Yay, a WALL-E with treads showed up at the thrift store!
I've had a WALL-E before, but it was a cheapy where the treads didn't move. In that one, instead, there were motorized wheels sticking out under the treads. I thought: even if this doesn't work, I can hack a new microcontroller, etc., into it just for fun. But after playing with it, I found that the tread movement is a huge part of what makes it fun, so I ended up sending that one back into the stream.
But this one, wow, this one is the real deal.
It's a WALL-E made by Thinkway. There isn't an obvious version or date on it. Under the front lip, "T90209" appears to have been heat-pressed into it as a serial number. The label reads, "I'm a Thinking Toy" and "THINKWAY TOYS".
Here he is.
WALL-E arrived dirty but promising. There was a serious amount of grime all over, as if the previous owner had it in a dusty garage. Amazingly, the battery compartment still had its cover, and the inside contacts did not show evidence of corrosion. So often, I find toys in the thrift stream where the battery cover has been removed so that batteries don't end up going into the waste. That's great, sure, but then they don't put the battery cover back.
He did not have his remote control. But, there is a "play" button on the top of the front chest area.
After doing some internet searches, I found that this tread-type WALL-E may be one of the earlier models they made. It's likely a "U-Command" WALL-E. Those sell for $100 to $500 on e-Bay.
With some fresh batteries in place, I wasn't getting the kind of behavior I'd hoped for, so I decided to take him apart and do some fixing.
There are at least a couple of good WALL-E repair videos.
This one shows a WALL-E that takes three horizontal batteries, whereas mine takes 4 AA vertically. Perhaps that's a US / European difference.
(Fix It Ade, partial fix) https://www.youtube.com/watch?v=B3MTJLdCoYo
(1nformatica) https://www.youtube.com/watch?app=desktop&v=kq-pym-aO4c
The latter is closer to the one I have, and exhibits some of the problems I found. But mine had other problems (some of which I may have been responsible for).
In the end, mine had these problems
- Poor soldering / strain relief caused complete power loss. Easy fix.
- The IR system was failing intermittently, possibly due to bad soldering.
- The front IR detector had poor soldering / strain relief, causing the signal loss.
- The IR transmitter was missing.
- The port-side axle gear was broken
- The starboard side axle-to-gear connection was stripped
Disassembly
Much of the disassembly follows what was shown in the 1nformatica video.
Prep
Set up a holding bin for parts
- The battery panel
- Tamper-resistant/security screws from the back panel
- Four base screws (holding wheel assembly to torso assembly)
- Three screws each for the wheel assemblies
- Two screws holding the rear IR board
- Two decorative side flaps above the rear axles
- Possibly seven flanged screws for the motor mounts
- Possibly two more screws for the front IR board, and two more that make it hard to get to
Power off
Set the power switch to Off
Remove the battery panel
Remove at least one of the batteries
Back panel removal, disconnect, and return
Remove the four security screws to release the back panel
Gently open the back panel, and remove the screws holding the rear IR board. Tuck it inside the torso cavity. Put the back panel back in place with one or two of the top security screws.
Base-torso separation
Place WALL-E face down
Remove the four base screws
Remove the rear-most screw from each wheel assembly. This screw is in a recessed area, and holds the wheel assembly to the torso. The other two screws in the wheel assembly hold the wheel assembly itself together, and they do not need to be removed.
Gently pull away the base from the torso in front, and you'll see that you can't get the base off unless you gently bend the wheel assembly off its side pin.
Once you do that, each side will be released. With each wheel assembly pulled away from its mounting pin, separate the base assembly from the torso assembly.
With the assemblies separated, you should be able to rest the wheel assembly flat (treads on the ground, as normal) with the torso-head portion still face-down.
Base-torso wiring separation
At this point, the main motor board will be visible. Disconnect the connectors for M2 (arms/eyes motor, 2 pins), M1 (wheel motor, 2 pins),
and power.
Then, gently remove the 8-pin sensor connector from its header.
Be very careful when prying the connectors out of their headers. Don't just pull them. I did that and it pulled the header out (but left the pins in place, luckily.) If you have to repair these, I think they're probably PH2.0 headers for M1 and M2, and an XH 2.54 header for power.
With these removed, the torso assembly is now completely disconnected from the base assembly.
(For possible future reference, this is the back side / underside of the
main board. I removed it to look at the connections to see if anything
appeared fried, but it looked good.)
The torso's inner connector board
We had disconnected the rear IR board earlier. Inside the torso, you'll find another connection board. That has two headers. One is a 3-pin, where the rear IR board's wires click in. The other is a 2-pin that is for the Play button.
(Torso connector board with 2-pin Play button connector removed)
The board also has hard-wired to it the 3 wires for the front IR board, and those are connected in parallel with the rear IR. The eight-pin ribbon from the base assembly also has its individual wires hard-soldered into this board.
(Back side of Torso connector board. Note fat traces connecting the two sets of IR sensor wires, 3 each, together. Also note: possibly insufficient solder in one of the set of IR sensor connectors, and excess wire sticking past 8-wire solder points, potentially risking shorts.)
All in all, the 8 wires are for
- 2 Play momentary switch / button. (Any biasing of signal is done on the main board.)
- 3 for eye LEDs (Presumably these are: two signal, the other either power or ground; each signal line probably can be or is PWM-controlled and has appropriate inline current resistance. I don't think there are SMD resistors on the board, so I suspect any inline resistors are on the main board.)
- 3 IR signals (in parallel from both sensors). (Presumably, power, ground, and signal, though each IR board uses the terms R-, R+, and R0. I didn't check to see which is which.)
Remote control
My initial focus was on remote control. If I were to run WALL-E just using the Play button, the movement would be haphazard. I wanted to find out if I could cause it to just go Forward or Reverse.
After some searching, I found a few promising documents online.
First, there's this Arduino page
https://forum.arduino.cc/t/u-command-wall-e-who-has-one-looking-for-ir-code/552285/11
that leads to an http (not https) page
http://www.hifi-remote.com/forums/viewtopic.php?t=12137
Its August 01, 2003, post has a listing of timings and IR codes. It says there's a lead-in of 6000 microseconds (us), then 4 data bits, 8 data bits, a repeat of the 8 data bits, and a lead-out of 2000us. Apparently, that sequence repeats three times for all commands except joystick fwd and reverse.
Each bit is represented by +1500 / -500 for a "1", and +500 / -1500 for a "0".
For all the "+" values (the lead-in, lead-out, and some portion of the 1 and 0 bits) we're supposed to generate a 44 kHz pulse wave. This is provided in Arduino code by the IRremote library. We can call IRsend's sendRaw() function, passing in an array of 16-bit integers for on and off times.
The behavior we'll be seeing is similar to what is shown in EEVblog #506
https://www.youtube.com/watch?v=BUvFGTxZBG8
I found another set of WALL-E data in the github repository Lucaslhm/Flipper-IRDB (in Toys > Thinkway_Toys > Wall-E_Robot.ir). If you look at the codes there, they're mostly similar, but not all the same. And, importantly the Lucaslhm ones show 38 kHz frequency with 0.33 duty cycle, whereas the hifi-remote ones were for 44 kHz with no duty cycle specified. Also, the identifying code appears to be 1010 for both hifi-remote and Lucaslhm/Flipper-IRDB.
IRSend build
Lacking a remote control, I made a first attempt at building one. Mine was comprised of
- A breadboard and a typical smattering of hook-up wires
- Two 330 ohm resistors in parallel (computes to 166 ohms)
- A scavenged IR remote from a discarded TV remote control
- An Arduino Micro
- A visible light LED
I hooked up Arduino pin 3 (default IRremote library output pin) to the resistors to the LED cathode, and then LED anode to ground. IR LEDs are pretty clear, so you can easily see which side has more metal inside, and that's the side that connects to Ground.
I measured the forward voltage of the IR LED at around 1.1V, and figured I had about 4.5V from power, so the current through the LED woulc be (4.5 - 1.1) / 166, or about 20.5 mA. This is within the safe range for an Arduino pin, and also within the typical safe current range for LEDs. For a visible LED, if the forward voltage is 0.7V, then the end result is about 23 mA.
I wrote a chunk of code that would then send the intended pulses out using irsend.sendRaw().
With WALL-E's power switch in the Play position (slide switch closest to the middle of the base) and a fresh set of batteries, I tried aiming the IR emitter towards him and... nothing. No reaction from WALL-E. Sad.
At this point, I had a number of possible problems or causes
- The IR LED isn't working
- The IR LED isn't bright enough or is facing other light interference
- There's dirt on the sensor or transmitter
- I got the frequency is wrong
- I got the code wrong
- The IR receiver isn't working
- The inner workings aren't processing the IR signal correctly
Is the IR LED transmitter working?
I re-ran my sketch in a darkened room, and watched the IR LED using my phone camera. When it ran, I could see a pink-purple light show up. That was some proof that I got the circuit right.
Is there other light interference?
I separated out the rear IR sensor, and reflowed the solder for it.
Then, I made my own jumper header using PH2.0 header and pins, some wire from an old ethernet cord, and an ethernet break-out board. That allowed me to have the sensor outside of the main cavity with no other plastic in the way. Still, nothing.
Is there dirt in the way?
I thought this might be a problem prior to knowing that the two IR sensors were wired in parallel. So, hoping that maybe the rear sensor failed, but the front sensor was good and simply dirty, I cleaned up the exterior plastic of the front sensor. Still, nothing.
The frequency is wrong, or I got the code wrong, or I got the bits wrong.
At this point, I looked at a variety of possible causes in the software side of things.
First, I rebuilt the sketch so that upon receiving a command "9" through the Serial interface, it would start looping through all 256 data byte values, and send them. I didn't know how long a given operation might take, but just delayed a little while after each transmission.
I also noticed that the AI-generated Arduino code (given to me by searching "how to use IRsend") was passing in an array of unsigned int values. But on a deeper inspection of the code, I found it was expecting to receive uint16_t, specifically, so I switched over to that. (In the end, it didn't matter, but this is one of those C coding tripping points that a proper IDE would have flagged.)
I ran the code at 38 kHz across all 256 codes, and got nothing. But then I tried again at 44 kHz and got a response, finally!
I added code that would emit the sequence and code to the Serial Monitor, and tied the input to a typical Arduino button, pulled low (with an extra 50ms delay for debouncing). I also noticed that reactions didn't start happening until the last 128 codes, which was in line with what the posted timings showed.
That allowed me to click the button for each of the data byte values 128..255, and test for reactions. Sure enough, I got reactions for specific codes. Of those, only one was identifiable -- the Dance action -- because I'd seen it in the 1nformatica video.
In the end, I ended up with this table of U-Command WALL-E IR codes
hifi-remote.com (44 kHz)
+6000, -2000, 1010, data, data, +2000
codes: fe, fd, ae, ac, ad, ab, a4, a6, a5, a7
Lucaslhm/Flipper-IRDB (38 kHz, 0.33 duty cycle)
+6000, -2000, 1010, data, data, +2000
codes: fe, fd, ae, ac, ad, ab, a4, a6, a5, a7
Mine found through experimentation
44 kHz
+6000, -2000, 1010, data, data, +2000
codes: fd, fe, df, ae, ac, ad, ab, a4, a6, a5, a7, maybe c8
So in the end, my WALL-E's codes and frequency match up better with hifi-remote, but not exactly, and there are mysterious extra codes.
Fixing movement
At this point, I was very happy to have a set of reactive IR codes, but couldn't really tell what each would do, because the treads weren't moving properly.
I had removed the treads and placed WALL-E on a lift so that the wheels could spin freely. (Note: I'd already made a pretty big mistake in doing this. I thought a gentle way to remove treads would be to slide a piece of paper in, and rotate the wheels so that the paper would get under a wheel. That worked, and I could slide the tread off easily. But in doing this, I was forcing the rear wheel to turn when its axis position was fixed, and you NEVER want to do that.)
When running my IR codes to WALL-E, I could see one wheel spin, but with the tread on, it would not.
At that point, I got on to disassembling and removing wheel assemblies from the gearboxes on each side.
Getting the entire assembly of both wheels out from the torso is the first step. The 1nformatica Youtube shows how to do that.
In between each gear box, there is a grey plastic cylinder with hex holes on each end. That can be pulled off easily, leaving you with each side: a gearbox and a tread assembly. One side also has the DC motor. Its gearbox is more intricate.
I'm not quite sure how it works, but the system is able to make WALL-E turn with only one DC motor. I think it's using momentum to cause a gear disengagement to happen.
This is what the starboard axle gear looks like. There may be a hairline fracture showing in the second picture (right side around the 3:00 position). But, I think the problem with it was that the axle knurling had ground a groove inside the hub of the gear.
This is an example of the 3d-printed gear, top and bottom. Variations can be seen in the teeth and infill. But, it does ok. This is actually one of two spares I ended up with. The others are in the gearboxes now. It'd be really nice to have a resin printer to make these. Using my current set-up (Artillery X1), they vary greatly in quality from layer to layer and on different teeth, even though I'm using a randomized layer alignment (Prusa Slicer). That said, I wonder what material choices I'd have with a resin printer. The model's pretty good. The printed result is mediocre.
Port gearbox with motor and motor board
Starboard gearbox (with replacement 3d-printed gear in place)
Axle with knurled area
Getting the tread part separated from the gearbox is a pretty brutal step, and doing it can be damaging. It turns out that each wheel assembly is probably press-fit onto the shaft as one of the final assembly steps. I think even the tread installation is done before the tread assembly's axle is pushed in. It was never really meant to be undone. The axle comes out of a tread assembly, and goes into the gearbox. The knurling near the end of the axle joins to a kind of spur gear. It is that same gear that is apt to break in two ways, both of which I experienced:
- The gear cracks due to age, thereby losing its friction grip on the axle. The 1nformatica video showed a clever repair for that, reusing the existing gear. Another youtube I saw suggested a similar problem with a zip-tie solution.
- The axle rotates while the gear is fixed, or vice versa. The nylon gives way, leaving the knurled area spinning freely. If this happens, the original gear is in one piece, but is not usable. Perhaps with clever melting, it could be coaxed back to a working state, but it'd likely be fragile. For this reason, NEVER ROTATE THE WHEELS MANUALLY!
Either way, to repair this, you have to get into the gearbox and replace the gear. Due to the way they placed the gearbox's four screws, getting into the gearbox forces you to separation the axle.
Why, oh why? The screws go into the gearbox from the tread side. It would have been much better if the screws went in from the opposite side, so the screw heads faced the center of WALL-E rather than facing the port and starboard walls.
Also, why press-fit? If they didn't have press-fit wheels, and instead used a "set screw" way of holding the gear and shaft together, it would have been better. But if that were to be done, the person assembling it would have needed access to the insides of the gearbox, and I think for "clean manufacturing" purposes, grease exposure/loss, etc., they made it so the gearbox was a closed system, only to be used once.
Well, so, we end up having to forcefully un-press the axle out of the gear. This took some force, but I used a few extra random pieces of metal (US 25c coins) to prevent direct pressure against the plastic, and very careful turning/prying with a flathead screwdriver to get things apart.
The gear itself is a 12-tooth, 5mm-thick spur (?) gear with another 4.5mm-4.7mm cylinder on top. It has a 4mm bore. I printed one based on a Thingiverse model ("Thinkway U-command Wall-E gear & cap" by "Zadakleader635"), and then re-built it myself based on measurements
I re-modeled it using Autodesk Fusion and a gear generator plug-in. (It's either that, or generate the gear vectors in Inkscape.) I wanted to control the dimensions, mainly to compensate for material shrinkage of the inner hub and teeth. I also liked adding my own chamfering.
3d-printed PET-G gears. (For my printer, I find I get globs if I print only one. so I printed sets of 4.)
The gear hole size is tricky. It has to be sufficiently large to slide on the axle. But also tight enough to engage the knurled section of the axle. By my measurements, that's only a 0.15mm diameter tolerance.
In the end, mine was designed with a 4.2mm hole that shrank to smaller than 4.0mm. After printing, I drilled out the hole with a 0.1575" (or, 4.0mm) drill bit. I printed it with PET-G, which is closer to nylon in behavior and more resilient than PLA.
I also had to size up the gear dimensions a bit, to compensate for shrinkage.
After printing, I finished the teeth with a hobby file. Then, I chucked the cylindrical part of the gear in a portable drill, and spun it at high speed and ran it against some sandpaper to remove the sharper edges.
Putting it all back together again
For me, both gearboxes had an axle gear problem, but that was all. I opened each gearbox very carefully so as to keep gears from popping out everywhere, removed each bad gear, put a new one in, and closed it back up again. I didn't add any grease.
Then, the two gearboxes and middle shaft went together, and the whole assembly went back into the base box.
Remember, at this point, to put the decorative cover plates back onto the base box.
After that, each tread assembly can be pressed back onto the gearbox. But, you have a conundrum here, because you know that eventually the base and torso assemblies have to go back together again, and doing that means you have to pull the treads away from the body a bit. Pressing the treads on now gives you more control over the pressing process. But, pressing them on later more likely yields better axle-to-axle gear engagement. Or, put another way, pressing them on early risks un-pressing them during the body reassembly, thus risking axle-to-axle gear disengagement/damage.
With the treads and gearboxes and decorative panels back in place, the rest of the reassembly can be done:
- Re-attach motor mount screws.
- If removed prior, reattach Play button and rear IR connectors to torso midboard.
- Connect M1, M2, and Power wires and 8-wire ribbon connector.
- Gently re-connect torso to base, "bending" each tread assembly outward a bit so that mounting pins can be fit together.
- Re-attach base screws
- Re-open back panel safety screws
- Re-attach rear IR panel
- Re-re-attach back panel safety screws
- Ensure power switch is in OFF position
- Add batteries
- Close battery panel
- Ensure arms are not in a conflicting position
- Switch power to Play position. Should result in a sound.
- Click Play button and you should see WALL-E dance around a bit and make sounds.
But wait, it's still not working!
At this point, I tried firing up my IR transmitter to see what commands would do what, and again I got no response from WALL-E! Oh, no! It was interesting that the Play button would work, but the IR would not. That showed that I hadn't (yet) fried the main board, and the motors were working.
So... more internal surgery. Not trusting the IR mechanisms, I checked all wire connections (made sure wires in ribbons were ok, checked for shorts, etc.). I re-flowed the solder points for the IR sensors. I inspected the ribbon cable hardwire/hard-solder points and found that they were suspect, so I re-flowed all of them and trimmed excess wire better than they had done.
At this point, in testing, WALL-E started responding to IR commands again, but only with the rear sensor!
Well, that's a step closer. Understanding that the IR sensors were set up in parallel (joined by thick traces on the torso connector board), and given that the rear IR sensor was working but the front one was not, it meant that the front IR sensor or its wiring had to be busted.
So, then I opened up our patient for surgery one more time. I found that when I messed with the front IR sensor, I must have broken the solder connections for one or two of its three wire solder points. In fact, as I pulled it out this time, the wires just pulled away from the board, clearly disconnected.
Not wanting to do this again, I pulled out a different 24awg 3-wire ribbon, and soldered that cleanly to the original front sensor board.
(Cleaned off old connection)
(Tinned new ribbon. Connect blue=brown, red=red, green=yellow)
(Insert ribbon wires through holes, solder to pads, trim afterward)
A dollop of hot glue would have been good at this point, but I didn't do that.
Then, I butt-welded those to the original wires after cleaning them up, and closed up with heat-shrink.
This gave a lot more play to the wire, so I taped things in place to avoid further metal fatigue.
He works!
After all that, WALL-E is mostly working again, finally! His right-side tread moves better than the left side.
Overall findings
Overall, this has been a very interesting foray into the world of a vintage WALL-E toy.
From a quality and build perspective
- It's great that the toy is mostly repairable
- The way they managed the motors and expressiveness in the eyes and arms and head is great.
- The use of triangle-shaped security screws is almost a joke.
- There are lots of places inside where extra wire would have reduced strain during repair.
- There are at least two places where the original build used clear, cellophane tape to hold wires in place and thereby prevent strain: the battery pack power lines, and another place where wire bundles were taped together.
- In many cases, the original soldering was functional, but didn't stand the test of time, and certainly didn't survive my manipulation. It was far too easy to break the wire-solder connection points. Granted, it's probably a 22-year-old toy now. In many cases, the wires were soldered to pads, and a rubber glue was added.
- The gearbox assembly -- oh, it would have been much better if they could have had the screws go in from the other side.
- The press-fit axle-to-gear connection is susceptible to damage, especially if you turn the wheels manually (NEVER do that!)
- The pin connection on the sides -- the one that holds the tread assembly in place -- also forces you to bend the tread assembly a bit during assembly, and makes it so that repairing the gears can cause damage to the gears. It would have been nice if the one stabilizing pin (on each side) that matters most could have been made removable from the outside (but perhaps they had to be super careful about potentially having kids swallow small parts that might come loose).
