Update on Arrakis 3.1- a New Mode of Operation?

Arrakis 3.1 has been working well after I sorted out some issues with the mechanism and the electronics. It is fast and quiet and not too ugly for the living room. I did run into one problem. I used 2040 V-slot material for the sandbox frame which put the glass just 28 mm over the sand. When the ball goes fast it kicks up sand and some of it ends up stuck to the bottom side of the glass. That means I have to clean it more often than I like, and the glass is heavy, so it a bit troublesome to clean it.

There are a couple ways to solve the problem. I can run the table slow, so the ball doesn't kick up the sand, but that's no fun. Or I can make a sandbox that will hold the glass a little higher off the sand. I ordered some 2060 t-slot to replace the old 2040 sandbox frame. Now the glass is 48 mm above the sand so there's much less of a problem of the sand sticking to the glass. I also printed some TPU end caps for the 2060 v-slot that work better than the hard plastic caps I had purchased for the 2040 v-slot that liked to fall out every time I looked at the table sideways.

2060 v-slot cut end. I cut the piece 3mm short at each end to allow room for the printed TPU end caps.

The 2060 end cap, printed in black TPU on UMMD. The crush ribs ensure a tight fit into the 2060 v-slot sandbox frame.

End cap in place at the corner of the sandbox frame. Unlike the crummy plastic parts I bought, these TPU parts don't fall out easily.

I had previously used a couple cheap vacuum handles to move the glass in Arrakis 2.0. They were supposed to be used as handles for old and infirm people to use in a shower/bath situation. While I was moving the glass one of them let go and the glass fell and shattered and I had to replace it at great expense.

Cheapo vacuum handles on the Arrakis 2.0 glass that failed. DON'T USE THESE!

Vacuum handle stuck to the glass. USE THESE!

The glass top I used for Arrakis 3.0 was 8mm thick and beveled. I got it cheaply via Craig's List because I wanted to try painting the LED frame on the underside of the glass and wasn't sure if I would like it. The painted LED frame was good, but the top surface of the glass had some pretty big scratches that were visible in daylight, but not visible when the table was lit up at night. 

I decided to get a new glass top for the sand table so I wouldn't have to see the scratches, at least until I put some new ones in it. I ordered a 12 mm thick, beveled, low-iron glass table top. It weighs over 23 kg (~50 lbs), so I also ordered some surprisingly cheap ($35 for a pair) vacuum handles to make dealing with the glass easier. Each of them can hold about 350 lbs, so they are very secure. The only problem with using them is that when the glass is set back on the sandbox, removing the suction handles creates a triboelectric charge that induces the opposite charge on the underside (the sand side) of the glass, and that charge picks up some of the sand and sticks it to the glass. Doh!

Sand stuck to underside of the glass after I removed the vacuum handle. This is exactly the opposite of what I want!

But what to do with the old sandbox frame?

Back when I first tried servomotors in "The Spice Must Flow" sand table I replaced the magnet with a blue LED, set the mechanism on its side in the dark, pointed my camera at it, opened the shutter, ran a drawing on the mechanism, moving the LED around, and when it finished the drawing, I closed the camera's shutter. I got pictures that look like this:

Light painting made using the Spice Must Flow sand table mechanism with servomotors. The controller applies acceleration and deceleration to the motion so whenever the LED is going to change direction the painting looks brighter because the LED was there longer.

The only way to see this light painting is to look at the photo. But I had another idea. What if I lit up a glow in the dark surface with the LED? 

The most common glow-in-the-dark stuff contains zinc sulfide. That's what you'll find in GITD 3D printer filaments, paints, and most GITD toys because it's very cheap. The glow isn't very bright and doesn't last very long, but there's something better. A compound called Alumane Dysprosium Europium Oxidanylidene Strontium (commonly called strontium aluminate) glows brighter and lasts longer. 

I painted the old glass with strontium aluminate paint for the light drawing. Then I mounted a UV laser on the magnet carriage, pointing up at the GITD paint. 

Painting a 3rd coat on the underside of the glass table top.

Daylight visibility of lines drawn with UV (405nm) laser pointer. The dimmer lines were drawn about 3 minutes before the brighter lines. The lines were drawn in about 2-3 seconds, so high speed drawing will be possible. Note, the laser pointer was hand-held maybe 300-400 mm from the paint surface.

Night time visibility of lines after about 10 minutes.

Tests look very promising, as you can see in the photos. 

There are some important differences between drawing in sand and drawing with light. One of the more important differences is that when drawing with light, there's no sand being thrown around and sticking to the glass, obscuring the view of the drawing. That means I can crank up the speed on the patterns so they will finish very quickly. Also, there's no need to erase the light drawing- it fades away after several minutes.

That's all very nice, but once I mounted the laser on the magnet carriage, I realized there was a problem. When the laser hits the painted surface and makes it glow very brightly, there are internal reflections in the glass top that create a halo around the spot lit by the laser beam. That halo is not as bright as the spot hit by the beam, but it is bright enough to cause the surrounding paint to glow. That wrecks the contrast and makes drawings look fuzzy- not exactly the effect I was going for.

It's also uncomfortably bright to watch it drawing in a dark room.

The halo problem. I've tried several types of internally baffled tubes, etc., to remove the halo until I realized it's caused by the glass table top, and not the laser optics.

You can see the bright halo around the laser beam. The red and blue light comes from the electronic status LEDs on the controller board which I later fixed with some gaffer's tape.

3D printed laser module mounted in the magnet carriage. There's a 500 mAH LiPo battery and a NC reed switch (with a disc magnet attached to keep the laser off). 

Unfortunately, this was a failed experiment. I may look into using laser galvanometers to steer the beam to draw on a wall painted with the GITD paint. There won't be any halo problems that way, and it will be a small, quiet box that does the drawing. I'll have to see if there's some hardware/software out there that can translate gcode in the drawing files to voltages to drive the galvanometers. I may have to come up with that myself.

Wiim Amp Ultra for the Quads

A while back I bought a Wiim Amp Pro (WAP) to replace the electronics in my bedroom stereo system. After I realized the bass management can be used as a crossover, I decided to try it in my living room system, driving the Quad ESL-63s that I recently rebuilt, and an SVS 3000 Micro subwoofer. I was impressed with the Wiim amp's performance and decided to put the WAP back in the bedroom and I bought a Wiim Amp Ultra (WAU) for the Quads/sub (20% off on Black Friday!). The WAU is similar to the WAP, but uses a different DAC and has a higher power output- 100W/ch at 8 Ohms and 200W/ch at 4 Ohms. It also has a touch screen I don't really have a use for.

Wiim Amp Ultra sitting on top of SVS 3000 Micro subwoofer.

The WAU has been reviewed ad nauseum:

WAU at Audio Science Review

WAU at Soundstageaccess

WAU at Darko Audio

But why?

The Quads, as wonderful as they are, are not perfect. The upper end of the audio spectrum (beyond my ability to hear it) is rolled off a little (easily corrected with an equalizer in the WAP or WAU), and they are incapable of producing low bass. Some people also complain of limited "dynamics" and maximum SPL - to me those are the same thing- if maximum SPL is limited, of course, "dynamics" will be limited.

The Quad's poor bass performance can be blamed on two things. First, they are bipolar radiators which means they emit sound from both the front and back sides. The front side and back side radiation are 180 degrees out of phase, so when the sound wavelength they produce is long compared to the size of the speaker, the front and back radiation tend to cancel and the result is weak bass. Second, the spacing between the diaphragms and stators in the drivers limits the maximum diaphragm excursion, and thus maximum sound pressure level (volume). The greatest excursion is demanded when reproducing low frequency sounds, so if you can prevent those low frequencies from going to the Quads, they can play everything else louder. This also addresses the "limited dynamics" some people complain of. 

The bass management in the WAP/WAU can be configured to keep the lows out of the speakers and send them directly to the subwoofer. The crossover is a Linkwitz-Riley 4th order type which means the low frequency roll-off in the Wiim amp will be 24 dB per octave below the crossover frequency. One octave is a doubling (or halving) of frequency, so one octave below 90 Hz is 45 Hz. So at 45 Hz, the signal going to the Quads from the amp will be 24 dB below what it would have been without the crossover being turned on. That means the Quads will be able to play much louder than they would if the full range signal were being sent to them. 

Linkwitz-Riley crossover response curves for 2nd (12 dB/oct), 3rd (18 dB/oct), and 4th order (24 dB/oct) implementations. The WAU/WAP use the 4th order curves. This image comes from the Rane site linked above. No, there won't be a suck-out at the crossover frequency. When the outputs of the drivers are summed acoustically, in your room, the response at the crossover frequency will be flat.

The SVS 3000 Micro sub has an 800 W amplifier built in and can produce lower bass at much higher output than the Quads ever could, so configuring the amp and speakers this way is a win-win situation. I get the lows that the Quads can't produce well, and the whole system can play louder. This is going to be true of any speakers you use that have poor low frequency output/response, including about 99% of all bookshelf speakers.

System Setup: The Subwoofer

The Wiim amp's bass management and the subwoofer can both be configured via my phone or tablet. It's a little bit of messing around, but once the configuration is done, it doesn't need to be done again. 

Connect the sub output on the Wiim Amp Pro to the LFE input on the subwoofer. Open the SVS app on your phone and switch to LFE mode - that tells the sub's DSP that it will only receive low frequencies at its input so it doesn't need to run a low pass filter of its own. In my system I set the subwoofer output to -20 dB, but your system may need a different setting. This can be changed later, as needed.

Much more detail here.

System Setup: The WAP or WAU

Connect the speakers to the jacks on the back of the amp and the sub output to the subwoofer's LFE input. Open the Wiim Home app on your phone. Select the amplifier and open its settings menu. Select the "Sub Out" item. Turn Sub Output on, set the level (start at 0dB and increase later if needed), set the crossover frequency (the default is 80 Hz which is pretty good for most speakers, including the Quads), you'll set the phase later, so don't worry about it for now, switch "Subwoofer Bypass Mode" off, and "Main Speakers Output Bass" off.

Play some music that has some low bass. As you listen, flip the phase switch back and forth between 0 and 180 degrees. At one setting, the bass response will experience a dip and at the other setting, a peak. You can usually hear the difference pretty distinctly. 

The final setting to make is to use the Wiim Amp's bass sync feature. The subwoofer and Wiim amp both have some delay resulting from the signal processing that goes on in their DSPs. There may also be a different "time of flight" between the main speakers and the sub and your listening position. Ideally, you want the subwoofer to be time aligned with the main speakers. The Wiim amps have the ability to perform the synchronization built in, but it's less than ideal. It uses a mic built into the amp to pick up test signals generated in the amp. It will then report the delay and adjust the signal(s?) to the main speakers and sub to be properly time-aligned. It would be better to use a mic at the listening position (as RoomFit uses), so right now, to do it right you have to put the amp at the listening position when you run the bass sync. Fortunately, using a mic at the listening position is possible in the latest beta release of the firmware for the amps, so in the near future you'll be able to use the same mic you use when you run RoomFit at the listening position.

More details here.

Listening

When I want to play music, I select it via Tidal Connect or LMS and touch the play button on the screen. The Wiim Amp Pro wakes up and starts playing which triggers the sub to start playing within a few seconds. I can start, stop, select new music, skip ahead, skip back, and control volume, all from my phone, so there are no remote controls to hunt for or juggle and I don't have to go to the amp and sub to turn them on. They just start when I want to hear music, and when the music finishes, they both drop back into standby mode. 

The amp can be placed out of sight, under furniture or in a cabinet, because there's no need to touch it. Do give it a little ventilation though because as efficient as class D amps are, they do generate a little heat:

WAU top side thermal image after playing for an hour at moderate volume, with low frequencies routed to the SVS 3000 Micro sub.

WAU bottom side thermal image after playing for about an hour at moderate volume.

This sort of operation won't satisfy you if you prefer to handle records or CDs and/or you like to twiddle knobs and flip switches, but for me it's ideal.

And in case you are wondering, yes, it sounds great! If you want more flowery superlatives, see some of the video reviews linked above. I don't have the vocabulary that those guys do. 

Arrakis 3.0 Updates- Now It's Arrakis 3.1

Update 3/31/26

The mechanism started making noise again, so I opened the table up and found that the belt was climbing on the corner pulleys, and that was wearing the edges of the belt and leaving bits of black rubber from the belt everywhere. The noise was occurring when the belts would climb the pulleys, increasing tension, and then snap back toward the center of the pulleys. 

I redesigned and printed the corner pulleys on the end of the stable opposite the motors to have a flat profile instead of the concave profile the other pulleys have. For some reason it only seems to be a problem at that end of the table. It's been running quietly again for several hours. I'm going to wait a while and see if I should replace the other pulleys with the flat profile type.

Now back to the original post: 

After living with Arrakis 3.0 for a few months, a few problems have led to some changes. 

1) The cheapo 6A rated power switch welded itself in the "on" position about the fifth time I switched the table on. The table has a 350W power supply - 6A should be plenty of capacity. 

2) I noticed that quite a few of the patterns started with the ball rocketing from the home position to one of the opposite edges of the drawing area, drawing a straight line across the table before the actual pattern started to be drawn. Sometimes the pattern was dense enough to wipe out that line, but very often it wasn't. I found that really annoying.

3) I noticed that the table started making some small noises only a couple months after I finished building it. That suggested parts were wearing.

4) I run the table in random mode most of the time, where the table randomly selects a pattern to draw from over 200 patterns stored in the controller's memory. Sometimes it draws a nice pattern that I'd like to keep on the table for a day or two. The only way to do that is to switch off power, which means the LEDs are also switched off. I'd prefer to be able to have the LEDs on without the table drawing any new patterns.

Changes

1) After checking the LRS-350-24 power supply specs and finding inrush current rated at 60A maximum (!), I replaced the 6A switch with one rated for 10A, and added a 16 Ohm 5A NTC thermistor and 3 second time delay relay (found in a box at the makerspace, with contacts rated for 7A @ 250V) to switch power to the table. Now when power is switched on, the thermistor is in series with the power line going into the power supply, limiting surge current to a maximum of about 10.6A (peak voltage on the power line is 170, so maximum current will be 170V/16 Ohms=10.6A, but only if I happen to flip the switch at the exact peak of the voltage). After 3 seconds, the relay closes and shorts out the thermistor. During that 3 seconds, the thermistor gets about 10C warmer than ambient temperature. This seems to have solved the power switch/surge current problem.

3 second time delay relay (the gray box) and NTC thermistor (the gray disc). If the relay ever fails, the thermistor will remain in the circuit and the table should continue to operate.

This is how the relay and thermistor are wired. Theoretically, I shouldn't need the relay at all as the thermistor resistance drops when it heats up, but I'd prefer not to have it sitting there hot when the table is powered up. If the relay ever fails, the thermistor will still be in the circuit and it should continue to operate with the thermistor sitting at an elevated temperature.

2) I manually edited all 223 pattern files and eliminated the lines that went across the table at the start of many of the patterns. I also eliminated some odd back-and-forth-along-the-edges motion that occurred at the start of some patterns. I'll be more careful to remove such lines before uploading new patterns to the table.

3) Arrakis 3.0 had wheeled carriages running in v-slot aluminum for the Y axis. I suspected that this is where the noise was coming from. When I opened up the table to inspect the mechanism I found that the Y axis carriage wheels on the outside of the mechanism frame were wearing out and the wheels on the inside of the frame weren't even touching the frame. The wear pattern is the result of the tension on the belts pushing the carriages inward while the attachment to the X axis linear guide is loose (by intention, clearly a mistake), allowing the inward motion and some tilting. It was a very bad design!

For the MakerFest in Elkhorn, Wi. on Feb.14th, I installed wheels that were made of some mystery material that was supposed to be "self-lubricating", and it restored the table to quiet operation. When I got the table home I opened it up to look at the mechanism and think about what I could do to improve it, and found this:

Self lubricating?

Arrakis 2.0 uses UHMW blocks sliding in the aluminum slots and has been 100% reliable for years. I decided to try converting the wheeled carriages in Arrakis 3.0 to use UHMW blocks instead of wheels, and to solidly attach the X axis linear guide to the two carriages. 

I found some UHMW in my stash that I was able to carve into sliding blocks to replace the wheels in the Y axis carriages. I tried it and it just didn't work out- the UHMW blocks fit loosely in the v-slots and the result was that the X axis tilted, making a noise each time the Y axis movement reversed direction. Even if they had fit well, as the mechanism wore in, it would have started making noise. Ugh.

The real problem with Arrakis 3.0 was that the X axis wasn't rigid enough. Each of the wheeled carriages could tilt due to the belt tension and that caused uneven pressure on the wheels resulting in uneven and excessive wear. Another problem is the limited accuracy of the construction of the mechanism and the use of just two eccentric adjusters at each of the wheeled carriages.

The solution to the first problem was to make the entire X axis more rigid, which I did by mounting the X axis guide rail on a piece of 1/4" aluminum tooling plate that spanned the width of the X axis. The 1/4" plate is flat and rigid, so it will flex much less (especially when screwed tightly to the X axis guide rail) than the two original 1/8" carriage plates that were loosely screwed to the ends of the guide rail.

Set up for drilling and milling the tooling plate for the X axis. Hole positions were critical, the milling was for weight reduction and cosmetic reasons (though it's inside the table and most people will never see it).

The X axis linear guide mounted on the tooling plate. 

The second problem was solved by using eccentric spacers for all the wheels. The first end is adjusted so all four wheels fit the V-slot rail, then, at the other end of the X axis, the adjusters for all four wheels are positioned to comfortably hold both sides of the V-slot rail. The holes in the eccentric spacers are offset from center by about 0.79 mm, so allow for about 1.58mm adjustment, which is more than sufficient. I would really prefer to spring load some of the wheels so they are always held in contact with the rails, regardless of imperfect spacing or flexing that may occur when the table is operating. 

I'm not sure about using 4 wheels at both ends of the X axis. If I just put 4 wheels (3?) at one end, the X axis assembly will follow the rail that it is clamped to by those wheels. Adding wheels at the other end over-constrains the motion. If the two Y axis rails aren't absolutely parallel, the second set of wheels may cause the mechanism to bind. This is why it would be good to have one set of wheels spring-loaded so they can allow for the rails to be out of parallel. Arrakis 2.0 has a spring loaded UHMW block at one end of the X axis for this reason and it works perfectly. In the end, I used 4 wheels at each end of the X axis. I will reexamine it when I see how well the wheels hold up in the new configuration.

Arrakis 3.1 left side of x axis. Yes, I know the screws holding the wheels are short- they will be replaced with longer screws when new wheels arrive and get installed.

Right end of X axis.

The tooling plate raised the X axis guide rail, belt clamps on the magnet carriage, and pulleys by 3.175 mm. That meant I also needed to raise all the corner pulleys and motors/drive pulleys, and endstop sensors the same amount. I had to modify and reprint all the motor mounts, end-stop sensor mounts, and corner pulley stand-offs. Raising the X axis guide rail also raised the magnet, so I reprinted the upper belt clamp part of the magnet carriage to lower the magnet so it stays at the its original vertical position and doesn't scrape the bottom of the sandbox.

Magnet carriage. The green parts is the new, thinner version that was needed to lower the magnet and prevent it from scraping the bottom of the sandbox. The gaffer's tape wrapped around the magnet keeps it from lifting up 

Note: the 3D printed PETG concave pulley flanges and motor mounts appear to be holding up just fine, as are the belts. There are some rubber crumbs around some of the pulleys, but that's to be expected. Eventually the belts will wear out and have to be replaced, but it won't be any time soon. I tried moving the twists in the belts from the short segments between the drive pulleys and the corner pulley blocks to the very long segments between the corner pulley blocks and opposite ends of the table. I was originally concerned about noise from the belt teeth hitting the pulleys at the corner blocks and with belt clearance along the long sides of the mechanism. Neither seems to be a problem, so I left the twists in the new positions. 

I may take another crack at using sliding UHMW blocks instead of wheels, depending on how well the wheels hold up with the new X axis configuration.

4) The solution to the LED problem was to install a switch that allows me to cut 24V power to the controller board. There are now three switches located on the underside of the table, one for main power, one to control power to the controller board, and one to switch power to the floor lights on the underside of the table. 

Here's the mechanism running:

3D Printed Knobs For My LG Stove

I've had a wonderful LG stove with induction cooktop and convection oven for a couple years. Everything about it is great, except the control knobs. The knobs are chrome plated plastic with black text stenciled on. The chrome plating reflects light in such a way that the thin black text is difficult to read in almost any light. The advantage of using a chrome finish is that it's easy to clean, but being able to read the text on the knobs is kind of important, too.

The front of my LG stove with the original, hard-to-read knobs. 

I recently got a Bambu Lab H2C 3D printer so I decided to try to print some new knobs for the stove with easier to read text using the multicolor capability of the printer. I thought it would be nice to use contrasting colors for the bulk of the knob and the text, and even to raise the text above the knob surface. And of course, I will want to be able to put the knobs in the dishwasher once in a while, so printing with PLA is not an option.

As you can see, it is pretty hard to read the thin, black text on the chrome knobs. This is true under almost any lighting situation. Also note the positions of the burner maps on the stove panel. If you study it hard enough, you tell which knob goes with which burner!

Designing New Knobs

I started by making some measurements of the existing knobs, and the control shafts on the stove. The shafts are round with a single flat surface on one side, commonly called a "D" shaft. It will be critical to get the knobs to fit tightly on the shafts, but not so tightly that the knobs can't be installed and removed without breaking the controls. The original knobs have steel inserts lining the holes and providing spring pressure against the shaft.

There are multiple ways to approach the design, and naturally, I started with a bad way. I attempted to print the knobs to fit on the D shafts on the stove and just couldn't get the consistency I wanted. They were either too tight or too loose.

The underside of one of the knobs. You can see the large steel weight used to make the knobs feel less cheesy, and the thin steel insert that grabs the control shaft inside the hole. I was unable to extract that insert. Note the white plastic around the shaft hole and apparently under the steel disc. 

Some failed test prints used when designing the new knobs. Don't try to print knobs that directly fit on the D shafts of the controls. It doesn't work.

I took another look at the original knobs and found that beneath the steel weight used to make the plastic knobs feel less cheesy, there is a white plastic insert glued into each knob. The glue they used was a sort of soft and rubbery, probably polyurethane. After a little prying with a screwdriver, the glue bond broke and the insert popped right out.

The insert is glued inside the knob shell with some sort of soft white glue.

The insert and steel weight extracted from the knob. I modeled this in just enough detail to design a knob to fit around it.

I put the insert assembly on the stove to check the clearance between it and the surface of the stove. In order to turn a burner on, you have to push in the knob 2 mm and then rotate it. I measured the clearance between the steel disc and the stove surface with the insert pushed in to figure out how far my knob design could extend toward the stove surface.

I went to work on modeling the insert in just enough detail to design a knob to fit around it. I printed a single color prototype and it fit perfectly on the first attempt.

The inside of the successful prototype knob with a cut-out designed to fit the insert and steel weight.

Single color prototype knob on the stove. I designed it so it sits 4mm closer to the stove control surface than the original knobs - it looks like it was specifically designed for this stove that way.

Multicolor Painting

I decided I wanted to use white text on a black background for maximum readability. Then I added some purple/blue color for the body of the knobs, and finally a little green, as you'll see below.

Coloring the print is done using a painting tool in the slicer. The knob is going to have white text against a black background. You can either start with a black knob and paint every surface of the text white (very tedious!), or you can start with a white knob and add the black color around the text- much easier! 

Here's how you turn a CAD model into a multicolor print in Bambu Studio:

Bring the part into Bambu Studio as a white object...

Flip it over... That white tower is there because I have the smooth time-lapse feature enabled. It wouldn't be there otherwise.

Select the paint bucket and black color...

Mouse over the area to be painted black- note that the edge detection finds the outer edges of the text and the edges of the polygon.

Click to paint it black. Now you still have to fill in the holes in the text, such as the center of the letter "O", and the corners of the polygon, but that's a lot faster and easier than coloring the text itself.

When you start with a white knob and paint the black areas, the inside of the knob will be mostly white.

Start with a black part and paint the text white, and you'll get a mostly black interior.

Coloring the bulk of the knob- positioning the mouse pointer highlights the boundaries.

One click of the mouse fills in the color.

Roll the mouse pointer over the target area and the boundaries get highlighted.

One click and the color is applied. Repeat 11 more times going around the knob.

More Details

There are three different kinds of knobs on this stove. There's one for the oven, that has a LOT of text, one for the warming burner, and four identical knobs for the induction burners. The oven knob was the most challenging- I had to fit a lot of text on it but the printer managed to produce easily readable text even with just 3 mm high characters.

The inserts weren't just glued into the original knob shells. They have concavities that mate with retention features inside the shells. There's also an index notch in the insert so that it can only go into the knob in the correct orientation (the flat side of the control shaft is down while the "off" text is up when the knob is on the stove). 

Concavities used to retain the inserts in the outer shells.

Retention bumps that snap into the concavities in the inserts. I'm sure these have a name, but I don't know what it is. On the left side you can also see the indexing bump that prevents the insert from being installed in the wrong orientation.

I made a test print with added retention bumps and an indexing bump in the interior of the knob design. The inserts with attached steel weight snap in tightly and don't wiggle at all. They can be easily popped out with a screw driver so the printed covers can be cleaned in the dishwasher.

While I was at it, I also reduced the bulk of plastic on the interior of the knobs so they'd print a little faster and use a little less filament (instead of the bulk of the interior printing, it has to print larger tree supports).

CAD render of the interior of my knob design. You can see two of the four circular bumps that fit the concavities in the inserts. Toward the right you can see the index bump that ensures the knob fits on the insert in the correct orientation. The walls are about 3 mm thick - more than strong enough for the job that the knobs need to do.

Finally, I added burner maps to the front surfaces of the knobs so it's easy to tell which knob controls which burner. I marked the target burners with green filament which matches the green sparklies in the purple/blue filament. 

Printing

This is a situation where printing all the knobs at once saves a lot of time and reduces wasted filament. There are 292 filament changes and only one prime tower (about 20g of filament) for the full set, which requires about 21 hours to print - a little over 3 hours per knob. If you print the knobs individually, the printer performs 292 filament changes and prints a 20g prime tower for each knob. So it takes about 5.5 hours to print a single knob.

The knobs are printed in 0.2 mm layers, with tree supports, 4 perimeters, 3 mm top and bottom thickness. The knobs print solid- there's no sparse infill. The filaments were Bambu Lab Black PETG-HF, some random white and transparent green PETG filaments I am still trying to use up, and the sparkly stuff is Stronghero 3D Mirror Chrome PETG in Wizard Voodoo Purple/Blue that I first opened in October of 2023. My hot girlfriend pointed out that it's a nice match for the blue porcelain interior of the oven.

Sliced set of LG stove knobs in Bambu Studio. The full set uses about 420g of PETG filament - about $10 worth. The 292 filament changes add about 2.5 hours to the total print time of 21 hours.

The Results

I printed one knob first as a test of the fit and finish, was satisfied with the result, so I printed the remaining five on one plate:

The remaining five knobs took about 18 hours to print. The four burner control knobs go into the slicer as identical parts. The green color that makes them unique gets added in the slicer.

The printer has cameras built in and it's easy to make a time-lapse video:

And here they are on the stove. The purple/blue filament looks different depending on the light source and viewing angles. It's also full of green sparkles!

It's so much easier to read the knobs and to identify the burners being controlled.