Light-Tracking BEAM Robot Head

Hey all! Super excited to finally be sharing this robot with the internet. It is a light-seeking robot that is entirely free-formed using BEAM style analog logic, that means no microcontrollers, and it runs entirely on solar power. It is a bit long but definitely watch my 30 minute explanation video if you want more detailed information on how everything works.

When I was very very young and just starting to get interested in electronics, these analog BEAM circuits were already going the way of the dinosaurs to make way for these newfangled PICAXE controllers or even an Arduino if you were lucky enough! As such, I only ever saw the tail end of the BEAM movement but something about those circuits still tickles me to this day. There is something about manipulating the movement of electrons at the most base level, no abstracting it with 1s and 0s, in order to make a pile of silicone... intelligent!

This circuit is much more finicky, took a tonne more work, and is nowhere near as powerful/versatile as a microcontroller based robot that would do the same thing, but I hope you will agree that there is something special about him. Maybe I am getting too philosophical here but there is something about his analogue nature that makes him more lifelike, gives him more personality than even a complicated digital program. Indeed, he was out of commission for just over a week while I waited for time to fix his motors and I really did miss him sitting there on my desk.In the demonstration videos you will even see a couple of animals/insects who I caught curiously inspecting the robot, I don't think they could tell if it was alive or not either ;).

In any case, I think that all of this philosophy is secondary to the main function of this robot which is... to look nice. I definitely think of this robot as an electronic sculpture and the "life" that the circuitry brings all just adds to the art.

All of my Instructables fall under 2 camps.

1. Follow along with the instructions as I have shown.
2. Mostly just inspiration and not something to follow exactly.

This Instructable falls cleanly in the second category and you would definitely need to understand the whole schematic and do lots of prototyping and testing if you were to attempt this project or something similar.

I will try not write as much as I tend to do as this is already a very large project so if I miss anything just ask in a comment.

Finally, some of this is out of order as I was jumping ahead and making sections while waiting for other parts. So I have presented it in the most logical order build wise but there may be some continuity errors.

EDIT: For anyone curious, my marble machine and tiny photopopper are both still going strong!

Tools

My 2 favourite hand tools are files and a jewellers saw. My 2 favourite power tools are my belt grinder and drill press. Alongside sandpaper and blowtorches for soldering these are the only tools I used to make the whole mechanical assembly.

For the free-forming I used a chisel tip soldering iron (only bringing out the fine tip for corrections I had to make to already densely populated sections of the free-forming). I also used a set of needle nose pliers from my local electronics hobby store and a pair of side cutters for all of the wire shaping.

Components

I cant list all of the components here so I will just list the important ones:

Solar panel: SM531K12L from digikey, 1.37W at 8.29V. It is quite pricey for a single panel but definitely the most efficient indoor solar panel I have come across at this size.

Motors: 15RPM 3V N20 motor (45RPM works also and were the motors I originally chose but they were too fast. I bought a lot of N20 motors to test and it would seem that the 45RPM and 15RPM use the same speed motor with different gearing, while the 30RPM motors use a weaker motor with the 45RPM motor gearing. So while 30RPM would probably be the better choice, the 30RPM motors I bought did not work with the circuit I designed.

ICs: 74HC240s for the octo inverters and a 74HC14 for the hex inverter for the audio processing.

Capacitors: There are many caps in this build but I wanted to mention that I generally chose long life electrolytics and high(ish) quality ceramics for a longer lifespan.

Resistors: I tried to use only those "half-size" 1/8W resistors to save space

LEDs: For all the LEDs (bar the hexagonal one) I used super high brightness LEDs so that even with the 10s of uA they are running on they are still visible.

Supercaps: There are 2 super capacitors on this model, one is a 7.5F 5.4V capacitor and the other is a 330mF 5.5V capacitor.

The introduction had a short demonstration video and a long explanation of the entire model. I also have three more videos here, the first just showing the robot off in a bunch of different conditions and the second is the full, uncut, 4 minute video I took of him getting excited to sit in the sun which I think is a really cute video.

The last video I made for a university competition amongst my cohort. It is essentially a very rushed version of my 30 minute explanation video as we were only given 8 minutes to show off our projects. I do think the robot looks nicer in this video as I filmed it the day after first finishing the bot so it is still super sensitive and using faster motors.

Before we start I have a couple of general tips for free-forming, at least the way I do it.

1. Clean soldering tips, flux, and good quality solder are probably the most helpful things for free-forming
2. Routing the components is not something that I can really tell you how to so, it is something that needs to be learnt but it is very similar to routing a PCB or building on a breadboard. Start with the biggest components, or the components that physically need to be somewhere such as the ICs, trimpots, etc. and then build around those. For this build in particular I tended to place large capacitors and the LEDs last. And finally, having a positive and ground rail is super helpful to keep everything logical and easy to build.
3. For actually placing components it is often helpful to leave small connections as the lead wires for each component instead of trying to solder very short lengths of brass (as if you try to solder 2 connections with a short length of brass it is going to want to melt both connections and fall straight off). For holding components I actually rarely use helping hands as it is too slow to set up with this many components so I mostly use blutac to hold one section, tweezers/pliers to hold another (most often a wire or small component), and then solder with the remaining hand.
4. Also in regard to placing the actual components I find the best approach is to sort of "tack weld" a component with just one connection and holding everything in place by hand. It will likely be quite messy and will probably be a cold solder joint. Then you can make the next connection nicer. And then when everything is securely soldered at at least 2 points, you can go back and clean up the joints.

Incredibly, I would estimate that prototyping took at least half of the total time of this project. Thanks to the finicky nature of these kind of analog circuits, almost none of the circuits worked off the bat and needed large adjustments to component values. Comparing my final Dual Slope Sampling Solar Engine to Wilf's original circuit we can see probably the biggest differences, many things had to be changed (not sure if that is because of my modern HC240, diodes, or something else).

I ended up buying the parts I knew I would need alongside a previous digikey order. Then I prototyped as much as I could and did another order to order anything else and anything I had forgotten, mostly that was those 1/8W resistors and long life caps.

Also I would just like to say that my breadboarding is usually quite neat I swear! But this was literally all the jumper wires I had from previous prototyping and I wanted to cut as few of those wires as I could because I didn't have access to any more during lockdown =P.

I have a couple of things to show in this step, first is the schematic which I go over in more detail in my explanation video.

Next are some initial drawings I did trying to decide how to lay out all of the circuit blocks.

Then we have some examples of how I plan out the component placement for each block. These photos are for the circuits built around the inverter ICs and in other steps I show what I drew for circuits without an IC to build around. These drawings stay right by my side throughout the free-forming process to make sure I dont make any mistakes so you can see some flux stains and such on the paper.

And finally I wanted to show off my initial sketch for this idea because I am impressed how close the final robot ended up being to that first sketch.

Last thing before getting to the build, I wanted to show my workspace while making this in case you think you need a massive makerspace with tonnes of tools and space. This was it for me plus a drill press and belt sander.

I started by building the mechanical parts of the bot, starting with the lower (left/right movement) motor mount. I started with a 10mm wide strip of 2mm thick brass sheet. Then I marked and drilled the holes for the motor mounting. When I first attempted to bend the mount I made a silly error and forgot about the bend radius so the bend was way too close to my mounting holes (which you can see in the pictures). To make the bends I used a triangular needle file to make space for the bends and then bent them in a vice. I annealed with a blowtorch a couple times while bending to make sure the bends did not crack. I then soldered the joints (with silver solder in this case but Im sure electrical solder would be enough) and filed the joints to make them look pretty. Finally I countersunk the screw holes as I forgot to do this earlier!

Next the M1.6 screws had to be cut to size. This can be done in many ways but I always use a jewellers saw and then round the tip with some files and sandpaper. Alternatively, buy the right size screws =P.

The upper bracket is made in much the same way but with one fewer bend. What I am showing in the photos here is that while the size of the lower bracket can be almost arbitrary, some considerations must be made for the length of the upper bracket so I did a test with a bent piece of aluminium. Namely, we must make sure that the switches for up/down cut off (will be covered later) have space to be activated. I made it such that the head would be able to look down just slightly however it never needs to in practice, and having the head sit a little further away from the up/down motor makes a longer lever arm making it necessary to use a larger/longer counterbalance. So it is definitely advisable to make this bracket as short as possible.

Both the finished brackets looking lovely and shiney!

The solar panel frame was a decision I made early on in the design process as it is generally helpful with these freeform circuits to have a large part of the structure be ground. This just helps with routing the wires and providing more mechanical stability to the freeforming. It is unfortunate in terms of electromagnetic interference that it is a large loop but this was never an issue and the benefits outweigh the problems. I also thought the brass frame would look great and I was beyond correct in this regard.

I considered using K&S brass channel for this frame but it only came in imperial and I decided that 1/8 inch would not hold the 2mm panel quite as securely as 3mm channel so I had to make my own. Hence I used a belt grinder to remove most of one wall of some 3mm brass square tube, then I hand sanded on a flat surface to remove the last of the wall.

I also think that sanding down one side, and making the exposed edge on the front of the bot 2.5mm (instead of the 3.8mm edge if I had used 1/8 inch channel) makes it look much better and more elegant from the front.

Now we must start putting the frame together. I created the 45 degree joints very carefully by hand and they all ended up quite tight. Then 3 sides of the frame must be soldered together with some triangular gussets cut from 0.5mm brass sheet. To do the soldering for the frame I "pre-tinned" the parts to be soldered and then held things together with a heavy steel bar. Then I heated everything with a small butane blowtorch with a sacrificial piece of wood underneath.

In order to solder to the frame without burning the panel, and to have access to more sides of the circuitry when free-forming, I decided to make the final side removable so that the panel can slide in and out of the frame. It is hard to explain in words how I did it but I think it is rather clear from the pictures.

There is likely a proper way to do this, but I coloured the parts I did not want to wet with solder in sharpie. It worked well but the sharpie was a little hard to remove after heating.

I also added a slight divot with a centre punch to hold the sections of brass together which started out way too tight and is now way too loose so there has to be a better way to do this also.

Soooo pretty!!! Cant go wrong with matte black and brass.

Next, a brace was placed across the middle of the frame to connect the motor mount and switches to.

The upper motor mount was filed to fit and the whole thing was soldered together. At this point there are a lot of solder joints in close proximity and I needed to pump a lot of heat into these final joints as they are quite large and on thick brass. So I covered a lot of the close joints with wet paper towels while soldering which I have done in the past and works well enough but you still have to be careful not to melt the whole frame back apart

Note: If I were to do this again I would try to find some mechanical method (screws etc) to hold this bracket onto the frame as I am very worried these solder joints will break and then I don't know what I will do ='(. I made sure to make everything mechanically repairable but did not think about this part and it is pretty key to the whole structure.

These round parts were fun to make as I don't have a lathe =D. To approximate a lathe with my drill press I bolted a cheap vice to my drill press and then used the chuck to place a drill bit in the vice, perfectly perpendicular to the chuck. Then I can put the workpiece in the chuck and now we have a terrible lathe! As you can see from the chips it creates, its not so bad as long as the bit is sharp.The rest of the process is a gain, best explained through the photos. I needed 2 sections with a hole through the centre so it made sense to make them both together and then cut them apart. The first section, that I am showing most of the process of in this step was the more complex part and needs a couple holes cross drilled and a threaded hole drilled for a screw and a grub screw.

The second part section of this part just needs the hole drilled right the way through and then I used a rotary tool to create a curved cut out. This worked much much better than I expected

Then I could join the two sections together with an M3 nut and countersunk bolt (I ended up changing these to brass nuts and bolts later on). Then I can put a grub screw into the threaded hole and line it up with the flat of the motor shaft. I originally bought 3mm long M3 grub screws but 4mm was a better choice in the end.

Now I can put together what we have so far. I know I have said it before but wow... beautiful =D! In the final photos you can see me starting to test where everything can actually fit on the final free-forming.

EDIT: Completely forgot to mention that between each of the surfaces pivoting against each other when the motors turn I stacked 3 phosphor bronze washers that I have from making knives. This supports the joint and created a good enough bearing surface.

I went to a lot of effort to make the mechanical parts of this project replaceable in the hopes that this project lasts for decades. Unfortunately, in the end, it is really hard to screw the switch on and off so this was a little bit of a fail. In any case the plates are 2mms thick and the bolts are M2 bolts.

The base is a bell jar from Ikea and the brass mount was made in the same way as the other motor shaft holder. I drilled the hole out with a 10mm forstner bit and it actually came out really clean.... and then I mangled the edge trying to put a chamfer on the hole to make putting the brass rod in easier >=(. I was initially going to make a brass plate to put around the base which would have covered it but I ended up liking the minimal base and in the end, the poor chamfering is hardly noticeable thanks to the brass rod's shadow

It took really long for the bell jar to arrive with lockdowns here in Sydney and when it did, I was quite a way into the project. I thought from my drawings that everything would fit but I did not know the exact dimensions of the glass. I also had to make the panel sit reasonably far away from the up/down motor pivot point so that I could use the switches. In the end everything is a suuuuuuper tight fit and I was so lucky to get away with this.

Two of the components needed some preparation.

1. The jumper on the back of the panel must be cut to turn the panel into 2 panels that we can electronically connect in series/parallel
2. I cut the tops off the MOSFETs with a jewellers saw as I was not planning on passing much current through any of them so no need for the extra thermal dissipation, and I needed the space.

Honestly, this is probably the most confusing part of the free-forming to follow so it is unfortunate that it is up first. When you are free-forming around an IC (as most of the rest of the sections are) everything stays a lot clearer whereas this is quite all over the place. The main goal was to keep everything as small as possible as I didn't know how big everything else would be in the end. To start with I played around with the components a little then drew out a logical way to arrange and connect the physical components. Then I just started following that drawing however felt most logical. I can't explain my specific process of free-forming but hopefully the progress pictures give some idea.

I initially wanted the power resistor to sit away from the rest of the circuitry but in the end neither the MOSFET or resistor get too too hot even in the brightest sun and with the motors disconnected.

This is an easy one to test, just add an LED to the output and hook it up to a variable power supply. Definitely important to test this one though as the circuit is poorly designed and super sensitive to component variation meaning the resistor in series with the zener diode will need to be tuned.

In terms of mounting to the rest of the circuit, this was also an important consideration for this block as it was on the side with the removable channel for sliding the panel in and out. As such, I would not be able to mechanically anchor the circuit to the frame anywhere along the side so everything had to be connected more rigidly. To connect the grounds I filed into some 1/16th brass tube that fit over the 0.032" rod and to join the positive supply I found a place on the solar panel logic that was quite close so I could use a short piece of wire and keep it rigid.

I like to test a lot. It takes a little time to test so often but it i quicker than having to troubleshoot a problem amongst a dozen sections that weren't tested individually.

Somehow I managed to not buy any of the small, 1/8W, 100ohm resistors so this is the only standard sized resistor on the model. I justify it by telling myself it is the only LED that I run at full power so it needed to be a 1/4W... right =') ?

I didn't know if I would have space for a large capacitor on the solar logic circuit so I just built it up as I saw fit and then added the largest one I could which happened to be 100uF. This cap was not needed but it did help keep everything stable when the solar logic was switching between parallel and series so I wanted to add it if I could.

To connect the 7.5F capacitor to the overvoltage discharging circuit I used 1/16 brass wire. Not so much for the current handling but just because it was quite a large gap to span so the extra rigidity helped. You will see me use this trick a lot but I place 10nF caps between positive and ground in a few places. This is because our brass structure is grounded so all the positive connections would be mechanically not very well connected to the structure. Placing a small 10nF capacitor provides a great mechanical connection to the frame and harms nothing. In this circuit I don't think noise ended up being too relevant but it should even help with noise reduction.

Next up is free-forming the DSSSE. You may notice a couple of these values are different throughout the build as I was experimenting with resistor values to get the correct on and off threshold voltages.

I figured this one would be easiest to test in situ so I connected up the positive and negative then checked one of the outputs on my oscilloscope while playing with the capacitor voltage. I had to mess around with the values quite a bit to get the hysteresis how I wanted. In any case, I ended up needed a ~50k resistor for the larger tuning resistor and I didnt buy any 1/8W resistors around 50k so I just stacked 2 100ks in parallel.

It is super hard to see from the crazy confusing pictures but this is also when I added the connection from the DSSSE to the capacitor switching MOSFET as seen in the final photo.

I did leave the flashing circuit until after I had verified the DSSSE was working as the DSSSE is super finicky so I wanted to test it as soon as there was enough built up to test. It was quite tricky to fit all the flashing components in the space so this part was a little messy

At this stage I actually hooked the solar panel up to the circuit properly and made sure all the charging worked as expected.

Next step was to add a positive rail from the 330mF capacitor. Again I used 1/16 brass rod and mechanincally supported it with 10nF caps to ground.

I don't know why I have so many photos of this step but I think I was excited it was finally starting to look like my vision.

This is something I completely forgot to consider in my original design and as such I did not take too many photographs of me making it as I didnt know what would work. I started by making this rod out of 3/16 (4.76mm) brass with a M3 thread tapped into either side. I again used my drill press as a lathe to get the holes perfectly centred (well not at all perfect actually but more than good enough). Then I used some blutack to estimate how much weight I would need to counterbalance the solar panel and other circuitry. Finally I cut some rough circles out of some 3 and 2mm thick brass sheet and rounded them by chucking them in my hand drill on a bolt and running them against my belt sander. Trying to make circular parts like this is terrible and always ends up with an oblong part but for aesthetic purposes these circles are more than good enough and really do look like they were done on a lathe.

Finally I had to drill and countersink a hole in the frame which was super nerve-racking and you can see this hole in later photos.

I took the audio negative supply from the DSSSE and routed it down to where the audio processing was going to happen. I held it in place mechanically with some 10nF caps.

Incredibly, this relatively simple section of free-forming is controlling all the movement of the head. Indeed, only 2 of the 74hc240's inverters are even used for the logic and the other 6 are just to be able to output higher power to the motors. As soon as the logic was done, I test the circuit by adding it to my breadboarded prototype, and then finished connecting the output buffers and made a second circuit for the other axis control.

A total of 3 connections (besides the motor and photodiodes) need to be made to attach the PSHs. First we need to join the enable pins with a very bent piece of brass. Then we need to connect the Vcc and GND of each 74hc240 to the frame and power rail on the final model. Finally I connected the enable pins to the last output of the DSSSE.

After all this effort we can finally connect the motors and photodiodes to the circuit and see if it moves!

I next built up the microphone amplifier. At this stage I realised I would be quite tight on space for this section so I really had to start squeezing things together. Again you can see me testing the circuit quite often, in this case just using an oscilloscope to test that the microphone signal was getting amplified.

Nice and easy, just needed to bring over the positive from the 330mF and negative from the DSSSE. I could have used this just as an enable signal but I decided to use it as the entire ground rail for this section so as not to waste any extra power while charging.

Honestly, I put this part off for way too long because while the nearby connections all have logical paths, I had to make a connection across the entire model for each set of eyes. Really this should have been done first but with how tight everything ended up being I think it ended up better doing this toward the end and snaking the connections around everything else. I did end up needing some heat shrink in places by this stage

This photodiode bridge is just temporarily connected to the PSH which you can see in step 40 but it will be connected better in step 47.

These were a little simpler as the connection between the diodes doesnt need to snake around and the diodes are very close to convenient positive and negative supplies.

I completely forgot to take a photo of connecting the photodiode bridge to the PSHs but I managed to crop a photo to show this connection.

Yes more testing but at this stage it is ready for a full test! Just without the final microphone stuff.

I rushed the photodoide stuff in between the audio circuitry there because I wanted to show this off in a competition for one of my university groups and I didn't know if I would get it done in time so I focused on getting the light-seeking working and then went back to audio. Luckily I did get it all done in the end though.

In any case, I made quite a few mistakes forming this section because it needed to be compact and I was rushing. This was actually the only place I made mistakes free-froming so I am quite proud of that.

As with the rest of the circuit modules, first step to add this section was to connect the positive and negative supplies. Then a final 10nF capacitor is used to connect the output of the amplifiers to the audio processing.

Pretty hard to notice but I screwed up a connection in the audio processing section and had to fix it now which was fun.

Adding these 3 LEDs that show the state of the audio processing was also a bit of a problem with how compact everything needed to be here so I had to place a lot of the circuitry on top of what was already built. Luckily there weren't too many crazy connections to make.

Perhaps the simplest section of circuitry to free-form is these 3 oscillators that make the robot dance. I played around with the values of the resistors to make the oscillations a little random.

Attaching them is also super simple. Again just connect the positive and negative and then connect the output of the audio processing circuit and connect it to the oscillators' enable pins (I started making other connections before taking that last picture so there is some continuity error, sorry).

The final major connections to make are the connections between the dance oscillators and the PSHs. I used this long connection to more cleanly connect the left/right photodiode bridge to the PSHs.

I had this cool pink, hexagonal LED that I used for the flashing dance LED which I really like.

Limit switches were then soldered together. I actually thought I wouldn't need any insulation on this connection but I did end up using a little electrical tape to protect from shorting the frame. The 10k resistors are just in case something goes wrong and both switches are pressed at once, we don't get those massive super capacitors shorting through these tiny switches. Then we need to make a small, winding connection to the PSHs.

Finally we can slide in the solar panel! With everything added in the end, the panel was actually very hard to press in and you may see in the final videos and photos that I actually drilled two small holes in the frame to help push out the solar panel in case I ever need to.

In any case, the connections that need to be made to the solar logic circuit are quite convoluted without really looking at the free-froming and schematic. Just know that 4 soldered connections need to be made between the panel and the circuitry and I used tome very flexible silicone wire for this.

Omg it is done.

I also found a place to sneak in this extra 1000uF capacitor. It is not strictly necessary but playing around with other circuits I found that trying to control a motor with quick pulses out of a super capacitor can sometimes be a problem thanks to their relatively high ESR. As such, I add 100-4700uF capacitors to all my circuits with a motor and super capacitor for that initial burst of energy, even if it doesn't appear to need it. Ideally this would be as close as possible to the motors but I wanted to hide it out of the way.

With all of these circuits I find that adding extra trim pots to tune the circuit for left right balance and such is unnecessary, even on a circuit this size, as I can tune these things by physically blocking out light. The final tuning of this model ended up being a really tough task and took me many weeks. It was oscillating in bright sun but I could not manage to limit the sensitivity of the system with resistors. I also had no way to tune the left right balance. So I killed two birds with one stone by adding these little "covers" over the bots "eyes". This worked alright but I really wish I didn't have to cover up the photodiodes as I really like the way they looked. Oh well. I ended up with 1mm and 1.25mm holes depending on how the bot needed to be tuned.

I have undoubtably made mistakes and left things out of this instructable so feel free to ask any questions