10 ms·

> It costs $10,000 USD to build one of these > The ten thousandth one costs $15 As a software engineer, I'm used to being able to build essentially whatever I

by miav 5y ago

> It costs $10,000 USD to build one of these > The ten thousandth one costs $15 As a software engineer, I'm used to being able to build essentially whatever I want, with the only cost being my time. I'm often unpleasantly reminded that software is an outlier in that regard and as soon as physical items are concerned, especially hardware, it turns out that manufacturing even a simple thing is prohibitively expensive if it's done at a small scale.

desireco42 5y ago
Yes that is true, but only thanks to huge reliance on OPEN SOURCE. :) Without it, no tools, no infrastructure, no nothing.
mastazi 5y ago
In the software world something that works roughly similarly (i.e. where a prototype is differentiated from production) is devops. If you are deploying a toy-project once, maybe you are not setting up a CI/CD pipeline, Terraform templates, Github team permissions, Github/Jira integration, etc etc. - you do all of these things only for "real" products. But yeah, going from prototype to production with hardware is much, much harder, I agree with you.
- mike_hock 5y ago
  And the macro architecture of the project. As a one man project, the only thing it makes sense to build is a monolith.
  - mastazi 5y ago
    Yeah as a solo freelance dev I couldn't see how stuff like microservices would remotely be helpful to me in any way. But a few years later I worked in a large-ish team where at some point we hit a wall due to the project being a monolith[1], then I realised what that stuff was about. [1] the project changed scope and we had to do a large refactoring of the codebase as a result, if it had been based on microservices the refactoring would have been an order of magnitude simpler. Luckily the monolith was relatively well architected with separated service providers so it wasn't super bad.
eyegor 5y ago
It's actually not too bad to order custom pcbs now ($10 or less), but you do get raked over the coals if you need quality components in small quantities. Individual bulk caps can be several dollars.
eternityforest 5y ago
You could probably build one at home for a few hundred if you really wanted and did not care at all about time. You would just need to 3D print a case and somehow cast your own silicone keypad. DIY electronics is actually not too bad cost wise.
- Gracana 5y ago
  I've been doing some test runs with HP multijet fusion prints, and I think that would be a great technology to use for case like this. You likely would need a couple-few tries to get it right, but it cleans up very nicely. I really don't like the look of FDM printed parts, so that aspect of it really appeals to me.
- outworlder 5y ago
  Keypad could probably be done with a flexible filament too.
- tdeck 5y ago
  You could probably design a very similar one of tbese wirh a standard enclosure ($5) tactile switches, and a custom PCB for the top. Custom PCBs are ridiculously cheap now and even pick and place has gotten affordable for small runs.
  - Teknoman117 5y ago
    You can even run 4-layer impedance controlled boards that support BGAs for ~$60 and get 10 of them. Problem is I now have a bunch of unpopulated PCBs I don't really intend to use. I made a few PCIe x1 to SMA adapters for an FPGA devkit - even building a single one and considering the rest of the PCBs as write-offs was less than a quarter the price of premade ones.
    - aidenn0 5y ago
      Can you verify good flow of solder on a BGA without an x-ray? Do you just reflow it for the prescribed time and see if it works?
adamddev1 5y ago
So true. I recently helped a friend build a little bit of a backyard carpentry project and I was struck by the fact that I got the same creative satisfaction from building little pieces of software. With one big difference: that 2-day carpentry project cost around $1000 in materials. People wanting to create these kinds of things have to pay a big price. Whereas with code (and lots of free-tier cloud services etc) I can bang out endless ideas and useful projects for free.
- moistly 5y ago
  The other big difference is that you generally can’t make mistakes in woodworking. Made a mistake in your code? No problem, you fix the bug. But not in woodworking. Cut a piece too short? You’re hooped. Drilled a hole in the wrong place? You’re hooped. Misalignment during glue-up? Hooped. Woodworking scratches many of the same itches as programming for me, with one key difference: it’s unforgiving of mistakes, so requires a lot more care.
  - lazide 5y ago
    Just wait until it’s machining! Wood working at least most of the time tolerances are +- 1/8”. Throwing away a thousand dollar part because you were .003” off is lame. At least if you have a grinder, welding can sometimes be more forgiving.
    - lkois 5y ago
      Oof, I used to work with one of the few workshops in the world that could produce small batches of parts down to 0.0001mm or something a bit ridiculous. Their pricing and supply times factored in a pretty significant failure rate, and we were basically asking them for highly precise screws and nuts for a couple thousand each.
      - andi999 5y ago
        For what application would one need such a precision?
krm01 5y ago
Really depends. I started a small (consumer electronics) hardware startup. Bootstrapped and producing small quantities. There are ways to make it work.
joemi 5y ago
Most of that "$10,000" is because of paying someone else for custom parts. If you use existing parts or make the custom parts yourself (if you have the ability/tools), it's less than $100. So the software equivalent would be if you oursourced a large chunk of your project to someone else, and in that case, I think $10,000 might not be too far off.
- 6510 5y ago
  Interesting take, I guess trees can be pretty cheap if you grow them yourself.
  - bee_rider 5y ago
    I'm fairly decent with clay, always wondered if I could make something like a raspberry pi case with sculpey. But not curious enough to ruin my toaster...
    - webmaven 5y ago
      > I'm fairly decent with clay, always wondered if I could make something like a raspberry pi case with sculpey. The short answer is yes you can, but it's a PITA. Among other problems you have to deal with shrinkage and drooping. I'd only use that approach for a quick and dirty prototype that just needed to hold parts together and survive rough handling. A better option is to build a case from parts you cut out of a sheet of plastic (anywhere from 1mm on up) that you then glue together. Once built, rounding the edges will make it a lot nicer to hold, but it will still basically be a box, and the glues you use are generally pretty toxic so don't breath in the fumes OR get it on your hands. More involved would be making the case by vacuum forming a heated sheet. The forms (one each for the top and bottom instead of the two you need for injection molding) are typically hand sculpted or machined from wood, filler and primer rather than metal, and vacuum forming is something a hobbyist can accomplish at home with a DIY setup. You can get a case nearly indistinguishable from an injection molded part, at least from the outside. The more complex the shape the more labor intensive the fit and finish will be (for example, getting the edges of the case halves to match up). However you make the case, there would be a lot of keyboard holes you'll need to carve out with something like a dremel rotary tool. It might be easier to make the case with a single large hole that you glue a thinner insert into that will be easier to cut holes in with a hobby or utility knife, but I've sliced my hands up a lot when doing this sort of work. A rotary tool seems scarier if you're not already familiar, but with as little as 30m of practice you'll probably have much better control than you would over a knife and lower risk of injury (as long as you use eye protection, etc. You really don't want the plastic shavings flying around getting into your eyes).
umvi 5y ago
> it turns out that manufacturing even a simple thing is prohibitively expensive if it's done at a small scale You can overcome some of the scaling costs with 3D printers. It's pretty cheap to make fairly robust plastic parts with a 3D printer if you can do basic CAD.
grishka 5y ago
I've seen people build one-of-a-kind custom electronics for themselves. There are PCB manufacturers that specialize on small orders with no setup costs (I think you can get as little as 3 copies?) and you can 3D print or laser cut the case if you need one. Though the thing that does stick out to me as a software engineer is that you need parts and thus can't iterate on your ideas nearly as quickly. You can't just pull in a library or something, you have to order a part and wait for it to physically get to you. And the cost of any mistake is in actual real money because wiring something incorrectly can destroy things.
- SethTro 5y ago
  https://oshpark.com/ https://oshpark.com/ is my favorite for low volume runs. You get 3 copies for $5/in^2
- allendoerfer 5y ago
  > And the cost of any mistake is in actual real money That's the same in software engineering. You are losing your own time. If you are just breaking your prototypes and not the machines to build them, your time lost is probably multiple times more valuable than the material. I guess, once you have billed a certain number of hours to clients, the relation between time and money just ingrains itself so deep inside your mind, that you might as well start telling the time in USD.
  - alexb_ 5y ago
    This is true, but it gets worse with hardware - it's not like your code fries itself because you forgot a character.
    - richardfey 5y ago
      But your customer's data can, with equally destructive results.
  - grishka 5y ago
    That's when you're working for money. If it's your own project, your mistakes cost you nothing but you learn something from every one of them.
hffftz 5y ago
Jenda_ 5y ago
Maybe it's including time invested into development? The F1C100s is $5 for one piece on AliExpress, all the parts seem to be solderable by hand/manual reflow in an hour (the passives can be populated by JLC for like $30 for a batch of 5 boards, saving your time), the board would be $2/piece for a batch of 5 at JLCPCB (seems to be 2-layer), the enclosure can also be 3D-printed on a professional printer at JLC for tens of $, the only thing I don't have experience with is the conductive-rubber keyboard.
zw123456 5y ago
I've been down this road myself a time or three. It's $200 for the parts BOM, PCB, $10K for my time. Then if you want to actually make a business out of it, another $500K for Lawyers, overhead, advertising, manufacturing contracts, headaches and so on. But, the computer is very cute, and nice work. I was halfway expecting YARPIP (Yet another Raspberry PI project) Very cool it was all done from scratch. Done that myself for a tablet prototype and it is a shit ton of work.
hugo82 5y ago
Amen. They used to say the first Space Shuttle launch of the year cost $3.5B, since that was the annual carrying cost of the non-unit costs. Got a lot cheaper after that, though never that "cheap".

by miav 5y ago

desireco42 5y ago

Yes that is true, but only thanks to huge reliance on OPEN SOURCE. :) Without it, no tools, no infrastructure, no nothing.

mastazi 5y ago

In the software world something that works roughly similarly (i.e. where a prototype is differentiated from production) is devops. If you are deploying a toy-project once, maybe you are not setting up a CI/CD pipeline, Terraform templates, Github team permissions, Github/Jira integration, etc etc. - you do all of these things only for "real" products. But yeah, going from prototype to production with hardware is much, much harder, I agree with you.

mike_hock 5y ago
And the macro architecture of the project. As a one man project, the only thing it makes sense to build is a monolith.
- mastazi 5y ago
  Yeah as a solo freelance dev I couldn't see how stuff like microservices would remotely be helpful to me in any way. But a few years later I worked in a large-ish team where at some point we hit a wall due to the project being a monolith[1], then I realised what that stuff was about. [1] the project changed scope and we had to do a large refactoring of the codebase as a result, if it had been based on microservices the refactoring would have been an order of magnitude simpler. Luckily the monolith was relatively well architected with separated service providers so it wasn't super bad.

eyegor 5y ago

It's actually not too bad to order custom pcbs now ($10 or less), but you do get raked over the coals if you need quality components in small quantities. Individual bulk caps can be several dollars.

eternityforest 5y ago

You could probably build one at home for a few hundred if you really wanted and did not care at all about time. You would just need to 3D print a case and somehow cast your own silicone keypad. DIY electronics is actually not too bad cost wise.

Gracana 5y ago
I've been doing some test runs with HP multijet fusion prints, and I think that would be a great technology to use for case like this. You likely would need a couple-few tries to get it right, but it cleans up very nicely. I really don't like the look of FDM printed parts, so that aspect of it really appeals to me.
outworlder 5y ago
Keypad could probably be done with a flexible filament too.
tdeck 5y ago
You could probably design a very similar one of tbese wirh a standard enclosure ($5) tactile switches, and a custom PCB for the top. Custom PCBs are ridiculously cheap now and even pick and place has gotten affordable for small runs.
- Teknoman117 5y ago
  You can even run 4-layer impedance controlled boards that support BGAs for ~$60 and get 10 of them. Problem is I now have a bunch of unpopulated PCBs I don't really intend to use. I made a few PCIe x1 to SMA adapters for an FPGA devkit - even building a single one and considering the rest of the PCBs as write-offs was less than a quarter the price of premade ones.
  - aidenn0 5y ago
    Can you verify good flow of solder on a BGA without an x-ray? Do you just reflow it for the prescribed time and see if it works?

adamddev1 5y ago

So true. I recently helped a friend build a little bit of a backyard carpentry project and I was struck by the fact that I got the same creative satisfaction from building little pieces of software. With one big difference: that 2-day carpentry project cost around $1000 in materials. People wanting to create these kinds of things have to pay a big price. Whereas with code (and lots of free-tier cloud services etc) I can bang out endless ideas and useful projects for free.

moistly 5y ago
The other big difference is that you generally can’t make mistakes in woodworking. Made a mistake in your code? No problem, you fix the bug. But not in woodworking. Cut a piece too short? You’re hooped. Drilled a hole in the wrong place? You’re hooped. Misalignment during glue-up? Hooped. Woodworking scratches many of the same itches as programming for me, with one key difference: it’s unforgiving of mistakes, so requires a lot more care.
- lazide 5y ago
  Just wait until it’s machining! Wood working at least most of the time tolerances are +- 1/8”. Throwing away a thousand dollar part because you were .003” off is lame. At least if you have a grinder, welding can sometimes be more forgiving.
  - lkois 5y ago
    Oof, I used to work with one of the few workshops in the world that could produce small batches of parts down to 0.0001mm or something a bit ridiculous. Their pricing and supply times factored in a pretty significant failure rate, and we were basically asking them for highly precise screws and nuts for a couple thousand each.
    - andi999 5y ago
      For what application would one need such a precision?

krm01 5y ago

Really depends. I started a small (consumer electronics) hardware startup. Bootstrapped and producing small quantities. There are ways to make it work.

joemi 5y ago

Most of that "$10,000" is because of paying someone else for custom parts. If you use existing parts or make the custom parts yourself (if you have the ability/tools), it's less than $100. So the software equivalent would be if you oursourced a large chunk of your project to someone else, and in that case, I think $10,000 might not be too far off.

6510 5y ago
Interesting take, I guess trees can be pretty cheap if you grow them yourself.
- bee_rider 5y ago
  I'm fairly decent with clay, always wondered if I could make something like a raspberry pi case with sculpey. But not curious enough to ruin my toaster...
  - webmaven 5y ago
    > I'm fairly decent with clay, always wondered if I could make something like a raspberry pi case with sculpey. The short answer is yes you can, but it's a PITA. Among other problems you have to deal with shrinkage and drooping. I'd only use that approach for a quick and dirty prototype that just needed to hold parts together and survive rough handling. A better option is to build a case from parts you cut out of a sheet of plastic (anywhere from 1mm on up) that you then glue together. Once built, rounding the edges will make it a lot nicer to hold, but it will still basically be a box, and the glues you use are generally pretty toxic so don't breath in the fumes OR get it on your hands. More involved would be making the case by vacuum forming a heated sheet. The forms (one each for the top and bottom instead of the two you need for injection molding) are typically hand sculpted or machined from wood, filler and primer rather than metal, and vacuum forming is something a hobbyist can accomplish at home with a DIY setup. You can get a case nearly indistinguishable from an injection molded part, at least from the outside. The more complex the shape the more labor intensive the fit and finish will be (for example, getting the edges of the case halves to match up). However you make the case, there would be a lot of keyboard holes you'll need to carve out with something like a dremel rotary tool. It might be easier to make the case with a single large hole that you glue a thinner insert into that will be easier to cut holes in with a hobby or utility knife, but I've sliced my hands up a lot when doing this sort of work. A rotary tool seems scarier if you're not already familiar, but with as little as 30m of practice you'll probably have much better control than you would over a knife and lower risk of injury (as long as you use eye protection, etc. You really don't want the plastic shavings flying around getting into your eyes).

umvi 5y ago

> it turns out that manufacturing even a simple thing is prohibitively expensive if it's done at a small scale You can overcome some of the scaling costs with 3D printers. It's pretty cheap to make fairly robust plastic parts with a 3D printer if you can do basic CAD.

grishka 5y ago

I've seen people build one-of-a-kind custom electronics for themselves. There are PCB manufacturers that specialize on small orders with no setup costs (I think you can get as little as 3 copies?) and you can 3D print or laser cut the case if you need one. Though the thing that does stick out to me as a software engineer is that you need parts and thus can't iterate on your ideas nearly as quickly. You can't just pull in a library or something, you have to order a part and wait for it to physically get to you. And the cost of any mistake is in actual real money because wiring something incorrectly can destroy things.

SethTro 5y ago
https://oshpark.com/ https://oshpark.com/ is my favorite for low volume runs. You get 3 copies for $5/in^2
allendoerfer 5y ago
> And the cost of any mistake is in actual real money That's the same in software engineering. You are losing your own time. If you are just breaking your prototypes and not the machines to build them, your time lost is probably multiple times more valuable than the material. I guess, once you have billed a certain number of hours to clients, the relation between time and money just ingrains itself so deep inside your mind, that you might as well start telling the time in USD.
- alexb_ 5y ago
  This is true, but it gets worse with hardware - it's not like your code fries itself because you forgot a character.
  - richardfey 5y ago
    But your customer's data can, with equally destructive results.
- grishka 5y ago
  That's when you're working for money. If it's your own project, your mistakes cost you nothing but you learn something from every one of them.

hffftz 5y ago

Jenda_ 5y ago

Maybe it's including time invested into development? The F1C100s is $5 for one piece on AliExpress, all the parts seem to be solderable by hand/manual reflow in an hour (the passives can be populated by JLC for like $30 for a batch of 5 boards, saving your time), the board would be $2/piece for a batch of 5 at JLCPCB (seems to be 2-layer), the enclosure can also be 3D-printed on a professional printer at JLC for tens of $, the only thing I don't have experience with is the conductive-rubber keyboard.

zw123456 5y ago

I've been down this road myself a time or three. It's $200 for the parts BOM, PCB, $10K for my time. Then if you want to actually make a business out of it, another $500K for Lawyers, overhead, advertising, manufacturing contracts, headaches and so on. But, the computer is very cute, and nice work. I was halfway expecting YARPIP (Yet another Raspberry PI project) Very cool it was all done from scratch. Done that myself for a tablet prototype and it is a shit ton of work.

hugo82 5y ago

Amen. They used to say the first Space Shuttle launch of the year cost $3.5B, since that was the annual carrying cost of the non-unit costs. Got a lot cheaper after that, though never that "cheap".