Robo-Dragon Autopsy Performed by Christopher Thomas in August 2008.
One fine Sunday, I found a very cool-looking robot dragon in Toys-R-Us. It
was inexpensive (and cool-looking), so I bought it. After playing with its
(minimal) available features, I took it apart. This page documents the
results of this autopsy.
Meet the "Wireless Remote-Control
Dragon"
The box certainly looked impressive:
Out of the box, so did the dragon. They certainly put effort into the
artwork and detailing:
Unfortunately, to call the feature set "limited" would be to give it too
much credit. The dragon can walk forwards, and roar. Nothing else. Not
even turn or walk in reverse. This gave it a distinct disadvantage when
duelling the Robo-Raptor.
Because I like to save the best for last, the first thing I took apart was
the controller. It wasn't complicated; just two buttons and an infrared
LED:
PCB front; not complicated.
Blank PCB back, LED, and patch header.
This controller takes AA batteries, which is nice, unlike the
Robo-Raptor's AAA.
The most interesting bit, however, is the patch header. There's a similar
header inside the robot, and when it's in the box, a cable connects them.
What it does is feed power, ground, and one other line (probably the IR
modulation signal) from the controller, into the robot. When the box is in
the store, customers can test out the robot by playing with the
controller. Only the controller has batteries pre-installed.
The controller provdes 4.5 V, which is enough to power the speaker and
electronics but not the motors (there's at least one diode drop involved).
This is good, because the dragon's limbs can't move while it's boxed, and
the motors do Bad Things when they stall (see below).
Modding potential: Small to none. The best thing to do is to stick
a scope to the control line or the LED trace, and reverse-engineer the
control signals. Then, build your own controller or load the signals into
a PDA with IR capability.
The really cool potential is in the control header itself. This provides
an easy way to piggyback a controller inside the robot without having to
rebuild its control board. Of course, you're still stuck with a robot that
can't turn.
The Tail
The tail is the lynchpin of the dragon robot: once it's on, it has to be
removed before the body case can be opened (it overlaps). The problem is,
it snaps on, and getting it off is a royal pain.
The tail, shown in the left hand image below, is pretty simple once
disassembled. It has a number of internal partitions, mounts for screws
that hold it together, a tail dart that mounts on a screw hole, and
several iron weights that keep the robot balanced (otherwise it'd tip
forwards due to the weight of the motors). It looks easy enough to take
apart.
Tail, disassembled.
Tail mounting point.
Turns out, it's not. The screws hold the tail together during
manufacturing, but by the time it exits the factory, the tail seams are
bonded together with glue and/or plastic solvent. I had to use a cutoff
wheel on the underside seam, and a screwdriver, prying, and cursing on the
top seam before it came apart. The cutoff wheel made a mess, and while a
wire brush wheel took care of the mess, the cutoff wheel also left a
groove. This is visible on close inspection after the tail is reassembled;
you'd need modelling putty and paint to fully restore it.
The mounting point itself is pretty elegant. There are plastic ratchet
tabs that lock it into place once it's inside the receptacle in the tail
body. The tail moves when the robot walks, thanks to a set of gears shown
below. These gears are actuated by a the pink tab in the right side of the
image, which is in turn tied to the cam that moves the legs. Leverage is
poor, but as long as the tail isn't obstructed by anything, it works well
enough.
Modding potential: Moderate. There's a bit of extra space
in the tail for packing widgetry, but getting at it is such a pain that I
really doubt I'd bother on a new robot. The most useful modification to
make would be to file off the ratchets (the triangle-profile stubs closest
to the end of the mounting point) before assembly, and install
screws to attach the tail instead. This allows much easier assembly and
disassembly, even if it does add two more screws to the tail's
surface.
Another area with potential is the tail mounting point. Like the tail, it
has enough open space in it to be useful for storing electronics. Unlike
the inside of the tail, it's easy to get at without a cutting wheel. I'll
probably end up using both the mountpoint and the tail spaces, but for
future purchases, I'll use the filing trick and not bother with the tail
itself.
The Wings
The wings are each attached with a single screw. Their mounting points
penetrate the body case, so the wings need to be removed in order for the
case to be removed. The mounts are posable, with a ratchet holding them
into place once set. The wings are geared together, so that adjusting
either one moves them both (keeping them symmetrical).
Modding potential: Small. To make the wings themselves do anything
interesting, you'd have to completely replace them (they're not
articulated). There's potential for modifying the mounts to accept servo
control, but that would require installing a bulky servo in an area that
doesn't have much space.
The most interesting practical mod I can think of would be to tap the head
actuation servo to fold and unfold the wings as the dragon's head moves
during the roar cycle. This would be tricky, and require fabbing a
fairly complicated set of linkages, so it's low on my priority list.
Cracking the Case
Like the tail, the case is secured by both screws and glue. Unlike the
tail, the cutting wheel wasn't needed to open the case. Just lots of
careful prying and cursing. I say "careful" prying, because if you pry too
hard, the plastic itself will snap instead of the glue (modern plastic
cement is quite strong, and solvent makes a bond that's as strong as the
plastic itself).
An important note about the case: The neck housing is thin rubber, and is
glued to the case using a glue that's stronger than the rubber itself.
Disconnecting it required very careful surgery with an exacto knife. If
done right, this cuts the glue, and leaves you with both the neck, and the
inner rubber flange that folds into the case. If done wrong, you cut
through the flange, and get a neck membrane that's very difficult to
reinstall.
The main objects of interest are, of course, the control board (lower
left), and the servos (middle). Before I talk about those, I'd like to
draw attention to a few other features in the right-hand image above.
First, we have the IR sensors (upper right and upper left, just below the
neck; only the right one is clearly visible in the picture). These are
housed in grey plastic domes that are molded to follow the texture of the
skin around them (lots of attention to detail in this model). Second, we
have the wiring harness. Wires go to the battery pack, the power switch,
the speaker, the head LED (that's the cable going up through the neck),
and the two servos. You can also see the control cable going from the
board to the receptacle near the battery opening (left side of the
image).
Important note: The wiring harness is easy to break. I ended up
having to re-solder the connections to the leg servo and to the power
switch, due to them breaking during handling. I'm convinced this is where
the failures mentioned in the reviews came from, too; when the motors
stall (if the legs are obstructed, usually), they draw a lot of current.
This heats up some of the motor and/or power solder joints enough that
they fail. The wiring harness is the biggest place in the robot where they
skimped on quality, and this seriously harms an otherwise-cool toy.
The third feature I'd like to draw attention to is the attachment point
for the hind leg, in the bottom of the image to the left below. The leg
cam's spindle enters the large hole, between the wire guides. As the cam
turns, the wire guides are forced left and right, moving the leg. The
screw in the middle is the leg's hinge. The screw on the bottom right is a
guide screw. The screw on the bottom left, though, controls the foot. This
is really cool, and is discussed in the next section.
The fourth feature I'd like to draw attention to is the speaker mounting
point and grille. The speaker is held on by a pink strap behind it. The
main points of note are that 1) it's loud (which is good!), 2) it has a
frequency response that almost doesn't suck (which is also good!), and 3)
it's strongly magnetic (which means it eats stray screws). Watch out for
that last bit if you remove it from its mounting point.
The last feature of note is the attachment point for the forelegs, shown
in the right image above. This is molded plastic, which snaps in. This
means the forelegs can't be taken off without wrecking the hinge, and that
actuating them would be a royal pain even if there was room for another
servo. On the other hand, their poses are pretty nice the way they are, so
changing this is low on my priority list.
Modding potential: High, for the case itself. Mostly this is
because there's quite a lot of room for extra widgetry. The electrical
system is also very simple, which lends itself to easy modification
(discussed in more detail below).
The Legs
The legs are one of this toy's most elegant features. The more
sophisticated Robo-Raptor toy tries to emulate a real walking gait by
rocking back and forth (to avoid needing an extra degree of freedom to
lift the feet off the ground). End result is that its feet don't come up
very high, and tend to have trouble with surfaces that are either too
slippery (like the lab's vinyl tile) or too rough (like carpet). The
"Remote Control Dragon" toy avoids these problems by putting small wheels
on the bottom of its feet:
The wheels can grip most surfaces, and only slide in one direction,
letting the Robo-Dragon pull itself along most surfaces just by shuffling
its feet. Inside the leg, there's a surprising amount of articulation:
The purpose of the linkage inside the leg is to keep the feet flat while
the legs shuffle back and forth. The peg at the top of this linkage is the
leftmost screw in the photo in the previous section. While the leg
moves, the top of the linkage is held pinned to one spot in the body,
providing a position reference which is translated to the feet.
The ratchet wheels in the feet bear a bit more mention, too. I'd expected
to find a springy piece of plastic keeping them rotating the right way,
but it turns out gravity and inertia alone will do it. One edge of the
wheel cutout is sharpened. Sliding the foot forwards pushes the wheels
back, where they can turn freely. Sliding the foot back pushes the wheels
forward, where their teeth lock on the sharpened edge, preventing spinning
and engaging traction. Colour me impressed.
Modding potential: Moderate. There's hollow space that can be used
to pack electronics or other goodies, but I honestly can't think of an
aspect of this design to improve. The leg could be articulated at more
points, but you'd have to do enough reworking that it would probably be
easier to build completely new legs.
Electronics and Servos
Next up, the wiring harness, electronics, and servos. Here's a view with
the first servo unmounted, showing connection to the tail actuation
linkage and the leg:
I'll say this again: The wiring harness is easy to break. If you
dismantle it, or even poke at it too much, or maybe even just look at it
funny, one of the motor wires will fail and you'll have to break out the
soldering iron. The crappy quality of the wiring harness is in my opinion
the worst design flaw of an otherwise-adequate toy. Replacing the motor
power wires with something heavier-gauge and better-soldered should be the
very first mod you consider if messing with the electronics.
The first interesting feature of the electronics package is the control
board. This is reasonably well-built, and fairly modern. All logic is
handled by the microprocessor under the epoxy dot. All discretes are
surface-mount where possible, and limited to things like resistors,
capacitors, and diodes. There are three transistors (high-current
bipolar), handling the two servos and the speaker. There's almost
certainly a surface-mount voltage regulator somewhere on the board, but
it's well-hidden. The two big capacitors will be connected to this
regulator. Reverse-engineering the board schematic wouldn't be very
difficult, but I'm lazy, so it'll wait until I feel like taking apart the
robocritter again.
Connections from the board lead to the demo header (power/ground/data),
the power switch, the two servos, the speaker, the two IR sensors (chest
and nape of the neck), and to the LED in the head (that's the wire running
up the neck). I'll stress again that many of these solder joints are poor;
especially the ones connecting to the motors.
The leg servo is very simple. In fact, calling it a "servomotor" at all is
a misnomer, as there's no feedback-based position control. It's a DC motor
with a set of reduction gears, and nothing else, driving a pair of
off-centre spindles (which in turn engage the leg mechanisms). This motor
has a couple of design flaws. First, it's the normal type of cheap DC
motor, which means torque varies depending on the position of the rotor.
What this means in practice, is that the motor can stall on a start
attempt if it's in the wrong position. Second, it's underpowered for the
load it has to drive. If anything obstructs the legs or tail, or the
screws are overtightened, or anything else goes wrong, the motor will be
much more prone to stalling. When it stalls, it'll heat up, and the solder
connections to it will fail. I'm almost positive this is what caused the
"dead on arrival" complaints in reviews of this toy.
A close look at the back of the motor shows small capacitors. These are
used to soak up the voltage spikes that occur when the motor is turned
off. At rest, and to some extent under other conditions as well, it's an
inductive load, so breaking the current loop powering it causes large
transient voltages. Because this is a brush-based motor, the current loop
is broken twice per revolution. Spiky indeed. The upshot for
disassembly is that I had to cut a small notch in one of the plastic
supports to allow a capacitor lead to pass through. Removing or
disconnecting the capacitors is a Bad Idea, as it'd do unkind things to
the electronics board.
The neck servo is very much like the leg servo, in that it's a motor and a
reduction gearbox. Instead of driving a pair of off-axis spindles, it uses
a cam to drive a spherical joint mounted off-axis. This causes the neck to
move around, and engages the jaw linkage, as described in the next
section.
Modding potential: High. At minimum, replace the motor wires with
something less likely to fail if you open this up. It'd also be pretty
easy to reverse-engineer the communications signals sent along the demo
header (there will only be two commands). This will let you keep the
baseline board as a "hindbrain", and piggyback a more intelligent
controller on top of it elsewhere in the case.
It'd be tempting to put in a proper H-bridge to get forward and reverse
motion in the legs, but that turns out not to work. The legs move forwards
and back as-is; the reason this translates into forward motion is the
ratchet system in the feet. In order to allow reverse motion, the feet
would have to be modified to either selectively freewheel, selectively
brake, or selectively reverse the ratchet direction (even harder). Doing
this with independent control of each foot would let you move forwards,
move backwards, and turn in either direction. The catch is that you'd be
making pretty major modifications to the feet, and you'd also need an
encoder of some sort on the leg cams to tell you when to lock and unlock
the wheels.
The only other really tempting modification to existing hardware is to put
better effects into the head. The main shortcoming of the existing system
is covered in the next section. Other modifications that don't involve
changing existing hardware (much) include adding support for radio
control, adding support for critter-to-critter networking over radio or
IR (it has the IR sensors already), and adding a controller smart enough
to handle autonomous behavior. Options are pretty open-ended, so knock
yourselves out.
The Neck and Head
With the possible exception of the tail, the neck and head are probably
the hardest parts of the dragon to take apart. The neck is a rubber
membrane; taking it apart cleanly takes an exacto knife and a steady hand.
It's important not to cut the rubber flange around the neck, if at
all possible. Cutting this makes it a lot harder to reassemble neatly (the
flange holds the neck in its proper position; otherwise there's a lot more
stress on it when it moves around). Remember that you want to cut the glue
holding the rubber to the plastic. You do not want to cut through
the rubber itself. There is also a wire pair running up the neck.
This is very easy to accidentally cut. Putting the neck back together
cleanly is even more of a pain (described in the next section). The skull
and lower jaw, on the other hand, are molded plastic (easy enough to
remove).
The guts of the neck and head are intricate, but impressively elegant in
function:
Getting the upper jaw off is a pain, but can be done with a bit of careful
effort. Inside, we have a spring-loaded lower jaw activated by a linkage
that leads to a tab at the bottom of the neck. The off-axis sphere ends up
pushing this linkage against a tab on the neck mount. This not only moves
the neck around (due to off-axis mounting), but makes the jaw open and
close during this process. Points for cleverness.
Negative points for style, though, come from the light-up eyes. I didn't
realize these were intended to light up until after I'd gutted the head
and neck. Checking the package, this is indeed advertised as a feature.
The problem is that the light source is one tiny 20-mA LED mounted inside
a very large piece of molded plastic that was probably intended to act as
a light-pipe, but in practice doesn't do much piping. If the toy was
activated in a darkened room, there might be a visible glow, but in a
room with the lights on, there's no discernable effect. Definitely
something to be upgraded if modding.
Modding potential: Moderate. Definitely replace the eye assembly
with something that actually works. My recommendation would be to keep the
outermost parts of the light-pipe, and glue LEDs directly behind them.
Rigging the electronics will be slightly fiddly, but only slightly. The
resistor for the LED is on the main printed circuit board, so in order to
get two LEDs working, you'll have to swap that out for a jumper wire, and
mount one resistor per LED, in the head. Resistor and LED are in series,
and the two resistor-plus-LED strings are in parallel with each other.
Other configurations are, of course, possible.
Per the previous section, it's also tempting to add effects to the head.
An orange LED giving indirect lighting to the mouth would add a nice fire
effect, if set up to be intermittent, for example. There's also a modest
amount of room to run more cabling or widgetry in the neck, under the
rubber membrane; just remember that the neck strut moves around inside it,
so there isn't quite as much free space as first appears (a misplaced
widget will show as a lump under the skin when the neck is in the wrong
position). When in doubt, keep widgets in the body.
Reassembly held only one major surprise: finding out exactly how
low-quality the wiring harness was. I ended up having to re-solder broken
connections twice, and take the dragon apart again at least three times.
The leg/tail servo is very sensitive to position (for both the servo in
the case, and the legs/tail when reinstalling the servo). A slight
misalignment will increase the resistance this mechanism encounters. It's
underpowered, which means it'll stall. The wiring is bad, which means
it'll fail under stall conditions. So, expect a lot of frustration before
the robot is finally assembled and working again.
The other irritation of note was expected, but still annoying. Taking off
the neck membrane involved carefully cutting away rubber from plastic,
while trying not to cut through the rubber. Reassembling the neck
requires gluing this rubber back to the plastic. First, only do this when
you've finished modding and have already fully tested the robot. If you
re-glue and then find out the robot doesn't work, you'll have to cut again
to take it apart (and glue again afterwards). Second, check very carefully
to make sure the neck membrane is lined up the way it was before you cut
it. The robot's paint job and molding are very detailed, and any mistake
you make will at minimum leave a visible discontinuity, and at worst leave
a yellow scar of unpainted plastic visible. For glue itself, super
glue (cyanoacrylate) is fine for tack-welding. For permanent
re-gluing, I'd use methylene chloride or a similar plastic solvent.
Annoyances notwithstanding, after reassembly, only a close inspection of
my dragon-bot would tell you it had been taken apart at all. As long as
proper care is taken, reassembly can be done without visible
artifacts.
Final Thoughts
Taking apart the "Wireless Remote-Control Dragon" was a fun diversion for
an afternoon. If, like me, you enjoy gutting electronic and mechanical
widgets to see how they work, I can definitely recommend the WRCD at list
price for that alone. Modders will have even more fun. It's just not worth
a whole lot as a toy.
