This guy's work and dedication is truly impressive. He posted a video [1] about his concept of a PCB drone 4 years ago and he and his methods have evolved so much. In addition to his highly specialized knowledge of designing, testing and iterating PCB coils, he also makes comprehensive and interesting videos out of it. I highly recommend checking out his channel.
I'm deeply unqualified in pretty much every way to comment on the efficacy of this design - but _wow_ am I impressed with both the creativity and the perseverance of this guy.
I love software, but there's just something undeniably cool about hardware.
> but there's just something undeniably cool about hardware
Personally I just like going lower level. I like Ben Eater making whatever out of a few chips. I've set my line at asm else I'd just end up making a whole computer out of transistors or sand.
I find Sam a little too dismissive sometimes. He has relatively wealthy parents who let him convert a garage into a home fab, and talks about dropping $30k on a second hand ESM like it's no big deal.
Don't get me wrong, he's a genius, but everytime he transitions into "anyone can do this stuff!" It rubs me the wrong way
Your comment reminds me of those woodshop youtubers with clickbait titles like "diy your own coffee table for only $50", only to find they are using about 10K or more of shop equipment.
> He has relatively wealthy parents who let him convert a garage into a home fab, and talks about dropping $30k on a second hand ESM like it's no big deal.
The envy I feel when I see these guys with such equipment is enormous. I really would love so much to have a loaded electronics work bench especially oscilloscopes, LAs etc. and of course the time to actually play with it. :-(
Check if you've got a local hackerspace! There are a lot of good ones out there. By and large you will find excellent people who want to enable you to do amazing things.
Analog engineering is really difficult because you need to solve differential equations that Wolfram is unable to solve. Most designs were made by very smart people in the 50s at RCA as far as I can tell. Please tell me about how wrong I am. I admit the analog circuit engineering world has looked totally unapproachable to me even as a reasonably competent working computer scientist.
the good news is you don't need exact solutions and there are pretty robust tools for numerical diffeq solving now. the best are Julia's ecosystem, but sundials is pretty decent also.
Something like making a new audio amplifier is pretty specialized, but you don't need any advanced math (Not even algebra) for some of it. As long as you're still pretty much thinking time domain and you don't need to calculate frequency responses and RF stuff there's lots to do.
But then again my only analog experience is just IO stuff for digital, I've never done anything interesting in pure analog, I think if you're a real analog enthusiast you might get bored of what you can do without math.
When I was looking at catalogs of various analog circuit designs, a lot of them were circuits RCA patented that now are public domain because the patents are now expired. They probably aren't the only ones, but they certainly contributed a lot.
I like how he avoided paying $150 for a development kit by whipping out Altium Designer, which I believe costs $4000/year. That must be a paid promotion.
Yes he is sponsored by Altium and regularly promotes them. I would prefer he used Kicad so everyone could easily use his work but this is the nature of earning money on YouTube sadly.
Phil's Lab is another great hardware design YouTuber who also produced some great Kicad content, until Altium paid him to partner with them too.
Sure, you can get a free trial through any of the altium.com/yt/<channelname> sponsorships, but if you want to continue to have access after the trial it's merely $12,000 for a perpetual license or $355/mo for a subscription. Just absurd for anyone who isn't using this professionally.
Some people pirate it - there are loads of Russian forums selling cracked license keys; be aware that the software with the cracked license will still 'phone home' and earn you a call from Altium legal in a few years with a lawsuit and your public IP/home network MAC addresses, especially if anyone with a legit license opens the files you produce. But it's just not better than Kicad, with the exception of the push router. It's sluggish, and constantly gets more sluggish... "Please wait a moment" is a blatant lie, it should say "I hope you saved recently". You get used to hitting Ctrl+Shift+Esc and force-quitting DXP.exe, because it hangs for 30+ seconds when you try to restart it the normal way. The scripting language is atrocious, but everything is built on layers of decades-old scripts. And when you run any tool that uses a script, you lose the ability to undo past that point. I hear it's no longer written entirely in Delphi, with the infamous "Ignore segfaults" option... but I've long since left for Kicad, when my education license ran out and resetting my VM stopped working I looked at the writing on the wall and ran.
I mostly do industrial automation - PLC, CNC, and robotics projects - and prefer to buy parts off the shelf for 10x the (actual) price of building my own PCB to avoid being the only person on the planet who can support that machine, but sometimes the shelf just doesn't have anything to do what you want. And Kicad has been perfect for that.
Yeah I have had the experience of really enjoying a Phil's Lab video, wanting to download the source files, and then seeing that they are in Altium format. That's so unfortunate for newbies who just want to learn!
I mean, you could s/Altium/KiCAD. Altium is also free for University students/faculty/staff, FWIW.
Lots of profession electrical engineers that I’ve met pay for personal copies of Altium for their moonlighting work, however. One short contract per year more than pays for it.
Hobbyists stick with KiCAD though, which would suffice for this work. Altium is mainly used by contractors for compatibility with their corporate counterparts.
Just passionate and talented hobbyist. It's not exactly [1] new invention, it's not being done because building proper engine is not all that more expensive, especially when you take into count suboptimal PCB coil design. Still need a bunch of custom elements too.
Kinda like coilguns, a fun hobby to fuck around and optimize but overall there is a reason it's not really commercially done all that much.
But helluva edition to resume if he ever decides to work in industry!
Because Ken Burns didn't make his living by making videos of him tossing a ball with his son. This guy is making a living by documenting his engineering via video. He is not a hobbyist, He is an engineer who makes money because people find his engineering projects interesting.
I didn't say professional engineer, I said professional. Not that distinguishment matters.
Him showing his engineering projects on Youtube and getting himself paid like that means he is a professional. The main difference between a hobbyist and a professional is simply whether you get paid for doing it.
I don't see how that is relevant, in fact it's not any different then any electrical engineering working in a company. No company is paying for engineering. Companies are paying for a product/service/feature. Engineering is just a means to an end. Exactly how it is for Youtube in this case. The engineering in the video entertains thousands making Youtube money.
Exactly - he is entertaining and maybe educating. So, fine - he's a professional entertainer or educator. If he were being evaluated and paid based on the quality of his electrical engineering output, then he would be a professional electrical engineer.
He is being evaluated and paid based on the quality of his electrical engineering output. If it's not good it's not entertaining for his particular niche.
But that is beside the point. Whether he is paid for "quality of his electrical engineering output" doesn't matter. He is paid for doing engineering, which means he is a professional.
Is he ? You are right that he isn't a hobbyist since I'm pretty sure he has formal training, but I thought he actually worked in an engineering related position. Did he make the transition to full time YouTube?
Great work this, I wonder if the end goal is a drone why it uses such a strange rotor design rather than something more conventional (which also would obviate the need for such extreme RPMs).
The size of most electric motors is proportional to the torque they produce - so if you can spin it faster, you can get more power for the same weight. There are some other factors that make high speed worse - for example, bearing and core losses. The PCB motor is a coreless motor though, so it has no core loss, biasing it even harder towards high speed operation.
Propellers generally get less efficient with higher speeds though, which might be a bigger factor than any of this.
> Propellers generally get less efficient with higher speeds though, which might be a bigger factor than any of this.
Unless operated at lower air-pressure (higher altitude) then higher speed is generally more efficient.
Source: Musk interviews about why electronic very high altitude aircraft make sense in a lot of ways (faster, more energy efficient, potentially cheaper, less polluting). But only after battery densities are improved in the future.
Musk isn't correct though, or maybe you misunderstood or misquoted him. Higher speed is only more efficient if the craft itself moves at a higher speed.
To explain a bit further: a prop is at its most efficient when it has clean air to work with and that only works if the craft moves forward at least as much as one prop's worth of air in the direction of motion. Less than that and the prop will encounter it's own backwash. This is the reason why variable props exist, to ensure that the prop has enough 'bite' rather than that it just churns the local air. So at low revs you run a higher angle than when you go faster and towards the tips of the prop the angle gets flatter as well.
Which more or less defines the range of a variable prop, once the tips are nearly flat there is nothing more to gain. Another important factor is blade count. A single blade is theoretically most efficient because it can run at the highest RPM before the blade encounters it's own wake again, but there are balancing issues and vibration issues with low (<3) blade count props. And in practice the efficiency gains are offset by complexity, weight (a single blade needs a counterweight) and drag of that weight. Two is common enough though because it is easy to make a sturdy two blader prop. Three is optimal from a longevity and maintenance point of view, and a offers very good efficiency.
And I think you meant 'electric', not 'electronic'.
> Musk isn't correct though, or maybe you misunderstood or misquoted him. Higher speed is only more efficient if the craft itself moves at a higher speed.
He's got it completely backwards from what Musk said in that interview, see my reply.
Hm, I don't entirely know how this works, but for RC planes we use slower motors and higher prop pitch for efficiency, faster motors and lower prop pitch for acceleration.
If you're going to cite sources like that at least get it remotelyclosetoright.
"For aircraft, or just generally, you want to move a large mass of air slowly. So you can reduce the velocity component of kinetic energy, which goes as the square. You want to move a large amount of mass slowly, not a small amount of mass fast... So, the way you make aircraft engines more efficient is you move a lot of air slowly. Like big fans, basically, big slow fans work great. Small, tiny. fast-moving jets, are very inefficient." - your favorite billionaire, in the exact interview you're inversely misquoting. (JRE #1609 last ~12 minutes is electric planes)
Huh? No way, the energy needed goes up with the cube of the speed, but the mass moved only goes up linearly-- the reason is that you are having to accelerate a small mass a lot rather than a big mass a little. The most efficient is a big slow propeller, but size and mass constraints put limits on how big and slow you can go.
A coreless motor has more losses, not less. That's because the field isn't focused as nicely. It's just that the losses aren't in the core but in the air.
Core losses by the way are typically a different kind of loss, they are eddy currents resulting from the fact that the stator laminates are not infinitely thin.
Yeah, I'm surprised he didn't try attaching some actual drone propellers to see how much thrust they'd give. There's no shortage of small plastic props available for 'Tiny Whoop' style micro-quadcopters, with 2-5 blades and in various small sizes.
He mentions that the goal is a PCB drone, but also mentions laptop cooling as a possible application. The initial design of the motor is a low heat one, as the early iterations had a problem with overheating at low RPM.
Is the end goal a drone? I've watched a bunch of his videos on his PCB motor project and I don't recall him mentioning it was for a drone at all. The title here on HN itself is editorialized to include drone so who knows.
I've been thinking about building a tiny, (relatively) cheap haptic knob. I wonder if this makes enough torque for it to feel right? Maybe with a big stator and a small knob.
Can't help myself: going to arm chair quarterback this.
Since sine waves work better at low speed consider a DAC driving small power amps. A dual output 8 bit DAC could easily generate the frequencies involved and allow tailoring the curve with RPM. Closing the loop would likely be easier and more effective with an optical pickup of some sort: that would do away with back EMF problems at different speeds and increase precision for better tuning, which is crucial for efficiency.
At some point off-the-shelf motor drivers are insufficient.
A brushless motor controller is basically a DAC anyway - so no need for separate components.
A brushless motor is essentially 3 large inductors. They act as a filter to PWM drive signals, which means that the resultant current is fairly continuous - much like a class D audio amplifier uses the inductance of a speaker to create smooth sine waves.
There's no need to add an intermediate smooth analog stage (which is what a true DAC would give you) - because then either you drive a linear amplifier and have massive power losses in the drive transistors, or drive a class D type amplifier (which turns your nice continuous signal back into PWM to the drive transistors) and you've just added an unnecessary digital -> analog -> PWM sequence.
When you want good low speed control of a brushless motor, the gold standard is a high resolution encoder (often magnetic) coupled with "field oriented control" - which is essentially using your 6 drive transistors to create a magnetic field which is exactly 90 degrees ahead or behind your permanent magnet field. You can use entirely digital PWM switching to create that field, as the natural filtering effect of the motor coils smooths it out, and the resultant driver & motor is highly efficient.
The much simpler "6 step commutation" discussed in the video can give almost as good control at low speeds, but the magnetic field you create in the coils isn't necessarily perfectly aligned with your motor's magnetic field, which means some of the magnetic force generated is just pulling on your bearings and not driving the motor. That means overall efficiency is slightly less and the torque at low speeds is variable depending on rotor position.
Would work, but I'd estimate that it would require 4x the circuit area that this fella is working with. Plus adding some Z axis height for the optical encoder.
For what he wants to do (real small, real light PCB based BLDC motor), I'd say his approach is right on track.
A DAC would involve a larger PCB. A small DDS (si5341) would be pretty good though at 3x3mm. Three phased channels could drive the six inductors far beyond any feasible speed here.
> Plus adding some Z axis height for the optical encoder.
A SMT photodiode and IR LED could do this off the edge with some fiddling, which appears to be in ample supply here.
Huh? That DDS is a silly choice. For one thing, those chips are $27 a pop and only available in QFN44 and QFN64. (May have been smaller FF at some other point, but not any more.) The final BLDC driver he selected is like $1.50.
Not to mention that you can't adjust the drive phases of a DDS with I2C fast enough to be of any use for BLDC. Once you add in a sensor/feedback element, it definitely wouldn't have the latency performance you'd need for BLDC.
You'd usually want 3 phases for a brushless DC motor, right? Though I'm sure theres a cheap Chinese 3/4 channel DAC on LCSC since the output quality won't matter.
I haven't considered using a DAC with a motor before, but that strikes me as extremely inefficient, which is one of the design considerations. You're basically going from a simple class D amp to a class A one. The motor has plenty of inductance so you may as well PWM it for efficiency.
I have always wanted an ergonomically laid out electrocapacative keyboard. Maybe this person will one day move on to that project. Definitely a big impact on society.
I'm not sure how big the real impact on society of that one is, but it's the kind of thing you could probably do pretty easily.
Not sure what exactly you have in mind for it so I couldn't give any specific numbers, but I would guess $150 to $500 max if I was doing it, maybe a bit more if you want any super fancy addons, and that's including making the case.
But you'd probably have to specify a layout yourself, otherwise that would be the hard part to actually get that one excellent. The rest is all pretty trivial well known engineering, making the layout actually feel nice to a touch typist especially would probably need someone who's very tactile and in tune with the subtitles of doing things by touch.
Capacitive touch sensors aren't horribly complex. I'm betting this is something you could learn about and implement in a fairly short amount of time (6 months with zero electronics knowledge?).
For some I would think more like 1 month if you're dedicated.
Every chip has capsense peripherals these days. Even the ones that don't have capsense peripherals secretly do if they have an ADC, that can sometimes be easier to make more reliable than the hardware(Although those might be better for larger areas).
[1] https://www.youtube.com/watch?v=mS_qUbPTYfk