FulanoDetailNow I will make the Voice Coil actuators.
I’m thinking of making the voice coil actuators using flexible resin, so they can be easier to handle and can act more like muscles.
For the reverse motion I will keep the pulley system.
(you add another pulley for reverse motion)
I will put the pulleys in the bearing section.
I said previously that I had the intention of making the voice coil actuators with 30cm of length, 14cm of outer diameter and 10cm of inner diameter, but if you insert these values in a hollow cylinder volume calculator, it gives 2.5 times more volume than the now 0.9kg version of the Xangô-33.
So, now you would make it with 11cm of inner diameter, 3mm of thickness, 11.6cm of outer diameter and 30cm of length.
With the 2kg version it would have around 8.6mm of thickness, 11cm of inner diameter and 12.7cm of outer diameter.
Of course, this wouldn’t be able to support 3 to 8 tons of weight, so I don’t really know what to do now…
The first idea I had was to put the aluminum wire on the inner diameter and then increase the thickness using structural material like glass fiber + resin.
But what guarantee do I have that the aluminum coils won’t be ripped from the structure?
Another idea I had was to mix the iron or ferrite powder with the resin and keep the coil inside of it, but we go back again to the problem of weight.
In any manner, I will 3D model it, because otherwise I can’t make the Mech’s hands work with the brushless motors… They are too big. But after that I will at least try to do a super compact version of the motors specifically for this case.
Anyway, 3D modeling the voice coils:
110mm of inner diameter, 150mm of outer diameter, 300mm of length.
Well, that was easy, lol
Kidding, I will add the cooling channels.
The idea is that the honeycomb design saves weight while allowing for cooling of the coils. I divided the outer shell in two disks 1mm apart so the aluminum coil can fit in (and also because it will be easier to make a mold out of it).
I did the same thing for the piston rod/moving part inside of it.
300kw on the left, 100kw on the right.
By the way, if you select the inner vertices of the cylinders and rotate, you can make the honeycomb pattern rotate too. Dunno if it will make it stronger or not.
In voice coil actuators the soft magnetic core surrounds the coil inside and out, I will add this option, but I don’t know if it will be strong enough…
The 100 kilowatt voice coil (right) is just the scaled down model of the 300 kw because I’m lazy.
I remade the brushless motor to be a super light motor, but now I have to do the same with the voice coils.
Assuming that it was made with 80% epoxy resin + 20% iron powder resin it would have a density of 2.5 g/cm³ and in the end the 300 kilowatt voice coil’s structure ended up weighing 7.9046kg alone.
The density of silicone rubber can reach 2.3 g/cm³ (not good with iron powder’s density) and polyurethane density is around 1.4 g/cm³ (which would increase to 2.5 g/cm³ with iron powder), so it would be the same thing anyway.
My biggest concern is that unlike the motor, the voice coil will support all the weight, and I don’t really see how it can survive the loads with something as thin as the electric motor.
So I don’t think I will reduce the weight even further, to be honest. I will only use the 100 kw voice coil for the fingers and/or the shoulders (and/or the ankles).
The 100kw voice coil actually ended up with 0.775kg.
Actually, I think I will make a new cleaner version of the voice coil, I will do the same thing to the axial flux actuators.
For some reason the lack of a plain version of them is itching my brain.
There.
Now I will add the encoders.
Oh boi, at the moment I have no fricking idea how I will make them.
Although I will 3D model the encoders, depending on the type you use, you will have to alter its dimensions. There are already a lot of open source designs (and cheap off-the-shelf encoders too) out there, so really, using my designs is always optional.
How to pick the right resolution encoder | US Digital Encoder Support
What is the difference between resolution, precision and accuracy? | Encoders 101
Why use Gray code for an optical shaft encoder?
Binary and Gray Shaft Encoders
What is the Difference between Absolute and Incremental Encoders?
How Encoders Monitor Position - Physics (Position Activity)
Programming Encoders - Boolean Logic | Arduino - Ep 8
Magnet Motor concept and How magnetic shielding works
https://www.thingiverse.com/search?q=binary+encoder+&page=1#google_vignette
3D Printed Absolute Encoder Is Absolutely Wonderful | Hackaday
openDog Dog Robot #15 | Encoders & Hardware Upgrades | James Bruton
MechSense:A Design & Fabrication Pipeline for Integrating Rotary Encoders into 3D Printed Mechanisms
High Precision Rotery Encoders
9.5 encoder strategies for open source bldc control
i designed an encoder for my robot
Precision motion control: ODrive Servo? Trinamic Stepper? Chinese Hybrid?
HACKED!: Using an HDD Motor as a Rotary Encoder?!
DIY Encoded Pulley and Guide to Make Encoder With Hall Sensors & Magnets : 7 Steps - Instructables
How to use encoders (Optical, Hall Effect, Quadrature)
Draw Wire Sensor for Length Measurement by atzensepp - Thingiverse
DIY draw-wire position sensor by André Colatino | Download free STL model | Printables.com
I intend on making them multi-layered and purposefully miss-aligned, so the encoder can know both its location, speed, direction of rotation and redundantly know each-other’s position. That would be an incremental-absolute hybrid encoder, no?
In essence, encoders are like a mix of limit-switches with punched tape readers.
Anyway, my designs:
Although I intend on making all of these, it doesn’t mean that the actuators will have space into them for that.
- First: resistor encoder = multi-layer cone-like taper across the sides.
There are many types of resistive encoders:
Rotary Switches & Mechanical Encoders
- Second: optical encoder = There will be a multi-layer optical encoder with the number of degrees from 1 to 360 in decimals in binary code.
There are other types of optical encoders, such as transmissive, reflective and interferential.
I just saw this mechanical speed reader, and I think that this could be an interesting option for optical encoders (and resistors too).
- Third: electro-mechanical switch encoder = similar to optical encoder there will be bumps/holes that will align with the switches, turning them on or off. Just like a music box.
- Fourth: I will make each encoder scalable, so I either fit them into the motors, or I make them a draw wire encoder/string potentiometer.
Magnetic encoders and inductive encoders are also interesting, but they may be manually exhausting to build/add in this specific setup. I mean, assuming you would add hall sensors through the surface of the motor so you could precisely detect where the magnetic fields are.
Resistive encoders done. I did 3 different patterns using the “Array” modifier, increasing them from 5mm to 10mm and then the “Curve” modifier to turn them into a circle. The already existing curves don’t have enough resolution, but you could just as easily add a circle using the “Shift+A” command and change its resolution in the creation box to 200. Finally, you can click with the right mouse button to convert it into a curve.
I couldn’t fit the 360 signals with decimals, it would be too small, unfortunately. Only 360 dots.
In any manner, you (and me) will need to redo the whole process in order to fit certain objects into the motor.
The optical encoder was a pain in the butt to do in blender, but essentially, I had to download an open source binary font, add that to Blender in “text mode” and then copy-paste 1 to 360 in a vertical column.
The problem is that the font doesn’t exactly translate every number to binary, it only replaces each number to its binary counterpart.
Also, every row has more or less 2mm of thickness and the width of the full pattern tape is around 30mm
Right now I would make the electromechanical encoder with the switches, but you can just use the optical binary encoder pattern as the pattern for the switches.
Just like that.
Of course, the problem will be to fit these patterns into the motor…
On this side there is the resistor encoder and the electromechanical encoder, on the opposite there is the optical encoder.
Optical encoders are normally made with holes to see through, I think It could be more practical for this specific application to coat the encoder with a reflective paint and position both the optical sensors and LED’s in the same base.
Unfortunately, I will need to 3D model the rest of the structure in order to properly position the base for the encoders.
I actually didn’t like how the whole binary symbol is based on a font that may or may not be available after a few years, so I redid it.
In this case I simply asked ChatGPT to count from 1 to 360 in binary code with 9 digits (conventional binary also converts each different number on its own binary sequence) then I just used the “Ctrl+F” in a document and then the option “Localize and replace” to replace all the “1” with “▀” and all the “0” with “█”.
So all the protrusions (I forgot the name) are “0” and all the plain spaces are “1”.
I Didn’t want to use a symbol for 1, but using spaces (and any other symbol) always misaligned the text on Blender, and I don’t know how to fix it. It starts at “000000001” (1) and ends at “101101000” (360), which in turn would be “████████▀” (1) and “▀█▀▀█▀███” (360) in the 9 digit binary code.
… Now that I think About it, how would the computer/encoder/program know which are the numbers in between?
I will add a document to the archives.
Like the idiot I am, I painstakingly removed each single “▀” by hand out of the 3D model.
All of the encoders I designed can be exchangeable depending on its mode of operation.
The optical ones can be made using a conductive paint, the resistive encoders can have a reflective layer that can be used for the optical sensors etc.
You can also just make a smaller encoder and connect it to the motor using draw wires, pulleys, gears or all of the options. This way you could have a compact encoder with extremely high resolution, since a single rotation of the whole motor would apply hundreds or even thousands of rotations to the encoder.
Also, it would be able to even use the optical encoder with the hall sensor. You would need to use a ferromagnetic paint on the optical encoder, then apply a really weak electromagnetic field behind the rotary plate and the hall sensors would be on the other side.
The ferromagnetic paint would block the super weak electromagnetic field, breaking the signal just like in an optical encoder.
Since the encoders were made with a 250cm diameter and divided into 360 parts, they would have around 2.11 mm of resolution.
Not that great…
But the bigger the reduction ratio of the rotation of the motor and the encoder, the bigger the resolution you will achieve with the encoder.
So, if you add a reduction of 1000:1, it would have a resolution of 0.78mm, or 780 micrometers. Depending on the size of the optical disk of the optical encoder, the resolution would be multiplied. Example, the optical disk has 10cm of diameter, it is divided into 360 parts, then it would have 0.87mm of resolution, but since the encoder is attached to the motor with a reduction ratio of 1000:1, then it would be 0.87/1000 = 0.0008 mm of resolution. Or 0.8 micrometers, or 800 nanometers.
But what about 1 billion rotations? 1 billion rotations of encoder per rotation of the motor = 1 nanometer resolution.
Ultimate Reduction: ultra-compact billion to one gearing, 1.000.000.000:1 !!! (this one gearbox achieved 1 billion to 1 gear ratio)
But that would mean if you reached 100 rpm at the motor, that you would reach 1,000,000,000 rpm at the encoder.
Wouldn’t that absolutely obliterate the gear itself?
1 000 000 RPM !!! TourneBille Ultra High-Speed (18000 tours par seconde)
I think the only viable way to do that would be to use magnetic gear reductions, where the final rotating object would be a small permanent magnet rotor. Just like in the videos, the hall sensor would detect how many times the small magnet rotated.
How To Make Invisible Gears With Eddy Currents
Since the 1 billion to 1 gear uses 3 cycloidal drives with 1000:1 gear ratio, you could use fixed electromagnetic cycloidal gearboxes that end up in a free rotating magnet with a hall effect encoder. Of course, the first stage of 1000:1 would be the connection between the motor and the encoder.
So, the circumference of the 250mm rotor (785.398mm), divided by 1000, then further 1,000,000 is 0,7 nanometers.
Of course, assuming that everything is perfectly aligned (which won’t be) and that the magnet rotor will perfectly stop and/or revert its direction of rotation based on the movement of the motor (which also won’t happen).
I think I will also try to 3D model these small encoders, except the supposedly sub-nanometer encoder.
It would be too complex and I’m not so sure it would work properly.
Should I add a gray code optical encoder option?
Added it just for the sake of it, lol.
By the way, I will stick with a belt/pulley system to keep the encoder compact.
But since it will achieve speeds of 100,000 rpm, you would need to make DIY magnetic bearings (which would be easier due to how small and light the rotating system needs to be) or you could buy small high speed bearings, which are somewhat cheap in quantities.
The first stage of the high speed encoder is a pulley disk with 25mm of diameter connected to the motor at a distance of 250 mm from the center, resulting in a 10:1 reduction ratio. Then it will be connected to a cylinder needle with 0.25mm of diameter connected to the optical encoder, resulting in a 100:1 reduction ratio, with the previous reduction it will result in 1000:1.
I added the dimensions of the encoder disk with its 100k rpm in a tip speed calculator, and its tip would be reaching mach 1.5. Soooooo, maybe it would be a good idea to reduce its size so the plastic film doesn't just tear apart.
Here is the abomination:
Again, you could/should use conventional and accessible encoders instead of this thing that may or may not work reliably. On top of that, a 1000:1 reduction ratio is overkill and you could just use a 100:1 or 10:1 reduction ratio for a conventional encoder that you can buy online.
I also forgot that you could use small cheap brushless motors as encoders themselves.
The movement of the permanent magnets on the rotor would generate electricity just like in a dynamo, the frequency and intensity of the electric waves generated could be used to estimate the position of the motor.
You can see this happening on those mirrored brushless motors:
BLDC Motor: sensorless haptic control
Low-Speed / Fine Positioning Brushless Motor Controller
Software on Paper - 1985 Cauzin Softstrip (maybe you could use something like this, using a laser encoder to make the strip and a scanner, even the laser itself)
About the locking mechanism:
The first Idea I had was to use a mechanism similar to that of car brakes, using a mechanical (hydraulic or electromagnetic) locking mechanism to keep everything in place with a clutch.
How hydraulic and/or pneumatic car brakes work.
THis is an eddy-current brake, but if you allow the electromagnet to complete the circuit with the disk, it becomes a holding electromagnet.
Another idea was to have a hydraulic/pneumatic/damper/shock-absorber cylinder that would have solenoid valves that would close once the thing stops moving. You can always make it electric.
I also thought of the possibility of adding a really small (and slow) hydraulic pump to also act as a hydraulic press.
But these aren’t the only options.
Variable Stiffness Composites:
Personally, I never gave enough credit to this type of material simply because the ones that I found were always… Well, I say this because normally they use vacuum or pressurized air for said variable stiffness, which has the implicit disadvantage of failing instantly in case of any kind of puncture.
Variable Stiffness Links VSL - Toward Inherently Safe Robotic Bio‐Inspired Conformable and Helical Soft Fabric Gripper with Variable Stiffness and Touch Sensing | Request PDF Design of a Lightweight and Deployable Soft Robotic Arm
(Yes, you could make them segmented so one part wouldn’t make the whole fail, but still)
Some electrically activated variable stiffness materials use dielectric elastomers and/or low melting point materials as the method of choice to “lock” a material in place.
Sources: https://advanced.onlinelibrary.wiley.com/doi/full/10.1002/aisy.202300131 Variable stiffnes materials — Michelle C. Yuen
You could use this type of mechanism for a completely flexible and safe equipment to use around humans, while also making a high load capacity continuum robot viable since it is stiff and malleable all at the same time. The issue is quantifying the stiffness to flexibility ratio.
Algorithmic Design of Low Power Variable-Stiffness Mechanisms (this may help)
Also, you could use this variable stiffness as the lock mechanism for the limbs/actuators once they aren’t doing work.
Design Optimization and Reliability Analysis of Variable Stiffness Composite Structures
Adaptive Joints with Variable Stiffness
Variable Stiffness Actuators: Review on Design and Components
Design, modeling and testing of a compact variable stiffness mechanism for exoskeletons
Soft Robotic Manipulators: Designs, Actuation, Stiffness Tuning, and Sensing
A Scalable Variable Stiffness Revolute Joint Based on Layer Jamming for Robotic Exoskeletons
Layer jamming: Modeling and experimental validation - ScienceDirect
Electrostatic Layer Jamming Variable Stiffness for Soft Robotics | Semantic Scholar
A Soft, Controllable, High Force Density Linear Brake Utilizing Layer Jamming
Output-Based Control of Robots with Variable Stiffness Actuation
Improving the performance of industrial machines with variable stiffness springs
Variable Stiffness Devices Using Fiber Jamming for Application in Soft Robotics and Wearable Haptics
Variable stiffness and fast-response soft structures based on electrorheological fluids
Variable stiffness material and structural concepts for morphing applications
Miniaturized Variable Stiffness Gripper Locally Actuated by Magnetic Fields
https://advanced.onlinelibrary.wiley.com/doi/10.1002/adfm.202107662
A Variable Shape and Variable Stiffness Controller for Haptic Virtual Interactions | Research
Stopping to think about it, maybe I’m overthinking this whole variable stiffness (I think I overthink everything too much).
New King of Magnetic Power? | Electromagnet vs. Neodymium Magnet
A holding electromagnet consumes around a few watts just to hold hundreds if not thousands of kilograms with little to no issue. Wouldn't it be more practical to use them with magnetic materials?
(random sheet specs of a random holding electromagnet I found on aliexpress)
The strength of the magnetic field would determine the stiffness of said material in whatever section of the body is designed to bend or lock. It could even use the layer jamming thing or even magnetic powder jamming.
5 Tips To Make A Good Electromagnet / How To Calculate Electromagnet Force?
Making a Powerful Electro Magnet from a Transformer
Full article: Variable stiffness wires based on magnetorheological liquid metals
https://advanced.onlinelibrary.wiley.com/doi/full/10.1002/aisy.202200465
New universal gripper using MR alpha fluid
3D-Printed Ready-To-Use Variable-Stiffness Structures
Iron powder for variable stiffness backbone, ferrofluid for cooling…
How long until I decide to make a T-1000?
Discussions
Become a Hackaday.io Member
Create an account to leave a comment. Already have an account? Log In.