I mentioned back in August that one of the next projects on the list was the next version of the eSpinner.  Well, September wasn’t the month for it.  Mostly, I wasn’t really in the mood for CAD work and I knew there would be a lot of it.

The first week of December though – I was all over it. We had a brief cold spell and it seemed like a great time to take advantage of the time in.

I started the modifications on December 2nd and ended up doing the bulk of the CAD work over the next four days.  At the same time, I began the printing.  There were a couple of parts redesigned from scratch but most were modifications of the designs I’d created for the first version back in 2019 when I was learning CAD.  This of course has its own pitfalls because how I design has changed a lot and what I did when I was learning was very inefficient.  So in updating the files, I had a lot of things I needed to modify other than just dimensions of the drawings.

Here was the wish list for V2.0:

1. look less like the dimensionally accurate reproduction of the treadled/non-electric wheel I modelled it after
2. make the base smaller and the eSpinner more compact in general.
3. a way to easily reinstall the drive band when it comes off if you remove the flyer.
4. metal nuts to match up to the fasteners for longevity
5. a way to tension the driveband – right now, pony bead lacing does a great job though.
6. a better motor mount
7. move the power to the left
8. move the mother-of-all to the right so a proper brake knob can be installed
9. change the mount of the motor to be a little stronger

The result

So let’s look at how this wish list informed the design of v2.0.  I was more or less successful on all points but the first.

1. Looks: Uhm… fail. At the end of the day, I decided it wasn’t so bad for this eSpinner to look like what it was: a wheel that could use Lendrum bobbins.

2. Smaller: The box was shortened by 1″ in height and in width.  The length had to stay the same.  The printed sides of the box were printed much more densely though, so there’s no weight reduction.  In fact, I think it might be heavier.  I did that on purpose because with the motor mounted high on the right side, it seemed like there was a possibility that it may be unstable if I reduced the overall weight.

3. Driveband maintenance: Previously, there had been a round hole in the bottom panel I’d intended to be able to pull the band on through.  In practice though, the wires for the control panel on the front got in the way and this wasn’t usable, nor was the hole big enough to see into and manipulate the belt at the same time.  This was turned into a slot and the electronics were relocated.  Additionally, I added a “parking spot” on the front maiden to hold the belt under tension when I remove the flyer. I had to reinstall the belt today when it broke.  It’s been sorely abused though in the testing/prototyping/disassembly/reassembly/etc. It’s gone through in the past couple of years and especially past couple of weeks. It was possible to do this with the orifice hook and the slot at the bottom.

4. Hardware: Check.  Metal nuts installed in panels.  This is the one modification that lead to the most reprints. All of the large reprints were due to this change.  It’s deceptively complicated to do this.  The holes to insert the nuts have to be big enough to get the nuts in but no bigger so they can’t twist if they meet any resistance.  The nuts are after all far stronger than plastic and they will just push their way through the plastic.  Additionally, the holes for the screws must be big enough to provide no resistance whatsoever or they will take the nut and screw off course and strip the screw or nut and possibly trap the pieces together forcing you to break them apart to disassemble.  Yes, that’s the voice of experience here.  With 3mm hardware too, it’s not hard to mash a little bit of thread if anything at all goes wrong.

5. and 6. and 9. Motor related: These ones had me stumped for a while.  In the end, I designed a block to mount the motor brackets to and that block matched up to slots in the right side of the box that let me move the motor up and down.  Previously, the motor was mounted on the bottom of the lid. This also makes it easy to unscrew the lid and have access to the interior without having to flex the wires on the front control panel.

Which turns out to be a good thing, since I’ve had to replace the switch and re-solder 2 other wires on the pulse width modulator already.  That’s why throughout this post, the control panel (front) changes configuration and the switch changes to a round one!

The addition of a retention plate front and back means that there’s no way this motor is moving again – even if it gets warm enough to deform the plastic which it shouldn’t because I’m not using the same type of band that caused the heat to begin with.

7. Power Plug: Check.  Though I can’t at the moment remember why I wanted to do this, it did work well with the relocation of the rest of the electronics/electrical to the left side.

8. Brake Knob:  Making the overall width narrower had the effect of moving the mother of all to the right which gave better access to the tension knob. The tension knob and the orifice hook knob are going to be turned on the lathe at some point.  These are 2 parts that are weak as printed items and there will always be a wood flyer on the wheel so this seems the best way to handle it.

I’ll post a pic of this closer up when I turn the knobs.  I had an idea while I was writing this post. The one in the pics above is the original one anyway.

Now for the worst part of the project:  Testing.  No wait, did I say worst?  I mean the best ever!  I have a license to spin in every spare moment in the name of Quality Assurance. 🙂

Here’s the full pic I used for the header by the way:

So does that mean I’m finished with this build now?  Probably pretty close.  I need:

• non-skid grippies for the bottom
• to turn the orifice knob and tension knobs
• I’d like a foot control to turn it off and on. I can use the IKEA Tradfri wireless outlet when it’s plugged into mains but if I’m using a battery pack, that doesn’t work.  I need to ponder a little more. Only this one would require any CAD work potentially.
• I could also see about doing something to hide the hardware or change the Allen head screws to something less conspicuous but I’m not really that bothered about it at the moment.  That could change.

Today’s post title comes from here: Mechanical Heart – Beth Hart.

The Process

For those interested, here’s some of the process – including pitfalls and frustrations of 3d printing (spoiler: it’s likely not what you think), the benefits, and a little bit about what’s misunderstood about 3D printing.

1. Design something like a new motor mount in CAD. This might take me an hour or so – depending on complexity and how clear my vision what I need is –  including some revisions before I decide I need to hold the part and try to install it to see if it fits how I need it to.

The other thing I need to design for is the limitations of an FDM 3d printer.  They can’t print in midair and they print the weakest on the vertical axis.  So, I design so that the parts are laying flat on the axis they need the most strength. This and preparing the file for printing takes some experience and trial and error learning.

The motor mount took about an hour because I already had part of it completed and the rest I finally had a good idea of what I wanted to do after discussing it with Ryan then sleeping on it. This was a relatively simple part to design.  Even redesigning the side panels took me longer.

In general, I catch a lot of flaws in my CAD designs before printing – usually at the stage where I prep the file for printing – so I rarely print a part more than twice, with the odd challenging part a third time.  The average part I design may go through 10 revisions but only 1-3 prints.

This is a very good thing for a couple of reasons.

• • • It’s wasteful. While some printing filaments can be reclaimed and reused in new filaments, the one I use isn’t currently easily reused locally and shipping the waste somewhere else makes little sense.  So, it’s best to try not to create the waste in the first place.
    • It’s so frustrating to find a tiny but fatal error after waiting 4 hours for a print (like a side panel), make a 2 minute change in the CAD file (because “tiny error”) and wait another 4 hours for the new print.  That’s the better part of a day wasted over a small error.

2. Time to print it.  In the case of the motor mount and retention system, this print took about 2 hours on draft mode. I used the time to work on the CAD for other parts that were being updated.

3. Installation – During installation, I may find a couple of small tweaks to work the way I needed but if I can get it installed, I will do that and check that the clearances are what I need with the other components of the build. I’ll make notes.

4. Some modifications in CAD using the notes I made from the previous step

5. Print again. Another 2 hours.

6. Installation: Hopefully with the design changes, this can be installed in the final eSpinner. In the case of the motor mount – it takes about 10 minutes to assemble and install into the eSpinner. (Mostly thanks to very tight quarters). The entire eSpinner takes me a couple of hours to assemble.

Despite how long I wait between reprints, this is still rapid prototyping because I’m waiting hours and not days/weeks/months between iterations.  It’s just when you get fixated on a project like I was for this one and you’re “In the Zone”, a 15 minute print is challenging to wait for and disruptive.  4 hours – as was the case for the side panels of the box and then to find an error can be enough to walk away in frustration. At first, it was OK because I had so much CAD work to do, I could keep myself busy.

For the most part, I had designed and built the next iteration of this eSpinner in less than a week.  I started on a Friday night and was spinning on it by Tuesday evening.  After a few days spent testing it, I opted to make a few changes to 2 parts and did the CAD work the following Sunday.  The eSpinner was back together before dinner and I’ve been spinning on it since.

The Stats

So let’s look at how much time, materials and resources I have into the redesign of this project.

(If I had to guess, I probably had close to 100 hours into the initial design but a lot of that time was learning the software and how to design in general. )

• Between Friday (Dec 2) and Tuesday (Dec 6) I put in about 14+ hours of CAD time.  This Sunday, I did another hour or so.  The one fix was easy but the control panel saw a complete redesign.  Let’s call that 16 hours.
• The printer was running for nearly 53 hours – during waking hours only because sometimes they mess up and go all Mount Vesuvius circa 79 A.D.  If we’re asleep when that happens, the damage is likely to be far worse.
• This project used 1.83KG (just over 4 lbs) which is about 600m or almost a 1/2 mile of filament.

Of the time and weight, 38% of the time and weight was in redo prints.  The wasted filament was about 700g between failed prints (10g) and do overs and more than 20 hours. That means that the finished printed weight of the eSpinner is about 1100g. This is a much higher redo/failure rate than I usually have (usually around 5-10%) but this project is far more complex than any other I’ve done and needs to fit together not only with its own parts but also the electronics and electrical parts I’ve ordered to power it. Dimensionally accurate CAD is always the most challenging and the most prone to waste in the early prototyping.

I hear some people saying “If it’s just a ‘copy’ of an existing wheel, why not 3d scan it? Even phones can do it.”

So, yeah about that.  First, good scanners are extremely pricey (far more so than the average printer) and have a huge learning curve.  The learning curve is sometimes so great that people/maker spaces/libraries, etc. have bought them and they sit on a shelf because no one can use them well. A local library around here did exactly that.  They also don’t do all of the heavy lifting.  There’s still a bunch more work to be done after the scan is finished.

While the computational part of photogrammetry is an involved process, it is actually the easiest part for the user, who often only has to drag in their images and push a few buttons. The real work starts when the 3D model has been generated. In no case does photogrammetry deliver a ready-to-3D-print, watertight mesh model. There are usually floating artifacts, background noise, holes, and irregularities to clean up. The object will also need to be reoriented and rescaled, which is done quite arbitrarily by photogrammetry software.

What I’m talking about here are the really good scanners/software.  Phones are another story altogether. Some phones advertise that they can scan and send directly to a printer.  Here’s the thing though: If good/expensive photogrammetry setups can’t product a printable file (see the quote above) – why would a secondary feature on a few hundred dollar phone that has way less processing power and an inferior camera be able to? No, photogrammetry on phones is really little more than a gimmick at this point. It’s actually doing 3d printers a disservice by making people think it’s as simple as a little scan and pushing start.  There’s a lot more to getting a good print and when folks think printers work like that, it sabotages 3d printer development as a viable tool for the future when they learn otherwise.

At the end of the day, the issue is the amount of human labour required to get the model ready for printing and the expectation from so many that there are no humans involved or needed (and that those humans don’t need to eat).  In theory, bespoke work is where these printers could excel but the design time often makes it a moot point. If you could just say I want a Pfaff pattern cam and the printer could print it, it would be a whole new world.  Unfortunately, we’re a long long way from even being able to order Earl Grey tea with a voice command and have it appear in front of us in a cup at the exact right temperature with no human intervention.  Replicating parts that the (sometimes 8bit) computers onboard these printers have no instructions for?  No, that’s decades away.