How to make (almost) anything (2020) | Hacker News Reader

How to make (almost) anything (2020)

fab.cba.mit.edu

277 pointsakeck58 comments

Next weekend can't come soon enough :)

I'm in the high power rocketry hobby and got a Prusa i3 mk3 for myself for Christmas. It's pretty amazing to sketch a bit while on conference calls, then after work model my sketch in Fusion360, export the mesh, import to Prusa slicr, export gcode to an sdcard, start the printer and goto bed. Next morning, there's my part waiting for me.

Electronics is the next bridge to cross and i really hope this class helps with that. I know a guy in the hobby who went from nothing to selling a custom designed altimeter for rockets on his website in about 6 months which was pretty impressive to me. https://flightsketch.com/store/catalog/flightsketch-mini_1/

Next thing you should do is get a raspberry pi with octoprint! Then you can drop the gcode in from your web browser :)

Or even easier, push to octoprint directly from your slicer!

What annoyed me was that the slicer does not mantain hole diameters, somehow holes are smaller. In my case I was planning to prototype cnc milled stuff (which is kind of expensive), but this killed to use case.

The slicer does maintain hole diameters, but just like CNC milling HDPE, you should expect dimensions to change a little after printing. If you need to have a tight tolerance fit on something, expect to run a drill or chucking reamer through it after printing. If it's really tight tolerances, it's easy enough to melt or press a drill bushing or threaded insert into an oversized hole to get fits tighter than you're likely able to hit with a mill.

Your expensive parts likely go through many steps before and after manufacturing already that involve things like fixturing the material in place, rotating it and reacquiring datums, running taps through by hand, deburring sharp edges, etc; you should not expect 3D printing to be as easy as sending the STEP file to the machine and getting a complete part out. Sanding, acetone vapor polish, drilling, reaming, tapping, etc. are normal parts of the manufacturing and prototyping process.

Just an fyi drills are not good for tight tolerances. They cut trilobular holes, not circles.

Reaming or boring are the route to tight tolerances.

It is much tighter than 3D printing, though - good enough for most cases if you’re using a drill press

Which slicer do you use? It seems to be a known problem of cura.

I have had the issue with both Cura and Simplify3D.

For a circular hole, as far as I can tell the slicer commands the hotend path to trace the circle exactly, rather than tracing a slightly bigger circle to account for output thickness.

I imagine if you were to use smaller nozzles the issue would be lesser, but not gone.

I use Cura. It’s across FDM printing techniques and is (somewhat) akin to the errors in subtractive manufacturing. Subtractive errors/inaccuracies tend to be of the type that remove too much material. Additive errors/inaccuracies tend to add material.

My models for FDM often include a “hole size compensation” user parameter and I tweak it to introduce an offsetting bias in through-holes. (Non through holes tend to get heat-set brass inserts, which are more forgiving of the 0.1mm or less radial errors.)

You can print a clearance tolerance test to see how accurate your printer is. I use a Prusa mk3s and as accurate as it is, at best my tolerance is 0.1mm. Enough to stop your parts fitting together. I normally build in a 0.2mm tolerance for close fitting parts.

i've noticed things shrink a bit when the cool off ( at least with PLA ). What i do is print, test fit, adjust dimensions, then print again. It usually only takes one or two iterations to get it just right. I built a deck once, designed it, spec'd out all the materials i would need, optimized for minimal wastage, the whole 9 yards. I learned the hard way that in the real world dimensions are a matter of degree and what's on the label may or may not be what you get (to a ridiculous degree when it comes to lumber).

I'm use to software where it either compiles or it doesn't, it either works or it doesn't. In the real world with physical objects it's much more a matter of "is it close enough or not".

Lumber is regularly sold up to an inch off the label. But that is standard and expected. Meaning people in the trade know how much lumber to buy given the finished dimensions of what they buy.

I have no idea why this is standard practice and when building out http://MajikBus.co ran into minor issues.

Holes with interior dimensions close to the width of the nozzle are hard and need to be calibrated if you absolutely need to print them to size. Basically, the plastic is "spreading" a bit wider than the nozzle tip.

better solutions, rather than making exactly sized holes, is to increase the number of vertical walls, make the hole size slightly under, and use a precision drill and reamer to get to the target size. Also, make sure you're setting your scaling right (for example, I need to set my object size to be 100.6% of its designed size, because PLA shrinks when it cools).

This is, I think, partly due to how the slicer routes the filament paths. I've had this problem moving between nozzle diameters and the only reason I could think of was that it wasn't possible for the slicer to maintain both accurate hole diameters and the correct placement of filament relative to previous extrusions. The STL was identical, but the parts printed with a 0.8 mm nozzle were undersized and only the holes. My experience of this was with Cura and an Ultimaker S5. Generally you should always start printing loose holes and tightening them up as you go to see where the limit is.

The simplest dimension to control is the perimeter. That can, for most parts, be perfectly accurate because there are no constraints on where the nozzle needs to avoid. However if you print a hole close to the edge of a part then the printer somehow has to lay down both an accurate edge to the hole while maintaining suitable separation from previously printed stuff. Essentially your design is quantised by the nozzle diameter within the perimeter of the part (I may be wrong here, but this seems like a simple explanation). You may have better results if you tune your part to the printer/nozzle you're using so that the slicer doesn't have to worry about leaving/filling gaps that are non-integer multiples of the nozzle size. In theory this shouldn't matter, because the hole perimeter can also be printed first, as it's an "outside" edge, but it doesn't always seem to work that way.

Yeah this is just a thing you have to design around while using FDM printers, as others have mentioned. I really like melting in threaded inserts if you need threads, otherwise use a drill bit or reamer. If you need super tight tolerances (like CNC level tight) then a consumer grade 3D printer likely won't get you there though.

Prusa uses 3D printed parts and I'm a bit surprised just how good the prints are.

I own one and can confirm they are quite nice. Still though, you aren’t getting machining level tolerances (0.001”) on a FDM machine no matter how hard you try.

Might be worth checking out heat-set nut inserts: https://www.mcmaster.com/heat-set-inserts/ I've has good luck pushing these into holes in 3d printed parts with a soldering iron the past.

IIRC there's a param in the slicers I've used that can fine tune hole sizes.

Chordial deviations add up on smaller holes.

Minimum size line segments and other things come into play on holes less than about 0.2"

What kind of parts do you print? I imagine engine parts are right out, and I'm guessing fins might melt as well. Nose cones? Body or internal stuff?

my least favorite part of building a rocket is the motor mount. It requires a handful of parts, a decent epoxy system, and is one of the more expensive parts of the airframe. I've designed a modular mount that i can pull out of one airframe and plug into another. So once the motor mount is done i can use it in all my airframes of the same diameter and motor diameter without any bonding. To cover all my current airframe diameters and motor diameter combinations requires 5 prints and adding a new one is pretty easy.

my next thought is modular fin cans. The rule of thumb for trapezoidal fin dimensions are a function of airframe diameter. I would like to be able to make one fin can for each airframe diameter and then re-use those on new airframes. With my current inventory I need 4 fin cans ( 2", 3", 4" and 7.5" diameter airframes )

Not gp, but things you can definitely use 3d prints for related to rocketry: nose cones, fins, motor mounts, electronics mounts, and probably a whole slew of other stuff I'm forgetting at the moment.

I never took this back when I was at the Media Lab and that’s one of my main regrets from that time. This class is very famous, but I’m not sure how useful this is as an online class because as a normal person you won’t have access to the water jet cutters, plastic mould injection machines, etc.

Covid aside, hacker spaces were already dying out in SF. I wish there was more push for ordinary people to become makers rather than just consumers.

3dprinters are pretty accessible but you're right about other tools. Also, some of the tools can be pretty dangerous ( looking at you lathes/mills ).

IMO one of the greatest things about software development is you can goto walmart and get $500 laptop, sit in a coffee shop, and have access to all the tools needed to build cutting edge high performance software of any type. It's not so easy to do that in meatspace with physical objects.

As someone not much into this scene, how useful are 3D Printers actually? I get the sense that they're very useful to hobbyists or for creating prototypes but are the plastics used suitable for production use cases?

Somewhere between useless and a godsend. It depends upon your application. There are many choices in materials and printer set-up options that greatly change the results that you get. Depending on how demanding your application is, you may or may not find the result meets your needs. A friend that is a freelance mechanical designer for medical devices has a small, inexpensive (relatively) SLA printer, and she sometimes is running it flat-out for days doing prototypes to show clients. Instead of sending that to a service and billing it through, she gets to bill it directly, allowing for a nice margin and also giving client a discount. My friends that build hobby robots use FDM printers for a lot of functional parts -- and results vary depending mostly on how appropriately the part was designed. But a Prusa is a lot cheaper and a lot easier to run than a CNC milling machine, on the flip side a chunk of FDM plastic doesn't compete well with a chunk of aluminum or titanium. At my work, we 3D print a lot of functional enclosures and other low-mechanical-load parts for prototyping.

As you say, they are useful for hobbyists and prototypes. They don't scale well but will in future.

Today it is small/medium business territory but not a large part of megafactories just yet.

Production thoughts:

- speed. 3D printers are not speedy, the tradeoff of speed vs flexibility is clear. Iterate the design with a 3D printer, make the mould and produce with injection moulding.

- material. $$$ for capability. Cheap plastics <$500. ABS+nylon $1000-$2000. Metal is $50k - $1mil depending on what is desired. Workable for some businesses. IE specialist mechanics: https://www.youtube.com/watch?v=4jbn0ah3u9E

- Method affects end result. Resin printing is different to FDM printing etc. FDM printing works in layers that make a plastic end result not food safe. So no printing kitchen bowls/ladles etc But that leaves a lot of possibilities out there!

The hobbyist angle pros:

- not that expensive! Mine cost $300.

- handy. Keyholder hooks, odd angle shelf brackets, fixing/upgrading kids toys, etc etc

- can send model elsewhere to be printed in fancy materials. See Shapeways.com

- can print models from others. See Thingiverse.com

Hobbyist cons: - time for 3D modelling. Unless you already know how to 3D model then this will take you time to learn. I see it as a "mindfulness" hobby, like knitting/crochet etc too.

I think to some extent it's a frame of mind thing. For instance, my friend's vacuum broke for some reason or another, some plastic piece that wasn't super complicated, but was required for functionality. And rather than buying a new one, he was able to model and print the piece that broke and the vacuum is able to work again, but mileage in general probably varies.

It depends what you mean by "production." There are many kinds of plastics. Common filaments like PETG do have production uses.

A 3D printed part is unlikely to be better quality than something you can buy, assuming you can buy it. There are also lots of limitations; for example, watertight prints are difficult and limited. Also, making something food safe might require a special coating, so it's probably not worth it. (I haven't tried.)

As a consumer, there isn't likely to be anything you need that you can't order online. However, if you want to make something new, with just the right dimensions, 3D printing can be very useful for many random parts you need. Also, you can make improvements to things you buy.

I use mine to 3d print parts and then cast them in metal. They make special resins for this process, works great. I can go from Sketch - Fusion - Slicer - 3d Print - Cast in about 18 hours depending on the part size (long time to print complicated / big parts). If you're interested look in my bio there is a link to my youtube - watch the "casting gold balls" video - it basically runs through the whole process (albeit without explanation).

Without going in to it - there is a massive range of plastics and non plastics of varying quality, expense, and accessibility.

Yes, 3d printing is a legit method for production. Typically if you can have a die made for it you will be better off in the long run with casting. Dies are incredibly expensive however.

Smaller runs or highly specialised and difficult geometries benefit from 3d printing.

except for video games and deep learning

depends on the video game. If you target what AAA was doing 5 or 10 years ago in term of graphical quality or even specific style like low polygon, pixel art, 2d art..., you don't really need the cutting edge in terms of graphic cards.

Noisebridge is re-booting pretty successfully in San Francisco -- a break during early Covid was either well-timed or overdetermined, because they also had to move venue due to rent rises, but the new space is coming together, and there's a bunch of in-person events now kicking off to celebrate the space's fourteenth birthday:

https://www.noisebridge.net/wiki/Noisebridge

Happy to see they're back! The death of maker spaces in SF has been very sad for me.

I wonder, how can i make a good application to Media Lab? Would appreciate any insight, I am looking to apply under Hyperinstruments by Tom Machover

My suggestion would be to highlight what makes you unique, what you'd want to do if you were accepted, why you want to join that group and the ML generally, and interweave it all into some passion / vision for the future.

Here's a complete class listing

https://cba.mit.edu/classes

See also "How to grow (almost) anything": https://fab.cba.mit.edu/classes/S63.21/

Note that the title is fairly misleading, students don't actually do real practical labs. A random example: https://fab.cba.mit.edu/classes/S63.21/class_site/pages/micr.... Probably because this class was being held during COVID times.

There's no hands on biology work. Homework consists of:

- Find a research or journal article - Propose a technology - Propose a methodology

etc

The 2019 version looks a little better: http://fab.cba.mit.edu/classes/S66.19/S66.19/

If i am allowed to joke, I was expecting to read an article on Rust.

I think you misread the title as “how to remake (almost) anything”

Thank you for linking to this. There are a couple of topics here that I've tried to find a good source for for all COVID. Mostly I'm excited about the molding/casting. The various channels/sprues are difficult to design, and most sources are too general and unactionable. Yes I know I need them, now how do I compute them!

And if anyone has more sources, I'll happily take suggestions.

Someone should add KLayout to this list. Free software for chip design that is absolutely incredible given that it is a free piece of software...

https://klayout.de/

Where does one go to get a part fabricated, though? It isn't quite as easy as getting a PCB made.

MOSIS, CMP, CMC. Prices start around US$20k in theory but probably more like US$200k in most cases.

All the videos linked in the class show me a message "video is not rated", saying I am unable to watch?

Workaround: Downloading with yt-dlp gets around the age block (also works for YouTube). https://github.com/yt-dlp/yt-dlp

I wonder how one should host their educational videos to have them be available without ads or other nonsense like this. Any affordable solutions?

> without ads

and

> affordable

I think lots of smart people with lots of energy have spent lots of time trying to solve this problem with respect to getting customers and keeping them. It's a hard hard problem. Subscription supporting the free-with-ads IMHO seems to be the sweet spot.

PeerTube seems to work okay? More generally internet bandwidth is a lot cheaper now than it was 15 years ago.

you need to login into Vimeo, then you can watch

I would much rather not engage with the platform out of principle.

If I want to watch a makers video, why should I have to give information/identification, due to some idiotic monitoring scheme camouflaged as 'think of the children' regulation?

You will probably dismiss this as tin foil but it is exactly this inch by inch giving in that has put us all in this walled garden situation that you have on most of the internet at the moment.

What do you mean?

As in the video is not porn/violence rated and I am too anonymous for Vimeo to know if I am an inocent unsupervised child on the internet.

Gershenfeld helped coordinate a bunch of folks with COVID work too!

Reply on news.ycombinator.com