YT YouTube2Text

YouTube2Text → Transcripts → The Most Beautiful Ridiculous Thing I've Ever Made

The Most Beautiful Ridiculous Thing I've Ever Made — Transcript

by Clough42 · 6,359 words · 928 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00Well, it's that time of year again. The
0:02sun is bright, the weather is warm, and
0:04the shop is cool and inviting. The
0:07machines are making chips, and the
0:08spirit of gift-giving is in the air.
0:11Also, wildfire smoke, which is another
0:14great reason to stay in the shop. It's
0:16time for Tool Room Takeover 2026.
0:23Welcome back to Cloud 42. I'm James.
0:26Last year, Eric with a K assembled a
0:28motley crew of YouTube machinists to
0:31draw names and make gifts for one
0:33another. People seem to like that, so we
0:35are doing it again with a few new faces.
0:39This year, I'm making a gift for Rowan.
0:41He claims he's not an engineer, but I've
0:44seen the things that he builds, and I'm
0:46suspicious of that claim.
0:48So,
0:50what do you make for someone like that
0:51who has or can make anything that he
0:54wants?
0:55Well, that's a good question. I actually
0:57am a software engineer, but I'm not a
0:59machinist, I'm not a mechanical
1:01engineer, I'm not an electrical
1:03engineer, I'm not a designer, and I'm
1:05definitely not a welder, but I do know a
1:08little bit about all of those things.
1:10So, I think my best approach is to make
1:12something that combines as many of those
1:14disciplines as possible and try to make
1:16something that's truly unique.
1:19The budget this year is $200, which if
1:22you've priced aluminum lately, you know
1:23does not go as far as it used to. But,
1:26the rules allow me to use things I
1:28already have on hand, so I had a look
1:30through my junk bin for free stuff, and
1:33I found a bunch of these NEMA 14 stepper
1:36motors. These were shipped to me by
1:38mistake by a Chinese seller, and
1:41shipping them back was too expensive, so
1:43I agreed to hang on to them and ship
1:45them out to the next US customer who
1:47ordered some.
1:49That was 11 years ago, and I've heard
1:51nothing since, so I think these are fair
1:53game. Now, my first thought is to try to
1:56build a split-flap display with thick
1:59stainless steel flaps and just make it
2:01as loud and obnoxious as possible, but
2:04the finished project has to be shipped
2:05to Australia, so I think that's a
2:07terrible idea. I kind of like it though,
2:09so I'll put that idea on the back burner
2:12and think about it. In the meantime, we
2:14can get started on the electronics. If
2:17I'm going to build a mechanical display,
2:19it needs a purpose. Now, the size of the
2:22motors means the display is going to be
2:23fairly large, so it can't have very many
2:26digits if I want to ship it halfway
2:28around the world, and that
2:29pretty much means it needs to be a
2:31clock. And if I'm going to build a
2:33clock, I want it to be a real timepiece.
2:36It has to be accurate, and it needs to
2:38get its time from an outside source. So,
2:41here's the plan.
2:42The display will be driven by six
2:44stepper motors using TMC 2208 silent
2:48stepper drivers. The whole system will
2:50be controlled using a Raspberry Pi Pico
2:53microcontroller, and the time data will
2:56come from a U-blox GPS receiver. I
2:59thought seriously about trying to
3:00connect to Wi-Fi and get time from the
3:02network, but that introduces a lot of
3:05unknowns that I really don't want to
3:07deal with for this project. I will need
3:10a 24-V power supply for the motor
3:12drivers, a buck converter to step the
3:14voltage down to run the logic, and a
3:16small display to show the GPS status and
3:19set the time zone.
3:21I got a head start on the project in a
3:23video I published back in February, so
3:25go check that out if you want to see the
3:28details. I went into great depth about
3:30setting up the development environment
3:32for the Raspberry Pi Pico, developing
3:34the firmware, and debugging with a logic
3:37analyzer. I started many months early
3:40because I was worried I might have
3:42problems with the integration. I'm using
3:45all import electronic modules from
3:47Amazon and eBay, and they're way too
3:50cheap for what they are. That's
3:52great if you're a hobbyist, but it also
3:54means that you don't know exactly what
3:56you're getting. Is this a real U-blox
3:59NEO-6M GPS module?
4:03Probably not. Should I trust it? For
4:05this project, it should be fine, but I
4:08want to let it burn in on the bench for
4:09a while and make sure it's actually
4:11going to do what I need it to do.
4:14I also asked Rowan to check the GPS
4:16signal strength in his shop to make sure
4:19that that doesn't derail things at the
4:21end. Building something like this and
4:23sending it halfway around the world for
4:25him to try for the first time on camera
4:27is risky, but
4:29it's also kind of a cool idea, so I'm
4:31going to do it anyway.
4:33The end result is this gnarly
4:36contraption. It works, and it works
4:38well, but boy, does it have a lot of
4:41wires. I've had this running here on the
4:43breadboard
4:45for a couple of weeks now, and I'm
4:47satisfied that it's stable, so now it's
4:49time to actually figure out what I want
4:51to do with the mechanics. And I'll keep
4:54thinking about what to do with all of
4:56those wires.
4:58I spent a lot of time trying to come up
5:00with a split-flap display design that I
5:02like, but the scale of the motors makes
5:04that pretty challenging. If I put the
5:07motor inside the hub, the flaps have to
5:10be so large that they become too heavy
5:12to ship safely without additional
5:14support bearings, which drives the cost
5:16and complexity up. And if I put the
5:18motors outside the hubs, the digits end
5:21up too far apart, or I have to have a
5:23complex mechanical linkage that I'm also
5:26not convinced will survive the trip. So,
5:29after many hours of experimenting in
5:31CAD, this is what I came up with. The
5:34motors are mounted in opposing pairs,
5:36with each one driving a large wheel with
5:39the numbers showing through windows in
5:41the front of the clock.
5:43The wheels are skeletonized both to make
5:45them look cool and to reduce their mass.
5:48Keeping the mass as low as possible
5:50reduces the moment of inertia that each
5:52stepper motor has to control, and it
5:54also reduces the shock loads that will
5:56be transmitted into the motor shafts
5:58during shipping.
6:00They should be pretty easy to 3D print,
6:02and they should stand up well in this
6:04application.
6:06The frame of the clock will be machined
6:07from 6061 T6 aluminum and held together
6:11with M3 screws.
6:13The shiny aluminum should make a really
6:15nice contrast with the black and white
6:17plastic wheels, and the whole thing is
6:20open so you can see the mechanism while
6:22it's operating. There's enough metal to
6:24support everything and enough orthogonal
6:26shear planes to keep it square and
6:28rigid, and supporting the motor mounts
6:30on standoffs helps keep the weight down
6:33for shipping. Leaving it open does leave
6:35the wheels more exposed to shipping
6:37damage, and it'll also be a dust magnet,
6:40but that's Rowan's problem. I think it
6:43looks great, and I think it should be
6:45about $100 worth of aluminum, so I can
6:47keep this within budget.
6:49I'm 3D printing the wheels with ABS
6:51glass fiber filament. ABS-GF is strong
6:55and light, and it will withstand higher
6:58temperatures than PLA both in shipping
7:00and if it ends up sitting in a hot car
7:03or shop in the Australian summer. I know
7:05it's winter there now, but it won't be
7:08forever.
7:09Printing something this large without
7:11warping in ABS or an ABS composite
7:13material can be challenging, but the
7:16heated chamber on the H2D handles it
7:18well. I did add a modifier in the slicer
7:21to print the central hub with 100% solid
7:24concentric infill, and that provides as
7:28much strength as possible for the motor
7:29shaft and a set screw. It would be
7:32possible to add a heat set threaded
7:34insert for the screws, but it isn't
7:36really necessary as long as the fit on
7:38the motor shaft is good. The wheels are
7:41light, and the motors aren't going to
7:42generate enough torque to be any kind of
7:45a problem.
7:46Now, since the motors don't have any
7:48inherent feedback mechanism, we're going
7:50to need magnets in the wheels, so the
7:53microcontroller can find their
7:55positions. The usual approach for adding
7:57magnets to 3D printed parts is just to
8:00glue a magnet in a hole, but I want them
8:02completely enclosed both for aesthetics
8:05and also for durability. So, I designed
8:08the wheel with an internal cavity and
8:10programmed the slicer to stop the print
8:12at the right point, so I can insert the
8:14magnets. This can be a little bit risky
8:17if your printer has a steel nozzle,
8:19because the magnet can pop out of the
8:21hole and stick to the nozzle, and if
8:23that happens, you will have a bad day.
8:26Ask me how I know.
8:28Now, one solution to that is to orient
8:30the part the other way up, so that the
8:31magnet is inserted after the first few
8:34layers, so it'll be attracted to the
8:35steel bed sheet, holding it in place,
8:38but for these parts, I needed to print
8:40them the other way up because of the
8:42spokes, so I designed the pocket to be a
8:44close friction fit that's strong enough
8:46to keep the magnet from jumping out of
8:48the hole. The Hall effect sensors I'm
8:50using only respond to the south pole of
8:53the magnet, so I made a 3D printed
8:55plastic tool with a rare earth magnet in
8:57the end and the poles clearly marked. If
9:00I pick up the magnet with the marked
9:02north pole of the tool, then the magnet
9:05will automatically orient itself with
9:07the south pole up, and I can slip it
9:09into the print without any drama.
9:12To get the fit I wanted on the motor
9:14shaft, I designed the part so that the
9:16bore prints just a tiny bit undersize,
9:19and then I can ream it so that it fits
9:21the shaft exactly.
9:23I'm reaming these parts with the reamer
9:25held in my tapping arm, and I did that
9:28for the first batch, and there was a
9:30little bit of wobble in the motors cuz
9:31it wasn't perfectly square. So, for the
9:33second batch, I switched over and reamed
9:35the holes in the milling machine to
9:37guarantee that they are perfectly square
9:40to the rotation of the wheel.
9:42The tapping arm isn't a precision square
9:43reference and it doesn't take much for
9:45the wheels to visibly wobble.
9:48Machining the frame out of aluminum is
9:50going to be a big time and material
9:52investment, so I want to be really sure
9:54that everything is going to fit before I
9:56start that process. I have 3D printers
9:59and the parts will all fit on the H2
10:01series print beds, so I cranked out a
10:03set of frame parts in PLA CF. It prints
10:07quickly and cleanly and it's
10:09dimensionally stable, so it makes a good
10:11prototyping material for aluminum parts.
10:14Of course, M3 threads are too small to
10:173D print directly into the part, so that
10:19means I have 93 holes to tap before I
10:22can start assembling. The tapping arm is
10:25invaluable for stuff like this. It would
10:27take forever to do this by hand.
10:30Most of the holes in these parts are
10:32blind and some of them are really
10:34shallow, requiring like 4 mm of threads
10:38in a 5 mm deep hole, but I can just
10:41program the tapping arm to go to the
10:43exact depth that I need and reverse out
10:45automatically. That way you can break a
10:48tap by entering the wrong number and you
10:50don't have to do it by hand.
10:53Now, one thing that became clear almost
10:55immediately when assembling this was
10:58that I need more than just the screws to
11:00align the parts and keep them square.
11:03The wheels are large enough that even a
11:05close tolerance clearance hole isn't
11:07enough to keep them aligned. Even 0.2 mm
11:10of clearance lets the plates move around
11:12enough for the gaps between the wheels
11:15to look uneven or even for the wheels to
11:17touch one another. So, I redesigned the
11:20front and back plates with little
11:22registration cleats to keep everything
11:24aligned.
11:25Pockets would also have worked in the 3D
11:27printed parts, but they would be much
11:29harder to machine out of aluminum.
11:32Now, these only protrude about a
11:34millimeter and leaving relief in the
11:36corners provides ample clearance for an
11:39end mill, so I'm not making my life any
11:41harder when it comes time to program the
11:43CNC for the final parts. And with the
11:46cleats in place, everything goes
11:48together easily and aligns correctly,
11:50and the cleats will also make it much
11:52stronger if it takes a hard hit in
11:54shipping since the sheer loads won't be
11:57carried by the screws, they'll be
11:59carried by the cleats.
12:00With the frame sorted, it is time to
12:02talk about packaging the electronics.
12:04The prototype has a lot of wires. I
12:08briefly considered trying to add a box
12:10under the base or to the back of the
12:12clock to contain them, but that's way
12:15harder than it looks. It's bulky, it's
12:17heavy, and wires always expand to take
12:20up 10% more space than you actually have
12:22available, no matter how hard you try.
12:25They're also prone to intermittent
12:27failures, and they're difficult to
12:28diagnose when that happens.
12:31The solution is to integrate everything
12:33on a custom PCB.
12:35PCBs are cheap, and manufacturing is
12:38readily available now, and more
12:40importantly, they're flat and easy to
12:42package.
12:44For this project, I designed a PCB that
12:46covers the entire back surface of the
12:48clock, so it can provide mounting points
12:50for all the components, distribute power
12:52and signals, and position connectors
12:54right behind each motor, so the only
12:57exposed wiring will be a short jumper.
13:00Even the Hall effect sensors to sense
13:02the wheel magnets can be soldered
13:03directly to the board and then protrude
13:05through the back plate to pick up the
13:07magnets in the wheels. This is a
13:09four-layer board that isn't strictly
13:11necessary, but it doesn't cost much
13:13more, and it makes power routing much
13:16more convenient. The internal layers are
13:1824-V power and ground planes, and the
13:21top and bottom are flooded with 3.3-V
13:24and 5-V fills, so
13:26traces aren't needed at all for the
13:28power. And having the large power planes
13:31next to each other provides some bulk
13:33decoupling capacitance that helps to
13:35control noise in the power rails, and it
13:39also keeps the 3.3 volt data and control
13:41traces on the top surface separated and
13:44shielded from the high current motor
13:46power traces on the bottom. To keep the
13:49packaging tight, the board is mounted
13:50directly to the aluminum back plate of
13:53the clock with no enclosure needed. Now,
13:55there is a chance that the aluminum
13:57could rub through the solder mask and
13:59cause a short circuit, so I designed a
14:020.4 mm thick insulator printed in TPU to
14:05go between the PCB and the metal. It
14:08provides a soft bed for the board, it
14:10helps to seal out dust, and it protects
14:12the hall effect sensors where they
14:14protrude through the plate. It's
14:16intentionally designed with one
14:18completely flat side and only minor
14:21bridging required, so it can be easily
14:24printed on the H2D or or really any
14:26similar size 3D printer.
14:28Now, the drivers might be okay without
14:31heat sinks,
14:32maybe, but I went ahead and designed a
14:35plastic cover to protect the electronics
14:38and a pair of finned aluminum heat sinks
14:40to carry away the heat. It's probably
14:42overkill, but overkill is underrated and
14:45they are going to look awesome.
14:48I never cease to be amazed when I get
14:50boards back from the fab. I I know
14:52exactly how they're made, but it still
14:54just feels like magic to design
14:56something in the computer, click a few
14:57buttons, spend a few dollars, and get
14:59the boards back in the mail.
15:02Now, it turns out I'm human, and
15:04regardless of how careful I am, mistakes
15:07happen. So, when I'm assembling a board
15:09like this, I always start with the power
15:11supply and work my way through the
15:13circuit, actually powering it up and
15:16testing to make sure that the voltages I
15:18expect are present in the places I
15:20expect them as I go. Soldering
15:23everything down at once and then
15:24attempting a big bang integration is a
15:26good way to let the magic smoke out of
15:28something expensive. Now,
15:30nothing here is really that expensive,
15:32but when something does go wrong, the
15:34parts that aren't on the board yet are
15:36safe and it's easier to debug the issues
15:38with less stuff in play.
15:41The 24-V power seems to be going the
15:43right place and the 5-V buck converter
15:46is stepping down the 5-V rail correctly,
15:48so it's now safe to solder down the
15:51Pico. Now, I goofed up when I ordered
15:54this board and I left the vias
15:56uncovered, so I'm putting down a strip
15:58of Kapton tape to space the Pico
16:01slightly off the board and prevent any
16:03possible shorts with exposed pads on the
16:06underside. Now, this small gap does make
16:09it a little bit harder to get the solder
16:11to flow properly around the castellated
16:13pads and I of course made the pads
16:15uncomfortably small for hand soldering,
16:18so I will make them bigger next time if
16:22I remember.
16:23With the Pico soldered down and the
16:253.3-V output confirmed, it's now safe to
16:28solder everything else to the board. The
16:30GPS receiver appears to power up and the
16:33drivers aren't catching fire, but the
16:35microcontroller inputs for the buttons
16:38are not toggling the way they should.
16:40Now, looking everything over, it looks
16:42like I messed up the pin out when I
16:44defined the push button footprint. I
16:47thought pins one and two were shorted
16:48and got connected to three and four when
16:50the button was pressed, but it's the
16:52other way around. One and four are
16:54shorted and they connect to two and
16:56three when the button's pressed.
16:59I updated the PCB design so that if I
17:01ever make another one, it'll be right,
17:03but for this one, I can just clip off
17:05two of the pins from each button and
17:08make it work.
17:09The rest of the parts go on the back
17:11side of the board, the motor connectors
17:13and the hall sensors. I specifically
17:16chose through-hole hall sensors so they
17:19could reach through the back plate to
17:20sense the magnets in the wheels. To get
17:23them all in the right position, I can
17:25just insert them into the TPU insulator,
17:28screw the board down onto the prototype
17:30back plate, and push on the pins to seat
17:33the sensor at the end of the little TPU
17:35pocket. Now, the drivers don't have heat
17:38sinks on them yet, so I'll set the
17:40current limits on the drivers to a very
17:42low level just to make sure that they
17:44don't burn up. I'll keep an eye on the
17:47temperatures and readjust later based on
17:49how they perform.
17:51A quick test with the same magnet stick
17:52I used to assemble the wheels confirms
17:54that all of the Hall sensors are
17:57oriented correctly and working, and the
17:59Pico can see the signals.
18:02The last thing we need to make this work
18:03is a set of short cables to connect the
18:05motors to the board. These are JST-XH
18:10connectors, and while I do have a manual
18:12crimper and could spend an hour under
18:14the microscope swearing and crimping the
18:1648 pins I need, it turns out you can
18:19just buy pre-crimped wires for about the
18:22same cost. These came from Digikey for
18:2525 cents a piece, and all I have to do
18:27is snap them into the connectors, and
18:29I'm good to go.
18:31Now, as easy as that sounds, I still
18:34managed to screw it up. The motors have
18:37six-pin connectors, but since they have
18:39bipolar windings, only four are used.
18:42Now, I assumed the coils were between
18:44pins one and three, and between four and
18:46six. And when I hooked them up that way,
18:49the motors just buzzed and shattered,
18:51and when I looked closer at the factory
18:53cables, I discovered that the center two
18:55pins are swapped. The coils are actually
18:58between one and four, and three and six.
19:01So, I went back and updated the PCB
19:03layout for anyone who might use it in
19:05the future, but for now, I can just
19:08cross over the two middle wires in in
19:10cable, and it will work. All right.
19:12Sense of aesthetics is wounded because
19:14the wires now kind of bunch up instead
19:16of laying in nice parallel curves, but
19:19if you agree not to tell anyone, I won't
19:21tell anybody either.
19:23With the cable sorted and everything
19:25fully assembled, it actually works. I
19:28adjusted the motor current to provide
19:30enough torque and I had to slow down the
19:33acceleration curves in the firmware just
19:35because of the larger moment of inertia
19:37of the wheels. I previously had it tuned
19:39for the little wheels in my prototype
19:40earlier this year and these wheels have
19:43a lot more mass out on a much larger
19:46rim. But with those changes, it's pretty
19:49good. I'll probably do some more tuning
19:51later, but it feels good to have most of
19:54the integration risk behind me at this
19:56point.
19:57I did notice a couple of things that I
19:59would like to address before we go any
20:01further. Now that I see it all put
20:03together, I don't think the windows for
20:05viewing the numbers are large enough. I
20:07tried to keep them tight to make sure
20:09that only one number is visible, but if
20:11you view it from an angle at all, part
20:14of the number is hidden. So, I 3D
20:16printed a new version with larger
20:17windows and I like it a lot better. It
20:20feels more open and less constrained and
20:23it's easier to read from off-axis. I
20:25think my fears about multiple numbers
20:28being visible were really unfounded. The
20:30looser fit is definitely better.
20:33I was also entertaining myself with the
20:35time zone and 24-hour buttons as one
20:38does and noticed that the motors can
20:40lose steps if you press the buttons
20:42rapidly. I spent some time with a logic
20:45analyzer and figured out what was
20:46actually going on. When you press the
20:49button to change the time zone, it spins
20:51the wheels to show the new value, but it
20:53also saves the setting into the EEPROM
20:56on the Pico. And when you're writing the
20:59EEPROM, it only takes a few
21:00milliseconds, but while that's taking
21:03place,
21:04both processor cores pause and if a
21:06motor is in motion, the driver stops
21:08generating steps abruptly and the
21:11inertia of the wheel over drives the
21:12motor and makes it lose steps. This is
21:15most evident when rolling over midnight
21:17or switching between 12 and 24-hour mode
21:19because the wheels are spinning faster
21:22and making full rotations. If you press
21:24the button at the wrong time, you can
21:25make it lose track of the wheels and
21:27then you have to reset the clock to get
21:29it to recover. Now, the solution turns
21:32out to be simple. I just modified the
21:33firmware to just remember when the
21:36setting has changed and write it to the
21:38EEPROM later when all the wheels are
21:40stopped. And with that fix, I can frog
21:43the buttons as much as I want in any
21:44combination and the mechanical display
21:47stays in sync with the LCD.
21:49Now, intermittent bugs like this are
21:51always unsettling, at least they are for
21:53me, until I finally get them figured
21:56out. And it is really satisfying
21:59to be able to abuse the controls and
22:01just utterly fail to get the system to
22:03misbehave.
22:05With the design finalized and all the
22:07bugs that I know about worked out, it is
22:09finally time to start making some
22:11aluminum chips. The front face plate is
22:1414 in or 356 mm wide. So, I'm holding
22:18the stock in a pair of mod vices with
22:21jack screws under each end to try to
22:24keep it from ringing.
22:25To remove the bulk of the material, I'm
22:27using a 3/8 in roughing end mill from
22:29Lakeshore Carbide. This is a variable
22:32flute design which helps prevent
22:34resonance and it works pretty well.
22:37I have this dialed up about as fast as I
22:39dare push it. The mill has a 1
22:41horsepower motor and you can hear the
22:44RPM sag a little as it enters each cut.
22:47I think the mill has the rigidity to
22:49push a lot harder, but the stock motor
22:51is the limiting factor. I'd actually
22:54like to try replacing it with an AC
22:56servo at some point, but that is a
22:58project for a future video. I'm
23:01finishing the insides of the windows
23:02with an 8-in end mill just to make sure
23:05that the cutter diameter is
23:07significantly smaller than the inside
23:09radius just to prevent it from
23:11chattering. And to cut the large
23:13chamfers around the windows, I'm using a
23:163/8-in four-flute carbide chamfer mill.
23:19Was a little bit worried about this
23:21chattering, but this is a spiral flute
23:24chamfer mill also from Lakeshore Carbide
23:26and it is not even struggling. Now,
23:29Lakeshore Carbide doesn't sponsor me, by
23:31the way, but I'd be open to it.
23:34The inside of the front plate has
23:36three-dimensional curved cutouts to
23:37clear the wheels and I'm cutting these
23:40with a two-flute ball end mill.
23:43I ended up settling on a 1-mm step over
23:45with a 1/4-in or 6-mm tool and that got
23:49the machining time for the pockets down
23:51to 7 minutes and it left a finish that I
23:54could live with. However, after it was
23:56done and after I felt the sharp edges on
23:59the insides of the windows, I decided to
24:01come back with the same ball end mill
24:03and use a scallop path to cut a
24:05three-dimensional fillet on the edge. It
24:08added another 15 minutes of machining
24:10time, but the soft feel of the part in
24:12the end was definitely worth it.
24:15For many of the parts, I needed two
24:17identical pieces. So, I set them both up
24:20at the same time in two mod vices and
24:22use multiple work offsets to run them at
24:25the same time tool by tool. It's not
24:28something that I've ever tried before,
24:29but it worked great. 10 out of 10, no
24:32notes.
24:33I was a little worried about the heat
24:34sinks just because slotting is always a
24:37risky operation, but these slots are
24:39shallow, so it went fine. I designed the
24:42slot specifically to be 20 thou or a
24:45half millimeter wider than the cutter so
24:48that I could slot out the center and
24:49then come back and finish the sides and
24:51the bottoms without the sideways forces
24:54and chip clearance issues that sometimes
24:56cause slotting cuts to chatter. Whenever
24:59I'm CNC machining, I always program
25:02chamfers wherever I can. A half
25:04millimeter chamfer cleans up an edge and
25:06leaves a nice facet that catches the
25:08light, and a point one millimeter
25:10chamfer just touches the edge to remove
25:13burrs while leaving the edge looking and
25:15feeling like it's still square. It saves
25:18a ton of time cleaning up the parts by
25:21hand afterwards. Now, when you flip the
25:23parts over, you do have to have very
25:25good registration of the front to the
25:27back to get those chamfers to line up
25:29sometimes, but as long as you can manage
25:31that, it saves a ton of hand work.
25:34The motor mounting plates are small and
25:36flat, and the best tool for this job
25:38really would have been a laser, but I
25:40don't have one that will cut eighth inch
25:42aluminum.
25:43I experimented with the plasma cutter,
25:45but I really didn't like the results. In
25:47the end, I just bit the bullet and
25:50machined them. I drilled and bored and
25:52contoured them partway through from a
25:54single strip of stock, then I cut them
25:56apart and machined a fixture to hold
25:58them so that I could contour the other
26:00side. This took a long time. A laser
26:04would have cut six pieces in six
26:05minutes.
26:07I debated whether to do the edge
26:08drilling in a manual mill, but since
26:10there were so many parts with holes in
26:12the same places, it just made sense to
26:14set up a fixture in the CNC. That way, I
26:17can just clamp in a new part and hit the
26:19green button and get consistent holes
26:21quickly.
26:23Now, the Masso controller doesn't
26:25support rigid tapping, but to be honest,
26:28the tool change time probably wouldn't
26:30have been worth it to me anyway, and
26:31since I have a tapping arm, that was an
26:34easy way to go. Now, for these parts,
26:36I'm using a form tap that swages the
26:39threads into the part without removing
26:41any material. This forms the grain
26:44structure of the metal so that it flows
26:46around the threads instead of being cut
26:48away. It's a lot like forging, and it's
26:50fantastic when you need to tap to the
26:53bottom of a blind hole because there are
26:55no chips.
26:57These holes are M2 and I can do the same
27:00thing with them. I would be holding my
27:02breath trying to tap these by hand
27:04afraid that I was going to break off a
27:06tap and a part that now has a lot of
27:09hours invested.
27:11Now I just have to clean up the mess I
27:12made, but fortunately it is all inside
27:15the mill enclosure and I designed it so
27:17I can just sweep everything down into a
27:19plastic bin in the stand. This is the
27:22first time I've tried it and
27:24it's pretty good.
27:27The parts all turned out great. I
27:29actually made two sets so that I can
27:31have a clock for myself when the project
27:33is done and also so I will have a backup
27:36if the shipping company betrays me
27:38again.
27:40Now the hardest dimension to control on
27:42a two-op machining process like this is
27:44the thickness. It is really easy to lose
27:48a few thousandths of an inch when you
27:49flip the part over to machine the other
27:51side and there are lots of factors to
27:53conspire to make this hard. Large flat
27:56parts warp, vice jaws lift, the Z axis
27:59has backlash, tool length measurements
28:01aren't always perfect and probing is
28:03hard. Tiny errors can stack up and
28:06become a problem in combination. But
28:09once I figured out that I can just tap
28:11the parts down on parallels on the
28:13fixture plate for op two between the
28:15smooth sides of the mod vice jaws,
28:18this machine has been fabulous. These
28:20parts all came out within a thou of the
28:22target dimension and many were a lot
28:25closer than that. For this project, the
28:27only place that that matters is where
28:29the plates fit into the alignment cleats
28:31in the front and back plates.
28:34It took about an hour to carefully
28:35disassemble the clock and replace all of
28:37the plastic parts with aluminum. I took
28:40my time and I used Loctite 242 on all of
28:44the fasteners. And did that for a couple
28:45of reasons. I don't want the screws to
28:48loosen from vibration during shipping,
28:50but I'm also using stainless steel
28:52screws threaded into aluminum and I
28:55don't want them to friction weld and
28:56gall. The rolled threads will help
28:58prevent that, but the Loctite provides a
29:01barrier between the metals that will
29:03also help. Is this really necessary?
29:06Maybe not, but it certainly won't hurt
29:09and I'll sleep better knowing that I did
29:11the best job I could.
29:13The back cover for the PCB is also
29:15printed in ABS glass fiber and while the
29:18fiber content makes it pretty stable,
29:21large or long parts like this are still
29:24going to shrink a little bit. When
29:26mounted on other 3D printed parts which
29:29also shrink a little bit, it isn't
29:30noticeable, but when I swapped in the
29:33machine aluminum parts, the holes in the
29:35ends of the cover didn't line up
29:37properly. I ended up reprinting the part
29:39scaling it up in the slicer by about
29:42half of a percent and now it fits
29:44correctly and that's right in line with
29:46the expected shrink rate for ABS. If
29:49your whole project is 3D printed or if
29:51it's smaller, you won't really notice
29:53stuff like this, but when mixing printed
29:56parts with machine metal parts at this
29:58scale, it starts to become more evident.
30:01The last step is to cut some thermal
30:03pads and install the heat sinks. These
30:05are high performance 12.8 W per meter
30:08Kelvin pads. Is that overkill?
30:11Yeah, probably.
30:13I was a little bit worried about
30:15powering it up with the aluminum frame
30:16installed. Any small error in the
30:19machine parts, the PCB layout, or a
30:22tolerance stack-up issue could end in a
30:24short circuit, blue smoke, and sadness,
30:27but not today. Today everything is
30:30working and I have to say that it looks
30:33even better than I hoped it would. Let's
30:35just take a moment and appreciate the
30:39ridiculous thing that's sitting here on
30:41my bench.
30:57>> [music]
31:11[music]
31:29[music]
31:34>> Regardless of how good it looks sitting
31:35here on my bench, it still has to get to
31:38Australia in one piece.
31:40I've been thinking for a long time about
31:42how to package it so that it will
31:44survive. The exposed wheels are
31:47delicate, and if the weight of the frame
31:49gets transmitted through them in an
31:51impact, they will break or bend the
31:54motor shafts.
31:55The solution I came up with is to 3D
31:57print flexible end caps. These are
32:00printed in 95A TPU, and they're sliced
32:03with gyroid infill and zero top and
32:07bottom layers. I experimented a little
32:09bit with different infill percentages
32:11and decided that 10% gave me the mix of
32:14support and cushioning that I wanted. I
32:17also included some L-shaped pockets in
32:19the caps to accept heavy cardboard edge
32:22protectors. These are some that I saved
32:24when I unpacked my new barbecue grill a
32:26couple of months ago, and they should be
32:28plenty for this application. They're
32:30probably overkill.
32:32Any impacts to the sides of the box
32:33should be redirected by the cardboard
32:36corners into the end caps where the load
32:39can be safely transmitted to the
32:41aluminum frame of the clock
32:43bypassing the wheels and protecting
32:45them.
32:46With some plastic wrap to keep debris
32:48out and to subdue the power supply and
32:51GPS antenna so they won't leave the
32:53chat, everything goes into a cardboard
32:56shipping container with another layer of
32:58packing foam around the outside.
33:00And I think that is about the best that
33:02I can do without making the box a lot
33:05bigger.
33:06And it is a good thing that shipping
33:08isn't included in the project budget
33:10because this is going to cost more to
33:12ship to Australia than it cost to make.
33:16Rowan,
33:17I hope it still works when it gets to
33:18you and I hope you like it.
33:21Speaking of shipping, look at what just
33:23showed up from NBR Works. Let's open it
33:26and see what wondrous thing is inside.
33:30I am already liking what I'm seeing.
33:32Anything that comes in a
33:34Gridfinity-compatible box
33:36has to be good.
33:38Oh, wow. He said I would know exactly
33:42what to do with his gift and he is
33:44absolutely right. This is a ball
33:47burnishing tool. It has hardened balls
33:50in the tips of the arms that clamp down
33:53on a part spinning in the lathe to
33:55burnish out the tool marks and leave a
33:57nice smooth shiny finish. Presumably
34:00there are radial ball bearings to back
34:02up the balls and carry the swaging
34:04forces.
34:07I see he included some extra screws in
34:09the box, but I'm not totally sure what
34:12this part is for or the pins.
34:15That looks like a 1/4-in drive socket,
34:17so maybe it's a pin spanner.
34:21Ah, okay. It's a tool to remove the ball
34:23covers.
34:26And the other side fits the pivot pins.
34:38>> And yeah, the loose pins must be to
34:40remove the bearing covers.
34:42Now that I've partially taken it apart,
34:43let's put it back together and go try it
34:46on the lathe.
34:49The first victim is a piece of 6061
34:52aluminum. I'll take a cleanup cut and
34:55then we can try burnishing it.
34:59I assume I want some cutting oil on
35:00this, but I don't
35:02really know, and I also don't have any
35:04idea how much pressure's needed. So,
35:07I'll just navigate using the force and
35:09we will see what happens.
35:50The burnished surface is a lot smoother
35:53to the touch, but it was pretty good as
35:54machined. So, let's try something
35:57harder.
36:00This is 1144 stress-proof steel. It
36:03machines freely, but it usually leaves a
36:06pretty dull finish. Let's see what the
36:09burnisher does with this.
36:15Again, I'm just guessing on the
36:17pressure, but the knob's pretty
36:18sensitive and it is surprisingly easy to
36:21feel what's happening.
36:41>> That is a night and day difference. I'm
36:45going to have to play with this.
36:47I'll put a link to the NBR Works video
36:49showing how he made this
36:51somewhere. You should definitely go
36:53check that out. The workmanship on this
36:56is just
36:57gorgeous.
36:59This project has been in the works for a
37:01long time, and I've probably got more
37:03hours invested in it than anything I've
37:06done in a while. It was a lot of work,
37:09but I am happy with the results, and I
37:12hope you enjoyed watching it. If you
37:14enjoyed the video, give it a thumbs up,
37:16and feel free to subscribe. As always,
37:19the files for this project will be
37:21posted on Patreon. I'll include the CAD
37:23models, the printable parts, and the
37:25Gerber files for the PCB,
37:27in case you want to make one for
37:28yourself.
37:29Thank you for watching.
37:33>> [music]
37:41[music]

About this transcript

This page contains the full transcript of The Most Beautiful Ridiculous Thing I've Ever Made by Clough42, generated from the public captions YouTube serves with the video. The transcript has 6,359 words across 928 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to translate the transcript into 100+ languages.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, translation to 100+ languages, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.