Seeking Your Thoughts: Speeding Up PCB Prototyping with a Desktop Device

[Thorsten-PCB]

Hey folks,

I’ve been mulling over a problem that I’m sure many of you have faced: the painfully slow process of getting PCB prototypes made. From sending out the design files to finally getting the boards back and then dealing with assembly, it can easily take 1 to 3 weeks—or longer if something goes wrong.

I’m considering starting a startup to tackle this issue head-on. The idea is a device that would sit right on your desk and handle the entire PCB prototyping process—from creating the board to applying solder paste, picking and placing components, and reflowing the board. The goal is to get factory-grade PCB prototypes within a single day, without the hassle of dealing with external suppliers or waiting weeks.

I’d love to hear your thoughts, especially on the following:

Do you or your company struggle with the slowness of going from design to prototype?
If so, how does it impact you? Missed deadlines? Delays in product development? Lost enthusiasm for the project?

Have you tried anything to speed up the prototyping process?
What worked, what didn’t, and why?

How would it help you if you could get a high-quality PCB prototype within a single day at an affordable cost?
Would it change how you approach your design process?

Feel free to rant on anything—whether it's the idea, the process, or your frustrations with prototyping. All feedback is valuable, and I’m particularly interested in hearing about your experiences and any tools or techniques you’ve used to try to speed things up.

[globoy]

I think, for any desktop device claiming to speed prototype development, the devil is in the details.

• What kind of PCBs can this device prototype? Layer count, feature size, materials, thicknesses, etc
• What kind of components can it place?
• How many different components can it handle at a time?
• How does the soldering process work?
• How much does it cost?
• Etc, etc.

IIRC years ago there was some "desktop PCB printer" that used a conductive ink or something like that to "print" single-layer boards.  I remember thinking that it solved exactly no problem I had because it could build nothing that I design.  Unless your device is pretty capable it will also solve no problem that I have.

[woody]

In the twenty+ years I design prototypes for one-offs and small series I have gone from breadboarding via soldering stuff to proto (perf) boards, then using a "spray etching" device and finally a decent €10k PCB mill.

All these techniques were lacking on one or more points. Bad connections, dangerous chemicals, under etching, over etching, UV, finicky processes, 2-layer trouble, 4 layer impossible, broken router bits, G-codes, no solder masks, no thru-hole (vias), very difficult to use 'smd-only"'components. And if a prototype made it into a product I had to do the design all over again and then still send it to a PCB manufacturer with all their intricacies. Tedious.

Meanwhile outsourcing PCBs became easier (just send your board file and use their wysiwyg viewers), cheaper and faster every year.

Five years ago I decided to throw in the towel, I crated the PCB mill (if anyone is interested: make me an offer, pickup only as it weights 40kg ) and started using a (then new) local (for The Netherlands / Germany) PCB outfit. I never looked back. I now design projects immediately in the 'final form'. Order a couple of boards with a stencil, wait for these to be delivered, populate one or two by hand and reflow them (vapor phase). From submitting the file to connecting the power supply can be done in under a week.

The week it takes to actually produce a board is no showstopper. I might be a bit slow but for me cooking up a solution and designing a board takes much more time. Components have to be selected and ordered, solutions need to be toyed with, more components need to be ordered, firmware needs to be written. I explain to my clients that some time is involved and they understand that. So no, it would not really help me if I had a device that made a complete proto in one day.

That said, the one thing I might throw some cash at in the future is a small and affordable PNP machine (think YY1). But what held me back so far is the feeling that setting this up for one board probably takes more time than good old hand placement. And the prices and lead times for outsourcing assembly at my board house have come down in the past couple of years, while setting this up gets more automated (easy) all the time.

[ataradov]

One of the biggest issues with prototyping is organizing and managing  all the components that come from different suppliers.

And in the end all your device would do is add one more component to manage. If you have a well established process, then 2-3 weeks for a prototype is not a huge deal. Usually you have a pipeline anyway, otherwise you will have a lot of dead time.

[Kim Christensen]

The idea is a device that would sit right on your desk and handle the entire PCB prototyping process—from creating the board to applying solder paste, picking and placing components, and reflowing the board. The goal is to get factory-grade PCB prototypes within a single day, without the hassle of dealing with external suppliers or waiting weeks.

Nobody's desk is that large.   
Looks like someone has already thought of your idea. (Not an endorsement. Pick and place only) You'd still need a warehouse of parts, reflow oven, fume extraction, make the PCB, etc...

[thm_w]

Unless you can think of some revolutionary way to print copper and FR4, it doesn't make a ton of sense.
If you have money for priority shipping, you can get PCBs in a few days. If you don't have money, it won't be possible to afford this machine anyway.

The Boarditto looks cool, but its just pick and place, and I doubt its well supported in terms of updates.

[Thorsten-PCB]

I think, for any desktop device claiming to speed prototype development, the devil is in the details.

• What kind of PCBs can this device prototype? Layer count, feature size, materials, thicknesses, etc
• What kind of components can it place?
• How many different components can it handle at a time?
• How does the soldering process work?
• How much does it cost?
• Etc, etc.

IIRC years ago there was some "desktop PCB printer" that used a conductive ink or something like that to "print" single-layer boards.  I remember thinking that it solved exactly no problem I had because it could build nothing that I design.  Unless your device is pretty capable it will also solve no problem that I have.

Thanks a lot for the feedback! Fully agree that the devil is in the detail!

I’m thinking of a device that can make high quality multilayer (up to 8 or so) boards and pnp components at least as small as 0201. Realistically, throughput would be limited and it would maybe take a day to make a fully assembled (multilayer) PCB.

[Thorsten-PCB]

One of the biggest issues with prototyping is organizing and managing  all the components that come from different suppliers.

And in the end all your device would do is add one more component to manage. If you have a well established process, then 2-3 weeks for a prototype is not a huge deal. Usually you have a pipeline anyway, otherwise you will have a lot of dead time.

In your experience, how long does it generally take for the components that you need to arrive after ordering them? Do you stock components or do you always need to order components for a new prototype again?

What do you mean by the device adding one more component to manage?

[jmelson]

LPKF made a PC board milling machine that cut grooves between the desired copper lands on rigid PCB stock.  Somebody else made a machine that used a spark to vaporize thin copper on a flexible substrate, and it cut traces in a similar manner to the LPKF, but in a raster method as a drum spun with the substrate wrapped around it.
 These are part of the process.  I'm not sure there is a way for one machine to do everything including part placement.
Jon

[ataradov]

In your experience, how long does it generally take for the components that you need to arrive after ordering them? Do you stock components or do you always need to order components for a new prototype again?

No idea, I don't do procurement. I know that typically assembled boards take 2-3 weeks from the order date. This is usually fine because there is usually other stuff going on and I probably won't get to them in 2 weeks anyway. Usually there is enough hardware around to start development. It may be previous iterations of the boards or dev kits.

This wait time was never a problem for personal or professional projects. Now, your magic device may change how things work, I don't know. It is hard to tell without seeing it in action and knowing the limitations. All previous attempts at automation like this ended up with abominations that are not useful for more than a blinking LED.

What do you mean by the device adding one more component to manage?

Who is going to be doing the work of setting up and maintaining that machine? I'm a firmware engineer. I do assemble my own boards for hobby projects, but I have zero interest in doing that for work. I would just ask hardware guys to provide working boards. And hardware guys have enough to do on the design side that they are not interested in populating the boards themselves.

So, this machine would require a separate person to run it. And this will eat into any savings you can have.

[globoy]

I work for myself so I do everything.  In general, I have about a 7 day turn time (order boards to debugging them).  Sometimes I have to wait for something outside my control and that period stretches out.

I echo a bit of what Alex said, I usually have things to do between ordering the boards/parts and wanting to turn them on.  In all my years and with all my projects I have very rarely needed faster turn times.  I can theoretically understand how a very fast (1-2 day) turn time might be interesting but I don't feel a immediate need.  In fact I'm old enough to consistently be amazed that I can order high quality boards from China on a Monday and have them show up on Friday.

I hand assemble some boards using a stencil and reflow oven (or even directly soldering the parts).  I have an outside service assemble more complex boards for me.  That's generally a 3-4 week period so I schedule the project accordingly (e.g. firmware development while waiting).

I have a stockpile of parts.  Some I've ordered because they are commonly used and buying in qty (e.g. reels) makes them very cheap.  Some are left over from other projects.  For client work I always order parts for each project and invoice the client.  Extra parts stay with the project until, at some point after it's done I decide I won't have to ever revisit that project and they move into the general pile.  Lately I have also gone to culling the stockpile of stuff I'll never use.

You mention supporting 8-layer boards and 0201-sized parts.  Those seem like pretty advanced goals.  In addition it seems that any useful machine would have to be able to support a lot of part types (I just finished a design that had 59 different part types on the top and bottom of the PCB).  Just based on my exposure to the industry I think any machine supporting those would be pretty complex and consequently pretty expensive.  Obviously the the more expensive the harder it would be to justify.

What's your target price for this machine (kitted out to a reasonable level for a development engineer)?

Also, I am curious, how to you plan to actually fab PCBs?  We have models for how parts placement and soldering machines can be made to work, but fabbing PCBs is a very heavy-duty industrial process.

[Doctorandus_P]

Many PCB factories have an option for delivering prototypes within 24 hour. Some companies get some extra use out of this round mudball by distributing the workload. They send out files for a prototype just before the work day ends, to someone who just starts their workday at another world part.

With machines like the LPKF you can get quicker prototypes, maybe half an hour of actual milling. (And there are other brands of similar machines too).  And for some companies, such machines are very useful. Think about RF design and validating antenna's or those integrated PCB elements / filters.

Apparently some high frequency stuff (such as 60+GHz radars) can not tolerated the "rounded" edges of etched PCB's, and the straight lines of milling are necessary.

But for me, it's mostly DIY. I have a dislike for low pin count microcontrollers such as the attiny 84. I much rather use the same microcontroller for all my projects. and use 30 or so pin devices even if I only need 4 I/O pins, simply because I have the bigger ones always on stock. If one of my projects would ever get mass produced (100+), then it's easy to reconsider a smaller uC and/or respin a PCB.

And of the jellybean stuff, I have a decent supply at home, from resistor assortments to Diodes / LED's and transistors. But whenever a "serious" project pops up it's head you still always have to buy specialized parts.

[phil from seattle]

The process of going from a CAD design to an actual board is pretty complex and requires 3 general machine operations (each of which has many steps within) - PCB creation, component placement and soldering (reflow most likely).  While the first two have some similarities, the third is completely different. But I don't think it would be easy to mix all three into a single machine. Especially one that sits on a "desktop".

As was pointed out above, much of the pick and place process is about inventory procurement/management. It is highly likely that you would be waiting for parts to arrive.  It only takes one part that you don't have and your project's completion time is gated by ordering/shipping time. Thus, ruining the promise of "push a button, get your populated PCB right away".

Your biggest competition is JLCPCB. I can get a populated PCB (PCBA) in about a week (sometimes as short as 5 days) delivered to my home (USA).  Cost range is 40 to 100 USD, depending on a number of factors (chip types, layers, board material, controlled impedance, surface finishes, ...). No need to have component stocks, dedicated space for your machine, maintenance.  And, you get a professional board that can be shown as a credible prototype to management/customers/investors and put into full production fairly quickly. I think it's hard to compete against that.

[tszaboo]

And hardware guys have enough to do on the design side that they are not interested in populating the boards themselves.

So, this machine would require a separate person to run it. And this will eat into any savings you can have.

Exactly, I have zero interest spending time assembling things at work.
In fact you would slow things down. If I have to spend 2 days on making a prototype, that's two days that is wasted from other parts of the project.
Not to mention, we not only test the schematics when making prototypes. The schematics is usually the least problematic parts for experienced engineers. It's an end to end test, where I need the correct stackup, the final assembler making the boards, using the final parts, hopefully without any BOM mistakes.

[Sensorcat]

I worked for a very big company with only a dozen customers, each of them with much more purchasing power than itself. When deadlines are missed in a project, the customer being upset, they tend to bypass the project manager, instead calling the managing director. He then gets threatend that, if the issue is not fixed by yesterday, consequences will be felt by the shareholders - which means that the managing director will get in touch with the CEO over the issue. The ideal 'end customer' for the machine proposed by the OP is such a managing director, about one week after the call, when all the ugly follow-up meetings have been. If chance is high that the proposed machine could be useful to save time, the cost of the machine and the boards made by it do not matter much in this scenario.

The good news for the prospect start-up end here, the problem is this: The board from this machine must have identical properties to the production board. 'Factory-grade' is not enough! It must be identical, not better, not worse, in anything that counts: geometry, mechanical, electrical, thermal properties. The reason is that the board will be used in many industries for a lot of tests from EMI to NVH. And if it has different properties in any way, it is useless for the testing process. If it covers only some tests, like functional tests on the developers desk, it is useless for the project, because the functional tests would have to be repeated later, with the 'real' board, effectively creating additional workload.

In consequence, the conductor of the board must be solid copper, because no other material except silver has enough conductivity. Since all kinds of printable mud called 'ink' have failed to provide the properties of a solid layer, the task must be to get a pattern into a solid layer of copper or silver that is attached to a dielectric. Any additive manufacturing is ruled out by the requirements at this point. Silver does not solve a problem with getting the pattern into the layer, so we can use copper also. The dielectric has to match the standard properties found in various kinds of FR4 or FR5 (mechanical, thermal, electrical, DC to GHz), which almost certain leads to the necessity to use standard PCB material. This compound material is also not suitable for an alternative path of processing. Conclusion: the machine has to mimic a standard PCB manufaturing plant to deliver PCBs that are like standard PCBs. Time-saving is reduced to the mailing of the PCBs from manufacturer to customer. If you have the money in a project, better put it into express service from a local supplier and maybe a courier.

For this business idea to work, at least one of my above assertions must be wrong. I'm curious about it. 

[Doctorandus_P]

There are already quite a lot of machines that attempt to combine various aspects of PCB production in a single machine. The video below lists a few of such machines. It's mostly small CNC milling machines, but also a conductive ink printer.




 I have not seen the whole video, but I think I've seen machines that attempt to combine it with solder paste dispensing, Pick & Place and maybe even an integrated hotplate for soldering. There have been quite a lot of attempts at machines like this.

[SMTech]

There are already quite a lot of machines that attempt to combine various aspects of PCB production in a single machine. The video below lists a few of such machines. It's mostly small CNC milling machines, but also a conductive ink printer.




 I have not seen the whole video, but I think I've seen machines that attempt to combine it with solder paste dispensing, Pick & Place and maybe even an integrated hotplate for soldering. There have been quite a lot of attempts at machines like this.

The Voltera is the one that got the most press and went from Kickstarter to a product with actual distribution. IIRC it did have lofty ideas about including some kind of silly placement option. The reality of that however is it would be completely pointless, its too small do do that in any useful way. Pick and place machines get pretty big because they need to be to fit feeders on, 1m^2 plus work area is about as small as they can get and still be useful. So quite apart form engineers typically not wanting anything to do with manufacturing (some don't even want to make minor mods themselves), the chances of it just being a desktop device are pretty slim, so they also have to want to give up a fair bit of space.

For those with the space and a suitably big budget, the top tier solution would be Nanodimension Dragonfly -> Essemtec all in one ->Vapor phase. Unlike other solutions I think this  also lets you make things you couldn't even do another way. You could even call it compact for what it does.

Our engineer(s) typically have multiple projects in development, at various stages, there is always something else to do on one of them.  The only times we have pressure over turnaround is for 3rd party designs where they are trying to meet a test lab time limit or similar. In those instances the requirement is for a production level build so some kind of DIY effort would not suffice anyway. In the UK a lot of designs will end up needing parts that can only be sourced from Mouser/DK which adds 3 days from a completed BOM anyway.

[RoGeorge]

I’m thinking of a device that can make high quality multilayer (up to 8 or so) boards and pnp components at least as small as 0201. Realistically, throughput would be limited and it would maybe take a day to make a fully assembled (multilayer) PCB.

A DIY, 8 layers PCB, desktop machine? 
That's not possible with a single machine, that's a factory.



In practice, you might need 2 factories:  one factory to manufacture the PCB, and another factory to assemble the components, both with many workers, and the process might still take more than a day for one assembled prototype.

If you lower down the specs to only 1-2 layer PCBs, you may DIY that at home in a weekend.  By the time you'll learn to do that in only one day, you'll know enough to design good PCBs from the first try.  At that moment it will be more convenient to send the design and order PCBs from a PCB house + contract an assembly house if you want the PCB populated.  Won't be very cheap, but will be way cheaper than designing the machine you wish.

[Doctorandus_P]

A DIY, 8 layers PCB, desktop machine? 
That's not possible with a single machine, that's a factory.

I previously missed the goals of 8 layers and 0201 part size. I would not say it's not possible, but those are tough goals to meet. I do believe it' would be madness tho attempt this without a lot of experience from previous projects. Machines such as the voltera and other machines shown in that video are doable if you have some decent tools (lathe, milling machine) yourself. But building your first machine that does much more then those machines is very likely doomed to fail.

Some brainstorming ideas:

One thing I am curious about and think about every now and then is a prototype machine that uses enameled wire for the connections. It could start with dual layer FR4, keep one layer for GND, and the other for milling pads and some connections, but the majority of the connections is then made by a CNC machine that puts enameled wires though the holes, and cuts and bends the ends. As far back as the early '80-ies or maybe even the '70-ies there were automated wire wrapping machines and also some very unconventional other automated processes.

And I guess that answers my ramblings too. All those machines went away. When electronics became much smaller and frequencies increased, prices of outsourcing also decreased, and things like impedance matching became more important. Back then the only alternative was many hundreds of man hours (form real humans, which do not work without a salary (except for russian soldiers) in wire wrapping and these days you have to compete with reliable PCB prototyping services, that also create a PCB that is the same as a production PCB, so it can be used for EMC testing etc.

Another quite different process that may (or not) be adaptable to "desktop size" is LtCC (Low Temperature Cofired Ceramics). I never looked into the exact details of how that process works, but I do know that is has a very good resolution and accuracy, builds up the product one layer at a time and can have many layers (over 50).

There are also other less mainstream processes that may be adoptable. For example there are metal core PCB's that have multiple layers of conductors laminated on top. I also do not know how those are made. Maybe you can get inspiration on how flex PCB's are made.

Using some double sided sticky tape would be an intriguing idea. First you start with sticky layer, and then you push in bare (not enameled) copper wire into the glue, and you cover it with the next layer of tape. Maybe it's doable, but it would be tough to get this to work reliably, especially on fine pitch stuff, and there is still the detail of how to do pads.

But whatever direction you take, at such resolutions it will become a quite expensive machine and being made for a small niche market, and that is not a good combination.

[jmelson]

Using some double sided sticky tape would be an intriguing idea. First you start with sticky layer, and then you push in bare (not enameled) copper wire into the glue, and you cover it with the next layer of tape. Maybe it's doable, but it would be tough to get this to work reliably, especially on fine pitch stuff, and there is still the detail of how to do pads.

There was a patented process called Multiwire, where a board with DIP pads was made, then coated with some thick uncured resin, and then fine magnet wire was laid down by a CNC dispensing needle.  Then, the board was baked in an oven to cure the resin.  Finally, holes were drilled and the board was plated through those holes like a typical 2-4 later board.  The drill bored through the wires and the plating process connected the holes to the wires.  It seemed pretty fantastical that this process actually worked, but we have some stuff at work made by this process.  Using something like #30 AWG magnet wire, it was possible to put a LOT of wires on a board that would require many PCB layers if done with traditional PCB technology.
Jon