"Squish" and "Brontus"

That is exceedingly cool.  I'm still trying to figure out exactly how
the parts fit together to form those nifty 3D shapes.  The lower-leg pic
is enticing, but I think it doesn't show all the individual pieces
involved... I'd love to see that.

How did you cut the plywood?  It looks very smooth, like a laser cutter
might do, except that I thought the edges of wood would be charred under
such conditions.  Did you have to treat the edges in any way (e.g.
sanding)?

Why did you choose plywood for this project over, say, lexan?


Another great design.  Where do you get aluminum laser-cut?  I thought
you needed to go to a water-jet machine to cut through that.

Thanks,
- Joe

Thanks. Yes, there is a second 'side' part missing from the foot. It is
the same as the left side without the mounting holes, but including the
Servo case hole. I will take more pictures when I build a second one.


Laser cutting for wood would char the edges which can then not be glued
anymore. This is done with a little CNC machine and a 1.2mm cylindrical
cutter. Sanding is needed, but no big deal. This is alos nice because I
can go only half the depth and create even better connections between
the pieces.


Short answer: I had it laying around. Long answer: lexan and friends are
more difficult to work with on a CNC machine because it melts at high
cutting speeds. It's also somewhat harder to glue. But I am actually
looking for a nice plastic that I can use for this kind of work.


Thanks. Olaf puts days and nights into his designs :-). I had it cut at
a friends machine shop who usually builds huge racks and enclosures.
Most shops do not like to laser cut aluminum at all, because it is so
reflective and high energy is needed. Since my buddy cuts up to 30mm
steel, 1mm aluminum was in range. In the long term, it supposedly ruins
the laser though. I have not looked into water-jet. Do you have heard
anything about it, goofd or bad?

Matthias

So the whole foot is just 5 pieces?  I'm having some trouble
understanding how it fits together.  Take the three long notches on each
side of the side part.  These mate with the roughly cross-shaped pieces
you've laid out next to them.  But why do the notches have an extra
semicircular extension on each end?  I was thinking that this was used
somehow to make the pieces interlock -- but now I suspect that these are
just an artifact of the CNC cutting process, and end up as (harmless)
empty space inside the part when it's assembled.  Can you clarify?

OK, maybe I do understand it now.  Very clever design.  I love the
upper-leg servo bracket.  How strong is it?

Also, if you don't mind my asking, what CAD program do you use to define
these parts, and what process do you use to develop them?  E.g., if
doing some funky new part like that upper leg, do you prototype it first
with foamboard, or go straight to the computer?  In short, how should I
go about developing the same skill?  :)


Hmm, if you're going to sand the edges anyway, I wonder if a laser
cutter would work just as well.  Once you've sanded the "char" off, it
should be just as good -- but then, I have no idea how deep the charring
goes.


Yes, lexan works nicely on a laser cutter, but I agree it's harder to
glue.  ABS might actually be better in that regard -- it can be both
glued, and (carefully) baked in order to fuse parts together (which I
understand is how they set the models at Legoland).


No, I've only heard it recommended as an alternative for materials too
tough for a laser.  Apparently the process is similar.  
<http://www.dcwaterjet.com/>  is one such service.

Best,
- Joe

Correct, Matthias said he used a 1.2mm diameter cutter. You can't
cut a square inside corner with that, so the cutter runs past.


The water jet has a fine abrasive mixture added. The machines
I've seen use a 35KW pump that pushes a 1mm jet of water at
700m/s (Mach 2!), and can cut through 150mm granite, and also
works well on glass. Solid carbide nozzles only last 20 minutes,
but there are diamond composite ones now that last many hours.
The advantage over lasers are obvious - no heat and no reflection
problems.

Matthias, you might like to fix some unfortunate spelling errors
on your pages, for example "bugger" where you meant "bigger" :-).
Nice work anyhow!

Yeah, the German version comes first and then I write the English text.
By that point, I am very impatient. Oh bugger, I'll spell check it. ;-)

Matthias

Yes, Clifford is right. These are the path of the cutter. If I wouldn't
make that little dip, my software would create two dips instead to make
sure that this corner is 100% milled out.


Thanks. Quite strong. I have yet to break one. The interlocking helps,
but I was alos very generous with the glue ;-)


I am using QCad. It's shareware and works well. 2D only, but since the
CNC process is 2 1/2 D only as well, I have no troubles.

I create the pieces first by sitting on the sofa for hours, thinking
;-). Then I start by drawing the six sides of a cigar box with
overlapping corners. Next, I remove what I don't need for the shape, for
example the round corners on the foot and the holes. When all that is
done, I do the interlocking, creating a gap on one part and removeing
everything that won't fit into that gap from the other part.

I did do some 3D modelling on the PC a while ago which really helps me
to envision the finished part. Brain-Foam, I guess.


I'll give it a try if I come across someone who'd do if for free. The
post-CNC sanding is very very minimal: just bput the part flat on the
paper and go back and forth twice. This removes the splinters from the
milling process. With the laser, I'd have to sand the shaped sides which
sound like a lot more work. But I'd give it a try.

Matthias

That makes sense.  So we could leave those little dips out if we were
targeting a laser cutter instead of a milling machine.


Fair enough.  Wood glue bonds very strongly.  Working with other
materials might require some experimentation to find the best bonding
method.


Great!  I recently did a survey of CAD apps available on my platform,
and settled on QCad as well.  It does everything I need, and does it
very efficiently once you learn how it works.  Are you willing to post
the QCad files?


Sounds like a good process.  I may have to try my hand at it myself.  I
wish I could think of an easier way to prototype it with foamboard,
though.  1/8" foamboard is cheap and reasonably strong, and can be cut
with a knife, but such cutting is rather tedious -- I'm not sure I'd
have the patience to cut all the edges in parts such as these.  Hmm,
maybe my first project should be a CNC machine good enough to cut
foamboard!


Yeah, I was thinking about that.  A custom modeling program which
assumes all 2D parts, but lets you assemble them in 3D, might not be too
hard.  I don't have time to write it though, alas.


You're probably right: with a laser cutter, we'd be better off using
some material which doesn't char.

Best,
- Joe

P.S. I hope you don't mind all my questions.  Until this thread, I was
totally ignorant of your work, and as you can tell, I'm quite excited
about it!

Yes, that would be the way.


Stick around, I probably will in a few weeks or so. Depends on my other
projects.


Well, look at the edge of those yellow Lynxmotion parts: charred.
Plastic will probably always char.


Thanks for the questions. I hadn't really published anything yet on the
robots side. I always like to get good feedback. And if you are
interested in having those parts laser cut in an exciting material, then
it sure adds to the project.

Matthias

One more question, if you don't mind.  How exactly do you make the
attachment of the bracket to the non-horn side of the servo?

Thanks,
- Joe

I remove the bottom plate from the servo, then drill a hole and push a
Nylon bushing into that. The axle itself is a brass screw with its
thread removed form the top 1/8th inch.

Matthias

Thanks for the explanation.  That certainly gives me food for thought.  
I think I'll start a separate thread about this issue.

Best,
- Joe

I've been thinking a lot lately about the use of servos in robotics --
my previous experience has been mainly with standard motors.  So I've
spent a lot of time looking at various servo-based robots (mostly
walkers of some sort, often bipeds) to see how you take an output shaft
that's on only one side of the servo, and turn this into a nice
self-supporting motorized joint.

Here are the methods I've seen so far:

1. Don't Bother.  Some people just bolt a plate onto the servo horn and
let that be the sole connection to the next section of the
arm/leg/whatever.  This seems to work fine for lightweight models, but
it must put a lot of stress on the servo axle.

2. Wrap-Around Mounting Bracket.  LynxMotion's "Multi-Purpose" bracket
provides mounting holes on the top of the servo, and a hole for a pivot
on the bottom side, as seen here:
  <http://www.lynxmotion.com/images/html/build083.htm>
This holds an unmodified servo nice and snug and provides a bottom pivot
by means of the bracket.

3. Modified Bottom Plate.  In another thread, Matthias Melcher


...which works well, as you can see in his "Squish" robot:
   <http://www.robowerk.com/img/squish/leg_up_hi.jpg>

But of course this means modifying the servo bottom plate, as well as a
brass screw.  (Is there anywhere you can by pre-drilled servo bottom
plates?  Seems like this would be a common need...)

4. Specialized Servos.  The new HITEC robot servos (introduced with
Robonova-1) have optional "idler horns" opposite the regular horns,
providing a slim built-in solution to this problem, as seen in this
review:
   <http://www.botmag.com/articles/hitec_servo.shtml>
They also lose the standard mounting flange and instead offer, on the
side away from the horns, mounting holes (which seem to do double-duty
in holding the bottom plate on), and extenders which can make the width
on that side equal to the total width across the dual horns; this makes
it very convenient to bolt one servo to another using just a couple of
flat plates.

Also in the specialized servos category would be the Bioloid servos, as
well as the new LEGO NXT servos I suppose.


Are there any other solutions to this problem that I may be missing?  I
really like the look of the Robonova servos, but I prefer the wider
selection available with standard servos.  I'm a newbie to robotics, and
I can tell this is a problem that's been solved many times... if you
have any insight, I'd love to hear it.

Thanks,
- Joe

    A through-shaft is probably the way to go.  Anything that puts a heavy
side load on the bearings from one end is going to give trouble.  Hitec
apparently puts an "idler" opposite the output shaft on their new servo,
but it's just a bearing attachment point, not a true through-shaft.

    These gearboxes

    http://www.servocity.com/html/spg785_gearbox.html
    http://www.servocity.com/html/worm_gearbox.html

    are through-shaft devices, but they're rather bulky.

We're getting closer.

                John Nagle

Yes, I'm amazed that it's so hard to find servos that work this way; the
Bioloid ones are the only ones I'm aware of.

In fact, in a lot of ways, standard servos seem ill-suited to robotics
-- for example, the way the cable connects on the same end as the servo
horn, pretty much guaranteeing that it's going to get in the way of the
joint, and also preventing the servo from fitting through a
full-surround mounting hole in a frame.  I suspect that the form factor
was designed for their applications in RC models, and now there's just
too much pain in deviating from the standard (with a few exceptions).

But I digress.


Right, but at least the idler supports the load on both sides of the
servo; I think this must help substantially.  A through-shaft would
certainly be better of course.
 

Hmm, yes, those are neat though.  That second one (the Tamiya worm
gearbox) actually doesn't look much bulkier than a servo, but of course
it lacks position control.  If you could cram some sort of position
feedback into that thing, it might make a really interesting robotic
joint.

As for the first one, that suggests another one to add to my list:
replace the servo horn with a gear, which drives another gear connected
to a through-shaft.  This does widen the servo assembly by probably 50%,
but on the plus side, you could choose your own tradeoff of servo
range/speed vs. power just by changing the gears.  Hmm.

Anybody know where to find gears that connect securely to a servo shaft?

Thanks,
- Joe

      True, but we're finally starting to see a few devices that aren't
just variants of R/C airplane parts.  The important innovation
is that some of the newer servos will read back position and torque
to the controlling computer.  That advances control from blind positioning
to force feedback.


     http://www.servocity.com/html/mechanicals___accessories.html

     Servo City has a collection of collars, shafts, and other small
parts for connecting things to R/C servos.

                John Nagle

Yeah, but no gears (that fit on the servo shaft) that I could find.  
They do have sprockets and chains, which could serve the same purpose,
but in a bulkier and more Rube-Goldbergish way.

Best,
- Joe

Hmmmm. They used to sell gears that incorporates splines for attaching
to servos. Convenient though it's usually not hard to use a standard
servo disc and just mount the exact gear you want over it.

-- Gordon

    If you don't have the catalogs of

    Small Parts, Inc.
    Stock Drive Products
    Berg

you should.  They're full of gears, shafts, collars, and related
mechanical parts.  That's where you get such things.  Prices are
higher than in the R/C world, but the quality is higher.

                John Nagle

But none of these have servo splines, so there's still that issues, and
when we're talking about R/C servos, I wonder if it makes sense to spend
more on the gear than it does on the servo. You can get a nice nylon or
Delrin gear replacement Traxxas gear, for example, for $3-4 retail.

-- Gordon