Hey all,
A short while ago, I blathered some words about stamping and 3D printing and how manufacturing is changing and getting reallll interesting in some ways and other people posted CNC robots and other cool stuff I didn’t know existed.
This is not a chapter, but it’s a catch-up on where I am, and a follow-up on those thoughts.
As to where I am, I am deep deep deep in development work, with basically every day being build-a-board test-a-board tweak-the-board reorder-the-board or print-the-chassis tweak-the-chassis print-it-again or new-stuff-coming qualify-it-now tweak-if-needed retest-confirm-repeat.
If this sounds horrifying, it’s not. It’s actually what I love to do. I wouldn’t be designing things if I didn’t like it. So this is really, really good for me.
Frustratingly, we’re working on a bunch of very interesting stuff, most of which I can’t talk about, mainly because it’s wayyyy too early to think of it as a viable product. But I expect 2023 to be extremely surprising, and not just because we had to move both our big ideas out past the end of the year (one as our choice, one due to parts being late, ah well.)
In other news, parts availability is definitely getting better (though still no Analog Devices DSPs, sigh), but costs continue to be bonkers. We’re trying to hold the line where we can, but don’t be surprised if you see some more bumps. Basically, if you’re looking at buying, best to buy early and often. With demand softening, we’ll see if we can turn the screws, but don’t hold your breath.
Okay, on to comments about manufacturing.
Stamping. For a long time, we’ve been using stamping to produce most of our products. This was both very smart (because it costs less if you’re doing more than, say, about 250 pieces), and very dumb, because we just handed over our CNC stuff and had them do the same thing.
A couple of points that should not be missed.
- 250 pieces is nothin. Nothin at all. Literally everything we do is cheaper when tooled and stamped rather than CNCed. CNC is fine when you’re making 50 pieces of a $1200 1” thick front panel (not exaggerating, I’ve seen it), but not a great idea when doing big numbers.
- Stamping is not CNC. CNC parts you pay per hole, per milled edge, per slot, per cut, per complex shape, and when you want to do something like a dimple or something, you get to do a tool anyway and deal with the whining and crying from a company not used to bending metal, and what happens when you bend metal. Seriously, you should see what the first dimpled bottom for Folkvangr looked like. It was a taco. We fixed it, but oh what a pain.
And the biggest point about stamping is this: if you’re not paying for every hole, not paying for every complex feature, not paying for debosses and dimples and swales and stuff,
then why not use it to make your product better? You have to get out of the mindset of CNC. And it’s hard to unlearn, when that’s all you’ve ever done.
Bottom line, when you treat stamping like stamping, you get much better results. You’ll be seeing some of what we’ve learned in the flesh here in a few weeks, and I think you’ll be shocked at how big a difference it makes.
3D Printing. For a long time, we’ve been doing 3D printing for prototypes and such. That’s also both very smart and very dumb. Very smart because it does help us avoid boneheaded mistakes, but very dumb because we were unaware of how far 3D printing has progressed. Blame it on being immersed in the industry since the 3D Systems days, when a stereolithography printer with a modest build envelope was $250,000. You can get a LCD resin printer with roughly the same capabilities for about $500 now.
So. Yeah. 3D printing. We recently bought a carbon fiber FDM machine, mainly for doing larger parts, and also some functional prototypes (brackets and stuff you can use). That’s been a learning process when coming from FormLabs SLA printers.
But it’s also fairly amazing what you can do. The larger protos mean that fit and form are much less guess-based now, and the ability to iterate some wacky ideas in real time mean that you’re gonna be amazed at what some of the new products look like.
The real breakthrough, though, started as a disaster.
I designed a new product to use our standard light pipes—injection molded pieces that are done standard 5-up, cut ‘em down for less, stack ‘em up for more, hope you like the spacing deal. The standardized spacing never thrilled me, and the fact we have products with 6 LEDs in a row never thrilled me either. But neither was disastrous until I did this new product…and, due to a change in where we mounted the board, the light pipes were too tall. They literally scraped the surface mount LEDs right off the PCB when installed.
This sucked. Big time. Because that meant changing the injection molding tool. Or throwing away a bunch of metal. Both are big $$$ and a ton of wasted time.
But that’s what we were gonna do—change the injection mold and run new parts.
Of course, during the course of changing the 3D CAD model, I needed to 3D print some prototypes to make sure they fit. They printed fine on clear resin, they fit fine…and
hell, they looked pretty much exactly like the production parts. With the white LED light going through them, you couldn’t tell which was injection molded and which were 3D printed.
This led to a crazy thought: what if we just 3D printed shorter light pipes?
I mean, it would be great in some ways. It would save us from changing the tool, save a bunch of time, and hell, we could even print the exact number we needed, rather than cutting them down from 5 places.
But it also sucked in some ways. I mean, resin printers are sloooooow. And kinda pricey in terms of resin.
Right?
Weeeeeeeeeeeeelllll…maybe not. I mean, we had a flexible build plate for the Formlabs printer; we could build directly on it with no supports. So I threw together a “high density” layout of light pipes and ran it. 3 hours for 140 parts. Not wonderful, but tolerable for a short run. Unfortunately, disaster struck and the Formlabs printer dispensed a bunch of clear resin into its guts, and the high density layout proved to be a bit too high density, the parts grew together.
Aside: yes, this can sometimes still the state of 3D printing, even on one of the most user-friendly machines out there. Beware.
But…there was promise. The parts that didn’t grow together worked well.
And I knew there were bigger, cheaper printers out there…printers that used resin that was literally 6-8x cheaper as well. But how would they do? Would these new printers (based on LCD screens masking UV LEDs, typically 6-8K resolution monochrome screens) be able to do more parts, faster?
I went ahead and picked up one of those new big cheap printers (from Amazon so we could send it back if it was garbage, natch). After a fairly easy (but much more DIY) setup, I laid out a 270-up light pipe layout and let ‘er rip. 19 minutes later, we had 270 perfect parts, without a trace of the growing-together problem of the other printer.
Yes, you read that right. 6X faster, 2x the number of parts, resin a 1/6 the cost.
That’s when Tyler commented, “Maybe we can just make all the light pipes this way.”
“Oh no,” I told him. “3D printing is a pain. Even the screens are consumable. You’ll need to have someone here run it, clean it, it stinks, it’s not really ready for prime time.”
But after one printer did a couple thousand parts in a day, I even started wondering. And I ran the numbers. Here’s the completely unexpected takeaway: even when factoring in consumables, operator time, cleaning and supplies, and unexpected breakdowns, it was literally 5x cheaper to print the light pipes. Hell, the raw resin cost was 1/13 that of the injection molded parts! As in, the injection molded parts were $0.16, the resin was $0.012.
Holy moly, we really could do it, if we wanted to.
Jury is still out, but we now have two of these large printers, and two wash stations, both cranking away doing prototypes…and maybe soon more. For those keeping score, that brings us up to 5 3D printers, three in Valencia and two in Corpus.
Sometimes desperation is the mother of invention. More on that later.
All the best,
Jason