This is seriously cool. Yes it is "early" and I expect it to evolve but I like where they are going.
A number of boards I've done recently were essentially PCB wiring harnesses for a bunch of breakout boards. Why? Because it was easy and quick and I could get from idea to something that could be shown to work quickly. Also for a lot of things most of the circuits are already out there and its just a matter of putting things together.
Things I would like to see in the future:
1) A "box" form factor so that things could easily be put into a chassis. Ideally the box would be an STL file so you could print it to the height as needed.
2) A "ui" mode which is basically the board design tools but with all indicators and controls, and a constraint that once you have everything laid out the way you want it has a single wiring harness between it and that "main" board.
This is also something that Crowd Supply could knock out I suspect if they were inclined to do so.
Have you seen the M5 stack and its various add ons? Its absolutely hit my sweet spot of quickly prototype but also very clean and professional looking.
I had not, but looking at it I can see the appeal.
Typically I use Cortex-M4's and Cortex-M7's, in part because I can stick 16MB of DDR on them pretty easily and for the application space that is important.
The number one question we’ve received about ALC is: “Why can’t I move the blocks around?” While we could absolutely design a tool that allows you to poke and prod down to the mil, there are already piles of tools that do that
I tried to spec out a quick small handheld console with two analog joysticks and four buttons and it was basically impossible to make something useful because of this constraint. I understand auto-layout for non-input components but it makes input useless except for complete debug components.
It's a huge problem for sensors and connectors too. Both can be subject to physical constraints, like "the temperature sensor needs to be located away from components that generate heat" or "this connector goes on the back".
Strangely close to the overall concept and selling points from my project a few years back. I talked with Nate in depth about it as I was getting ready to launch it and he seemed super interested, even going so far as to post it in their blog. Guess he really was interested.
They justify it as they have spent a lot of money in R&D, but I feel almost $1,000 on top of the manufacturing charge is pretty high given there is almost no human cost/marginal cost for each new order. They then charge an extra $150 for the source files. Every iteration is considered a new design so the fees would rack up.
But on the other hand they seem to guarantee the board is electrically sound.
Well, I tried selling to the hobby market in the $150-$200 range because I figured hobbyists would be more forgiving if it didn't have advanced features. The market was completely uninterested (which may have more to do with my lack of marketing skills at the time, but I think there is still some truth that you can't immediately go for low cost when doing something like this). After reflecting on how mine turned out, I agree that the fees need to be higher. There is always room to come down later once the software is more mature.
Setting a low price puts you in the hobby market, and hobby engineers are some of the most penny-pinching people out there. They will spend weeks working on something they could buy for $30 and at the end say they saved money.
If you are selling to businesses, those problems mostly go away and you can get real money as long as you make sure you deliver stuff that works, and in the time you said you would. Having an engineer make a new board revision will easily cost more than $1000 for anything but the most trivial boards.
I have a small side business writing arduino code and building simple controller prototypes. A lot of the code I write is in that price range and is written as I sit in front of the TV. The most interesting thing I've discovered is that hobbyists don't mind paying for code, but they really don't want to pay anything but the lowest possible prices for hardware even after it's been explained why the dirt cheap Arduino peripherals are often crap and spending just a few dollars more for better designed stuff is worth it. AliExpress has firmly anchored the price of hardware at close to $0.
The only ones who take me up on my offer to write all the code and assemble the hardware are businesses. Hobbyists will struggle for weeks to save a few bucks.
My best customers are small or micro businesses who need a simple thing automated but the project size is below the threshold that engineering services firms want to deal with. Many of them are engineers themselves who just want to farm the task out and get it over with.
My original idea was that I saw a hole in the market that could be filled by code on the level of PLC programming, but on more flexible hardware. I'm still sure that the hole exists, but it's not clear just how scalable my approach is.
I ordered from Sparkfun multiple times and everything worked out great. I wonder if they are slowly going to expand their reach to be a la protolabs for all the B2B PCB design/prototype deals. Protolabs did a quarter billion sales last year I think.
Clearing $1k from semi automated work sounds a lot less tedious than packing 500 bags with a $2 sensor. I always feel like they lose money on sub $10 orders.
SparkFun staff answered my questions about hardware thoroughly, probably a few lectures worth within 5-10 replies. That moment is always in my heart! I don’t do hardware much though, I was curious about making a ring. Litho.cc made something similar to what I wanted to create.
I assumed there would be some "person behind the curtain" to this, as well (design checking, touching up connections, making the board look a bit nicer).
IMO this is not a service to be used to make shippable product. It is basically an advanced reference design tool. They do the work of making all the right component connections and getting everything on a board so that you can focus on software and basic functionality testing.
If your alpha design works as intended, and you realize you didn't leave anything out or make other simple common mistakes you can get the $150 files that you can import into your design system and then nudge components around as desired. Or, you can send it off to a more boutique board house to layout, knowing that you are not going to have to come back later and make major disruptive changes.
I have prototyped a device that consists of 1 uno WiFi, 4 megas, 16 8 channel relays, and 104 actuators. How do I go about finding a “person behind the curtain?”
I have no idea if what I built even makes sense. It works and it’s the way I was able to solve each problem I faced. I could use a service like this (I think? Not clear to me) because I’d love to have this thing be on a board instead of a massive ball of yarn. Is this a service to accomplish this?
If I have to hire someone, what profile am I looking for? EE seems too generic
To me at least, it seems priced out of reach if you want just one or two of your "massive balls of yarn", but if you've tested/validated your electrical design with your yarnballs and want another 20 or 50 of them, it becomes much more attractive.
I suspect there's a sweet spot somewhere in there 20-1000 unit range, where this'd be "the right choice".
Fewer than that, the time spent assembling them myself is not worth a grand US (to me, for hobby stuff at least).
More than 1000 I'd be leaning much more towards a bespoke design/layout and getting them manufactured in China (which the proviso that "Hardware is Hard!" and that it's a risk getting the right place over there to work with. Been there, done that, keep telling myself "Never again!!!" - until next time a really great sounding idea appears...)
In between, and where the constraints around exact layout/packaging are not a showstopper, I'd jump on this.
Thanks for the feedback. My application is low volume, probably would order in runs of ~100 depending on pricing. But it is revenue generating so I’m not swayed with these dollar amounts. I’m already at ~$300 before the actuators and my impression is that I could get that down a bit by improving the design. I’m also concerned with serviceability and I’m not great at soldering. I’ve come this far on breadboards.
Seems like I might hire a design service then utilize JLCPCB for production. I can’t tell if this sparkfun service is essentially combining those two actions. Although I think the DIY requirement is still a bit high based on my interpretation of the service description. I’m in a spot where I want to show someone what I’ve built and let them be a professional/specialist to turn it into a real device/controller
The ESP32 module being away from the edge of the board, with the ground plane completely surrounding its antenna, is the first thing that jumped out at me. Typically RF modules like that are intended to be placed overhanging the edge of the board, or abutting the edge with a large copper-free area to the left and right of the antenna.
I don't think that's indicative of an unfixable issue per se, but it's not a great look in the promo photo.
Yep, similar sentiments from me. And the ESP being away from the edge/enclosed in copper? Very easily fixable if this is done by an algorithm. I would think the micro block is the first thing being place anyway.
Not only is a design fee of $949 pretty high, but they're also upcharging significantly for the components they're placing. They charge $20 for an ESP32, for instance; the modules are $2.99 at Digi-Key.
Where are you seeing them on Digi-Key? The Microcontrollers section[0] only has other μCs with 32 in the name, and I couldn't find anything other than devkits.
I was hoping for something more along the lines of Seeed's Fusion PCB/PCB assembly service where you could get a quick prototype board with parts from a list of available components. I'm impressed by Sparkfun's software though and I hope that as the service evolves it gets cheaper.
Hi! Thanks for the comments - we are so excited to get this into the wild. If you plan to try it, be sure to use the promo code ALCSPARKFUN50 for 50% off the design fee.
I think you are asking exactly what efforts are put into the board once someone designs it in ALC. To be honest: it primarily goes through the same general steps as most of our boards - ALC Operations Team designs it; PCBs are ordered; once PCBs are received, it goes through the pick & place, reflow, inspection, and packaging; finally, the boards are sent using our standard fulfillment/shipping. Does that help?
I was more asking about "ALC Operations Team designs it" - the obvious goal would be for you to automate as much of this as possible, I just wondered how much work (if any) has gone into that so far, even if it's not being used yet on customer boards.
What is the license for the finished pcb design if you use this service?
I believe most Sparkfun designs are "Creative Commons Share-Alike" [0], but they are a little confusing as to their interpretation of that license [1]. My reading of that license is that if I base my commercial product (consumer, not for the hobbyist market) on one of their designs but make changes to the pcb design (such as additions and the shape) I would have to release my design to the public (which I would not want to do)?
But here [1] they seem to contradict that and suggest all you have to do is attribute some of the design to them?
I ask because this seems to be a really good way to get to a mk1+n prototype but obviously before going to mass production you are likely going to want to make changes, potentially large ones. If your final (very different) design has a lineage back to a CCSA licensed design then technically your final design would have to be shared for free.
Obviously the CCSA license is important for them to protect their core business, the hobbyist market, and ensure competitors don't just steel their designs.
Just to add to this having had a play with the service, for it to be useful more precise positioning of key components such as connectors/buttons and the mechanical design of the PCB (shape and fixing points) would be needed.
If I am paying nearly 1000$ for a prototype PCB it would be to put into a product prototype with mechanical design considerations. It’s that intermediate prototyping stage, after breadboarding but before full design for manufacturing. A form and fit prototype to test ux and consumer fit.
Who are you finding who can give you a PCB for a form and fit prototype from a list of preassembled breadboards you stuck together, for just 1k?
Afaik, 1k wouldn't even pay for someone to take your ball of breadboards and figure out what the a pared down component list looks like
This is great for taking a maker's prototype of multiple boards hooked together with jumper wires and random connectors in a project box... and giving them a single board that's still meant to be used in an exposed state, but is "moderately scalable". No where near being ready for "form and fit"
From the article:
> Cut and stripped wires, point-to-point soldered breakout boards, anchored everything to the proverbial breadboard, and after four hours sat back to realize you only had five more devices to build?
> Found yourself with a great idea, but unsure of how to bring it to life?
> Had an amazing prototype but needed to create a demo ready proof-of-concept?
Proof of concept is well before what you're talking about
- The copyrightability of PCB layout designs is somewhat ambiguous since they are ostensibly purely functional. (I'm sure someone can chime in here with precedent.)
- If Sparkfun created the original CC-SA design, CC-SA does not prohibit them as the copyright owners from later licensing their work under different terms as well.
So I have just done a little reading. It seems a circuit itself cannot be copyrighted, but a visual representation of it can be. It can however be patented.
I think this means that sparkfuns files are copy-write with a CCSA license. However the circuit design itself is not.
You are correct. A circuit design cannot be copyrighted.
A PCB design can be, however. The PCB design is the actual layout of the copper traces on the board. In the industry, it is referred to as "artwork" and is copyrightable as such.
IC production masks have their own special category of IP protection, "mask work rights", precisely because it was uncertain whether they could be copyrighted.
It does look interesting, although the $949 design fee is steep. In all honesty, you can get pretty far with KiCAD, some data sheets, Google, OSH Park & DigiKey. Basically the cost to you is your time & beer money.
Having said that, to go from prototype to production you need to have a professional engineer properly (re-)design the board (e.g. so it doesn't catch fire or generate interference).
From a different perspective: I was looking into having this done at my company in 2015 and the cheapest startup cost I could find was $10,000 (someplace in S.E.A). We have a few dozen customers that we supply custom boards with, but we send them Gerber file and a DigiKey BOM, this is within the realm of making some key customers happy without having to wince at our budget sheet.
I'd be one of those customers. $1k versus $10k means I stop spending time on the weird boards for manufacturing fixtures, and get back to working on something that my customers actually care about at the end of the day.
To my knowlege there is no practical way to automate placement AND routing of a PCB design (so basicaly only the a netlist as input). Yes, there are autorouters, but they work under the assumption that there is already a sensible placement of the compontents. Is this still true?
At least Mentor Graphics & Altium don't seem to have any solutions for this.
Not sure if it's true, but SparkFun's platform still appears to require manual design, just it looks like their staff will do it for you. To test-drive it I selected an ESP32 with a humidity/temperature sensor, and it quoted a board cost of ~$35, with a one-time ~$1k design fee.
So presumably they give the component list to a real human who does the board layout for you.
It can be automated if your requirements are limited and your expectations are low. I think that's the approach SparkFun are taking here -- aside from power, the only connections needed are between the components and the microcontroller, so placement is kind of arbitrary, and they leave themselves enough room that even a terrible autorouting job will work.
Per board fee - priced per component including installation
It seems like you are paying someone to manually design the board and link any components to the controller. I guess it's limited to only simple designs where each component just connects to the controller. I like how the software keeps track of pins and power available
> Per board fee - priced per component including installation
"over-priced per component" might be a better description.
As someone else mentioned, they charge you $20 for an ESP32 "block". That's a 33% markup on their ESP32 MicroMod board, and about 1000% markup on retail pricing for the ESP32 bare component. It's compatible price wise if your spiders nest hairball has a whole ESP32Thing in it, but there's almost certainly gonna be doubling up on connectors, pushbuttons, USB ports/controllers, voltage regulators, etc - that it'd be nice if the ALC magic at least kept the component price as low as the least expensive version they'll sell you as a module/breakout board.
Great to see this project go public! Seems pretty good if your requirements fit inside the spec.
$949 is a decent chunk of change, but it is half of what the very similar Geppetto service costs and Sparkfun seems willing to also sell you the native CAD files (for an extra $150).
Very curious to see how much real use this will get.
I'd guess the price is initially also to give them room for review and as a rate-limit. Do fewer projects guaranteed well at first, adjust price with experience and capacity.
Even then, it was pretty limited: 100x80mm board size maximum, only 2 layers, 2 schematic pages, for personal use only, but Kicad was pretty hard to get configured. Altium, Cadence etc. were still the industrial behemoths.
But now Eagle is among the behemoths, owned by Autodesk. It requires a subscription and an Internet connection, the subscription is free for one year, and then costs $495/year to continue to get access to your files on Autodesk's cloud. One hazard I remember from projects at school in 2010 was that the school had an education/commercial license from Cadsoft (the original manufacturer) and you were essentially unlimited in the schematics and PCBs you could create. However, if you created a design on your laptop in the freeware version, opened it with a school computer using their paid license, you could no longer open it on your freeware license - that's a pretty old bug (or feature, depending on your definition. I trust that when Sparkfun sends back Eagle files, they're able to be opened by people with a "Fusion 360 Personal" subscription Eagle license.
Kicad is much, much easier to use today, especially after the big development push from CERN. And it's open source, so whether Sparkfun would want to extend it with Python scripting, write a plugin, or modify the software itself, something like this could be trivially automated.
The choice of Eagle seems to be obsolete by half a decade, when you're looking for a open-source-hardware electronics CAD package, especially if you'd want it to be scriptable and extensible, as I'm sure Sparkfun ALC is (I can't imagine they're paying someone in Colorado to route traces for $950). On the other hand, there's a lot of institutional inertia when all of Sparkfun's early designs were created in Eagle. Sparkfun is one of the big drivers of hobbyist Eagle adoption. I wonder if any early discussions of how to implement ALC were guided by an in-house Autodesk rep...
Both have an autorouter that does a reasonable job, though many PCB designers scoff at using one for most tasks. Kicad has a nice push-and-shove manual router, which Eagle didn't, though looking back I see they added one a few years back.
Both placement systems just drop all your parts into a grid sorted by reference designator, Eagle doesn't (didn't?) have automatic placement optimization for minimizing ratsnest length, Kicad does have that, plus various plugins to lay out your board like the schematic, auto-place similar groups of components from the schematic hierarchy, etc. Both have trace length matching and can add a meander to reach a given length automatically, Kicad has a built-in impedance calculator.
A critical difference is that Kicad isn't limited to 2-layers, and for small-run high-speed digital work a 4-layer board makes a lot of sense these days and significantly simplifies routing. When all your power and ground signals and decoupling caps just go to a via and then to the internal plane, there's a lot less work to do, and a lot more space in which to do it!
Just mentioning for those interested, but dissuaded by the $949 design fee comments in this thread, SparkFun is providing a temporary discount:
> To encourage you to give ALC a try, we’re offering 50% off the design fee on all new orders through 11:59 p.m. on 12/31/2020 with promocode: ALCSPARKFUN50
Disclaimer: Not at all affiliated with SparkFun, but I'm a fan of fun with sparks (magic-smoke, !smokey-the-bear).
I'm really wondering who this is aimed at that a 50% discount would make any difference. As an individual hobbyist, this is a service that provides maybe $50-100 worth of value to me at most. If I was looking to start some kind of businesses in my garage, my very first priority in putting the business together would be making sure either I or a partner knew how to do this ourselves, so the value add would be ~$0. If I'm a cog in some big corporation, I can probably put $949 on a corporate credit card just as easily as $475.
Probably researchers working with grant money, or people in corporate R&D with tight deadlines, but not so tight budgets.
They also save you from having to learn how to actually design the board yourself, so if you don't know how to do that with modern tooling, then you can pay $1000 for someone else to do it per board, rather than hire someone out of your project budget to do it.
The $950 design fee seems steep now, but it's totally plausible that they'll be able to automate the board layout, so we can expect this number to drop relatively quickly once competitors apepar, towards the marginal cost of providing the service (which would be basically 0).
One potential market this seems to corner is software engineers without much hardware background willing to dip their toes in product design or proving some concept for a startup (think of adding a touch of interactivity to museum exhibit or kickstarting another thermostat company).
With current PCB production prices I also sometimes just order a one-off PCB that wires together a few modules, as it's just more efficient use of time.
OT: my snarky side would present this news as: geez, JavaScript developers will get their way and be able to get a solution that has sensor X and LED on it without learning anything about hardware :)
But it's a good thing- tinkering with Arduino when it just came out did set the course for career change I wanted, and maybe this will do the same for someone else.
I reckon another potential market will be Burningman-style art/hobby projects. The number of times I've been sat in camp re-soldering broken shit where if it'd been properly assembled it wouldn't have been nearly as likely to fail... If I wanted to take 50 or more of something out to the Playa, I'd totally pay up just for the increased reliability.
I'm confused on the wording about the design fee. Is the design fee for purely design so the manufacturing setup cost is embedded in the margin of the PCBA or is the design fee design + manufacturing fee combined?
My immediate thought would be how cool it would be to make programming/test jigs with this service. Doesn't seem super feasible if you're not allowed to place components, though.
> Geppetto is a ground-breaking cloud-based design tool that allows anyone to design electronics, you don’t need to know about resistors, capacitors, inductors, routing paths or anything EE to design your own board. Seeed has embedded the Geppetto electronic design tool into its platform, so users can go straight from creating a design to manufacturing in one session.
> Geppetto make it possible to create an custom Arduino board for under $50, including design to pcb manufacture and assembly.
I haven't looked at both closely, but it looks like ALC is a broad component design service, and Geppetto is constrained to a set menu of components that readily fit together.
No cortex-M microcontroller will run Linux without several megabytes of external RAM, and their lack of an MMU makes it difficult even then.
You can program them with normal C code or thumb-2 assembly, but the peripherals on SAMD chips aren't very intuitive if you want to program them bare-metal.
can't comment on the specifics of their implementation but the idea behind it is a great one. I currently have a pretty impressive project that is still trapped and works in-breadboard and I want to make a custom arduino mega shield out of it -- but I need more activation energy or a lower bar, I this project helps lower the bar.
I can find no details on the site about where the board manufacturing and assembly happens. Simply from a standpoint about caring about labour practices, and workers rights, this is concerning. Is this happening using US prison labour? Chinese labour under a government that is actively committing genocide against the Uyghurs? I hope not, but with that info obscured, there's no way to know if using this service passes a basic ethics smell test.
There is likely quite a bit of extra logic on top of what a traditional autorouter does and a human in the loop to correct / QA what the software does as well as a sort of insurance in case a built board fails and has to be reworked.
1)Marketing stunt: `For businesses, this is much lower than you’ll find with comparable do-it-yourself board designers, definitely less expensive than hiring a contractor or employee to design the board for you, and clearly more efficient than trying to do it on your own. We’ve been using it to run new product ideas and it’s amazing how much it saves us in schematic and hardware mistakes.`
Let's start from this.
Do you guys really serious about positioning yourself VS significant part of your customers base?
So you mean you make service that supposed to make EE unemployed and you think EE will keep buying your stuff and sponsoring your "activities"?
2)While it is fun and i enjoyed Sparkfun modules for prototyping, i would not consider them to second stage hardware development for functional prototype.
Let's talk about your MOSFET "block". You claim it is 50A, but thats it. No MOSFET model, means you can't really know absolute maximums, you can't know required Vgs to open it properly so it wont release smoke on load. Does it have any ESD protection? It claims to drive solenoids, but i dont see necessary protection for that.
And where gate pin is routed, to microcontroller? Or i need to break out header? If on MCU - which GPIO? How the hell it will be controlled?
I am supposed to play lottery and checkout (after many warnings of "NO REFUND"), pay $949 (or discounted 50%) and then hope for best?
And terminals offered in this block, looking to similar on your modules - can handle max 11A, nothing close to claimed 50A.
3)This "autoplacement optimized thingy that you can't change" placed GPS module right next to the MCU with Wifi /PLL. Deaf GPS module is no good.
Worse, on demo picture you provide on your own website, i can see module with RF antenna in the middle of board surrounded with solid polygons. This is embarassing.
Also i was able to add 16 potentiometers, while ESP32 have only 10 usable ADC inputs (second ADC will be off as soon as wifi turns on). This means customer dont really know where they will be routed even single one and chance that even single potentiometer will be connected to ADC input that wont work together with wifi.
Sure i am going to buy a cat in a sack and hope i can figure out how to make such board work.
And as you claimed in some tweets, "you did research" and you are positioning your device for stage on burning man, replacing $20k consultant.
Well, news for you, most likely in this $20k most are paid for not screwing up stage performance. And your ToS pretty clear about liabilities.
And if someone really want "constructor", everybody know there is way more superior service, Geppetto from Gumstix, way more flexibility and choice, for way more complex boards.
Your market research is ... bad.
P.S. Sorry for the angry post, but i cannot be kind when someone claims that he wants to take job away from my colleagues and probably me.
Instead of cooperating with industry engineers, you claim from the start that you are their competitor.
You literally shot yourself in the leg.
A number of boards I've done recently were essentially PCB wiring harnesses for a bunch of breakout boards. Why? Because it was easy and quick and I could get from idea to something that could be shown to work quickly. Also for a lot of things most of the circuits are already out there and its just a matter of putting things together.
Things I would like to see in the future:
1) A "box" form factor so that things could easily be put into a chassis. Ideally the box would be an STL file so you could print it to the height as needed.
2) A "ui" mode which is basically the board design tools but with all indicators and controls, and a constraint that once you have everything laid out the way you want it has a single wiring harness between it and that "main" board.
This is also something that Crowd Supply could knock out I suspect if they were inclined to do so.