupped me.
Also, I have found that when I do things on a circle (like teeth on a
gear) I do NOT go from 0 to 360. 360 and 0 are the same position, i.e.
that tooth is created twice. I typically do for (i=[0:5:359])so in
your case for(i=[0:360/62:359]).
Good Point. Thanks.
Now as you want to make several gears, instead of 5+ comments for
optional code it just is: for ([i=[0:360/N:359]) where N is the
number of teeth desired
I'm learning.
Line 31: The magic "54.05". I've done this myself, tweaking something
until it fits. Then when I want to reuse the model 3 months later,
with a bigger gear, f.ex, I can't get it to work, because I can't
remember why the "54.05" was there, or what it represents. Here I can
surmise it is the offset from center to the edge of the ring. I tried
translate([InnerDia/2, 0, 0]) but that was too far. Then I realized
that I had to move it inward by the radius of the tooth.
yes, the tooth is center=true.
But that was
too far inwards. And then I realized that because you create a
triangle of a circle (with the $fn=3) the base of the triangle is
inside the circle and I got tired of working out where the edge of an
equilateral triangle is in terms of the radius when inscribed in a
circle. Nor did you, so you just nudged it around until it touched the
surface the way you wanted it. Works, but is hellishly hard to handle
when sizes change.
yes, dd1 and dd2 could be an automatic, The larger fits inside the output
cup, held in place by friction of a snug fit and some goop. The innerrd
could be expanded to divorce the ring from the teeth, but is sized to
just include the base of the tooth via the union. And a third difference
ring could be sized to just clip the inner (or outer) points of the
teeth, which in turn would allow a slightly fatter tooth that actually
meshed better. Such niceties will probably be added in a build about 2
more fine tunings hence to get the finals. As is, the tips of the teeth
are catching hell when the rotational loading is heavy.
Back to the drawing board. Here you too might find the solution to
your problem ... you do not want equilateral triangle teeth, but
isosceles - which you realized yourself already as you ask for how
much to squeeze the equilateral triangle, thus making it an
isosceles one. Using linear_extrude(height=ToothHeight) { Polygon= [
[-TW,0], [TW,0], 0,ToothHeight] ] ) ; } to generate a triangle of the
right shape instead of the cylinder ( ..$fn=3). Now that the baseline
of the triangle is known, the magic translation "54.05" can be
replaced by translate([InnerDia,0,0]). "TW" for ToothWidth is the
circumference of your inner surface divided by the number of teeth and
then - and that was your question, I think - scale it a tiny amount to
give you the clearance you want.
I attach my mangled version of your code. You can change ToothCount
and everything else falls into place. You can change the ToothHeight
and it will NOT affect ToothWidth. ToothWidth I have "tweaked" with
0.95 - remove it or change to suit.
Thank you very much.
Note that all changes are just "better programming", not lots of math
(apart from calculating the tooth width to fit to the circle). Your
fundamental problem is using triangle teeth. Involute tooth shapes are
better, they "roll", not "slide". Triangular teeth work if there is
enough clearance and you're not transmitting lots of force. It is easy
to change the polygon shape to something that approximates a involute
tooth, f.ex just make it a trapezoid (chop of the tip of triangle),
and that would be a large improvement.
Something I've considered, but not done yet.
Well, that escalated quickly. I just want to write a few notes, but
now I've spent (an enjoyable) 2 hours with this post. Hope it helps.
It does indeed.
Msquare
// new example, using a low ratio harmonic drive floating
// splined belt with 60 teeth.
// s/b 30/1 when finished.
InnerDia=111.4 ;
OuterDia=117.8;
ToothCount=58;
ToothHeight=6.95; // was 69.00000000 a blond one too big
ToothWidth = InnerDia*PI/ToothCount *0.95 ; // the "0.95" is a fudge
to give clearance
GearHeight=9.75 ;
// overall span of armature plus bearings equals 96.6mm's
// Ring part
difference() { // the outside smoother
cylinder(d=OuterDia, h=GearHeight, $fn=360); //outer is fussy
cylinder(d=InnerDia, h=GearHeight, $fn=180); //but the inner isn't
};
// Teeth
for( TP = [ 0: 360/ToothCount : 359 ]) {
//cylinder(h=11.75, d=8, center=true);
rotate([0,0, TP]) // position of tooth round ring
translate([InnerDia/2, 0, 0]) // mv x only
linear_extrude(height=GearHeight) {
polygon( [ [0,-ToothWidth/2],
[0,ToothWidth/2],
[-ToothHeight,0] ] ) ; }// make thick triangle
}; // switch h=9.500 for outer rings
I'll give this a shot, after setting the latest sizes or corrections
based on the half finished one on the build plate now.
On Thu, 6 May 2021 at 23:40, Gene Heskett gheskett@shentel.net wrote:
On Thursday 06 May 2021 15:25:57 Jeff Meeks wrote:
Are you using the involute gear code available with the BOSL2
package? https://github.com/revarbat/BOSL2/wiki/gears.scad
https://github.com/revarbat/BOSL2/wiki/gears.scad
This I downloaded, but there is no INSTALL or README, so I wasn't able to
make openscad AppImage aware of it. A question I've been meaning to ask.
Not that I know of. Its all self contained in the .scad file, one of
which I'll attach. This one generates a 60 tooth spline, but
changing the comment markers of the for loops will generate either a
62 or a 58 tooth version.
On 5/6/2021 7:12 AM, Gene Heskett wrote:
Greetings to all the math wizards here;
I have redesigned this to use the elipse of the internal rotor
as the shaping ring driver. Interposing 3 ball bearing
assemblies to make the tooth bearing ring into the elipse
desired
But I am seeing, when I hold the parts up to the light to gauge
the engagement, that the tipping of the teeth by 6 degrees a
tooth as a circle of 60 are composed, the effects on the
effective shape of the tooth, placing the tips of the teeth
faceing out at a wider spacing than those facing inward, than
the outer rings corresponding teeth facing inward, and this is
creating a condition where the teeth do not slide on their
mating faces, but the tips contact first, creating a high
pressure, rapid wear point. A wear point that will, as it wears
in, create undesirable backlash in the assembly. So the
equilateral triangles I'm using aren't quite the ideal shape.
In composing the tooth, I'm doing it at 10x x-y then scaling it
.1 x-y to generate the tooth. That "scale" statement then is
the ideal place to put the correction values. My question then
is how much correction would be mathematically correct given
that each tooth needs only the x tweek, the rotation is later.
For rotation angles of 6 degrees for 60 teeth, and 5.8064516129
degrees for the 62 tooth, thereby restoring the equalatteraly
that the triangles rotation destroys. FWIW. all rotations re
done on center=true
And equal but opposite would apply when the tooth is rotated 180
degrees to make its facing mate. So I'd need some +x for the
teeth facing out and some -x for the teeth facing in. But is it
that simple? IDK.
Thanks for any insight.
Thanks, Jeff Meeks
Take care and stay well Michael Möller.
Cheers, Gene Heskett
"There are four boxes to be used in defense of liberty:
soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
If we desire respect for the law, we must first make the law respectable.
- Louis D. Brandeis
Genes Web page http://geneslinuxbox.net:6309/gene
If you have the full BOSL2 distribution you should have README.md at the top
level, which contains installation instructions.
Gene Heskett wrote
Are you using the involute gear code available with the BOSL2
package? https://github.com/revarbat/BOSL2/wiki/gears.scad
<https://github.com/revarbat/BOSL2/wiki/gears.scad>
This I downloaded, but there is no INSTALL or README, so I wasn't able to
make openscad AppImage aware of it. A question I've been meaning to ask.
--
Sent from: http://forum.openscad.org/
On Thursday 06 May 2021 22:08:34 adrianv wrote:
If you have the full BOSL2 distribution you should have README.md at
the top level, which contains installation instructions.
I must have gotten the wrong stuff, its named MCAD-master
Gene Heskett wrote
Are you using the involute gear code available with the BOSL2
package? https://github.com/revarbat/BOSL2/wiki/gears.scad
<https://github.com/revarbat/BOSL2/wiki/gears.scad>
This I downloaded, but there is no INSTALL or README, so I wasn't
able to make openscad AppImage aware of it. A question I've been
meaning to ask.
--
Sent from: http://forum.openscad.org/
Cheers, Gene Heskett
"There are four boxes to be used in defense of liberty:
soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
If we desire respect for the law, we must first make the law respectable.
- Louis D. Brandeis
Genes Web page http://geneslinuxbox.net:6309/gene
On Thursday 06 May 2021 22:08:34 adrianv wrote:
If you have the full BOSL2 distribution you should have README.md at
the top level, which contains installation instructions.
Gene Heskett wrote
Are you using the involute gear code available with the BOSL2
package? https://github.com/revarbat/BOSL2/wiki/gears.scad
<https://github.com/revarbat/BOSL2/wiki/gears.scad>
This I downloaded, but there is no INSTALL or README, so I wasn't
able to make openscad AppImage aware of it. A question I've been
meaning to ask.
found it at the first link, put it away like it said, looks like it
worked. Thank you Adrian
--
Sent from: http://forum.openscad.org/
Cheers, Gene Heskett
"There are four boxes to be used in defense of liberty:
soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
If we desire respect for the law, we must first make the law respectable.
- Louis D. Brandeis
Genes Web page http://geneslinuxbox.net:6309/gene
Here's the link for BOSL2.
https://github.com/revarbat/BOSL2
Once you're there, click on 'Wiki" and bookmark it. There are a ton of great
explanations and examples.
Then download and install the code according to the README.md
Gene Heskett wrote
On Thursday 06 May 2021 22:08:34 adrianv wrote:
If you have the full BOSL2 distribution you should have README.md at
the top level, which contains installation instructions.
I must have gotten the wrong stuff, its named MCAD-master
--
Sent from: http://forum.openscad.org/
On Friday 07 May 2021 00:23:17 lar3ry wrote:
Here's the link for BOSL2.
https://github.com/revarbat/BOSL2
Once you're there, click on 'Wiki" and bookmark it. There are a ton of
great explanations and examples.
Then download and install the code according to the README.md
WoW! Just rounding is many many pages. Sweet. Thank you.
Gene Heskett wrote
On Thursday 06 May 2021 22:08:34 adrianv wrote:
If you have the full BOSL2 distribution you should have README.md
at the top level, which contains installation instructions.
I must have gotten the wrong stuff, its named MCAD-master
--
Sent from: http://forum.openscad.org/
Cheers, Gene Heskett
"There are four boxes to be used in defense of liberty:
soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
If we desire respect for the law, we must first make the law respectable.
- Louis D. Brandeis
Genes Web page http://geneslinuxbox.net:6309/gene