Re: 3D Printers

I think you have made a good point that there doesn't always have to be a specific need to buy a piece of equipment. For years I had the use of a workshop with a machinist; when he died and wasn't replaced, I bought a lathe and taught myself to use it. There was no particular job that required it but the lack of it meant that many small jobs had to be done in a roundabout way or adapted to use non-optimal off-the-shelf parts.

Of course the first year of ownership was spent in makng or collecting accessories for the lathe itself - then it became a really useful resource. Now I would use the lathe in the same way as I would pick up a pen or pencil (or soldering iron).

The purchase of task-specific items is driven by specific needs, the purchase of general-purpose tools is more speculative.

A lot of mills are CNC and the parameters for various tools are pre-programmed, so the programmer doesn't need to worry about them. Once a program has been written, the operator just loads and unloads the workpieces. The main difference between a mill and a 3D printer is that a badly programmed mill can do a lot of damage whereas a badly-programmed 3D prioter just makes a mess.

Printing is standalone - there's no 'computing power needed to drive it'. You need computing power to generate the files for it, but they are just files. You put them on the printer (wifi, SD card, USB stick, ...) and it goes. Many of the older ones run from 8 bit microcontrollers; more modern versions run from an STM32 or similar. Fancier ones have a processor running Linux inside them for network connectivity features.

Many printers have power loss protection so they will restart from where they were when the power went out. That may cause a visible line on the part (the previous layer of material was cold rather than warm) but otherwise should be good to carry on. They will also stop and prompt you to change filament reel when it runs out, which is a similar process.

As Liz says, once you have it as a tool you find uses where it's easier to

As long as you have a machine that 'just works', you can focus on the application rather than debugging the machine. You may well find your requirements aren't actually what you think they might be to begin with (eg a large format machine is useful, but if large prints take multiple days maybe you wouldn't in practice print anything too large?)

One thing you can do without having a printer is playing with the software like Orcaslicer, to understand the workflow. yeggi.com is a useful search engine of pre-made designs.

Exactly. Although watch UV stability for that application - PLA isn't, I hear ASA is better. Or paint it afterwards. Also I wouldn't be printing long runs of 4" pipe (very slow) but perhaps you can print the fittings and then add them to standard pipe.

Theo

you still need to know what tool to use for each operation, how to hold the part to able to mill it and it if you have to flip the part between operation which features to use for realignment

I would imagine a similar set of skills is required to programe a 3D printer unless you are just running someone elses program to make demonstration pieces. The mill is more flexible than the 3D printer, so you would need to use a larger skill set if you wanted to do all the things it is capable of, but making a trinket using a downloaded program is not going to need any more skill on a mill than on a printer.

There may be a few settings like temperature you can override on the front panel of the printer, but for most things you need to go back and reslice. eg printing slower might not entirely be a linear scaling operation - to avoid artifacts at certain points you do a lot of accelerating and decelerating and that might not scale linearly (will still work, but might not prevent the artifacts).

The generated gcode files describe the tool path so, while they're plain text, they're opaque in the same way a PCL file going to a laser printer is. If you want to tweak something you go back to the source document and reprint (reslice for a 3D printer).

No reason, as long as you know it'll be days before you get the result. And if it goes wrong you'll have to start over. Meanwhile you can't print anything else, unless you have a second machine.

Theo

Most of the things in that list were not available commercially. For the rest I could get something close, but not quite exactly what I needed, or I needed it ASAP and didn't want to wait for shipping.

I was responding to the "curios" part of your rant, not the "need" part. But most of the things I print are things I can't buy on Amazon, for example.

It had a custom oil fill, so yes. No, I don't know why. He couldn't buy one that fit. No, I don't know why that either. Car people can be funny that way.

Fridge part: a few cents worth of printing. OEM? A few hundred dollars to buy the complete module, individual parts not available, with the same badly designed part that would have just broken again.

Printing something because you need it now and you want to see if your design works... typically, in my case, leads to learning that your print is good enough and you just use it forever ;-)

I got one of those printers as my fourth printer, because my other three weren't big enough. I've got it configured with two different nozzle sizes and a couple different types and colors of filament, so it's always ready to print "whatever". It's nice to be able to print panels and boxes with the labels already in them, with supports that cleanly break away because they're a different chemistry.

For a custom LED display that takes advantage of these features, see Clough42's

The file is just too big and it's too difficult to map the bazillion of tiny things its commanding to what you get in the real world, unlike programming a CNC machine. The file is so big, in fact, that most printers support a compressed binary format *just* to get it to the printer faster.

Example: I just printed a small L-shaped bracket for my wire stripper (a stop so I'd strip a consistent amount off). The raw ascii gcode without comments was 1,522 commands. The biggest gcode I still have on disk is

So while you can (and I have) manually edit gcode, you really don't want to. It's like choosing to hack the postscript being sent to your printer instead of just fixing that typo in the word processor instead.

And no, most 3D printers do not support complex gcode constructs like looping or conditionals, so manually crafting gcode for them is tedious. Even for custom hacks you end up writing software to generate the gcode for you.

Prusa prints a lot of the parts that are used to make Prusa printers, on Prusa printers, so sometimes it does work even when you need 100k of them. It depends on how quickly you need those 100k, how often you plan on iterating the design, and what the capabilities of the final part are. how expensive the tooling is, etc.

Heck, I once printed an M14 nut for my electric motorcycle conversion because I didn't happen to have enough on hand and I needed to mount the battery box for testing. Then I forgot about it for a couple of months, during which I got the bike up to a few live tests. It worked *well enough* for its purpose (although I did swap it for a metal one when I realized, duh ;)

But, of course, we each have our own needs...

The point is that both PCL and Gcode are sufficiently low level that you would not want to edit them. You edit the input to the printer driver and 'print' again.

I started off with a 2019 machine that was basic by today's standards (CR10S Pro). It had a few upgrades from the base CR10 (2017) which made it a bit more usable. It worked fine for a bit, but started to struggle. I think this was partly due to a leaky extruder but also due to damp filament (I wasn't doing any moisture control, and often print in the cold and damp ~80%RH).

I did quite a few upgrades which improved things (bed sensor, hotend, CPU), but never quite managed to get things set up right - I think the new hotend was overpowered which meant it couldn't control extrusion accurately enough.

In the meantime the Centauri Carbon came along and it was cheap enough to just replace the whole unit, rather than spend any more time tinkering. This time I've also been trying to keep filament dry to address the moisture problems.

Thus far it's been a much smoother experience - no tinkering needed, which is what you want when it's a tool not a hobby. The Centauri has an annoying habit of chatting to the internet for no reason, so I've put it on an isolated VLAN and not upgraded the firmware - it works fine locally.

PLA is good and cheap, but not heat resistant PETG has more heat resistance ABS or ASA are resistant and require high temperatures, also emit nasty fumes TPU is rubbery - good for gaskets and flexible parts Nylon might be good for mechanical things like printed bolts or hinges

There are other filaments with things in them like carbon fibre, glass fibre, wood, ... which might be good for structural things. I haven't tried anything but PLA yet.

One other thing if you can print multiple filaments is you can use a water soluble filament for support structures (ie the 'scaffolding' by which you can print overhangs), which you can wash away afterwards. Although Orcaslicer prints supports in the base PLA which are fairly easy to break off so that's been sufficient for me.

Theo

have you ever used a cnc mill ?

I've repaired a few, does that count?

Print-in-place translucent patches are something modern 3D printers can do.

But I think they'd fall short of "clear". You can see light through them; you can't read through them.

Big CNC machines, like those running linuxcnc, have a crude form of variables and expressions in them, but big CNC mills tend to have smaller programs that are still useful.

It's like limiting Postscript to "moveto" and "lineto" and trying to print modern text with it.

does repairing an airplane count as being e pilot? :P

The way they work, they put down all of filament A in the layer, then all of filament B, etc, then go back to A for the next layer. Filament switches are considered expensive.

You would print light barriers (opaque) into the back of the panel to guide the light to where you wanted it.

Cheap 8-bit microcontrollers, plus no need.

Yes and no. Inter-layer strength is much below cross-layer strength, but the old layer is still hot enough that the heat from the new extrusion is enough to fuse them. The software adjusts temperatures to make sure this happens, so having it happen for intra-layer adhesion is a given.

For some prints and plastics, the problem is cooling it fast enough, not having it cool too fast. Most printers have part cooling fans. Printers and slicers are really good at managing thermal loads.

Yup, if they were up against the clear plastic.

Printed walls can be 1mm or less in width, and arbitrarily tall. My printer can print 0.25mm walls.

I don't think it's a material issue, unless you can find a plastic with the same reflective constants as air ;-)

It's so easy to regenerate a gcode file in the slicer these days, that the main factor is the "no need" part. Modern printers use fast 32-bit processors and nobody has bothered making the gcode smarter. Faster and smaller, sure, but not smarter. The parameters you want are in the slicer, just slice it again.

The non-smooth top would distort the image the further away it was.

Note: that test was on my first printer; I haven't re-tested with clear PETG on my newest printer with it's "ironing" settings, which are designed to smooth the top layer.

The test I did was for a lens, which implies some polishing and smoothing. What were were testing for was if the inside was "lens-y" enough.

I don't know.

Typically you'd ship the *model* to the service bureau, and they would slice it with recipies known to work well on their printers with their filament. They might even need custom gcode for their farm's automation.

It's rare to share gcode because everyone has different printer and filament combinations, but it does happen in forums where there's consistency. Bambu and Printables share gcode for their specific printers, for example, but sharing the model is much more common.